diff options
-rw-r--r-- | net/core/Makefile | 1 | ||||
-rw-r--r-- | net/core/netfilter.c | 737 | ||||
-rw-r--r-- | net/netfilter/Makefile | 4 | ||||
-rw-r--r-- | net/netfilter/core.c | 216 | ||||
-rw-r--r-- | net/netfilter/nf_internals.h | 39 | ||||
-rw-r--r-- | net/netfilter/nf_log.c | 165 | ||||
-rw-r--r-- | net/netfilter/nf_queue.c | 273 | ||||
-rw-r--r-- | net/netfilter/nf_sockopt.c | 132 |
8 files changed, 829 insertions, 738 deletions
diff --git a/net/core/Makefile b/net/core/Makefile index f5f5e58943e8..630da0f0579e 100644 --- a/net/core/Makefile +++ b/net/core/Makefile | |||
@@ -12,7 +12,6 @@ obj-y += dev.o ethtool.o dev_mcast.o dst.o \ | |||
12 | 12 | ||
13 | obj-$(CONFIG_XFRM) += flow.o | 13 | obj-$(CONFIG_XFRM) += flow.o |
14 | obj-$(CONFIG_SYSFS) += net-sysfs.o | 14 | obj-$(CONFIG_SYSFS) += net-sysfs.o |
15 | obj-$(CONFIG_NETFILTER) += netfilter.o | ||
16 | obj-$(CONFIG_NET_DIVERT) += dv.o | 15 | obj-$(CONFIG_NET_DIVERT) += dv.o |
17 | obj-$(CONFIG_NET_PKTGEN) += pktgen.o | 16 | obj-$(CONFIG_NET_PKTGEN) += pktgen.o |
18 | obj-$(CONFIG_NET_RADIO) += wireless.o | 17 | obj-$(CONFIG_NET_RADIO) += wireless.o |
diff --git a/net/core/netfilter.c b/net/core/netfilter.c deleted file mode 100644 index 98cc61e79fea..000000000000 --- a/net/core/netfilter.c +++ /dev/null | |||
@@ -1,737 +0,0 @@ | |||
1 | /* netfilter.c: look after the filters for various protocols. | ||
2 | * Heavily influenced by the old firewall.c by David Bonn and Alan Cox. | ||
3 | * | ||
4 | * Thanks to Rob `CmdrTaco' Malda for not influencing this code in any | ||
5 | * way. | ||
6 | * | ||
7 | * Rusty Russell (C)2000 -- This code is GPL. | ||
8 | * | ||
9 | * February 2000: Modified by James Morris to have 1 queue per protocol. | ||
10 | * 15-Mar-2000: Added NF_REPEAT --RR. | ||
11 | * 08-May-2003: Internal logging interface added by Jozsef Kadlecsik. | ||
12 | */ | ||
13 | #include <linux/config.h> | ||
14 | #include <linux/kernel.h> | ||
15 | #include <linux/netfilter.h> | ||
16 | #include <net/protocol.h> | ||
17 | #include <linux/init.h> | ||
18 | #include <linux/skbuff.h> | ||
19 | #include <linux/wait.h> | ||
20 | #include <linux/module.h> | ||
21 | #include <linux/interrupt.h> | ||
22 | #include <linux/if.h> | ||
23 | #include <linux/netdevice.h> | ||
24 | #include <linux/inetdevice.h> | ||
25 | #include <linux/proc_fs.h> | ||
26 | #include <net/sock.h> | ||
27 | |||
28 | /* In this code, we can be waiting indefinitely for userspace to | ||
29 | * service a packet if a hook returns NF_QUEUE. We could keep a count | ||
30 | * of skbuffs queued for userspace, and not deregister a hook unless | ||
31 | * this is zero, but that sucks. Now, we simply check when the | ||
32 | * packets come back: if the hook is gone, the packet is discarded. */ | ||
33 | #ifdef CONFIG_NETFILTER_DEBUG | ||
34 | #define NFDEBUG(format, args...) printk(format , ## args) | ||
35 | #else | ||
36 | #define NFDEBUG(format, args...) | ||
37 | #endif | ||
38 | |||
39 | /* Sockopts only registered and called from user context, so | ||
40 | net locking would be overkill. Also, [gs]etsockopt calls may | ||
41 | sleep. */ | ||
42 | static DECLARE_MUTEX(nf_sockopt_mutex); | ||
43 | |||
44 | struct list_head nf_hooks[NPROTO][NF_MAX_HOOKS]; | ||
45 | static LIST_HEAD(nf_sockopts); | ||
46 | static DEFINE_SPINLOCK(nf_hook_lock); | ||
47 | |||
48 | /* | ||
49 | * A queue handler may be registered for each protocol. Each is protected by | ||
50 | * long term mutex. The handler must provide an an outfn() to accept packets | ||
51 | * for queueing and must reinject all packets it receives, no matter what. | ||
52 | */ | ||
53 | static struct nf_queue_handler_t { | ||
54 | nf_queue_outfn_t outfn; | ||
55 | void *data; | ||
56 | } queue_handler[NPROTO]; | ||
57 | |||
58 | static struct nf_queue_rerouter *queue_rerouter; | ||
59 | |||
60 | static DEFINE_RWLOCK(queue_handler_lock); | ||
61 | |||
62 | int nf_register_hook(struct nf_hook_ops *reg) | ||
63 | { | ||
64 | struct list_head *i; | ||
65 | |||
66 | spin_lock_bh(&nf_hook_lock); | ||
67 | list_for_each(i, &nf_hooks[reg->pf][reg->hooknum]) { | ||
68 | if (reg->priority < ((struct nf_hook_ops *)i)->priority) | ||
69 | break; | ||
70 | } | ||
71 | list_add_rcu(®->list, i->prev); | ||
72 | spin_unlock_bh(&nf_hook_lock); | ||
73 | |||
74 | synchronize_net(); | ||
75 | return 0; | ||
76 | } | ||
77 | |||
78 | void nf_unregister_hook(struct nf_hook_ops *reg) | ||
79 | { | ||
80 | spin_lock_bh(&nf_hook_lock); | ||
81 | list_del_rcu(®->list); | ||
82 | spin_unlock_bh(&nf_hook_lock); | ||
83 | |||
84 | synchronize_net(); | ||
85 | } | ||
86 | |||
87 | /* Do exclusive ranges overlap? */ | ||
88 | static inline int overlap(int min1, int max1, int min2, int max2) | ||
89 | { | ||
90 | return max1 > min2 && min1 < max2; | ||
91 | } | ||
92 | |||
93 | /* Functions to register sockopt ranges (exclusive). */ | ||
94 | int nf_register_sockopt(struct nf_sockopt_ops *reg) | ||
95 | { | ||
96 | struct list_head *i; | ||
97 | int ret = 0; | ||
98 | |||
99 | if (down_interruptible(&nf_sockopt_mutex) != 0) | ||
100 | return -EINTR; | ||
101 | |||
102 | list_for_each(i, &nf_sockopts) { | ||
103 | struct nf_sockopt_ops *ops = (struct nf_sockopt_ops *)i; | ||
104 | if (ops->pf == reg->pf | ||
105 | && (overlap(ops->set_optmin, ops->set_optmax, | ||
106 | reg->set_optmin, reg->set_optmax) | ||
107 | || overlap(ops->get_optmin, ops->get_optmax, | ||
108 | reg->get_optmin, reg->get_optmax))) { | ||
109 | NFDEBUG("nf_sock overlap: %u-%u/%u-%u v %u-%u/%u-%u\n", | ||
110 | ops->set_optmin, ops->set_optmax, | ||
111 | ops->get_optmin, ops->get_optmax, | ||
112 | reg->set_optmin, reg->set_optmax, | ||
113 | reg->get_optmin, reg->get_optmax); | ||
114 | ret = -EBUSY; | ||
115 | goto out; | ||
116 | } | ||
117 | } | ||
118 | |||
119 | list_add(®->list, &nf_sockopts); | ||
120 | out: | ||
121 | up(&nf_sockopt_mutex); | ||
122 | return ret; | ||
123 | } | ||
124 | |||
125 | void nf_unregister_sockopt(struct nf_sockopt_ops *reg) | ||
126 | { | ||
127 | /* No point being interruptible: we're probably in cleanup_module() */ | ||
128 | restart: | ||
129 | down(&nf_sockopt_mutex); | ||
130 | if (reg->use != 0) { | ||
131 | /* To be woken by nf_sockopt call... */ | ||
132 | /* FIXME: Stuart Young's name appears gratuitously. */ | ||
133 | set_current_state(TASK_UNINTERRUPTIBLE); | ||
134 | reg->cleanup_task = current; | ||
135 | up(&nf_sockopt_mutex); | ||
136 | schedule(); | ||
137 | goto restart; | ||
138 | } | ||
139 | list_del(®->list); | ||
140 | up(&nf_sockopt_mutex); | ||
141 | } | ||
142 | |||
143 | /* Call get/setsockopt() */ | ||
144 | static int nf_sockopt(struct sock *sk, int pf, int val, | ||
145 | char __user *opt, int *len, int get) | ||
146 | { | ||
147 | struct list_head *i; | ||
148 | struct nf_sockopt_ops *ops; | ||
149 | int ret; | ||
150 | |||
151 | if (down_interruptible(&nf_sockopt_mutex) != 0) | ||
152 | return -EINTR; | ||
153 | |||
154 | list_for_each(i, &nf_sockopts) { | ||
155 | ops = (struct nf_sockopt_ops *)i; | ||
156 | if (ops->pf == pf) { | ||
157 | if (get) { | ||
158 | if (val >= ops->get_optmin | ||
159 | && val < ops->get_optmax) { | ||
160 | ops->use++; | ||
161 | up(&nf_sockopt_mutex); | ||
162 | ret = ops->get(sk, val, opt, len); | ||
163 | goto out; | ||
164 | } | ||
165 | } else { | ||
166 | if (val >= ops->set_optmin | ||
167 | && val < ops->set_optmax) { | ||
168 | ops->use++; | ||
169 | up(&nf_sockopt_mutex); | ||
170 | ret = ops->set(sk, val, opt, *len); | ||
171 | goto out; | ||
172 | } | ||
173 | } | ||
174 | } | ||
175 | } | ||
176 | up(&nf_sockopt_mutex); | ||
177 | return -ENOPROTOOPT; | ||
178 | |||
179 | out: | ||
180 | down(&nf_sockopt_mutex); | ||
181 | ops->use--; | ||
182 | if (ops->cleanup_task) | ||
183 | wake_up_process(ops->cleanup_task); | ||
184 | up(&nf_sockopt_mutex); | ||
185 | return ret; | ||
186 | } | ||
187 | |||
188 | int nf_setsockopt(struct sock *sk, int pf, int val, char __user *opt, | ||
189 | int len) | ||
190 | { | ||
191 | return nf_sockopt(sk, pf, val, opt, &len, 0); | ||
192 | } | ||
193 | |||
194 | int nf_getsockopt(struct sock *sk, int pf, int val, char __user *opt, int *len) | ||
195 | { | ||
196 | return nf_sockopt(sk, pf, val, opt, len, 1); | ||
197 | } | ||
198 | |||
199 | static unsigned int nf_iterate(struct list_head *head, | ||
200 | struct sk_buff **skb, | ||
201 | int hook, | ||
202 | const struct net_device *indev, | ||
203 | const struct net_device *outdev, | ||
204 | struct list_head **i, | ||
205 | int (*okfn)(struct sk_buff *), | ||
206 | int hook_thresh) | ||
207 | { | ||
208 | unsigned int verdict; | ||
209 | |||
210 | /* | ||
211 | * The caller must not block between calls to this | ||
212 | * function because of risk of continuing from deleted element. | ||
213 | */ | ||
214 | list_for_each_continue_rcu(*i, head) { | ||
215 | struct nf_hook_ops *elem = (struct nf_hook_ops *)*i; | ||
216 | |||
217 | if (hook_thresh > elem->priority) | ||
218 | continue; | ||
219 | |||
220 | /* Optimization: we don't need to hold module | ||
221 | reference here, since function can't sleep. --RR */ | ||
222 | verdict = elem->hook(hook, skb, indev, outdev, okfn); | ||
223 | if (verdict != NF_ACCEPT) { | ||
224 | #ifdef CONFIG_NETFILTER_DEBUG | ||
225 | if (unlikely((verdict & NF_VERDICT_MASK) | ||
226 | > NF_MAX_VERDICT)) { | ||
227 | NFDEBUG("Evil return from %p(%u).\n", | ||
228 | elem->hook, hook); | ||
229 | continue; | ||
230 | } | ||
231 | #endif | ||
232 | if (verdict != NF_REPEAT) | ||
233 | return verdict; | ||
234 | *i = (*i)->prev; | ||
235 | } | ||
236 | } | ||
237 | return NF_ACCEPT; | ||
238 | } | ||
239 | |||
240 | int nf_register_queue_handler(int pf, nf_queue_outfn_t outfn, void *data) | ||
241 | { | ||
242 | int ret; | ||
243 | |||
244 | if (pf >= NPROTO) | ||
245 | return -EINVAL; | ||
246 | |||
247 | write_lock_bh(&queue_handler_lock); | ||
248 | if (queue_handler[pf].outfn) | ||
249 | ret = -EBUSY; | ||
250 | else { | ||
251 | queue_handler[pf].outfn = outfn; | ||
252 | queue_handler[pf].data = data; | ||
253 | ret = 0; | ||
254 | } | ||
255 | write_unlock_bh(&queue_handler_lock); | ||
256 | |||
257 | return ret; | ||
258 | } | ||
259 | |||
260 | /* The caller must flush their queue before this */ | ||
261 | int nf_unregister_queue_handler(int pf) | ||
262 | { | ||
263 | if (pf >= NPROTO) | ||
264 | return -EINVAL; | ||
265 | |||
266 | write_lock_bh(&queue_handler_lock); | ||
267 | queue_handler[pf].outfn = NULL; | ||
268 | queue_handler[pf].data = NULL; | ||
269 | write_unlock_bh(&queue_handler_lock); | ||
270 | |||
271 | return 0; | ||
272 | } | ||
273 | |||
274 | int nf_register_queue_rerouter(int pf, struct nf_queue_rerouter *rer) | ||
275 | { | ||
276 | if (pf >= NPROTO) | ||
277 | return -EINVAL; | ||
278 | |||
279 | write_lock_bh(&queue_handler_lock); | ||
280 | memcpy(&queue_rerouter[pf], rer, sizeof(queue_rerouter[pf])); | ||
281 | write_unlock_bh(&queue_handler_lock); | ||
282 | |||
283 | return 0; | ||
284 | } | ||
285 | |||
286 | int nf_unregister_queue_rerouter(int pf) | ||
287 | { | ||
288 | if (pf >= NPROTO) | ||
289 | return -EINVAL; | ||
290 | |||
291 | write_lock_bh(&queue_handler_lock); | ||
292 | memset(&queue_rerouter[pf], 0, sizeof(queue_rerouter[pf])); | ||
293 | write_unlock_bh(&queue_handler_lock); | ||
294 | return 0; | ||
295 | } | ||
296 | |||
297 | void nf_unregister_queue_handlers(nf_queue_outfn_t outfn) | ||
298 | { | ||
299 | int pf; | ||
300 | |||
301 | write_lock_bh(&queue_handler_lock); | ||
302 | for (pf = 0; pf < NPROTO; pf++) { | ||
303 | if (queue_handler[pf].outfn == outfn) { | ||
304 | queue_handler[pf].outfn = NULL; | ||
305 | queue_handler[pf].data = NULL; | ||
306 | } | ||
307 | } | ||
308 | write_unlock_bh(&queue_handler_lock); | ||
309 | } | ||
310 | |||
311 | /* | ||
312 | * Any packet that leaves via this function must come back | ||
313 | * through nf_reinject(). | ||
314 | */ | ||
315 | static int nf_queue(struct sk_buff **skb, | ||
316 | struct list_head *elem, | ||
317 | int pf, unsigned int hook, | ||
318 | struct net_device *indev, | ||
319 | struct net_device *outdev, | ||
320 | int (*okfn)(struct sk_buff *), | ||
321 | unsigned int queuenum) | ||
322 | { | ||
323 | int status; | ||
324 | struct nf_info *info; | ||
325 | #ifdef CONFIG_BRIDGE_NETFILTER | ||
326 | struct net_device *physindev = NULL; | ||
327 | struct net_device *physoutdev = NULL; | ||
328 | #endif | ||
329 | |||
330 | /* QUEUE == DROP if noone is waiting, to be safe. */ | ||
331 | read_lock(&queue_handler_lock); | ||
332 | if (!queue_handler[pf].outfn) { | ||
333 | read_unlock(&queue_handler_lock); | ||
334 | kfree_skb(*skb); | ||
335 | return 1; | ||
336 | } | ||
337 | |||
338 | info = kmalloc(sizeof(*info)+queue_rerouter[pf].rer_size, GFP_ATOMIC); | ||
339 | if (!info) { | ||
340 | if (net_ratelimit()) | ||
341 | printk(KERN_ERR "OOM queueing packet %p\n", | ||
342 | *skb); | ||
343 | read_unlock(&queue_handler_lock); | ||
344 | kfree_skb(*skb); | ||
345 | return 1; | ||
346 | } | ||
347 | |||
348 | *info = (struct nf_info) { | ||
349 | (struct nf_hook_ops *)elem, pf, hook, indev, outdev, okfn }; | ||
350 | |||
351 | /* If it's going away, ignore hook. */ | ||
352 | if (!try_module_get(info->elem->owner)) { | ||
353 | read_unlock(&queue_handler_lock); | ||
354 | kfree(info); | ||
355 | return 0; | ||
356 | } | ||
357 | |||
358 | /* Bump dev refs so they don't vanish while packet is out */ | ||
359 | if (indev) dev_hold(indev); | ||
360 | if (outdev) dev_hold(outdev); | ||
361 | |||
362 | #ifdef CONFIG_BRIDGE_NETFILTER | ||
363 | if ((*skb)->nf_bridge) { | ||
364 | physindev = (*skb)->nf_bridge->physindev; | ||
365 | if (physindev) dev_hold(physindev); | ||
366 | physoutdev = (*skb)->nf_bridge->physoutdev; | ||
367 | if (physoutdev) dev_hold(physoutdev); | ||
368 | } | ||
369 | #endif | ||
370 | if (queue_rerouter[pf].save) | ||
371 | queue_rerouter[pf].save(*skb, info); | ||
372 | |||
373 | status = queue_handler[pf].outfn(*skb, info, queuenum, | ||
374 | queue_handler[pf].data); | ||
375 | |||
376 | if (status >= 0 && queue_rerouter[pf].reroute) | ||
377 | status = queue_rerouter[pf].reroute(skb, info); | ||
378 | |||
379 | read_unlock(&queue_handler_lock); | ||
380 | |||
381 | if (status < 0) { | ||
382 | /* James M doesn't say fuck enough. */ | ||
383 | if (indev) dev_put(indev); | ||
384 | if (outdev) dev_put(outdev); | ||
385 | #ifdef CONFIG_BRIDGE_NETFILTER | ||
386 | if (physindev) dev_put(physindev); | ||
387 | if (physoutdev) dev_put(physoutdev); | ||
388 | #endif | ||
389 | module_put(info->elem->owner); | ||
390 | kfree(info); | ||
391 | kfree_skb(*skb); | ||
392 | |||
393 | return 1; | ||
394 | } | ||
395 | |||
396 | return 1; | ||
397 | } | ||
398 | |||
399 | /* Returns 1 if okfn() needs to be executed by the caller, | ||
400 | * -EPERM for NF_DROP, 0 otherwise. */ | ||
401 | int nf_hook_slow(int pf, unsigned int hook, struct sk_buff **pskb, | ||
402 | struct net_device *indev, | ||
403 | struct net_device *outdev, | ||
404 | int (*okfn)(struct sk_buff *), | ||
405 | int hook_thresh) | ||
406 | { | ||
407 | struct list_head *elem; | ||
408 | unsigned int verdict; | ||
409 | int ret = 0; | ||
410 | |||
411 | /* We may already have this, but read-locks nest anyway */ | ||
412 | rcu_read_lock(); | ||
413 | |||
414 | elem = &nf_hooks[pf][hook]; | ||
415 | next_hook: | ||
416 | verdict = nf_iterate(&nf_hooks[pf][hook], pskb, hook, indev, | ||
417 | outdev, &elem, okfn, hook_thresh); | ||
418 | if (verdict == NF_ACCEPT || verdict == NF_STOP) { | ||
419 | ret = 1; | ||
420 | goto unlock; | ||
421 | } else if (verdict == NF_DROP) { | ||
422 | kfree_skb(*pskb); | ||
423 | ret = -EPERM; | ||
424 | } else if ((verdict & NF_VERDICT_MASK) == NF_QUEUE) { | ||
425 | NFDEBUG("nf_hook: Verdict = QUEUE.\n"); | ||
426 | if (!nf_queue(pskb, elem, pf, hook, indev, outdev, okfn, | ||
427 | verdict >> NF_VERDICT_BITS)) | ||
428 | goto next_hook; | ||
429 | } | ||
430 | unlock: | ||
431 | rcu_read_unlock(); | ||
432 | return ret; | ||
433 | } | ||
434 | |||
435 | void nf_reinject(struct sk_buff *skb, struct nf_info *info, | ||
436 | unsigned int verdict) | ||
437 | { | ||
438 | struct list_head *elem = &info->elem->list; | ||
439 | struct list_head *i; | ||
440 | |||
441 | rcu_read_lock(); | ||
442 | |||
443 | /* Release those devices we held, or Alexey will kill me. */ | ||
444 | if (info->indev) dev_put(info->indev); | ||
445 | if (info->outdev) dev_put(info->outdev); | ||
446 | #ifdef CONFIG_BRIDGE_NETFILTER | ||
447 | if (skb->nf_bridge) { | ||
448 | if (skb->nf_bridge->physindev) | ||
449 | dev_put(skb->nf_bridge->physindev); | ||
450 | if (skb->nf_bridge->physoutdev) | ||
451 | dev_put(skb->nf_bridge->physoutdev); | ||
452 | } | ||
453 | #endif | ||
454 | |||
455 | /* Drop reference to owner of hook which queued us. */ | ||
456 | module_put(info->elem->owner); | ||
457 | |||
458 | list_for_each_rcu(i, &nf_hooks[info->pf][info->hook]) { | ||
459 | if (i == elem) | ||
460 | break; | ||
461 | } | ||
462 | |||
463 | if (elem == &nf_hooks[info->pf][info->hook]) { | ||
464 | /* The module which sent it to userspace is gone. */ | ||
465 | NFDEBUG("%s: module disappeared, dropping packet.\n", | ||
466 | __FUNCTION__); | ||
467 | verdict = NF_DROP; | ||
468 | } | ||
469 | |||
470 | /* Continue traversal iff userspace said ok... */ | ||
471 | if (verdict == NF_REPEAT) { | ||
472 | elem = elem->prev; | ||
473 | verdict = NF_ACCEPT; | ||
474 | } | ||
475 | |||
476 | if (verdict == NF_ACCEPT) { | ||
477 | next_hook: | ||
478 | verdict = nf_iterate(&nf_hooks[info->pf][info->hook], | ||
479 | &skb, info->hook, | ||
480 | info->indev, info->outdev, &elem, | ||
481 | info->okfn, INT_MIN); | ||
482 | } | ||
483 | |||
484 | switch (verdict & NF_VERDICT_MASK) { | ||
485 | case NF_ACCEPT: | ||
486 | info->okfn(skb); | ||
487 | break; | ||
488 | |||
489 | case NF_QUEUE: | ||
490 | if (!nf_queue(&skb, elem, info->pf, info->hook, | ||
491 | info->indev, info->outdev, info->okfn, | ||
492 | verdict >> NF_VERDICT_BITS)) | ||
493 | goto next_hook; | ||
494 | break; | ||
495 | } | ||
496 | rcu_read_unlock(); | ||
497 | |||
498 | if (verdict == NF_DROP) | ||
499 | kfree_skb(skb); | ||
500 | |||
501 | kfree(info); | ||
502 | return; | ||
503 | } | ||
504 | |||
505 | int skb_make_writable(struct sk_buff **pskb, unsigned int writable_len) | ||
506 | { | ||
507 | struct sk_buff *nskb; | ||
508 | |||
509 | if (writable_len > (*pskb)->len) | ||
510 | return 0; | ||
511 | |||
512 | /* Not exclusive use of packet? Must copy. */ | ||
513 | if (skb_shared(*pskb) || skb_cloned(*pskb)) | ||
514 | goto copy_skb; | ||
515 | |||
516 | return pskb_may_pull(*pskb, writable_len); | ||
517 | |||
518 | copy_skb: | ||
519 | nskb = skb_copy(*pskb, GFP_ATOMIC); | ||
520 | if (!nskb) | ||
521 | return 0; | ||
522 | BUG_ON(skb_is_nonlinear(nskb)); | ||
523 | |||
524 | /* Rest of kernel will get very unhappy if we pass it a | ||
525 | suddenly-orphaned skbuff */ | ||
526 | if ((*pskb)->sk) | ||
527 | skb_set_owner_w(nskb, (*pskb)->sk); | ||
528 | kfree_skb(*pskb); | ||
529 | *pskb = nskb; | ||
530 | return 1; | ||
531 | } | ||
532 | EXPORT_SYMBOL(skb_make_writable); | ||
533 | |||
534 | /* Internal logging interface, which relies on the real | ||
535 | LOG target modules */ | ||
536 | |||
537 | #define NF_LOG_PREFIXLEN 128 | ||
538 | |||
539 | static struct nf_logger *nf_logging[NPROTO]; /* = NULL */ | ||
540 | static DEFINE_SPINLOCK(nf_log_lock); | ||
541 | |||
542 | int nf_log_register(int pf, struct nf_logger *logger) | ||
543 | { | ||
544 | int ret = -EBUSY; | ||
545 | |||
546 | /* Any setup of logging members must be done before | ||
547 | * substituting pointer. */ | ||
548 | spin_lock(&nf_log_lock); | ||
549 | if (!nf_logging[pf]) { | ||
550 | rcu_assign_pointer(nf_logging[pf], logger); | ||
551 | ret = 0; | ||
552 | } | ||
553 | spin_unlock(&nf_log_lock); | ||
554 | return ret; | ||
555 | } | ||
556 | |||
557 | void nf_log_unregister_pf(int pf) | ||
558 | { | ||
559 | spin_lock(&nf_log_lock); | ||
560 | nf_logging[pf] = NULL; | ||
561 | spin_unlock(&nf_log_lock); | ||
562 | |||
563 | /* Give time to concurrent readers. */ | ||
564 | synchronize_net(); | ||
565 | } | ||
566 | |||
567 | void nf_log_unregister_logger(struct nf_logger *logger) | ||
568 | { | ||
569 | int i; | ||
570 | |||
571 | spin_lock(&nf_log_lock); | ||
572 | for (i = 0; i < NPROTO; i++) { | ||
573 | if (nf_logging[i] == logger) | ||
574 | nf_logging[i] = NULL; | ||
575 | } | ||
576 | spin_unlock(&nf_log_lock); | ||
577 | |||
578 | synchronize_net(); | ||
579 | } | ||
580 | |||
581 | void nf_log_packet(int pf, | ||
582 | unsigned int hooknum, | ||
583 | const struct sk_buff *skb, | ||
584 | const struct net_device *in, | ||
585 | const struct net_device *out, | ||
586 | struct nf_loginfo *loginfo, | ||
587 | const char *fmt, ...) | ||
588 | { | ||
589 | va_list args; | ||
590 | char prefix[NF_LOG_PREFIXLEN]; | ||
591 | struct nf_logger *logger; | ||
592 | |||
593 | rcu_read_lock(); | ||
594 | logger = rcu_dereference(nf_logging[pf]); | ||
595 | if (logger) { | ||
596 | va_start(args, fmt); | ||
597 | vsnprintf(prefix, sizeof(prefix), fmt, args); | ||
598 | va_end(args); | ||
599 | /* We must read logging before nf_logfn[pf] */ | ||
600 | logger->logfn(pf, hooknum, skb, in, out, loginfo, prefix); | ||
601 | } else if (net_ratelimit()) { | ||
602 | printk(KERN_WARNING "nf_log_packet: can\'t log since " | ||
603 | "no backend logging module loaded in! Please either " | ||
604 | "load one, or disable logging explicitly\n"); | ||
605 | } | ||
606 | rcu_read_unlock(); | ||
607 | } | ||
608 | EXPORT_SYMBOL(nf_log_register); | ||
609 | EXPORT_SYMBOL(nf_log_unregister_pf); | ||
610 | EXPORT_SYMBOL(nf_log_unregister_logger); | ||
611 | EXPORT_SYMBOL(nf_log_packet); | ||
612 | |||
613 | #ifdef CONFIG_PROC_FS | ||
614 | struct proc_dir_entry *proc_net_netfilter; | ||
615 | EXPORT_SYMBOL(proc_net_netfilter); | ||
616 | |||
617 | static void *seq_start(struct seq_file *seq, loff_t *pos) | ||
618 | { | ||
619 | rcu_read_lock(); | ||
620 | |||
621 | if (*pos >= NPROTO) | ||
622 | return NULL; | ||
623 | |||
624 | return pos; | ||
625 | } | ||
626 | |||
627 | static void *seq_next(struct seq_file *s, void *v, loff_t *pos) | ||
628 | { | ||
629 | (*pos)++; | ||
630 | |||
631 | if (*pos >= NPROTO) | ||
632 | return NULL; | ||
633 | |||
634 | return pos; | ||
635 | } | ||
636 | |||
637 | static void seq_stop(struct seq_file *s, void *v) | ||
638 | { | ||
639 | rcu_read_unlock(); | ||
640 | } | ||
641 | |||
642 | static int seq_show(struct seq_file *s, void *v) | ||
643 | { | ||
644 | loff_t *pos = v; | ||
645 | const struct nf_logger *logger; | ||
646 | |||
647 | logger = rcu_dereference(nf_logging[*pos]); | ||
648 | |||
649 | if (!logger) | ||
650 | return seq_printf(s, "%2lld NONE\n", *pos); | ||
651 | |||
652 | return seq_printf(s, "%2lld %s\n", *pos, logger->name); | ||
653 | } | ||
654 | |||
655 | static struct seq_operations nflog_seq_ops = { | ||
656 | .start = seq_start, | ||
657 | .next = seq_next, | ||
658 | .stop = seq_stop, | ||
659 | .show = seq_show, | ||
660 | }; | ||
661 | |||
662 | static int nflog_open(struct inode *inode, struct file *file) | ||
663 | { | ||
664 | return seq_open(file, &nflog_seq_ops); | ||
665 | } | ||
666 | |||
667 | static struct file_operations nflog_file_ops = { | ||
668 | .owner = THIS_MODULE, | ||
669 | .open = nflog_open, | ||
670 | .read = seq_read, | ||
671 | .llseek = seq_lseek, | ||
672 | .release = seq_release, | ||
673 | }; | ||
674 | |||
675 | #endif /* PROC_FS */ | ||
676 | |||
677 | |||
678 | /* This does not belong here, but locally generated errors need it if connection | ||
679 | tracking in use: without this, connection may not be in hash table, and hence | ||
680 | manufactured ICMP or RST packets will not be associated with it. */ | ||
681 | void (*ip_ct_attach)(struct sk_buff *, struct sk_buff *); | ||
682 | |||
683 | void nf_ct_attach(struct sk_buff *new, struct sk_buff *skb) | ||
684 | { | ||
685 | void (*attach)(struct sk_buff *, struct sk_buff *); | ||
686 | |||
687 | if (skb->nfct && (attach = ip_ct_attach) != NULL) { | ||
688 | mb(); /* Just to be sure: must be read before executing this */ | ||
689 | attach(new, skb); | ||
690 | } | ||
691 | } | ||
692 | |||
693 | void __init netfilter_init(void) | ||
694 | { | ||
695 | int i, h; | ||
696 | #ifdef CONFIG_PROC_FS | ||
697 | struct proc_dir_entry *pde; | ||
698 | #endif | ||
699 | |||
700 | queue_rerouter = kmalloc(NPROTO * sizeof(struct nf_queue_rerouter), | ||
701 | GFP_KERNEL); | ||
702 | if (!queue_rerouter) | ||
703 | panic("netfilter: cannot allocate queue rerouter array\n"); | ||
704 | memset(queue_rerouter, 0, NPROTO * sizeof(struct nf_queue_rerouter)); | ||
705 | |||
706 | for (i = 0; i < NPROTO; i++) { | ||
707 | for (h = 0; h < NF_MAX_HOOKS; h++) | ||
708 | INIT_LIST_HEAD(&nf_hooks[i][h]); | ||
709 | } | ||
710 | |||
711 | #ifdef CONFIG_PROC_FS | ||
712 | proc_net_netfilter = proc_mkdir("netfilter", proc_net); | ||
713 | if (!proc_net_netfilter) | ||
714 | panic("cannot create netfilter proc entry"); | ||
715 | pde = create_proc_entry("nf_log", S_IRUGO, proc_net_netfilter); | ||
716 | if (!pde) | ||
717 | panic("cannot create /proc/net/netfilter/nf_log"); | ||
718 | pde->proc_fops = &nflog_file_ops; | ||
719 | #endif | ||
720 | } | ||
721 | |||
722 | EXPORT_SYMBOL(ip_ct_attach); | ||
723 | EXPORT_SYMBOL(nf_ct_attach); | ||
724 | EXPORT_SYMBOL(nf_getsockopt); | ||
725 | EXPORT_SYMBOL(nf_hook_slow); | ||
726 | EXPORT_SYMBOL(nf_hooks); | ||
727 | EXPORT_SYMBOL(nf_register_hook); | ||
728 | EXPORT_SYMBOL(nf_register_queue_handler); | ||
729 | EXPORT_SYMBOL(nf_register_sockopt); | ||
730 | EXPORT_SYMBOL(nf_reinject); | ||
731 | EXPORT_SYMBOL(nf_setsockopt); | ||
732 | EXPORT_SYMBOL(nf_unregister_hook); | ||
733 | EXPORT_SYMBOL(nf_unregister_queue_handler); | ||
734 | EXPORT_SYMBOL_GPL(nf_unregister_queue_handlers); | ||
735 | EXPORT_SYMBOL_GPL(nf_register_queue_rerouter); | ||
736 | EXPORT_SYMBOL_GPL(nf_unregister_queue_rerouter); | ||
737 | EXPORT_SYMBOL(nf_unregister_sockopt); | ||
diff --git a/net/netfilter/Makefile b/net/netfilter/Makefile index c41caebc4a7c..b3b44f8b415a 100644 --- a/net/netfilter/Makefile +++ b/net/netfilter/Makefile | |||
@@ -1,3 +1,7 @@ | |||
1 | netfilter-objs := core.o nf_log.o nf_queue.o nf_sockopt.o | ||
2 | |||
3 | obj-$(CONFIG_NETFILTER) = netfilter.o | ||
4 | |||
1 | obj-$(CONFIG_NETFILTER_NETLINK) += nfnetlink.o | 5 | obj-$(CONFIG_NETFILTER_NETLINK) += nfnetlink.o |
2 | obj-$(CONFIG_NETFILTER_NETLINK_QUEUE) += nfnetlink_queue.o | 6 | obj-$(CONFIG_NETFILTER_NETLINK_QUEUE) += nfnetlink_queue.o |
3 | obj-$(CONFIG_NETFILTER_NETLINK_LOG) += nfnetlink_log.o | 7 | obj-$(CONFIG_NETFILTER_NETLINK_LOG) += nfnetlink_log.o |
diff --git a/net/netfilter/core.c b/net/netfilter/core.c new file mode 100644 index 000000000000..1ceb1a6c254b --- /dev/null +++ b/net/netfilter/core.c | |||
@@ -0,0 +1,216 @@ | |||
1 | /* netfilter.c: look after the filters for various protocols. | ||
2 | * Heavily influenced by the old firewall.c by David Bonn and Alan Cox. | ||
3 | * | ||
4 | * Thanks to Rob `CmdrTaco' Malda for not influencing this code in any | ||
5 | * way. | ||
6 | * | ||
7 | * Rusty Russell (C)2000 -- This code is GPL. | ||
8 | * | ||
9 | * February 2000: Modified by James Morris to have 1 queue per protocol. | ||
10 | * 15-Mar-2000: Added NF_REPEAT --RR. | ||
11 | * 08-May-2003: Internal logging interface added by Jozsef Kadlecsik. | ||
12 | */ | ||
13 | #include <linux/config.h> | ||
14 | #include <linux/kernel.h> | ||
15 | #include <linux/netfilter.h> | ||
16 | #include <net/protocol.h> | ||
17 | #include <linux/init.h> | ||
18 | #include <linux/skbuff.h> | ||
19 | #include <linux/wait.h> | ||
20 | #include <linux/module.h> | ||
21 | #include <linux/interrupt.h> | ||
22 | #include <linux/if.h> | ||
23 | #include <linux/netdevice.h> | ||
24 | #include <linux/inetdevice.h> | ||
25 | #include <linux/proc_fs.h> | ||
26 | #include <net/sock.h> | ||
27 | |||
28 | #include "nf_internals.h" | ||
29 | |||
30 | /* In this code, we can be waiting indefinitely for userspace to | ||
31 | * service a packet if a hook returns NF_QUEUE. We could keep a count | ||
32 | * of skbuffs queued for userspace, and not deregister a hook unless | ||
33 | * this is zero, but that sucks. Now, we simply check when the | ||
34 | * packets come back: if the hook is gone, the packet is discarded. */ | ||
35 | struct list_head nf_hooks[NPROTO][NF_MAX_HOOKS]; | ||
36 | EXPORT_SYMBOL(nf_hooks); | ||
37 | static DEFINE_SPINLOCK(nf_hook_lock); | ||
38 | |||
39 | int nf_register_hook(struct nf_hook_ops *reg) | ||
40 | { | ||
41 | struct list_head *i; | ||
42 | |||
43 | spin_lock_bh(&nf_hook_lock); | ||
44 | list_for_each(i, &nf_hooks[reg->pf][reg->hooknum]) { | ||
45 | if (reg->priority < ((struct nf_hook_ops *)i)->priority) | ||
46 | break; | ||
47 | } | ||
48 | list_add_rcu(®->list, i->prev); | ||
49 | spin_unlock_bh(&nf_hook_lock); | ||
50 | |||
51 | synchronize_net(); | ||
52 | return 0; | ||
53 | } | ||
54 | EXPORT_SYMBOL(nf_register_hook); | ||
55 | |||
56 | void nf_unregister_hook(struct nf_hook_ops *reg) | ||
57 | { | ||
58 | spin_lock_bh(&nf_hook_lock); | ||
59 | list_del_rcu(®->list); | ||
60 | spin_unlock_bh(&nf_hook_lock); | ||
61 | |||
62 | synchronize_net(); | ||
63 | } | ||
64 | EXPORT_SYMBOL(nf_unregister_hook); | ||
65 | |||
66 | unsigned int nf_iterate(struct list_head *head, | ||
67 | struct sk_buff **skb, | ||
68 | int hook, | ||
69 | const struct net_device *indev, | ||
70 | const struct net_device *outdev, | ||
71 | struct list_head **i, | ||
72 | int (*okfn)(struct sk_buff *), | ||
73 | int hook_thresh) | ||
74 | { | ||
75 | unsigned int verdict; | ||
76 | |||
77 | /* | ||
78 | * The caller must not block between calls to this | ||
79 | * function because of risk of continuing from deleted element. | ||
80 | */ | ||
81 | list_for_each_continue_rcu(*i, head) { | ||
82 | struct nf_hook_ops *elem = (struct nf_hook_ops *)*i; | ||
83 | |||
84 | if (hook_thresh > elem->priority) | ||
85 | continue; | ||
86 | |||
87 | /* Optimization: we don't need to hold module | ||
88 | reference here, since function can't sleep. --RR */ | ||
89 | verdict = elem->hook(hook, skb, indev, outdev, okfn); | ||
90 | if (verdict != NF_ACCEPT) { | ||
91 | #ifdef CONFIG_NETFILTER_DEBUG | ||
92 | if (unlikely((verdict & NF_VERDICT_MASK) | ||
93 | > NF_MAX_VERDICT)) { | ||
94 | NFDEBUG("Evil return from %p(%u).\n", | ||
95 | elem->hook, hook); | ||
96 | continue; | ||
97 | } | ||
98 | #endif | ||
99 | if (verdict != NF_REPEAT) | ||
100 | return verdict; | ||
101 | *i = (*i)->prev; | ||
102 | } | ||
103 | } | ||
104 | return NF_ACCEPT; | ||
105 | } | ||
106 | |||
107 | |||
108 | /* Returns 1 if okfn() needs to be executed by the caller, | ||
109 | * -EPERM for NF_DROP, 0 otherwise. */ | ||
110 | int nf_hook_slow(int pf, unsigned int hook, struct sk_buff **pskb, | ||
111 | struct net_device *indev, | ||
112 | struct net_device *outdev, | ||
113 | int (*okfn)(struct sk_buff *), | ||
114 | int hook_thresh) | ||
115 | { | ||
116 | struct list_head *elem; | ||
117 | unsigned int verdict; | ||
118 | int ret = 0; | ||
119 | |||
120 | /* We may already have this, but read-locks nest anyway */ | ||
121 | rcu_read_lock(); | ||
122 | |||
123 | elem = &nf_hooks[pf][hook]; | ||
124 | next_hook: | ||
125 | verdict = nf_iterate(&nf_hooks[pf][hook], pskb, hook, indev, | ||
126 | outdev, &elem, okfn, hook_thresh); | ||
127 | if (verdict == NF_ACCEPT || verdict == NF_STOP) { | ||
128 | ret = 1; | ||
129 | goto unlock; | ||
130 | } else if (verdict == NF_DROP) { | ||
131 | kfree_skb(*pskb); | ||
132 | ret = -EPERM; | ||
133 | } else if ((verdict & NF_VERDICT_MASK) == NF_QUEUE) { | ||
134 | NFDEBUG("nf_hook: Verdict = QUEUE.\n"); | ||
135 | if (!nf_queue(pskb, elem, pf, hook, indev, outdev, okfn, | ||
136 | verdict >> NF_VERDICT_BITS)) | ||
137 | goto next_hook; | ||
138 | } | ||
139 | unlock: | ||
140 | rcu_read_unlock(); | ||
141 | return ret; | ||
142 | } | ||
143 | EXPORT_SYMBOL(nf_hook_slow); | ||
144 | |||
145 | |||
146 | int skb_make_writable(struct sk_buff **pskb, unsigned int writable_len) | ||
147 | { | ||
148 | struct sk_buff *nskb; | ||
149 | |||
150 | if (writable_len > (*pskb)->len) | ||
151 | return 0; | ||
152 | |||
153 | /* Not exclusive use of packet? Must copy. */ | ||
154 | if (skb_shared(*pskb) || skb_cloned(*pskb)) | ||
155 | goto copy_skb; | ||
156 | |||
157 | return pskb_may_pull(*pskb, writable_len); | ||
158 | |||
159 | copy_skb: | ||
160 | nskb = skb_copy(*pskb, GFP_ATOMIC); | ||
161 | if (!nskb) | ||
162 | return 0; | ||
163 | BUG_ON(skb_is_nonlinear(nskb)); | ||
164 | |||
165 | /* Rest of kernel will get very unhappy if we pass it a | ||
166 | suddenly-orphaned skbuff */ | ||
167 | if ((*pskb)->sk) | ||
168 | skb_set_owner_w(nskb, (*pskb)->sk); | ||
169 | kfree_skb(*pskb); | ||
170 | *pskb = nskb; | ||
171 | return 1; | ||
172 | } | ||
173 | EXPORT_SYMBOL(skb_make_writable); | ||
174 | |||
175 | |||
176 | /* This does not belong here, but locally generated errors need it if connection | ||
177 | tracking in use: without this, connection may not be in hash table, and hence | ||
178 | manufactured ICMP or RST packets will not be associated with it. */ | ||
179 | void (*ip_ct_attach)(struct sk_buff *, struct sk_buff *); | ||
180 | EXPORT_SYMBOL(ip_ct_attach); | ||
181 | |||
182 | void nf_ct_attach(struct sk_buff *new, struct sk_buff *skb) | ||
183 | { | ||
184 | void (*attach)(struct sk_buff *, struct sk_buff *); | ||
185 | |||
186 | if (skb->nfct && (attach = ip_ct_attach) != NULL) { | ||
187 | mb(); /* Just to be sure: must be read before executing this */ | ||
188 | attach(new, skb); | ||
189 | } | ||
190 | } | ||
191 | EXPORT_SYMBOL(nf_ct_attach); | ||
192 | |||
193 | #ifdef CONFIG_PROC_FS | ||
194 | struct proc_dir_entry *proc_net_netfilter; | ||
195 | EXPORT_SYMBOL(proc_net_netfilter); | ||
196 | #endif | ||
197 | |||
198 | void __init netfilter_init(void) | ||
199 | { | ||
200 | int i, h; | ||
201 | for (i = 0; i < NPROTO; i++) { | ||
202 | for (h = 0; h < NF_MAX_HOOKS; h++) | ||
203 | INIT_LIST_HEAD(&nf_hooks[i][h]); | ||
204 | } | ||
205 | |||
206 | #ifdef CONFIG_PROC_FS | ||
207 | proc_net_netfilter = proc_mkdir("netfilter", proc_net); | ||
208 | if (!proc_net_netfilter) | ||
209 | panic("cannot create netfilter proc entry"); | ||
210 | #endif | ||
211 | |||
212 | if (netfilter_queue_init() < 0) | ||
213 | panic("cannot initialize nf_queue"); | ||
214 | if (netfilter_log_init() < 0) | ||
215 | panic("cannot initialize nf_log"); | ||
216 | } | ||
diff --git a/net/netfilter/nf_internals.h b/net/netfilter/nf_internals.h new file mode 100644 index 000000000000..6bdee2910617 --- /dev/null +++ b/net/netfilter/nf_internals.h | |||
@@ -0,0 +1,39 @@ | |||
1 | #ifndef _NF_INTERNALS_H | ||
2 | #define _NF_INTERNALS_H | ||
3 | |||
4 | #include <linux/config.h> | ||
5 | #include <linux/list.h> | ||
6 | #include <linux/skbuff.h> | ||
7 | #include <linux/netdevice.h> | ||
8 | |||
9 | #ifdef CONFIG_NETFILTER_DEBUG | ||
10 | #define NFDEBUG(format, args...) printk(format , ## args) | ||
11 | #else | ||
12 | #define NFDEBUG(format, args...) | ||
13 | #endif | ||
14 | |||
15 | |||
16 | /* core.c */ | ||
17 | extern unsigned int nf_iterate(struct list_head *head, | ||
18 | struct sk_buff **skb, | ||
19 | int hook, | ||
20 | const struct net_device *indev, | ||
21 | const struct net_device *outdev, | ||
22 | struct list_head **i, | ||
23 | int (*okfn)(struct sk_buff *), | ||
24 | int hook_thresh); | ||
25 | |||
26 | /* nf_queue.c */ | ||
27 | extern int nf_queue(struct sk_buff **skb, | ||
28 | struct list_head *elem, | ||
29 | int pf, unsigned int hook, | ||
30 | struct net_device *indev, | ||
31 | struct net_device *outdev, | ||
32 | int (*okfn)(struct sk_buff *), | ||
33 | unsigned int queuenum); | ||
34 | extern int __init netfilter_queue_init(void); | ||
35 | |||
36 | /* nf_log.c */ | ||
37 | extern int __init netfilter_log_init(void); | ||
38 | |||
39 | #endif | ||
diff --git a/net/netfilter/nf_log.c b/net/netfilter/nf_log.c new file mode 100644 index 000000000000..ec58c4d2c667 --- /dev/null +++ b/net/netfilter/nf_log.c | |||
@@ -0,0 +1,165 @@ | |||
1 | #include <linux/config.h> | ||
2 | #include <linux/kernel.h> | ||
3 | #include <linux/init.h> | ||
4 | #include <linux/module.h> | ||
5 | #include <linux/proc_fs.h> | ||
6 | #include <linux/skbuff.h> | ||
7 | #include <linux/netfilter.h> | ||
8 | #include <net/protocol.h> | ||
9 | |||
10 | #include "nf_internals.h" | ||
11 | |||
12 | /* Internal logging interface, which relies on the real | ||
13 | LOG target modules */ | ||
14 | |||
15 | #define NF_LOG_PREFIXLEN 128 | ||
16 | |||
17 | static struct nf_logger *nf_logging[NPROTO]; /* = NULL */ | ||
18 | static DEFINE_SPINLOCK(nf_log_lock); | ||
19 | |||
20 | int nf_log_register(int pf, struct nf_logger *logger) | ||
21 | { | ||
22 | int ret = -EBUSY; | ||
23 | |||
24 | /* Any setup of logging members must be done before | ||
25 | * substituting pointer. */ | ||
26 | spin_lock(&nf_log_lock); | ||
27 | if (!nf_logging[pf]) { | ||
28 | rcu_assign_pointer(nf_logging[pf], logger); | ||
29 | ret = 0; | ||
30 | } | ||
31 | spin_unlock(&nf_log_lock); | ||
32 | return ret; | ||
33 | } | ||
34 | EXPORT_SYMBOL(nf_log_register); | ||
35 | |||
36 | void nf_log_unregister_pf(int pf) | ||
37 | { | ||
38 | spin_lock(&nf_log_lock); | ||
39 | nf_logging[pf] = NULL; | ||
40 | spin_unlock(&nf_log_lock); | ||
41 | |||
42 | /* Give time to concurrent readers. */ | ||
43 | synchronize_net(); | ||
44 | } | ||
45 | EXPORT_SYMBOL(nf_log_unregister_pf); | ||
46 | |||
47 | void nf_log_unregister_logger(struct nf_logger *logger) | ||
48 | { | ||
49 | int i; | ||
50 | |||
51 | spin_lock(&nf_log_lock); | ||
52 | for (i = 0; i < NPROTO; i++) { | ||
53 | if (nf_logging[i] == logger) | ||
54 | nf_logging[i] = NULL; | ||
55 | } | ||
56 | spin_unlock(&nf_log_lock); | ||
57 | |||
58 | synchronize_net(); | ||
59 | } | ||
60 | EXPORT_SYMBOL(nf_log_unregister_logger); | ||
61 | |||
62 | void nf_log_packet(int pf, | ||
63 | unsigned int hooknum, | ||
64 | const struct sk_buff *skb, | ||
65 | const struct net_device *in, | ||
66 | const struct net_device *out, | ||
67 | struct nf_loginfo *loginfo, | ||
68 | const char *fmt, ...) | ||
69 | { | ||
70 | va_list args; | ||
71 | char prefix[NF_LOG_PREFIXLEN]; | ||
72 | struct nf_logger *logger; | ||
73 | |||
74 | rcu_read_lock(); | ||
75 | logger = rcu_dereference(nf_logging[pf]); | ||
76 | if (logger) { | ||
77 | va_start(args, fmt); | ||
78 | vsnprintf(prefix, sizeof(prefix), fmt, args); | ||
79 | va_end(args); | ||
80 | /* We must read logging before nf_logfn[pf] */ | ||
81 | logger->logfn(pf, hooknum, skb, in, out, loginfo, prefix); | ||
82 | } else if (net_ratelimit()) { | ||
83 | printk(KERN_WARNING "nf_log_packet: can\'t log since " | ||
84 | "no backend logging module loaded in! Please either " | ||
85 | "load one, or disable logging explicitly\n"); | ||
86 | } | ||
87 | rcu_read_unlock(); | ||
88 | } | ||
89 | EXPORT_SYMBOL(nf_log_packet); | ||
90 | |||
91 | #ifdef CONFIG_PROC_FS | ||
92 | static void *seq_start(struct seq_file *seq, loff_t *pos) | ||
93 | { | ||
94 | rcu_read_lock(); | ||
95 | |||
96 | if (*pos >= NPROTO) | ||
97 | return NULL; | ||
98 | |||
99 | return pos; | ||
100 | } | ||
101 | |||
102 | static void *seq_next(struct seq_file *s, void *v, loff_t *pos) | ||
103 | { | ||
104 | (*pos)++; | ||
105 | |||
106 | if (*pos >= NPROTO) | ||
107 | return NULL; | ||
108 | |||
109 | return pos; | ||
110 | } | ||
111 | |||
112 | static void seq_stop(struct seq_file *s, void *v) | ||
113 | { | ||
114 | rcu_read_unlock(); | ||
115 | } | ||
116 | |||
117 | static int seq_show(struct seq_file *s, void *v) | ||
118 | { | ||
119 | loff_t *pos = v; | ||
120 | const struct nf_logger *logger; | ||
121 | |||
122 | logger = rcu_dereference(nf_logging[*pos]); | ||
123 | |||
124 | if (!logger) | ||
125 | return seq_printf(s, "%2lld NONE\n", *pos); | ||
126 | |||
127 | return seq_printf(s, "%2lld %s\n", *pos, logger->name); | ||
128 | } | ||
129 | |||
130 | static struct seq_operations nflog_seq_ops = { | ||
131 | .start = seq_start, | ||
132 | .next = seq_next, | ||
133 | .stop = seq_stop, | ||
134 | .show = seq_show, | ||
135 | }; | ||
136 | |||
137 | static int nflog_open(struct inode *inode, struct file *file) | ||
138 | { | ||
139 | return seq_open(file, &nflog_seq_ops); | ||
140 | } | ||
141 | |||
142 | static struct file_operations nflog_file_ops = { | ||
143 | .owner = THIS_MODULE, | ||
144 | .open = nflog_open, | ||
145 | .read = seq_read, | ||
146 | .llseek = seq_lseek, | ||
147 | .release = seq_release, | ||
148 | }; | ||
149 | |||
150 | #endif /* PROC_FS */ | ||
151 | |||
152 | |||
153 | int __init netfilter_log_init(void) | ||
154 | { | ||
155 | #ifdef CONFIG_PROC_FS | ||
156 | struct proc_dir_entry *pde; | ||
157 | pde = create_proc_entry("nf_log", S_IRUGO, proc_net_netfilter); | ||
158 | #endif | ||
159 | if (!pde) | ||
160 | return -1; | ||
161 | |||
162 | pde->proc_fops = &nflog_file_ops; | ||
163 | |||
164 | return 0; | ||
165 | } | ||
diff --git a/net/netfilter/nf_queue.c b/net/netfilter/nf_queue.c new file mode 100644 index 000000000000..5586f843ed45 --- /dev/null +++ b/net/netfilter/nf_queue.c | |||
@@ -0,0 +1,273 @@ | |||
1 | #include <linux/config.h> | ||
2 | #include <linux/kernel.h> | ||
3 | #include <linux/init.h> | ||
4 | #include <linux/module.h> | ||
5 | #include <linux/proc_fs.h> | ||
6 | #include <linux/skbuff.h> | ||
7 | #include <linux/netfilter.h> | ||
8 | #include <net/protocol.h> | ||
9 | |||
10 | #include "nf_internals.h" | ||
11 | |||
12 | /* | ||
13 | * A queue handler may be registered for each protocol. Each is protected by | ||
14 | * long term mutex. The handler must provide an an outfn() to accept packets | ||
15 | * for queueing and must reinject all packets it receives, no matter what. | ||
16 | */ | ||
17 | static struct nf_queue_handler_t { | ||
18 | nf_queue_outfn_t outfn; | ||
19 | void *data; | ||
20 | } queue_handler[NPROTO]; | ||
21 | |||
22 | static struct nf_queue_rerouter *queue_rerouter; | ||
23 | |||
24 | static DEFINE_RWLOCK(queue_handler_lock); | ||
25 | |||
26 | |||
27 | int nf_register_queue_handler(int pf, nf_queue_outfn_t outfn, void *data) | ||
28 | { | ||
29 | int ret; | ||
30 | |||
31 | if (pf >= NPROTO) | ||
32 | return -EINVAL; | ||
33 | |||
34 | write_lock_bh(&queue_handler_lock); | ||
35 | if (queue_handler[pf].outfn) | ||
36 | ret = -EBUSY; | ||
37 | else { | ||
38 | queue_handler[pf].outfn = outfn; | ||
39 | queue_handler[pf].data = data; | ||
40 | ret = 0; | ||
41 | } | ||
42 | write_unlock_bh(&queue_handler_lock); | ||
43 | |||
44 | return ret; | ||
45 | } | ||
46 | EXPORT_SYMBOL(nf_register_queue_handler); | ||
47 | |||
48 | /* The caller must flush their queue before this */ | ||
49 | int nf_unregister_queue_handler(int pf) | ||
50 | { | ||
51 | if (pf >= NPROTO) | ||
52 | return -EINVAL; | ||
53 | |||
54 | write_lock_bh(&queue_handler_lock); | ||
55 | queue_handler[pf].outfn = NULL; | ||
56 | queue_handler[pf].data = NULL; | ||
57 | write_unlock_bh(&queue_handler_lock); | ||
58 | |||
59 | return 0; | ||
60 | } | ||
61 | EXPORT_SYMBOL(nf_unregister_queue_handler); | ||
62 | |||
63 | int nf_register_queue_rerouter(int pf, struct nf_queue_rerouter *rer) | ||
64 | { | ||
65 | if (pf >= NPROTO) | ||
66 | return -EINVAL; | ||
67 | |||
68 | write_lock_bh(&queue_handler_lock); | ||
69 | memcpy(&queue_rerouter[pf], rer, sizeof(queue_rerouter[pf])); | ||
70 | write_unlock_bh(&queue_handler_lock); | ||
71 | |||
72 | return 0; | ||
73 | } | ||
74 | EXPORT_SYMBOL_GPL(nf_register_queue_rerouter); | ||
75 | |||
76 | int nf_unregister_queue_rerouter(int pf) | ||
77 | { | ||
78 | if (pf >= NPROTO) | ||
79 | return -EINVAL; | ||
80 | |||
81 | write_lock_bh(&queue_handler_lock); | ||
82 | memset(&queue_rerouter[pf], 0, sizeof(queue_rerouter[pf])); | ||
83 | write_unlock_bh(&queue_handler_lock); | ||
84 | return 0; | ||
85 | } | ||
86 | EXPORT_SYMBOL_GPL(nf_unregister_queue_rerouter); | ||
87 | |||
88 | void nf_unregister_queue_handlers(nf_queue_outfn_t outfn) | ||
89 | { | ||
90 | int pf; | ||
91 | |||
92 | write_lock_bh(&queue_handler_lock); | ||
93 | for (pf = 0; pf < NPROTO; pf++) { | ||
94 | if (queue_handler[pf].outfn == outfn) { | ||
95 | queue_handler[pf].outfn = NULL; | ||
96 | queue_handler[pf].data = NULL; | ||
97 | } | ||
98 | } | ||
99 | write_unlock_bh(&queue_handler_lock); | ||
100 | } | ||
101 | EXPORT_SYMBOL_GPL(nf_unregister_queue_handlers); | ||
102 | |||
103 | /* | ||
104 | * Any packet that leaves via this function must come back | ||
105 | * through nf_reinject(). | ||
106 | */ | ||
107 | int nf_queue(struct sk_buff **skb, | ||
108 | struct list_head *elem, | ||
109 | int pf, unsigned int hook, | ||
110 | struct net_device *indev, | ||
111 | struct net_device *outdev, | ||
112 | int (*okfn)(struct sk_buff *), | ||
113 | unsigned int queuenum) | ||
114 | { | ||
115 | int status; | ||
116 | struct nf_info *info; | ||
117 | #ifdef CONFIG_BRIDGE_NETFILTER | ||
118 | struct net_device *physindev = NULL; | ||
119 | struct net_device *physoutdev = NULL; | ||
120 | #endif | ||
121 | |||
122 | /* QUEUE == DROP if noone is waiting, to be safe. */ | ||
123 | read_lock(&queue_handler_lock); | ||
124 | if (!queue_handler[pf].outfn) { | ||
125 | read_unlock(&queue_handler_lock); | ||
126 | kfree_skb(*skb); | ||
127 | return 1; | ||
128 | } | ||
129 | |||
130 | info = kmalloc(sizeof(*info)+queue_rerouter[pf].rer_size, GFP_ATOMIC); | ||
131 | if (!info) { | ||
132 | if (net_ratelimit()) | ||
133 | printk(KERN_ERR "OOM queueing packet %p\n", | ||
134 | *skb); | ||
135 | read_unlock(&queue_handler_lock); | ||
136 | kfree_skb(*skb); | ||
137 | return 1; | ||
138 | } | ||
139 | |||
140 | *info = (struct nf_info) { | ||
141 | (struct nf_hook_ops *)elem, pf, hook, indev, outdev, okfn }; | ||
142 | |||
143 | /* If it's going away, ignore hook. */ | ||
144 | if (!try_module_get(info->elem->owner)) { | ||
145 | read_unlock(&queue_handler_lock); | ||
146 | kfree(info); | ||
147 | return 0; | ||
148 | } | ||
149 | |||
150 | /* Bump dev refs so they don't vanish while packet is out */ | ||
151 | if (indev) dev_hold(indev); | ||
152 | if (outdev) dev_hold(outdev); | ||
153 | |||
154 | #ifdef CONFIG_BRIDGE_NETFILTER | ||
155 | if ((*skb)->nf_bridge) { | ||
156 | physindev = (*skb)->nf_bridge->physindev; | ||
157 | if (physindev) dev_hold(physindev); | ||
158 | physoutdev = (*skb)->nf_bridge->physoutdev; | ||
159 | if (physoutdev) dev_hold(physoutdev); | ||
160 | } | ||
161 | #endif | ||
162 | if (queue_rerouter[pf].save) | ||
163 | queue_rerouter[pf].save(*skb, info); | ||
164 | |||
165 | status = queue_handler[pf].outfn(*skb, info, queuenum, | ||
166 | queue_handler[pf].data); | ||
167 | |||
168 | if (status >= 0 && queue_rerouter[pf].reroute) | ||
169 | status = queue_rerouter[pf].reroute(skb, info); | ||
170 | |||
171 | read_unlock(&queue_handler_lock); | ||
172 | |||
173 | if (status < 0) { | ||
174 | /* James M doesn't say fuck enough. */ | ||
175 | if (indev) dev_put(indev); | ||
176 | if (outdev) dev_put(outdev); | ||
177 | #ifdef CONFIG_BRIDGE_NETFILTER | ||
178 | if (physindev) dev_put(physindev); | ||
179 | if (physoutdev) dev_put(physoutdev); | ||
180 | #endif | ||
181 | module_put(info->elem->owner); | ||
182 | kfree(info); | ||
183 | kfree_skb(*skb); | ||
184 | |||
185 | return 1; | ||
186 | } | ||
187 | |||
188 | return 1; | ||
189 | } | ||
190 | |||
191 | void nf_reinject(struct sk_buff *skb, struct nf_info *info, | ||
192 | unsigned int verdict) | ||
193 | { | ||
194 | struct list_head *elem = &info->elem->list; | ||
195 | struct list_head *i; | ||
196 | |||
197 | rcu_read_lock(); | ||
198 | |||
199 | /* Release those devices we held, or Alexey will kill me. */ | ||
200 | if (info->indev) dev_put(info->indev); | ||
201 | if (info->outdev) dev_put(info->outdev); | ||
202 | #ifdef CONFIG_BRIDGE_NETFILTER | ||
203 | if (skb->nf_bridge) { | ||
204 | if (skb->nf_bridge->physindev) | ||
205 | dev_put(skb->nf_bridge->physindev); | ||
206 | if (skb->nf_bridge->physoutdev) | ||
207 | dev_put(skb->nf_bridge->physoutdev); | ||
208 | } | ||
209 | #endif | ||
210 | |||
211 | /* Drop reference to owner of hook which queued us. */ | ||
212 | module_put(info->elem->owner); | ||
213 | |||
214 | list_for_each_rcu(i, &nf_hooks[info->pf][info->hook]) { | ||
215 | if (i == elem) | ||
216 | break; | ||
217 | } | ||
218 | |||
219 | if (elem == &nf_hooks[info->pf][info->hook]) { | ||
220 | /* The module which sent it to userspace is gone. */ | ||
221 | NFDEBUG("%s: module disappeared, dropping packet.\n", | ||
222 | __FUNCTION__); | ||
223 | verdict = NF_DROP; | ||
224 | } | ||
225 | |||
226 | /* Continue traversal iff userspace said ok... */ | ||
227 | if (verdict == NF_REPEAT) { | ||
228 | elem = elem->prev; | ||
229 | verdict = NF_ACCEPT; | ||
230 | } | ||
231 | |||
232 | if (verdict == NF_ACCEPT) { | ||
233 | next_hook: | ||
234 | verdict = nf_iterate(&nf_hooks[info->pf][info->hook], | ||
235 | &skb, info->hook, | ||
236 | info->indev, info->outdev, &elem, | ||
237 | info->okfn, INT_MIN); | ||
238 | } | ||
239 | |||
240 | switch (verdict & NF_VERDICT_MASK) { | ||
241 | case NF_ACCEPT: | ||
242 | info->okfn(skb); | ||
243 | break; | ||
244 | |||
245 | case NF_QUEUE: | ||
246 | if (!nf_queue(&skb, elem, info->pf, info->hook, | ||
247 | info->indev, info->outdev, info->okfn, | ||
248 | verdict >> NF_VERDICT_BITS)) | ||
249 | goto next_hook; | ||
250 | break; | ||
251 | } | ||
252 | rcu_read_unlock(); | ||
253 | |||
254 | if (verdict == NF_DROP) | ||
255 | kfree_skb(skb); | ||
256 | |||
257 | kfree(info); | ||
258 | return; | ||
259 | } | ||
260 | EXPORT_SYMBOL(nf_reinject); | ||
261 | |||
262 | int __init netfilter_queue_init(void) | ||
263 | { | ||
264 | queue_rerouter = kmalloc(NPROTO * sizeof(struct nf_queue_rerouter), | ||
265 | GFP_KERNEL); | ||
266 | if (!queue_rerouter) | ||
267 | return -ENOMEM; | ||
268 | |||
269 | memset(queue_rerouter, 0, NPROTO * sizeof(struct nf_queue_rerouter)); | ||
270 | |||
271 | return 0; | ||
272 | } | ||
273 | |||
diff --git a/net/netfilter/nf_sockopt.c b/net/netfilter/nf_sockopt.c new file mode 100644 index 000000000000..61a833a9caa6 --- /dev/null +++ b/net/netfilter/nf_sockopt.c | |||
@@ -0,0 +1,132 @@ | |||
1 | #include <linux/config.h> | ||
2 | #include <linux/kernel.h> | ||
3 | #include <linux/init.h> | ||
4 | #include <linux/module.h> | ||
5 | #include <linux/skbuff.h> | ||
6 | #include <linux/netfilter.h> | ||
7 | #include <net/sock.h> | ||
8 | |||
9 | #include "nf_internals.h" | ||
10 | |||
11 | /* Sockopts only registered and called from user context, so | ||
12 | net locking would be overkill. Also, [gs]etsockopt calls may | ||
13 | sleep. */ | ||
14 | static DECLARE_MUTEX(nf_sockopt_mutex); | ||
15 | static LIST_HEAD(nf_sockopts); | ||
16 | |||
17 | /* Do exclusive ranges overlap? */ | ||
18 | static inline int overlap(int min1, int max1, int min2, int max2) | ||
19 | { | ||
20 | return max1 > min2 && min1 < max2; | ||
21 | } | ||
22 | |||
23 | /* Functions to register sockopt ranges (exclusive). */ | ||
24 | int nf_register_sockopt(struct nf_sockopt_ops *reg) | ||
25 | { | ||
26 | struct list_head *i; | ||
27 | int ret = 0; | ||
28 | |||
29 | if (down_interruptible(&nf_sockopt_mutex) != 0) | ||
30 | return -EINTR; | ||
31 | |||
32 | list_for_each(i, &nf_sockopts) { | ||
33 | struct nf_sockopt_ops *ops = (struct nf_sockopt_ops *)i; | ||
34 | if (ops->pf == reg->pf | ||
35 | && (overlap(ops->set_optmin, ops->set_optmax, | ||
36 | reg->set_optmin, reg->set_optmax) | ||
37 | || overlap(ops->get_optmin, ops->get_optmax, | ||
38 | reg->get_optmin, reg->get_optmax))) { | ||
39 | NFDEBUG("nf_sock overlap: %u-%u/%u-%u v %u-%u/%u-%u\n", | ||
40 | ops->set_optmin, ops->set_optmax, | ||
41 | ops->get_optmin, ops->get_optmax, | ||
42 | reg->set_optmin, reg->set_optmax, | ||
43 | reg->get_optmin, reg->get_optmax); | ||
44 | ret = -EBUSY; | ||
45 | goto out; | ||
46 | } | ||
47 | } | ||
48 | |||
49 | list_add(®->list, &nf_sockopts); | ||
50 | out: | ||
51 | up(&nf_sockopt_mutex); | ||
52 | return ret; | ||
53 | } | ||
54 | EXPORT_SYMBOL(nf_register_sockopt); | ||
55 | |||
56 | void nf_unregister_sockopt(struct nf_sockopt_ops *reg) | ||
57 | { | ||
58 | /* No point being interruptible: we're probably in cleanup_module() */ | ||
59 | restart: | ||
60 | down(&nf_sockopt_mutex); | ||
61 | if (reg->use != 0) { | ||
62 | /* To be woken by nf_sockopt call... */ | ||
63 | /* FIXME: Stuart Young's name appears gratuitously. */ | ||
64 | set_current_state(TASK_UNINTERRUPTIBLE); | ||
65 | reg->cleanup_task = current; | ||
66 | up(&nf_sockopt_mutex); | ||
67 | schedule(); | ||
68 | goto restart; | ||
69 | } | ||
70 | list_del(®->list); | ||
71 | up(&nf_sockopt_mutex); | ||
72 | } | ||
73 | EXPORT_SYMBOL(nf_unregister_sockopt); | ||
74 | |||
75 | /* Call get/setsockopt() */ | ||
76 | static int nf_sockopt(struct sock *sk, int pf, int val, | ||
77 | char __user *opt, int *len, int get) | ||
78 | { | ||
79 | struct list_head *i; | ||
80 | struct nf_sockopt_ops *ops; | ||
81 | int ret; | ||
82 | |||
83 | if (down_interruptible(&nf_sockopt_mutex) != 0) | ||
84 | return -EINTR; | ||
85 | |||
86 | list_for_each(i, &nf_sockopts) { | ||
87 | ops = (struct nf_sockopt_ops *)i; | ||
88 | if (ops->pf == pf) { | ||
89 | if (get) { | ||
90 | if (val >= ops->get_optmin | ||
91 | && val < ops->get_optmax) { | ||
92 | ops->use++; | ||
93 | up(&nf_sockopt_mutex); | ||
94 | ret = ops->get(sk, val, opt, len); | ||
95 | goto out; | ||
96 | } | ||
97 | } else { | ||
98 | if (val >= ops->set_optmin | ||
99 | && val < ops->set_optmax) { | ||
100 | ops->use++; | ||
101 | up(&nf_sockopt_mutex); | ||
102 | ret = ops->set(sk, val, opt, *len); | ||
103 | goto out; | ||
104 | } | ||
105 | } | ||
106 | } | ||
107 | } | ||
108 | up(&nf_sockopt_mutex); | ||
109 | return -ENOPROTOOPT; | ||
110 | |||
111 | out: | ||
112 | down(&nf_sockopt_mutex); | ||
113 | ops->use--; | ||
114 | if (ops->cleanup_task) | ||
115 | wake_up_process(ops->cleanup_task); | ||
116 | up(&nf_sockopt_mutex); | ||
117 | return ret; | ||
118 | } | ||
119 | |||
120 | int nf_setsockopt(struct sock *sk, int pf, int val, char __user *opt, | ||
121 | int len) | ||
122 | { | ||
123 | return nf_sockopt(sk, pf, val, opt, &len, 0); | ||
124 | } | ||
125 | EXPORT_SYMBOL(nf_setsockopt); | ||
126 | |||
127 | int nf_getsockopt(struct sock *sk, int pf, int val, char __user *opt, int *len) | ||
128 | { | ||
129 | return nf_sockopt(sk, pf, val, opt, len, 1); | ||
130 | } | ||
131 | EXPORT_SYMBOL(nf_getsockopt); | ||
132 | |||