xref: /openbmc/linux/net/netfilter/core.c (revision 325f8ab7)
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  * This code is GPL.
8  */
9 #include <linux/kernel.h>
10 #include <linux/netfilter.h>
11 #include <net/protocol.h>
12 #include <linux/init.h>
13 #include <linux/skbuff.h>
14 #include <linux/wait.h>
15 #include <linux/module.h>
16 #include <linux/interrupt.h>
17 #include <linux/if.h>
18 #include <linux/netdevice.h>
19 #include <linux/netfilter_ipv6.h>
20 #include <linux/inetdevice.h>
21 #include <linux/proc_fs.h>
22 #include <linux/mutex.h>
23 #include <linux/mm.h>
24 #include <linux/rcupdate.h>
25 #include <net/net_namespace.h>
26 #include <net/netfilter/nf_queue.h>
27 #include <net/sock.h>
28 
29 #include "nf_internals.h"
30 
31 const struct nf_ipv6_ops __rcu *nf_ipv6_ops __read_mostly;
32 EXPORT_SYMBOL_GPL(nf_ipv6_ops);
33 
34 DEFINE_PER_CPU(bool, nf_skb_duplicated);
35 EXPORT_SYMBOL_GPL(nf_skb_duplicated);
36 
37 #ifdef CONFIG_JUMP_LABEL
38 struct static_key nf_hooks_needed[NFPROTO_NUMPROTO][NF_MAX_HOOKS];
39 EXPORT_SYMBOL(nf_hooks_needed);
40 #endif
41 
42 static DEFINE_MUTEX(nf_hook_mutex);
43 
44 /* max hooks per family/hooknum */
45 #define MAX_HOOK_COUNT		1024
46 
47 #define nf_entry_dereference(e) \
48 	rcu_dereference_protected(e, lockdep_is_held(&nf_hook_mutex))
49 
allocate_hook_entries_size(u16 num)50 static struct nf_hook_entries *allocate_hook_entries_size(u16 num)
51 {
52 	struct nf_hook_entries *e;
53 	size_t alloc = sizeof(*e) +
54 		       sizeof(struct nf_hook_entry) * num +
55 		       sizeof(struct nf_hook_ops *) * num +
56 		       sizeof(struct nf_hook_entries_rcu_head);
57 
58 	if (num == 0)
59 		return NULL;
60 
61 	e = kvzalloc(alloc, GFP_KERNEL_ACCOUNT);
62 	if (e)
63 		e->num_hook_entries = num;
64 	return e;
65 }
66 
__nf_hook_entries_free(struct rcu_head * h)67 static void __nf_hook_entries_free(struct rcu_head *h)
68 {
69 	struct nf_hook_entries_rcu_head *head;
70 
71 	head = container_of(h, struct nf_hook_entries_rcu_head, head);
72 	kvfree(head->allocation);
73 }
74 
nf_hook_entries_free(struct nf_hook_entries * e)75 static void nf_hook_entries_free(struct nf_hook_entries *e)
76 {
77 	struct nf_hook_entries_rcu_head *head;
78 	struct nf_hook_ops **ops;
79 	unsigned int num;
80 
81 	if (!e)
82 		return;
83 
84 	num = e->num_hook_entries;
85 	ops = nf_hook_entries_get_hook_ops(e);
86 	head = (void *)&ops[num];
87 	head->allocation = e;
88 	call_rcu(&head->head, __nf_hook_entries_free);
89 }
90 
accept_all(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)91 static unsigned int accept_all(void *priv,
92 			       struct sk_buff *skb,
93 			       const struct nf_hook_state *state)
94 {
95 	return NF_ACCEPT; /* ACCEPT makes nf_hook_slow call next hook */
96 }
97 
98 static const struct nf_hook_ops dummy_ops = {
99 	.hook = accept_all,
100 	.priority = INT_MIN,
101 };
102 
103 static struct nf_hook_entries *
nf_hook_entries_grow(const struct nf_hook_entries * old,const struct nf_hook_ops * reg)104 nf_hook_entries_grow(const struct nf_hook_entries *old,
105 		     const struct nf_hook_ops *reg)
106 {
107 	unsigned int i, alloc_entries, nhooks, old_entries;
108 	struct nf_hook_ops **orig_ops = NULL;
109 	struct nf_hook_ops **new_ops;
110 	struct nf_hook_entries *new;
111 	bool inserted = false;
112 
113 	alloc_entries = 1;
114 	old_entries = old ? old->num_hook_entries : 0;
115 
116 	if (old) {
117 		orig_ops = nf_hook_entries_get_hook_ops(old);
118 
119 		for (i = 0; i < old_entries; i++) {
120 			if (orig_ops[i] != &dummy_ops)
121 				alloc_entries++;
122 
123 			/* Restrict BPF hook type to force a unique priority, not
124 			 * shared at attach time.
125 			 *
126 			 * This is mainly to avoid ordering issues between two
127 			 * different bpf programs, this doesn't prevent a normal
128 			 * hook at same priority as a bpf one (we don't want to
129 			 * prevent defrag, conntrack, iptables etc from attaching).
130 			 */
131 			if (reg->priority == orig_ops[i]->priority &&
132 			    reg->hook_ops_type == NF_HOOK_OP_BPF)
133 				return ERR_PTR(-EBUSY);
134 		}
135 	}
136 
137 	if (alloc_entries > MAX_HOOK_COUNT)
138 		return ERR_PTR(-E2BIG);
139 
140 	new = allocate_hook_entries_size(alloc_entries);
141 	if (!new)
142 		return ERR_PTR(-ENOMEM);
143 
144 	new_ops = nf_hook_entries_get_hook_ops(new);
145 
146 	i = 0;
147 	nhooks = 0;
148 	while (i < old_entries) {
149 		if (orig_ops[i] == &dummy_ops) {
150 			++i;
151 			continue;
152 		}
153 
154 		if (inserted || reg->priority > orig_ops[i]->priority) {
155 			new_ops[nhooks] = (void *)orig_ops[i];
156 			new->hooks[nhooks] = old->hooks[i];
157 			i++;
158 		} else {
159 			new_ops[nhooks] = (void *)reg;
160 			new->hooks[nhooks].hook = reg->hook;
161 			new->hooks[nhooks].priv = reg->priv;
162 			inserted = true;
163 		}
164 		nhooks++;
165 	}
166 
167 	if (!inserted) {
168 		new_ops[nhooks] = (void *)reg;
169 		new->hooks[nhooks].hook = reg->hook;
170 		new->hooks[nhooks].priv = reg->priv;
171 	}
172 
173 	return new;
174 }
175 
hooks_validate(const struct nf_hook_entries * hooks)176 static void hooks_validate(const struct nf_hook_entries *hooks)
177 {
178 #ifdef CONFIG_DEBUG_MISC
179 	struct nf_hook_ops **orig_ops;
180 	int prio = INT_MIN;
181 	size_t i = 0;
182 
183 	orig_ops = nf_hook_entries_get_hook_ops(hooks);
184 
185 	for (i = 0; i < hooks->num_hook_entries; i++) {
186 		if (orig_ops[i] == &dummy_ops)
187 			continue;
188 
189 		WARN_ON(orig_ops[i]->priority < prio);
190 
191 		if (orig_ops[i]->priority > prio)
192 			prio = orig_ops[i]->priority;
193 	}
194 #endif
195 }
196 
nf_hook_entries_insert_raw(struct nf_hook_entries __rcu ** pp,const struct nf_hook_ops * reg)197 int nf_hook_entries_insert_raw(struct nf_hook_entries __rcu **pp,
198 				const struct nf_hook_ops *reg)
199 {
200 	struct nf_hook_entries *new_hooks;
201 	struct nf_hook_entries *p;
202 
203 	p = rcu_dereference_raw(*pp);
204 	new_hooks = nf_hook_entries_grow(p, reg);
205 	if (IS_ERR(new_hooks))
206 		return PTR_ERR(new_hooks);
207 
208 	hooks_validate(new_hooks);
209 
210 	rcu_assign_pointer(*pp, new_hooks);
211 
212 	BUG_ON(p == new_hooks);
213 	nf_hook_entries_free(p);
214 	return 0;
215 }
216 EXPORT_SYMBOL_GPL(nf_hook_entries_insert_raw);
217 
218 /*
219  * __nf_hook_entries_try_shrink - try to shrink hook array
220  *
221  * @old -- current hook blob at @pp
222  * @pp -- location of hook blob
223  *
224  * Hook unregistration must always succeed, so to-be-removed hooks
225  * are replaced by a dummy one that will just move to next hook.
226  *
227  * This counts the current dummy hooks, attempts to allocate new blob,
228  * copies the live hooks, then replaces and discards old one.
229  *
230  * return values:
231  *
232  * Returns address to free, or NULL.
233  */
__nf_hook_entries_try_shrink(struct nf_hook_entries * old,struct nf_hook_entries __rcu ** pp)234 static void *__nf_hook_entries_try_shrink(struct nf_hook_entries *old,
235 					  struct nf_hook_entries __rcu **pp)
236 {
237 	unsigned int i, j, skip = 0, hook_entries;
238 	struct nf_hook_entries *new = NULL;
239 	struct nf_hook_ops **orig_ops;
240 	struct nf_hook_ops **new_ops;
241 
242 	if (WARN_ON_ONCE(!old))
243 		return NULL;
244 
245 	orig_ops = nf_hook_entries_get_hook_ops(old);
246 	for (i = 0; i < old->num_hook_entries; i++) {
247 		if (orig_ops[i] == &dummy_ops)
248 			skip++;
249 	}
250 
251 	/* if skip == hook_entries all hooks have been removed */
252 	hook_entries = old->num_hook_entries;
253 	if (skip == hook_entries)
254 		goto out_assign;
255 
256 	if (skip == 0)
257 		return NULL;
258 
259 	hook_entries -= skip;
260 	new = allocate_hook_entries_size(hook_entries);
261 	if (!new)
262 		return NULL;
263 
264 	new_ops = nf_hook_entries_get_hook_ops(new);
265 	for (i = 0, j = 0; i < old->num_hook_entries; i++) {
266 		if (orig_ops[i] == &dummy_ops)
267 			continue;
268 		new->hooks[j] = old->hooks[i];
269 		new_ops[j] = (void *)orig_ops[i];
270 		j++;
271 	}
272 	hooks_validate(new);
273 out_assign:
274 	rcu_assign_pointer(*pp, new);
275 	return old;
276 }
277 
278 static struct nf_hook_entries __rcu **
nf_hook_entry_head(struct net * net,int pf,unsigned int hooknum,struct net_device * dev)279 nf_hook_entry_head(struct net *net, int pf, unsigned int hooknum,
280 		   struct net_device *dev)
281 {
282 	switch (pf) {
283 	case NFPROTO_NETDEV:
284 		break;
285 #ifdef CONFIG_NETFILTER_FAMILY_ARP
286 	case NFPROTO_ARP:
287 		if (WARN_ON_ONCE(ARRAY_SIZE(net->nf.hooks_arp) <= hooknum))
288 			return NULL;
289 		return net->nf.hooks_arp + hooknum;
290 #endif
291 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
292 	case NFPROTO_BRIDGE:
293 		if (WARN_ON_ONCE(ARRAY_SIZE(net->nf.hooks_bridge) <= hooknum))
294 			return NULL;
295 		return net->nf.hooks_bridge + hooknum;
296 #endif
297 #ifdef CONFIG_NETFILTER_INGRESS
298 	case NFPROTO_INET:
299 		if (WARN_ON_ONCE(hooknum != NF_INET_INGRESS))
300 			return NULL;
301 		if (!dev || dev_net(dev) != net) {
302 			WARN_ON_ONCE(1);
303 			return NULL;
304 		}
305 		return &dev->nf_hooks_ingress;
306 #endif
307 	case NFPROTO_IPV4:
308 		if (WARN_ON_ONCE(ARRAY_SIZE(net->nf.hooks_ipv4) <= hooknum))
309 			return NULL;
310 		return net->nf.hooks_ipv4 + hooknum;
311 	case NFPROTO_IPV6:
312 		if (WARN_ON_ONCE(ARRAY_SIZE(net->nf.hooks_ipv6) <= hooknum))
313 			return NULL;
314 		return net->nf.hooks_ipv6 + hooknum;
315 	default:
316 		WARN_ON_ONCE(1);
317 		return NULL;
318 	}
319 
320 #ifdef CONFIG_NETFILTER_INGRESS
321 	if (hooknum == NF_NETDEV_INGRESS) {
322 		if (dev && dev_net(dev) == net)
323 			return &dev->nf_hooks_ingress;
324 	}
325 #endif
326 #ifdef CONFIG_NETFILTER_EGRESS
327 	if (hooknum == NF_NETDEV_EGRESS) {
328 		if (dev && dev_net(dev) == net)
329 			return &dev->nf_hooks_egress;
330 	}
331 #endif
332 	WARN_ON_ONCE(1);
333 	return NULL;
334 }
335 
nf_ingress_check(struct net * net,const struct nf_hook_ops * reg,int hooknum)336 static int nf_ingress_check(struct net *net, const struct nf_hook_ops *reg,
337 			    int hooknum)
338 {
339 #ifndef CONFIG_NETFILTER_INGRESS
340 	if (reg->hooknum == hooknum)
341 		return -EOPNOTSUPP;
342 #endif
343 	if (reg->hooknum != hooknum ||
344 	    !reg->dev || dev_net(reg->dev) != net)
345 		return -EINVAL;
346 
347 	return 0;
348 }
349 
nf_ingress_hook(const struct nf_hook_ops * reg,int pf)350 static inline bool __maybe_unused nf_ingress_hook(const struct nf_hook_ops *reg,
351 						  int pf)
352 {
353 	if ((pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_INGRESS) ||
354 	    (pf == NFPROTO_INET && reg->hooknum == NF_INET_INGRESS))
355 		return true;
356 
357 	return false;
358 }
359 
nf_egress_hook(const struct nf_hook_ops * reg,int pf)360 static inline bool __maybe_unused nf_egress_hook(const struct nf_hook_ops *reg,
361 						 int pf)
362 {
363 	return pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_EGRESS;
364 }
365 
nf_static_key_inc(const struct nf_hook_ops * reg,int pf)366 static void nf_static_key_inc(const struct nf_hook_ops *reg, int pf)
367 {
368 #ifdef CONFIG_JUMP_LABEL
369 	int hooknum;
370 
371 	if (pf == NFPROTO_INET && reg->hooknum == NF_INET_INGRESS) {
372 		pf = NFPROTO_NETDEV;
373 		hooknum = NF_NETDEV_INGRESS;
374 	} else {
375 		hooknum = reg->hooknum;
376 	}
377 	static_key_slow_inc(&nf_hooks_needed[pf][hooknum]);
378 #endif
379 }
380 
nf_static_key_dec(const struct nf_hook_ops * reg,int pf)381 static void nf_static_key_dec(const struct nf_hook_ops *reg, int pf)
382 {
383 #ifdef CONFIG_JUMP_LABEL
384 	int hooknum;
385 
386 	if (pf == NFPROTO_INET && reg->hooknum == NF_INET_INGRESS) {
387 		pf = NFPROTO_NETDEV;
388 		hooknum = NF_NETDEV_INGRESS;
389 	} else {
390 		hooknum = reg->hooknum;
391 	}
392 	static_key_slow_dec(&nf_hooks_needed[pf][hooknum]);
393 #endif
394 }
395 
__nf_register_net_hook(struct net * net,int pf,const struct nf_hook_ops * reg)396 static int __nf_register_net_hook(struct net *net, int pf,
397 				  const struct nf_hook_ops *reg)
398 {
399 	struct nf_hook_entries *p, *new_hooks;
400 	struct nf_hook_entries __rcu **pp;
401 	int err;
402 
403 	switch (pf) {
404 	case NFPROTO_NETDEV:
405 #ifndef CONFIG_NETFILTER_INGRESS
406 		if (reg->hooknum == NF_NETDEV_INGRESS)
407 			return -EOPNOTSUPP;
408 #endif
409 #ifndef CONFIG_NETFILTER_EGRESS
410 		if (reg->hooknum == NF_NETDEV_EGRESS)
411 			return -EOPNOTSUPP;
412 #endif
413 		if ((reg->hooknum != NF_NETDEV_INGRESS &&
414 		     reg->hooknum != NF_NETDEV_EGRESS) ||
415 		    !reg->dev || dev_net(reg->dev) != net)
416 			return -EINVAL;
417 		break;
418 	case NFPROTO_INET:
419 		if (reg->hooknum != NF_INET_INGRESS)
420 			break;
421 
422 		err = nf_ingress_check(net, reg, NF_INET_INGRESS);
423 		if (err < 0)
424 			return err;
425 		break;
426 	}
427 
428 	pp = nf_hook_entry_head(net, pf, reg->hooknum, reg->dev);
429 	if (!pp)
430 		return -EINVAL;
431 
432 	mutex_lock(&nf_hook_mutex);
433 
434 	p = nf_entry_dereference(*pp);
435 	new_hooks = nf_hook_entries_grow(p, reg);
436 
437 	if (!IS_ERR(new_hooks)) {
438 		hooks_validate(new_hooks);
439 		rcu_assign_pointer(*pp, new_hooks);
440 	}
441 
442 	mutex_unlock(&nf_hook_mutex);
443 	if (IS_ERR(new_hooks))
444 		return PTR_ERR(new_hooks);
445 
446 #ifdef CONFIG_NETFILTER_INGRESS
447 	if (nf_ingress_hook(reg, pf))
448 		net_inc_ingress_queue();
449 #endif
450 #ifdef CONFIG_NETFILTER_EGRESS
451 	if (nf_egress_hook(reg, pf))
452 		net_inc_egress_queue();
453 #endif
454 	nf_static_key_inc(reg, pf);
455 
456 	BUG_ON(p == new_hooks);
457 	nf_hook_entries_free(p);
458 	return 0;
459 }
460 
461 /*
462  * nf_remove_net_hook - remove a hook from blob
463  *
464  * @oldp: current address of hook blob
465  * @unreg: hook to unregister
466  *
467  * This cannot fail, hook unregistration must always succeed.
468  * Therefore replace the to-be-removed hook with a dummy hook.
469  */
nf_remove_net_hook(struct nf_hook_entries * old,const struct nf_hook_ops * unreg)470 static bool nf_remove_net_hook(struct nf_hook_entries *old,
471 			       const struct nf_hook_ops *unreg)
472 {
473 	struct nf_hook_ops **orig_ops;
474 	unsigned int i;
475 
476 	orig_ops = nf_hook_entries_get_hook_ops(old);
477 	for (i = 0; i < old->num_hook_entries; i++) {
478 		if (orig_ops[i] != unreg)
479 			continue;
480 		WRITE_ONCE(old->hooks[i].hook, accept_all);
481 		WRITE_ONCE(orig_ops[i], (void *)&dummy_ops);
482 		return true;
483 	}
484 
485 	return false;
486 }
487 
__nf_unregister_net_hook(struct net * net,int pf,const struct nf_hook_ops * reg)488 static void __nf_unregister_net_hook(struct net *net, int pf,
489 				     const struct nf_hook_ops *reg)
490 {
491 	struct nf_hook_entries __rcu **pp;
492 	struct nf_hook_entries *p;
493 
494 	pp = nf_hook_entry_head(net, pf, reg->hooknum, reg->dev);
495 	if (!pp)
496 		return;
497 
498 	mutex_lock(&nf_hook_mutex);
499 
500 	p = nf_entry_dereference(*pp);
501 	if (WARN_ON_ONCE(!p)) {
502 		mutex_unlock(&nf_hook_mutex);
503 		return;
504 	}
505 
506 	if (nf_remove_net_hook(p, reg)) {
507 #ifdef CONFIG_NETFILTER_INGRESS
508 		if (nf_ingress_hook(reg, pf))
509 			net_dec_ingress_queue();
510 #endif
511 #ifdef CONFIG_NETFILTER_EGRESS
512 		if (nf_egress_hook(reg, pf))
513 			net_dec_egress_queue();
514 #endif
515 		nf_static_key_dec(reg, pf);
516 	} else {
517 		WARN_ONCE(1, "hook not found, pf %d num %d", pf, reg->hooknum);
518 	}
519 
520 	p = __nf_hook_entries_try_shrink(p, pp);
521 	mutex_unlock(&nf_hook_mutex);
522 	if (!p)
523 		return;
524 
525 	nf_queue_nf_hook_drop(net);
526 	nf_hook_entries_free(p);
527 }
528 
nf_unregister_net_hook(struct net * net,const struct nf_hook_ops * reg)529 void nf_unregister_net_hook(struct net *net, const struct nf_hook_ops *reg)
530 {
531 	if (reg->pf == NFPROTO_INET) {
532 		if (reg->hooknum == NF_INET_INGRESS) {
533 			__nf_unregister_net_hook(net, NFPROTO_INET, reg);
534 		} else {
535 			__nf_unregister_net_hook(net, NFPROTO_IPV4, reg);
536 			__nf_unregister_net_hook(net, NFPROTO_IPV6, reg);
537 		}
538 	} else {
539 		__nf_unregister_net_hook(net, reg->pf, reg);
540 	}
541 }
542 EXPORT_SYMBOL(nf_unregister_net_hook);
543 
nf_hook_entries_delete_raw(struct nf_hook_entries __rcu ** pp,const struct nf_hook_ops * reg)544 void nf_hook_entries_delete_raw(struct nf_hook_entries __rcu **pp,
545 				const struct nf_hook_ops *reg)
546 {
547 	struct nf_hook_entries *p;
548 
549 	p = rcu_dereference_raw(*pp);
550 	if (nf_remove_net_hook(p, reg)) {
551 		p = __nf_hook_entries_try_shrink(p, pp);
552 		nf_hook_entries_free(p);
553 	}
554 }
555 EXPORT_SYMBOL_GPL(nf_hook_entries_delete_raw);
556 
nf_register_net_hook(struct net * net,const struct nf_hook_ops * reg)557 int nf_register_net_hook(struct net *net, const struct nf_hook_ops *reg)
558 {
559 	int err;
560 
561 	if (reg->pf == NFPROTO_INET) {
562 		if (reg->hooknum == NF_INET_INGRESS) {
563 			err = __nf_register_net_hook(net, NFPROTO_INET, reg);
564 			if (err < 0)
565 				return err;
566 		} else {
567 			err = __nf_register_net_hook(net, NFPROTO_IPV4, reg);
568 			if (err < 0)
569 				return err;
570 
571 			err = __nf_register_net_hook(net, NFPROTO_IPV6, reg);
572 			if (err < 0) {
573 				__nf_unregister_net_hook(net, NFPROTO_IPV4, reg);
574 				return err;
575 			}
576 		}
577 	} else {
578 		err = __nf_register_net_hook(net, reg->pf, reg);
579 		if (err < 0)
580 			return err;
581 	}
582 
583 	return 0;
584 }
585 EXPORT_SYMBOL(nf_register_net_hook);
586 
nf_register_net_hooks(struct net * net,const struct nf_hook_ops * reg,unsigned int n)587 int nf_register_net_hooks(struct net *net, const struct nf_hook_ops *reg,
588 			  unsigned int n)
589 {
590 	unsigned int i;
591 	int err = 0;
592 
593 	for (i = 0; i < n; i++) {
594 		err = nf_register_net_hook(net, &reg[i]);
595 		if (err)
596 			goto err;
597 	}
598 	return err;
599 
600 err:
601 	if (i > 0)
602 		nf_unregister_net_hooks(net, reg, i);
603 	return err;
604 }
605 EXPORT_SYMBOL(nf_register_net_hooks);
606 
nf_unregister_net_hooks(struct net * net,const struct nf_hook_ops * reg,unsigned int hookcount)607 void nf_unregister_net_hooks(struct net *net, const struct nf_hook_ops *reg,
608 			     unsigned int hookcount)
609 {
610 	unsigned int i;
611 
612 	for (i = 0; i < hookcount; i++)
613 		nf_unregister_net_hook(net, &reg[i]);
614 }
615 EXPORT_SYMBOL(nf_unregister_net_hooks);
616 
617 /* Returns 1 if okfn() needs to be executed by the caller,
618  * -EPERM for NF_DROP, 0 otherwise.  Caller must hold rcu_read_lock. */
nf_hook_slow(struct sk_buff * skb,struct nf_hook_state * state,const struct nf_hook_entries * e,unsigned int s)619 int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state,
620 		 const struct nf_hook_entries *e, unsigned int s)
621 {
622 	unsigned int verdict;
623 	int ret;
624 
625 	for (; s < e->num_hook_entries; s++) {
626 		verdict = nf_hook_entry_hookfn(&e->hooks[s], skb, state);
627 		switch (verdict & NF_VERDICT_MASK) {
628 		case NF_ACCEPT:
629 			break;
630 		case NF_DROP:
631 			kfree_skb_reason(skb,
632 					 SKB_DROP_REASON_NETFILTER_DROP);
633 			ret = NF_DROP_GETERR(verdict);
634 			if (ret == 0)
635 				ret = -EPERM;
636 			return ret;
637 		case NF_QUEUE:
638 			ret = nf_queue(skb, state, s, verdict);
639 			if (ret == 1)
640 				continue;
641 			return ret;
642 		default:
643 			/* Implicit handling for NF_STOLEN, as well as any other
644 			 * non conventional verdicts.
645 			 */
646 			return 0;
647 		}
648 	}
649 
650 	return 1;
651 }
652 EXPORT_SYMBOL(nf_hook_slow);
653 
nf_hook_slow_list(struct list_head * head,struct nf_hook_state * state,const struct nf_hook_entries * e)654 void nf_hook_slow_list(struct list_head *head, struct nf_hook_state *state,
655 		       const struct nf_hook_entries *e)
656 {
657 	struct sk_buff *skb, *next;
658 	struct list_head sublist;
659 	int ret;
660 
661 	INIT_LIST_HEAD(&sublist);
662 
663 	list_for_each_entry_safe(skb, next, head, list) {
664 		skb_list_del_init(skb);
665 		ret = nf_hook_slow(skb, state, e, 0);
666 		if (ret == 1)
667 			list_add_tail(&skb->list, &sublist);
668 	}
669 	/* Put passed packets back on main list */
670 	list_splice(&sublist, head);
671 }
672 EXPORT_SYMBOL(nf_hook_slow_list);
673 
674 /* This needs to be compiled in any case to avoid dependencies between the
675  * nfnetlink_queue code and nf_conntrack.
676  */
677 const struct nfnl_ct_hook __rcu *nfnl_ct_hook __read_mostly;
678 EXPORT_SYMBOL_GPL(nfnl_ct_hook);
679 
680 const struct nf_ct_hook __rcu *nf_ct_hook __read_mostly;
681 EXPORT_SYMBOL_GPL(nf_ct_hook);
682 
683 const struct nf_defrag_hook __rcu *nf_defrag_v4_hook __read_mostly;
684 EXPORT_SYMBOL_GPL(nf_defrag_v4_hook);
685 
686 const struct nf_defrag_hook __rcu *nf_defrag_v6_hook __read_mostly;
687 EXPORT_SYMBOL_GPL(nf_defrag_v6_hook);
688 
689 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
690 u8 nf_ctnetlink_has_listener;
691 EXPORT_SYMBOL_GPL(nf_ctnetlink_has_listener);
692 
693 const struct nf_nat_hook __rcu *nf_nat_hook __read_mostly;
694 EXPORT_SYMBOL_GPL(nf_nat_hook);
695 
696 /* This does not belong here, but locally generated errors need it if connection
697  * tracking in use: without this, connection may not be in hash table, and hence
698  * manufactured ICMP or RST packets will not be associated with it.
699  */
nf_ct_attach(struct sk_buff * new,const struct sk_buff * skb)700 void nf_ct_attach(struct sk_buff *new, const struct sk_buff *skb)
701 {
702 	const struct nf_ct_hook *ct_hook;
703 
704 	if (skb->_nfct) {
705 		rcu_read_lock();
706 		ct_hook = rcu_dereference(nf_ct_hook);
707 		if (ct_hook)
708 			ct_hook->attach(new, skb);
709 		rcu_read_unlock();
710 	}
711 }
712 EXPORT_SYMBOL(nf_ct_attach);
713 
nf_conntrack_destroy(struct nf_conntrack * nfct)714 void nf_conntrack_destroy(struct nf_conntrack *nfct)
715 {
716 	const struct nf_ct_hook *ct_hook;
717 
718 	rcu_read_lock();
719 	ct_hook = rcu_dereference(nf_ct_hook);
720 	if (ct_hook)
721 		ct_hook->destroy(nfct);
722 	rcu_read_unlock();
723 
724 	WARN_ON(!ct_hook);
725 }
726 EXPORT_SYMBOL(nf_conntrack_destroy);
727 
nf_ct_set_closing(struct nf_conntrack * nfct)728 void nf_ct_set_closing(struct nf_conntrack *nfct)
729 {
730 	const struct nf_ct_hook *ct_hook;
731 
732 	if (!nfct)
733 		return;
734 
735 	rcu_read_lock();
736 	ct_hook = rcu_dereference(nf_ct_hook);
737 	if (ct_hook)
738 		ct_hook->set_closing(nfct);
739 
740 	rcu_read_unlock();
741 }
742 EXPORT_SYMBOL_GPL(nf_ct_set_closing);
743 
nf_ct_get_tuple_skb(struct nf_conntrack_tuple * dst_tuple,const struct sk_buff * skb)744 bool nf_ct_get_tuple_skb(struct nf_conntrack_tuple *dst_tuple,
745 			 const struct sk_buff *skb)
746 {
747 	const struct nf_ct_hook *ct_hook;
748 	bool ret = false;
749 
750 	rcu_read_lock();
751 	ct_hook = rcu_dereference(nf_ct_hook);
752 	if (ct_hook)
753 		ret = ct_hook->get_tuple_skb(dst_tuple, skb);
754 	rcu_read_unlock();
755 	return ret;
756 }
757 EXPORT_SYMBOL(nf_ct_get_tuple_skb);
758 
759 /* Built-in default zone used e.g. by modules. */
760 const struct nf_conntrack_zone nf_ct_zone_dflt = {
761 	.id	= NF_CT_DEFAULT_ZONE_ID,
762 	.dir	= NF_CT_DEFAULT_ZONE_DIR,
763 };
764 EXPORT_SYMBOL_GPL(nf_ct_zone_dflt);
765 #endif /* CONFIG_NF_CONNTRACK */
766 
767 static void __net_init
__netfilter_net_init(struct nf_hook_entries __rcu ** e,int max)768 __netfilter_net_init(struct nf_hook_entries __rcu **e, int max)
769 {
770 	int h;
771 
772 	for (h = 0; h < max; h++)
773 		RCU_INIT_POINTER(e[h], NULL);
774 }
775 
netfilter_net_init(struct net * net)776 static int __net_init netfilter_net_init(struct net *net)
777 {
778 	__netfilter_net_init(net->nf.hooks_ipv4, ARRAY_SIZE(net->nf.hooks_ipv4));
779 	__netfilter_net_init(net->nf.hooks_ipv6, ARRAY_SIZE(net->nf.hooks_ipv6));
780 #ifdef CONFIG_NETFILTER_FAMILY_ARP
781 	__netfilter_net_init(net->nf.hooks_arp, ARRAY_SIZE(net->nf.hooks_arp));
782 #endif
783 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
784 	__netfilter_net_init(net->nf.hooks_bridge, ARRAY_SIZE(net->nf.hooks_bridge));
785 #endif
786 #ifdef CONFIG_PROC_FS
787 	net->nf.proc_netfilter = proc_net_mkdir(net, "netfilter",
788 						net->proc_net);
789 	if (!net->nf.proc_netfilter) {
790 		if (!net_eq(net, &init_net))
791 			pr_err("cannot create netfilter proc entry");
792 
793 		return -ENOMEM;
794 	}
795 #endif
796 
797 	return 0;
798 }
799 
netfilter_net_exit(struct net * net)800 static void __net_exit netfilter_net_exit(struct net *net)
801 {
802 	remove_proc_entry("netfilter", net->proc_net);
803 }
804 
805 static struct pernet_operations netfilter_net_ops = {
806 	.init = netfilter_net_init,
807 	.exit = netfilter_net_exit,
808 };
809 
netfilter_init(void)810 int __init netfilter_init(void)
811 {
812 	int ret;
813 
814 	ret = register_pernet_subsys(&netfilter_net_ops);
815 	if (ret < 0)
816 		goto err;
817 
818 #ifdef CONFIG_LWTUNNEL
819 	ret = netfilter_lwtunnel_init();
820 	if (ret < 0)
821 		goto err_lwtunnel_pernet;
822 #endif
823 	ret = netfilter_log_init();
824 	if (ret < 0)
825 		goto err_log_pernet;
826 
827 	return 0;
828 err_log_pernet:
829 #ifdef CONFIG_LWTUNNEL
830 	netfilter_lwtunnel_fini();
831 err_lwtunnel_pernet:
832 #endif
833 	unregister_pernet_subsys(&netfilter_net_ops);
834 err:
835 	return ret;
836 }
837