xref: /openbmc/linux/net/ipv6/netfilter/ip6_tables.c (revision d63670d2)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Packet matching code.
4  *
5  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
6  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
7  * Copyright (c) 2006-2010 Patrick McHardy <kaber@trash.net>
8  */
9 
10 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
11 
12 #include <linux/kernel.h>
13 #include <linux/capability.h>
14 #include <linux/in.h>
15 #include <linux/skbuff.h>
16 #include <linux/kmod.h>
17 #include <linux/vmalloc.h>
18 #include <linux/netdevice.h>
19 #include <linux/module.h>
20 #include <linux/poison.h>
21 #include <linux/icmpv6.h>
22 #include <net/ipv6.h>
23 #include <net/compat.h>
24 #include <linux/uaccess.h>
25 #include <linux/mutex.h>
26 #include <linux/proc_fs.h>
27 #include <linux/err.h>
28 #include <linux/cpumask.h>
29 
30 #include <linux/netfilter_ipv6/ip6_tables.h>
31 #include <linux/netfilter/x_tables.h>
32 #include <net/netfilter/nf_log.h>
33 #include "../../netfilter/xt_repldata.h"
34 
35 MODULE_LICENSE("GPL");
36 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
37 MODULE_DESCRIPTION("IPv6 packet filter");
38 MODULE_ALIAS("ip6t_icmp6");
39 
40 void *ip6t_alloc_initial_table(const struct xt_table *info)
41 {
42 	return xt_alloc_initial_table(ip6t, IP6T);
43 }
44 EXPORT_SYMBOL_GPL(ip6t_alloc_initial_table);
45 
46 /* Returns whether matches rule or not. */
47 /* Performance critical - called for every packet */
48 static inline bool
49 ip6_packet_match(const struct sk_buff *skb,
50 		 const char *indev,
51 		 const char *outdev,
52 		 const struct ip6t_ip6 *ip6info,
53 		 unsigned int *protoff,
54 		 u16 *fragoff, bool *hotdrop)
55 {
56 	unsigned long ret;
57 	const struct ipv6hdr *ipv6 = ipv6_hdr(skb);
58 
59 	if (NF_INVF(ip6info, IP6T_INV_SRCIP,
60 		    ipv6_masked_addr_cmp(&ipv6->saddr, &ip6info->smsk,
61 					 &ip6info->src)) ||
62 	    NF_INVF(ip6info, IP6T_INV_DSTIP,
63 		    ipv6_masked_addr_cmp(&ipv6->daddr, &ip6info->dmsk,
64 					 &ip6info->dst)))
65 		return false;
66 
67 	ret = ifname_compare_aligned(indev, ip6info->iniface, ip6info->iniface_mask);
68 
69 	if (NF_INVF(ip6info, IP6T_INV_VIA_IN, ret != 0))
70 		return false;
71 
72 	ret = ifname_compare_aligned(outdev, ip6info->outiface, ip6info->outiface_mask);
73 
74 	if (NF_INVF(ip6info, IP6T_INV_VIA_OUT, ret != 0))
75 		return false;
76 
77 /* ... might want to do something with class and flowlabel here ... */
78 
79 	/* look for the desired protocol header */
80 	if (ip6info->flags & IP6T_F_PROTO) {
81 		int protohdr;
82 		unsigned short _frag_off;
83 
84 		protohdr = ipv6_find_hdr(skb, protoff, -1, &_frag_off, NULL);
85 		if (protohdr < 0) {
86 			if (_frag_off == 0)
87 				*hotdrop = true;
88 			return false;
89 		}
90 		*fragoff = _frag_off;
91 
92 		if (ip6info->proto == protohdr) {
93 			if (ip6info->invflags & IP6T_INV_PROTO)
94 				return false;
95 
96 			return true;
97 		}
98 
99 		/* We need match for the '-p all', too! */
100 		if ((ip6info->proto != 0) &&
101 			!(ip6info->invflags & IP6T_INV_PROTO))
102 			return false;
103 	}
104 	return true;
105 }
106 
107 /* should be ip6 safe */
108 static bool
109 ip6_checkentry(const struct ip6t_ip6 *ipv6)
110 {
111 	if (ipv6->flags & ~IP6T_F_MASK)
112 		return false;
113 	if (ipv6->invflags & ~IP6T_INV_MASK)
114 		return false;
115 
116 	return true;
117 }
118 
119 static unsigned int
120 ip6t_error(struct sk_buff *skb, const struct xt_action_param *par)
121 {
122 	net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
123 
124 	return NF_DROP;
125 }
126 
127 static inline struct ip6t_entry *
128 get_entry(const void *base, unsigned int offset)
129 {
130 	return (struct ip6t_entry *)(base + offset);
131 }
132 
133 /* All zeroes == unconditional rule. */
134 /* Mildly perf critical (only if packet tracing is on) */
135 static inline bool unconditional(const struct ip6t_entry *e)
136 {
137 	static const struct ip6t_ip6 uncond;
138 
139 	return e->target_offset == sizeof(struct ip6t_entry) &&
140 	       memcmp(&e->ipv6, &uncond, sizeof(uncond)) == 0;
141 }
142 
143 static inline const struct xt_entry_target *
144 ip6t_get_target_c(const struct ip6t_entry *e)
145 {
146 	return ip6t_get_target((struct ip6t_entry *)e);
147 }
148 
149 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
150 /* This cries for unification! */
151 static const char *const hooknames[] = {
152 	[NF_INET_PRE_ROUTING]		= "PREROUTING",
153 	[NF_INET_LOCAL_IN]		= "INPUT",
154 	[NF_INET_FORWARD]		= "FORWARD",
155 	[NF_INET_LOCAL_OUT]		= "OUTPUT",
156 	[NF_INET_POST_ROUTING]		= "POSTROUTING",
157 };
158 
159 enum nf_ip_trace_comments {
160 	NF_IP6_TRACE_COMMENT_RULE,
161 	NF_IP6_TRACE_COMMENT_RETURN,
162 	NF_IP6_TRACE_COMMENT_POLICY,
163 };
164 
165 static const char *const comments[] = {
166 	[NF_IP6_TRACE_COMMENT_RULE]	= "rule",
167 	[NF_IP6_TRACE_COMMENT_RETURN]	= "return",
168 	[NF_IP6_TRACE_COMMENT_POLICY]	= "policy",
169 };
170 
171 static const struct nf_loginfo trace_loginfo = {
172 	.type = NF_LOG_TYPE_LOG,
173 	.u = {
174 		.log = {
175 			.level = LOGLEVEL_WARNING,
176 			.logflags = NF_LOG_DEFAULT_MASK,
177 		},
178 	},
179 };
180 
181 /* Mildly perf critical (only if packet tracing is on) */
182 static inline int
183 get_chainname_rulenum(const struct ip6t_entry *s, const struct ip6t_entry *e,
184 		      const char *hookname, const char **chainname,
185 		      const char **comment, unsigned int *rulenum)
186 {
187 	const struct xt_standard_target *t = (void *)ip6t_get_target_c(s);
188 
189 	if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
190 		/* Head of user chain: ERROR target with chainname */
191 		*chainname = t->target.data;
192 		(*rulenum) = 0;
193 	} else if (s == e) {
194 		(*rulenum)++;
195 
196 		if (unconditional(s) &&
197 		    strcmp(t->target.u.kernel.target->name,
198 			   XT_STANDARD_TARGET) == 0 &&
199 		    t->verdict < 0) {
200 			/* Tail of chains: STANDARD target (return/policy) */
201 			*comment = *chainname == hookname
202 				? comments[NF_IP6_TRACE_COMMENT_POLICY]
203 				: comments[NF_IP6_TRACE_COMMENT_RETURN];
204 		}
205 		return 1;
206 	} else
207 		(*rulenum)++;
208 
209 	return 0;
210 }
211 
212 static void trace_packet(struct net *net,
213 			 const struct sk_buff *skb,
214 			 unsigned int hook,
215 			 const struct net_device *in,
216 			 const struct net_device *out,
217 			 const char *tablename,
218 			 const struct xt_table_info *private,
219 			 const struct ip6t_entry *e)
220 {
221 	const struct ip6t_entry *root;
222 	const char *hookname, *chainname, *comment;
223 	const struct ip6t_entry *iter;
224 	unsigned int rulenum = 0;
225 
226 	root = get_entry(private->entries, private->hook_entry[hook]);
227 
228 	hookname = chainname = hooknames[hook];
229 	comment = comments[NF_IP6_TRACE_COMMENT_RULE];
230 
231 	xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
232 		if (get_chainname_rulenum(iter, e, hookname,
233 		    &chainname, &comment, &rulenum) != 0)
234 			break;
235 
236 	nf_log_trace(net, AF_INET6, hook, skb, in, out, &trace_loginfo,
237 		     "TRACE: %s:%s:%s:%u ",
238 		     tablename, chainname, comment, rulenum);
239 }
240 #endif
241 
242 static inline struct ip6t_entry *
243 ip6t_next_entry(const struct ip6t_entry *entry)
244 {
245 	return (void *)entry + entry->next_offset;
246 }
247 
248 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
249 unsigned int
250 ip6t_do_table(void *priv, struct sk_buff *skb,
251 	      const struct nf_hook_state *state)
252 {
253 	const struct xt_table *table = priv;
254 	unsigned int hook = state->hook;
255 	static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
256 	/* Initializing verdict to NF_DROP keeps gcc happy. */
257 	unsigned int verdict = NF_DROP;
258 	const char *indev, *outdev;
259 	const void *table_base;
260 	struct ip6t_entry *e, **jumpstack;
261 	unsigned int stackidx, cpu;
262 	const struct xt_table_info *private;
263 	struct xt_action_param acpar;
264 	unsigned int addend;
265 
266 	/* Initialization */
267 	stackidx = 0;
268 	indev = state->in ? state->in->name : nulldevname;
269 	outdev = state->out ? state->out->name : nulldevname;
270 	/* We handle fragments by dealing with the first fragment as
271 	 * if it was a normal packet.  All other fragments are treated
272 	 * normally, except that they will NEVER match rules that ask
273 	 * things we don't know, ie. tcp syn flag or ports).  If the
274 	 * rule is also a fragment-specific rule, non-fragments won't
275 	 * match it. */
276 	acpar.fragoff = 0;
277 	acpar.hotdrop = false;
278 	acpar.state   = state;
279 
280 	WARN_ON(!(table->valid_hooks & (1 << hook)));
281 
282 	local_bh_disable();
283 	addend = xt_write_recseq_begin();
284 	private = READ_ONCE(table->private); /* Address dependency. */
285 	cpu        = smp_processor_id();
286 	table_base = private->entries;
287 	jumpstack  = (struct ip6t_entry **)private->jumpstack[cpu];
288 
289 	/* Switch to alternate jumpstack if we're being invoked via TEE.
290 	 * TEE issues XT_CONTINUE verdict on original skb so we must not
291 	 * clobber the jumpstack.
292 	 *
293 	 * For recursion via REJECT or SYNPROXY the stack will be clobbered
294 	 * but it is no problem since absolute verdict is issued by these.
295 	 */
296 	if (static_key_false(&xt_tee_enabled))
297 		jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
298 
299 	e = get_entry(table_base, private->hook_entry[hook]);
300 
301 	do {
302 		const struct xt_entry_target *t;
303 		const struct xt_entry_match *ematch;
304 		struct xt_counters *counter;
305 
306 		WARN_ON(!e);
307 		acpar.thoff = 0;
308 		if (!ip6_packet_match(skb, indev, outdev, &e->ipv6,
309 		    &acpar.thoff, &acpar.fragoff, &acpar.hotdrop)) {
310  no_match:
311 			e = ip6t_next_entry(e);
312 			continue;
313 		}
314 
315 		xt_ematch_foreach(ematch, e) {
316 			acpar.match     = ematch->u.kernel.match;
317 			acpar.matchinfo = ematch->data;
318 			if (!acpar.match->match(skb, &acpar))
319 				goto no_match;
320 		}
321 
322 		counter = xt_get_this_cpu_counter(&e->counters);
323 		ADD_COUNTER(*counter, skb->len, 1);
324 
325 		t = ip6t_get_target_c(e);
326 		WARN_ON(!t->u.kernel.target);
327 
328 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
329 		/* The packet is traced: log it */
330 		if (unlikely(skb->nf_trace))
331 			trace_packet(state->net, skb, hook, state->in,
332 				     state->out, table->name, private, e);
333 #endif
334 		/* Standard target? */
335 		if (!t->u.kernel.target->target) {
336 			int v;
337 
338 			v = ((struct xt_standard_target *)t)->verdict;
339 			if (v < 0) {
340 				/* Pop from stack? */
341 				if (v != XT_RETURN) {
342 					verdict = (unsigned int)(-v) - 1;
343 					break;
344 				}
345 				if (stackidx == 0)
346 					e = get_entry(table_base,
347 					    private->underflow[hook]);
348 				else
349 					e = ip6t_next_entry(jumpstack[--stackidx]);
350 				continue;
351 			}
352 			if (table_base + v != ip6t_next_entry(e) &&
353 			    !(e->ipv6.flags & IP6T_F_GOTO)) {
354 				if (unlikely(stackidx >= private->stacksize)) {
355 					verdict = NF_DROP;
356 					break;
357 				}
358 				jumpstack[stackidx++] = e;
359 			}
360 
361 			e = get_entry(table_base, v);
362 			continue;
363 		}
364 
365 		acpar.target   = t->u.kernel.target;
366 		acpar.targinfo = t->data;
367 
368 		verdict = t->u.kernel.target->target(skb, &acpar);
369 		if (verdict == XT_CONTINUE)
370 			e = ip6t_next_entry(e);
371 		else
372 			/* Verdict */
373 			break;
374 	} while (!acpar.hotdrop);
375 
376 	xt_write_recseq_end(addend);
377 	local_bh_enable();
378 
379 	if (acpar.hotdrop)
380 		return NF_DROP;
381 	else return verdict;
382 }
383 
384 /* Figures out from what hook each rule can be called: returns 0 if
385    there are loops.  Puts hook bitmask in comefrom. */
386 static int
387 mark_source_chains(const struct xt_table_info *newinfo,
388 		   unsigned int valid_hooks, void *entry0,
389 		   unsigned int *offsets)
390 {
391 	unsigned int hook;
392 
393 	/* No recursion; use packet counter to save back ptrs (reset
394 	   to 0 as we leave), and comefrom to save source hook bitmask */
395 	for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
396 		unsigned int pos = newinfo->hook_entry[hook];
397 		struct ip6t_entry *e = entry0 + pos;
398 
399 		if (!(valid_hooks & (1 << hook)))
400 			continue;
401 
402 		/* Set initial back pointer. */
403 		e->counters.pcnt = pos;
404 
405 		for (;;) {
406 			const struct xt_standard_target *t
407 				= (void *)ip6t_get_target_c(e);
408 			int visited = e->comefrom & (1 << hook);
409 
410 			if (e->comefrom & (1 << NF_INET_NUMHOOKS))
411 				return 0;
412 
413 			e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
414 
415 			/* Unconditional return/END. */
416 			if ((unconditional(e) &&
417 			     (strcmp(t->target.u.user.name,
418 				     XT_STANDARD_TARGET) == 0) &&
419 			     t->verdict < 0) || visited) {
420 				unsigned int oldpos, size;
421 
422 				/* Return: backtrack through the last
423 				   big jump. */
424 				do {
425 					e->comefrom ^= (1<<NF_INET_NUMHOOKS);
426 					oldpos = pos;
427 					pos = e->counters.pcnt;
428 					e->counters.pcnt = 0;
429 
430 					/* We're at the start. */
431 					if (pos == oldpos)
432 						goto next;
433 
434 					e = entry0 + pos;
435 				} while (oldpos == pos + e->next_offset);
436 
437 				/* Move along one */
438 				size = e->next_offset;
439 				e = entry0 + pos + size;
440 				if (pos + size >= newinfo->size)
441 					return 0;
442 				e->counters.pcnt = pos;
443 				pos += size;
444 			} else {
445 				int newpos = t->verdict;
446 
447 				if (strcmp(t->target.u.user.name,
448 					   XT_STANDARD_TARGET) == 0 &&
449 				    newpos >= 0) {
450 					/* This a jump; chase it. */
451 					if (!xt_find_jump_offset(offsets, newpos,
452 								 newinfo->number))
453 						return 0;
454 				} else {
455 					/* ... this is a fallthru */
456 					newpos = pos + e->next_offset;
457 					if (newpos >= newinfo->size)
458 						return 0;
459 				}
460 				e = entry0 + newpos;
461 				e->counters.pcnt = pos;
462 				pos = newpos;
463 			}
464 		}
465 next:		;
466 	}
467 	return 1;
468 }
469 
470 static void cleanup_match(struct xt_entry_match *m, struct net *net)
471 {
472 	struct xt_mtdtor_param par;
473 
474 	par.net       = net;
475 	par.match     = m->u.kernel.match;
476 	par.matchinfo = m->data;
477 	par.family    = NFPROTO_IPV6;
478 	if (par.match->destroy != NULL)
479 		par.match->destroy(&par);
480 	module_put(par.match->me);
481 }
482 
483 static int check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
484 {
485 	const struct ip6t_ip6 *ipv6 = par->entryinfo;
486 
487 	par->match     = m->u.kernel.match;
488 	par->matchinfo = m->data;
489 
490 	return xt_check_match(par, m->u.match_size - sizeof(*m),
491 			      ipv6->proto, ipv6->invflags & IP6T_INV_PROTO);
492 }
493 
494 static int
495 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
496 {
497 	struct xt_match *match;
498 	int ret;
499 
500 	match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
501 				      m->u.user.revision);
502 	if (IS_ERR(match))
503 		return PTR_ERR(match);
504 
505 	m->u.kernel.match = match;
506 
507 	ret = check_match(m, par);
508 	if (ret)
509 		goto err;
510 
511 	return 0;
512 err:
513 	module_put(m->u.kernel.match->me);
514 	return ret;
515 }
516 
517 static int check_target(struct ip6t_entry *e, struct net *net, const char *name)
518 {
519 	struct xt_entry_target *t = ip6t_get_target(e);
520 	struct xt_tgchk_param par = {
521 		.net       = net,
522 		.table     = name,
523 		.entryinfo = e,
524 		.target    = t->u.kernel.target,
525 		.targinfo  = t->data,
526 		.hook_mask = e->comefrom,
527 		.family    = NFPROTO_IPV6,
528 	};
529 
530 	return xt_check_target(&par, t->u.target_size - sizeof(*t),
531 			       e->ipv6.proto,
532 			       e->ipv6.invflags & IP6T_INV_PROTO);
533 }
534 
535 static int
536 find_check_entry(struct ip6t_entry *e, struct net *net, const char *name,
537 		 unsigned int size,
538 		 struct xt_percpu_counter_alloc_state *alloc_state)
539 {
540 	struct xt_entry_target *t;
541 	struct xt_target *target;
542 	int ret;
543 	unsigned int j;
544 	struct xt_mtchk_param mtpar;
545 	struct xt_entry_match *ematch;
546 
547 	if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
548 		return -ENOMEM;
549 
550 	j = 0;
551 	memset(&mtpar, 0, sizeof(mtpar));
552 	mtpar.net	= net;
553 	mtpar.table     = name;
554 	mtpar.entryinfo = &e->ipv6;
555 	mtpar.hook_mask = e->comefrom;
556 	mtpar.family    = NFPROTO_IPV6;
557 	xt_ematch_foreach(ematch, e) {
558 		ret = find_check_match(ematch, &mtpar);
559 		if (ret != 0)
560 			goto cleanup_matches;
561 		++j;
562 	}
563 
564 	t = ip6t_get_target(e);
565 	target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
566 					t->u.user.revision);
567 	if (IS_ERR(target)) {
568 		ret = PTR_ERR(target);
569 		goto cleanup_matches;
570 	}
571 	t->u.kernel.target = target;
572 
573 	ret = check_target(e, net, name);
574 	if (ret)
575 		goto err;
576 	return 0;
577  err:
578 	module_put(t->u.kernel.target->me);
579  cleanup_matches:
580 	xt_ematch_foreach(ematch, e) {
581 		if (j-- == 0)
582 			break;
583 		cleanup_match(ematch, net);
584 	}
585 
586 	xt_percpu_counter_free(&e->counters);
587 
588 	return ret;
589 }
590 
591 static bool check_underflow(const struct ip6t_entry *e)
592 {
593 	const struct xt_entry_target *t;
594 	unsigned int verdict;
595 
596 	if (!unconditional(e))
597 		return false;
598 	t = ip6t_get_target_c(e);
599 	if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
600 		return false;
601 	verdict = ((struct xt_standard_target *)t)->verdict;
602 	verdict = -verdict - 1;
603 	return verdict == NF_DROP || verdict == NF_ACCEPT;
604 }
605 
606 static int
607 check_entry_size_and_hooks(struct ip6t_entry *e,
608 			   struct xt_table_info *newinfo,
609 			   const unsigned char *base,
610 			   const unsigned char *limit,
611 			   const unsigned int *hook_entries,
612 			   const unsigned int *underflows,
613 			   unsigned int valid_hooks)
614 {
615 	unsigned int h;
616 	int err;
617 
618 	if ((unsigned long)e % __alignof__(struct ip6t_entry) != 0 ||
619 	    (unsigned char *)e + sizeof(struct ip6t_entry) >= limit ||
620 	    (unsigned char *)e + e->next_offset > limit)
621 		return -EINVAL;
622 
623 	if (e->next_offset
624 	    < sizeof(struct ip6t_entry) + sizeof(struct xt_entry_target))
625 		return -EINVAL;
626 
627 	if (!ip6_checkentry(&e->ipv6))
628 		return -EINVAL;
629 
630 	err = xt_check_entry_offsets(e, e->elems, e->target_offset,
631 				     e->next_offset);
632 	if (err)
633 		return err;
634 
635 	/* Check hooks & underflows */
636 	for (h = 0; h < NF_INET_NUMHOOKS; h++) {
637 		if (!(valid_hooks & (1 << h)))
638 			continue;
639 		if ((unsigned char *)e - base == hook_entries[h])
640 			newinfo->hook_entry[h] = hook_entries[h];
641 		if ((unsigned char *)e - base == underflows[h]) {
642 			if (!check_underflow(e))
643 				return -EINVAL;
644 
645 			newinfo->underflow[h] = underflows[h];
646 		}
647 	}
648 
649 	/* Clear counters and comefrom */
650 	e->counters = ((struct xt_counters) { 0, 0 });
651 	e->comefrom = 0;
652 	return 0;
653 }
654 
655 static void cleanup_entry(struct ip6t_entry *e, struct net *net)
656 {
657 	struct xt_tgdtor_param par;
658 	struct xt_entry_target *t;
659 	struct xt_entry_match *ematch;
660 
661 	/* Cleanup all matches */
662 	xt_ematch_foreach(ematch, e)
663 		cleanup_match(ematch, net);
664 	t = ip6t_get_target(e);
665 
666 	par.net      = net;
667 	par.target   = t->u.kernel.target;
668 	par.targinfo = t->data;
669 	par.family   = NFPROTO_IPV6;
670 	if (par.target->destroy != NULL)
671 		par.target->destroy(&par);
672 	module_put(par.target->me);
673 	xt_percpu_counter_free(&e->counters);
674 }
675 
676 /* Checks and translates the user-supplied table segment (held in
677    newinfo) */
678 static int
679 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
680 		const struct ip6t_replace *repl)
681 {
682 	struct xt_percpu_counter_alloc_state alloc_state = { 0 };
683 	struct ip6t_entry *iter;
684 	unsigned int *offsets;
685 	unsigned int i;
686 	int ret = 0;
687 
688 	newinfo->size = repl->size;
689 	newinfo->number = repl->num_entries;
690 
691 	/* Init all hooks to impossible value. */
692 	for (i = 0; i < NF_INET_NUMHOOKS; i++) {
693 		newinfo->hook_entry[i] = 0xFFFFFFFF;
694 		newinfo->underflow[i] = 0xFFFFFFFF;
695 	}
696 
697 	offsets = xt_alloc_entry_offsets(newinfo->number);
698 	if (!offsets)
699 		return -ENOMEM;
700 	i = 0;
701 	/* Walk through entries, checking offsets. */
702 	xt_entry_foreach(iter, entry0, newinfo->size) {
703 		ret = check_entry_size_and_hooks(iter, newinfo, entry0,
704 						 entry0 + repl->size,
705 						 repl->hook_entry,
706 						 repl->underflow,
707 						 repl->valid_hooks);
708 		if (ret != 0)
709 			goto out_free;
710 		if (i < repl->num_entries)
711 			offsets[i] = (void *)iter - entry0;
712 		++i;
713 		if (strcmp(ip6t_get_target(iter)->u.user.name,
714 		    XT_ERROR_TARGET) == 0)
715 			++newinfo->stacksize;
716 	}
717 
718 	ret = -EINVAL;
719 	if (i != repl->num_entries)
720 		goto out_free;
721 
722 	ret = xt_check_table_hooks(newinfo, repl->valid_hooks);
723 	if (ret)
724 		goto out_free;
725 
726 	if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
727 		ret = -ELOOP;
728 		goto out_free;
729 	}
730 	kvfree(offsets);
731 
732 	/* Finally, each sanity check must pass */
733 	i = 0;
734 	xt_entry_foreach(iter, entry0, newinfo->size) {
735 		ret = find_check_entry(iter, net, repl->name, repl->size,
736 				       &alloc_state);
737 		if (ret != 0)
738 			break;
739 		++i;
740 	}
741 
742 	if (ret != 0) {
743 		xt_entry_foreach(iter, entry0, newinfo->size) {
744 			if (i-- == 0)
745 				break;
746 			cleanup_entry(iter, net);
747 		}
748 		return ret;
749 	}
750 
751 	return ret;
752  out_free:
753 	kvfree(offsets);
754 	return ret;
755 }
756 
757 static void
758 get_counters(const struct xt_table_info *t,
759 	     struct xt_counters counters[])
760 {
761 	struct ip6t_entry *iter;
762 	unsigned int cpu;
763 	unsigned int i;
764 
765 	for_each_possible_cpu(cpu) {
766 		seqcount_t *s = &per_cpu(xt_recseq, cpu);
767 
768 		i = 0;
769 		xt_entry_foreach(iter, t->entries, t->size) {
770 			struct xt_counters *tmp;
771 			u64 bcnt, pcnt;
772 			unsigned int start;
773 
774 			tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
775 			do {
776 				start = read_seqcount_begin(s);
777 				bcnt = tmp->bcnt;
778 				pcnt = tmp->pcnt;
779 			} while (read_seqcount_retry(s, start));
780 
781 			ADD_COUNTER(counters[i], bcnt, pcnt);
782 			++i;
783 			cond_resched();
784 		}
785 	}
786 }
787 
788 static void get_old_counters(const struct xt_table_info *t,
789 			     struct xt_counters counters[])
790 {
791 	struct ip6t_entry *iter;
792 	unsigned int cpu, i;
793 
794 	for_each_possible_cpu(cpu) {
795 		i = 0;
796 		xt_entry_foreach(iter, t->entries, t->size) {
797 			const struct xt_counters *tmp;
798 
799 			tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
800 			ADD_COUNTER(counters[i], tmp->bcnt, tmp->pcnt);
801 			++i;
802 		}
803 		cond_resched();
804 	}
805 }
806 
807 static struct xt_counters *alloc_counters(const struct xt_table *table)
808 {
809 	unsigned int countersize;
810 	struct xt_counters *counters;
811 	const struct xt_table_info *private = table->private;
812 
813 	/* We need atomic snapshot of counters: rest doesn't change
814 	   (other than comefrom, which userspace doesn't care
815 	   about). */
816 	countersize = sizeof(struct xt_counters) * private->number;
817 	counters = vzalloc(countersize);
818 
819 	if (counters == NULL)
820 		return ERR_PTR(-ENOMEM);
821 
822 	get_counters(private, counters);
823 
824 	return counters;
825 }
826 
827 static int
828 copy_entries_to_user(unsigned int total_size,
829 		     const struct xt_table *table,
830 		     void __user *userptr)
831 {
832 	unsigned int off, num;
833 	const struct ip6t_entry *e;
834 	struct xt_counters *counters;
835 	const struct xt_table_info *private = table->private;
836 	int ret = 0;
837 	const void *loc_cpu_entry;
838 
839 	counters = alloc_counters(table);
840 	if (IS_ERR(counters))
841 		return PTR_ERR(counters);
842 
843 	loc_cpu_entry = private->entries;
844 
845 	/* FIXME: use iterator macros --RR */
846 	/* ... then go back and fix counters and names */
847 	for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
848 		unsigned int i;
849 		const struct xt_entry_match *m;
850 		const struct xt_entry_target *t;
851 
852 		e = loc_cpu_entry + off;
853 		if (copy_to_user(userptr + off, e, sizeof(*e))) {
854 			ret = -EFAULT;
855 			goto free_counters;
856 		}
857 		if (copy_to_user(userptr + off
858 				 + offsetof(struct ip6t_entry, counters),
859 				 &counters[num],
860 				 sizeof(counters[num])) != 0) {
861 			ret = -EFAULT;
862 			goto free_counters;
863 		}
864 
865 		for (i = sizeof(struct ip6t_entry);
866 		     i < e->target_offset;
867 		     i += m->u.match_size) {
868 			m = (void *)e + i;
869 
870 			if (xt_match_to_user(m, userptr + off + i)) {
871 				ret = -EFAULT;
872 				goto free_counters;
873 			}
874 		}
875 
876 		t = ip6t_get_target_c(e);
877 		if (xt_target_to_user(t, userptr + off + e->target_offset)) {
878 			ret = -EFAULT;
879 			goto free_counters;
880 		}
881 	}
882 
883  free_counters:
884 	vfree(counters);
885 	return ret;
886 }
887 
888 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
889 static void compat_standard_from_user(void *dst, const void *src)
890 {
891 	int v = *(compat_int_t *)src;
892 
893 	if (v > 0)
894 		v += xt_compat_calc_jump(AF_INET6, v);
895 	memcpy(dst, &v, sizeof(v));
896 }
897 
898 static int compat_standard_to_user(void __user *dst, const void *src)
899 {
900 	compat_int_t cv = *(int *)src;
901 
902 	if (cv > 0)
903 		cv -= xt_compat_calc_jump(AF_INET6, cv);
904 	return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
905 }
906 
907 static int compat_calc_entry(const struct ip6t_entry *e,
908 			     const struct xt_table_info *info,
909 			     const void *base, struct xt_table_info *newinfo)
910 {
911 	const struct xt_entry_match *ematch;
912 	const struct xt_entry_target *t;
913 	unsigned int entry_offset;
914 	int off, i, ret;
915 
916 	off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
917 	entry_offset = (void *)e - base;
918 	xt_ematch_foreach(ematch, e)
919 		off += xt_compat_match_offset(ematch->u.kernel.match);
920 	t = ip6t_get_target_c(e);
921 	off += xt_compat_target_offset(t->u.kernel.target);
922 	newinfo->size -= off;
923 	ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
924 	if (ret)
925 		return ret;
926 
927 	for (i = 0; i < NF_INET_NUMHOOKS; i++) {
928 		if (info->hook_entry[i] &&
929 		    (e < (struct ip6t_entry *)(base + info->hook_entry[i])))
930 			newinfo->hook_entry[i] -= off;
931 		if (info->underflow[i] &&
932 		    (e < (struct ip6t_entry *)(base + info->underflow[i])))
933 			newinfo->underflow[i] -= off;
934 	}
935 	return 0;
936 }
937 
938 static int compat_table_info(const struct xt_table_info *info,
939 			     struct xt_table_info *newinfo)
940 {
941 	struct ip6t_entry *iter;
942 	const void *loc_cpu_entry;
943 	int ret;
944 
945 	if (!newinfo || !info)
946 		return -EINVAL;
947 
948 	/* we dont care about newinfo->entries */
949 	memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
950 	newinfo->initial_entries = 0;
951 	loc_cpu_entry = info->entries;
952 	ret = xt_compat_init_offsets(AF_INET6, info->number);
953 	if (ret)
954 		return ret;
955 	xt_entry_foreach(iter, loc_cpu_entry, info->size) {
956 		ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
957 		if (ret != 0)
958 			return ret;
959 	}
960 	return 0;
961 }
962 #endif
963 
964 static int get_info(struct net *net, void __user *user, const int *len)
965 {
966 	char name[XT_TABLE_MAXNAMELEN];
967 	struct xt_table *t;
968 	int ret;
969 
970 	if (*len != sizeof(struct ip6t_getinfo))
971 		return -EINVAL;
972 
973 	if (copy_from_user(name, user, sizeof(name)) != 0)
974 		return -EFAULT;
975 
976 	name[XT_TABLE_MAXNAMELEN-1] = '\0';
977 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
978 	if (in_compat_syscall())
979 		xt_compat_lock(AF_INET6);
980 #endif
981 	t = xt_request_find_table_lock(net, AF_INET6, name);
982 	if (!IS_ERR(t)) {
983 		struct ip6t_getinfo info;
984 		const struct xt_table_info *private = t->private;
985 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
986 		struct xt_table_info tmp;
987 
988 		if (in_compat_syscall()) {
989 			ret = compat_table_info(private, &tmp);
990 			xt_compat_flush_offsets(AF_INET6);
991 			private = &tmp;
992 		}
993 #endif
994 		memset(&info, 0, sizeof(info));
995 		info.valid_hooks = t->valid_hooks;
996 		memcpy(info.hook_entry, private->hook_entry,
997 		       sizeof(info.hook_entry));
998 		memcpy(info.underflow, private->underflow,
999 		       sizeof(info.underflow));
1000 		info.num_entries = private->number;
1001 		info.size = private->size;
1002 		strcpy(info.name, name);
1003 
1004 		if (copy_to_user(user, &info, *len) != 0)
1005 			ret = -EFAULT;
1006 		else
1007 			ret = 0;
1008 
1009 		xt_table_unlock(t);
1010 		module_put(t->me);
1011 	} else
1012 		ret = PTR_ERR(t);
1013 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1014 	if (in_compat_syscall())
1015 		xt_compat_unlock(AF_INET6);
1016 #endif
1017 	return ret;
1018 }
1019 
1020 static int
1021 get_entries(struct net *net, struct ip6t_get_entries __user *uptr,
1022 	    const int *len)
1023 {
1024 	int ret;
1025 	struct ip6t_get_entries get;
1026 	struct xt_table *t;
1027 
1028 	if (*len < sizeof(get))
1029 		return -EINVAL;
1030 	if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1031 		return -EFAULT;
1032 	if (*len != sizeof(struct ip6t_get_entries) + get.size)
1033 		return -EINVAL;
1034 
1035 	get.name[sizeof(get.name) - 1] = '\0';
1036 
1037 	t = xt_find_table_lock(net, AF_INET6, get.name);
1038 	if (!IS_ERR(t)) {
1039 		struct xt_table_info *private = t->private;
1040 		if (get.size == private->size)
1041 			ret = copy_entries_to_user(private->size,
1042 						   t, uptr->entrytable);
1043 		else
1044 			ret = -EAGAIN;
1045 
1046 		module_put(t->me);
1047 		xt_table_unlock(t);
1048 	} else
1049 		ret = PTR_ERR(t);
1050 
1051 	return ret;
1052 }
1053 
1054 static int
1055 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1056 	     struct xt_table_info *newinfo, unsigned int num_counters,
1057 	     void __user *counters_ptr)
1058 {
1059 	int ret;
1060 	struct xt_table *t;
1061 	struct xt_table_info *oldinfo;
1062 	struct xt_counters *counters;
1063 	struct ip6t_entry *iter;
1064 
1065 	ret = 0;
1066 	counters = xt_counters_alloc(num_counters);
1067 	if (!counters) {
1068 		ret = -ENOMEM;
1069 		goto out;
1070 	}
1071 
1072 	t = xt_request_find_table_lock(net, AF_INET6, name);
1073 	if (IS_ERR(t)) {
1074 		ret = PTR_ERR(t);
1075 		goto free_newinfo_counters_untrans;
1076 	}
1077 
1078 	/* You lied! */
1079 	if (valid_hooks != t->valid_hooks) {
1080 		ret = -EINVAL;
1081 		goto put_module;
1082 	}
1083 
1084 	oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1085 	if (!oldinfo)
1086 		goto put_module;
1087 
1088 	/* Update module usage count based on number of rules */
1089 	if ((oldinfo->number > oldinfo->initial_entries) ||
1090 	    (newinfo->number <= oldinfo->initial_entries))
1091 		module_put(t->me);
1092 	if ((oldinfo->number > oldinfo->initial_entries) &&
1093 	    (newinfo->number <= oldinfo->initial_entries))
1094 		module_put(t->me);
1095 
1096 	xt_table_unlock(t);
1097 
1098 	get_old_counters(oldinfo, counters);
1099 
1100 	/* Decrease module usage counts and free resource */
1101 	xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1102 		cleanup_entry(iter, net);
1103 
1104 	xt_free_table_info(oldinfo);
1105 	if (copy_to_user(counters_ptr, counters,
1106 			 sizeof(struct xt_counters) * num_counters) != 0) {
1107 		/* Silent error, can't fail, new table is already in place */
1108 		net_warn_ratelimited("ip6tables: counters copy to user failed while replacing table\n");
1109 	}
1110 	vfree(counters);
1111 	return ret;
1112 
1113  put_module:
1114 	module_put(t->me);
1115 	xt_table_unlock(t);
1116  free_newinfo_counters_untrans:
1117 	vfree(counters);
1118  out:
1119 	return ret;
1120 }
1121 
1122 static int
1123 do_replace(struct net *net, sockptr_t arg, unsigned int len)
1124 {
1125 	int ret;
1126 	struct ip6t_replace tmp;
1127 	struct xt_table_info *newinfo;
1128 	void *loc_cpu_entry;
1129 	struct ip6t_entry *iter;
1130 
1131 	if (copy_from_sockptr(&tmp, arg, sizeof(tmp)) != 0)
1132 		return -EFAULT;
1133 
1134 	/* overflow check */
1135 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1136 		return -ENOMEM;
1137 	if (tmp.num_counters == 0)
1138 		return -EINVAL;
1139 
1140 	tmp.name[sizeof(tmp.name)-1] = 0;
1141 
1142 	newinfo = xt_alloc_table_info(tmp.size);
1143 	if (!newinfo)
1144 		return -ENOMEM;
1145 
1146 	loc_cpu_entry = newinfo->entries;
1147 	if (copy_from_sockptr_offset(loc_cpu_entry, arg, sizeof(tmp),
1148 			tmp.size) != 0) {
1149 		ret = -EFAULT;
1150 		goto free_newinfo;
1151 	}
1152 
1153 	ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1154 	if (ret != 0)
1155 		goto free_newinfo;
1156 
1157 	ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1158 			   tmp.num_counters, tmp.counters);
1159 	if (ret)
1160 		goto free_newinfo_untrans;
1161 	return 0;
1162 
1163  free_newinfo_untrans:
1164 	xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1165 		cleanup_entry(iter, net);
1166  free_newinfo:
1167 	xt_free_table_info(newinfo);
1168 	return ret;
1169 }
1170 
1171 static int
1172 do_add_counters(struct net *net, sockptr_t arg, unsigned int len)
1173 {
1174 	unsigned int i;
1175 	struct xt_counters_info tmp;
1176 	struct xt_counters *paddc;
1177 	struct xt_table *t;
1178 	const struct xt_table_info *private;
1179 	int ret = 0;
1180 	struct ip6t_entry *iter;
1181 	unsigned int addend;
1182 
1183 	paddc = xt_copy_counters(arg, len, &tmp);
1184 	if (IS_ERR(paddc))
1185 		return PTR_ERR(paddc);
1186 	t = xt_find_table_lock(net, AF_INET6, tmp.name);
1187 	if (IS_ERR(t)) {
1188 		ret = PTR_ERR(t);
1189 		goto free;
1190 	}
1191 
1192 	local_bh_disable();
1193 	private = t->private;
1194 	if (private->number != tmp.num_counters) {
1195 		ret = -EINVAL;
1196 		goto unlock_up_free;
1197 	}
1198 
1199 	i = 0;
1200 	addend = xt_write_recseq_begin();
1201 	xt_entry_foreach(iter, private->entries, private->size) {
1202 		struct xt_counters *tmp;
1203 
1204 		tmp = xt_get_this_cpu_counter(&iter->counters);
1205 		ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1206 		++i;
1207 	}
1208 	xt_write_recseq_end(addend);
1209  unlock_up_free:
1210 	local_bh_enable();
1211 	xt_table_unlock(t);
1212 	module_put(t->me);
1213  free:
1214 	vfree(paddc);
1215 
1216 	return ret;
1217 }
1218 
1219 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1220 struct compat_ip6t_replace {
1221 	char			name[XT_TABLE_MAXNAMELEN];
1222 	u32			valid_hooks;
1223 	u32			num_entries;
1224 	u32			size;
1225 	u32			hook_entry[NF_INET_NUMHOOKS];
1226 	u32			underflow[NF_INET_NUMHOOKS];
1227 	u32			num_counters;
1228 	compat_uptr_t		counters;	/* struct xt_counters * */
1229 	struct compat_ip6t_entry entries[];
1230 };
1231 
1232 static int
1233 compat_copy_entry_to_user(struct ip6t_entry *e, void __user **dstptr,
1234 			  unsigned int *size, struct xt_counters *counters,
1235 			  unsigned int i)
1236 {
1237 	struct xt_entry_target *t;
1238 	struct compat_ip6t_entry __user *ce;
1239 	u_int16_t target_offset, next_offset;
1240 	compat_uint_t origsize;
1241 	const struct xt_entry_match *ematch;
1242 	int ret = 0;
1243 
1244 	origsize = *size;
1245 	ce = *dstptr;
1246 	if (copy_to_user(ce, e, sizeof(struct ip6t_entry)) != 0 ||
1247 	    copy_to_user(&ce->counters, &counters[i],
1248 	    sizeof(counters[i])) != 0)
1249 		return -EFAULT;
1250 
1251 	*dstptr += sizeof(struct compat_ip6t_entry);
1252 	*size -= sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1253 
1254 	xt_ematch_foreach(ematch, e) {
1255 		ret = xt_compat_match_to_user(ematch, dstptr, size);
1256 		if (ret != 0)
1257 			return ret;
1258 	}
1259 	target_offset = e->target_offset - (origsize - *size);
1260 	t = ip6t_get_target(e);
1261 	ret = xt_compat_target_to_user(t, dstptr, size);
1262 	if (ret)
1263 		return ret;
1264 	next_offset = e->next_offset - (origsize - *size);
1265 	if (put_user(target_offset, &ce->target_offset) != 0 ||
1266 	    put_user(next_offset, &ce->next_offset) != 0)
1267 		return -EFAULT;
1268 	return 0;
1269 }
1270 
1271 static int
1272 compat_find_calc_match(struct xt_entry_match *m,
1273 		       const struct ip6t_ip6 *ipv6,
1274 		       int *size)
1275 {
1276 	struct xt_match *match;
1277 
1278 	match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
1279 				      m->u.user.revision);
1280 	if (IS_ERR(match))
1281 		return PTR_ERR(match);
1282 
1283 	m->u.kernel.match = match;
1284 	*size += xt_compat_match_offset(match);
1285 	return 0;
1286 }
1287 
1288 static void compat_release_entry(struct compat_ip6t_entry *e)
1289 {
1290 	struct xt_entry_target *t;
1291 	struct xt_entry_match *ematch;
1292 
1293 	/* Cleanup all matches */
1294 	xt_ematch_foreach(ematch, e)
1295 		module_put(ematch->u.kernel.match->me);
1296 	t = compat_ip6t_get_target(e);
1297 	module_put(t->u.kernel.target->me);
1298 }
1299 
1300 static int
1301 check_compat_entry_size_and_hooks(struct compat_ip6t_entry *e,
1302 				  struct xt_table_info *newinfo,
1303 				  unsigned int *size,
1304 				  const unsigned char *base,
1305 				  const unsigned char *limit)
1306 {
1307 	struct xt_entry_match *ematch;
1308 	struct xt_entry_target *t;
1309 	struct xt_target *target;
1310 	unsigned int entry_offset;
1311 	unsigned int j;
1312 	int ret, off;
1313 
1314 	if ((unsigned long)e % __alignof__(struct compat_ip6t_entry) != 0 ||
1315 	    (unsigned char *)e + sizeof(struct compat_ip6t_entry) >= limit ||
1316 	    (unsigned char *)e + e->next_offset > limit)
1317 		return -EINVAL;
1318 
1319 	if (e->next_offset < sizeof(struct compat_ip6t_entry) +
1320 			     sizeof(struct compat_xt_entry_target))
1321 		return -EINVAL;
1322 
1323 	if (!ip6_checkentry(&e->ipv6))
1324 		return -EINVAL;
1325 
1326 	ret = xt_compat_check_entry_offsets(e, e->elems,
1327 					    e->target_offset, e->next_offset);
1328 	if (ret)
1329 		return ret;
1330 
1331 	off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1332 	entry_offset = (void *)e - (void *)base;
1333 	j = 0;
1334 	xt_ematch_foreach(ematch, e) {
1335 		ret = compat_find_calc_match(ematch, &e->ipv6, &off);
1336 		if (ret != 0)
1337 			goto release_matches;
1338 		++j;
1339 	}
1340 
1341 	t = compat_ip6t_get_target(e);
1342 	target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
1343 					t->u.user.revision);
1344 	if (IS_ERR(target)) {
1345 		ret = PTR_ERR(target);
1346 		goto release_matches;
1347 	}
1348 	t->u.kernel.target = target;
1349 
1350 	off += xt_compat_target_offset(target);
1351 	*size += off;
1352 	ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
1353 	if (ret)
1354 		goto out;
1355 
1356 	return 0;
1357 
1358 out:
1359 	module_put(t->u.kernel.target->me);
1360 release_matches:
1361 	xt_ematch_foreach(ematch, e) {
1362 		if (j-- == 0)
1363 			break;
1364 		module_put(ematch->u.kernel.match->me);
1365 	}
1366 	return ret;
1367 }
1368 
1369 static void
1370 compat_copy_entry_from_user(struct compat_ip6t_entry *e, void **dstptr,
1371 			    unsigned int *size,
1372 			    struct xt_table_info *newinfo, unsigned char *base)
1373 {
1374 	struct xt_entry_target *t;
1375 	struct ip6t_entry *de;
1376 	unsigned int origsize;
1377 	int h;
1378 	struct xt_entry_match *ematch;
1379 
1380 	origsize = *size;
1381 	de = *dstptr;
1382 	memcpy(de, e, sizeof(struct ip6t_entry));
1383 	memcpy(&de->counters, &e->counters, sizeof(e->counters));
1384 
1385 	*dstptr += sizeof(struct ip6t_entry);
1386 	*size += sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1387 
1388 	xt_ematch_foreach(ematch, e)
1389 		xt_compat_match_from_user(ematch, dstptr, size);
1390 
1391 	de->target_offset = e->target_offset - (origsize - *size);
1392 	t = compat_ip6t_get_target(e);
1393 	xt_compat_target_from_user(t, dstptr, size);
1394 
1395 	de->next_offset = e->next_offset - (origsize - *size);
1396 	for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1397 		if ((unsigned char *)de - base < newinfo->hook_entry[h])
1398 			newinfo->hook_entry[h] -= origsize - *size;
1399 		if ((unsigned char *)de - base < newinfo->underflow[h])
1400 			newinfo->underflow[h] -= origsize - *size;
1401 	}
1402 }
1403 
1404 static int
1405 translate_compat_table(struct net *net,
1406 		       struct xt_table_info **pinfo,
1407 		       void **pentry0,
1408 		       const struct compat_ip6t_replace *compatr)
1409 {
1410 	unsigned int i, j;
1411 	struct xt_table_info *newinfo, *info;
1412 	void *pos, *entry0, *entry1;
1413 	struct compat_ip6t_entry *iter0;
1414 	struct ip6t_replace repl;
1415 	unsigned int size;
1416 	int ret;
1417 
1418 	info = *pinfo;
1419 	entry0 = *pentry0;
1420 	size = compatr->size;
1421 	info->number = compatr->num_entries;
1422 
1423 	j = 0;
1424 	xt_compat_lock(AF_INET6);
1425 	ret = xt_compat_init_offsets(AF_INET6, compatr->num_entries);
1426 	if (ret)
1427 		goto out_unlock;
1428 	/* Walk through entries, checking offsets. */
1429 	xt_entry_foreach(iter0, entry0, compatr->size) {
1430 		ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1431 							entry0,
1432 							entry0 + compatr->size);
1433 		if (ret != 0)
1434 			goto out_unlock;
1435 		++j;
1436 	}
1437 
1438 	ret = -EINVAL;
1439 	if (j != compatr->num_entries)
1440 		goto out_unlock;
1441 
1442 	ret = -ENOMEM;
1443 	newinfo = xt_alloc_table_info(size);
1444 	if (!newinfo)
1445 		goto out_unlock;
1446 
1447 	memset(newinfo->entries, 0, size);
1448 
1449 	newinfo->number = compatr->num_entries;
1450 	for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1451 		newinfo->hook_entry[i] = compatr->hook_entry[i];
1452 		newinfo->underflow[i] = compatr->underflow[i];
1453 	}
1454 	entry1 = newinfo->entries;
1455 	pos = entry1;
1456 	size = compatr->size;
1457 	xt_entry_foreach(iter0, entry0, compatr->size)
1458 		compat_copy_entry_from_user(iter0, &pos, &size,
1459 					    newinfo, entry1);
1460 
1461 	/* all module references in entry0 are now gone. */
1462 	xt_compat_flush_offsets(AF_INET6);
1463 	xt_compat_unlock(AF_INET6);
1464 
1465 	memcpy(&repl, compatr, sizeof(*compatr));
1466 
1467 	for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1468 		repl.hook_entry[i] = newinfo->hook_entry[i];
1469 		repl.underflow[i] = newinfo->underflow[i];
1470 	}
1471 
1472 	repl.num_counters = 0;
1473 	repl.counters = NULL;
1474 	repl.size = newinfo->size;
1475 	ret = translate_table(net, newinfo, entry1, &repl);
1476 	if (ret)
1477 		goto free_newinfo;
1478 
1479 	*pinfo = newinfo;
1480 	*pentry0 = entry1;
1481 	xt_free_table_info(info);
1482 	return 0;
1483 
1484 free_newinfo:
1485 	xt_free_table_info(newinfo);
1486 	return ret;
1487 out_unlock:
1488 	xt_compat_flush_offsets(AF_INET6);
1489 	xt_compat_unlock(AF_INET6);
1490 	xt_entry_foreach(iter0, entry0, compatr->size) {
1491 		if (j-- == 0)
1492 			break;
1493 		compat_release_entry(iter0);
1494 	}
1495 	return ret;
1496 }
1497 
1498 static int
1499 compat_do_replace(struct net *net, sockptr_t arg, unsigned int len)
1500 {
1501 	int ret;
1502 	struct compat_ip6t_replace tmp;
1503 	struct xt_table_info *newinfo;
1504 	void *loc_cpu_entry;
1505 	struct ip6t_entry *iter;
1506 
1507 	if (copy_from_sockptr(&tmp, arg, sizeof(tmp)) != 0)
1508 		return -EFAULT;
1509 
1510 	/* overflow check */
1511 	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1512 		return -ENOMEM;
1513 	if (tmp.num_counters == 0)
1514 		return -EINVAL;
1515 
1516 	tmp.name[sizeof(tmp.name)-1] = 0;
1517 
1518 	newinfo = xt_alloc_table_info(tmp.size);
1519 	if (!newinfo)
1520 		return -ENOMEM;
1521 
1522 	loc_cpu_entry = newinfo->entries;
1523 	if (copy_from_sockptr_offset(loc_cpu_entry, arg, sizeof(tmp),
1524 			tmp.size) != 0) {
1525 		ret = -EFAULT;
1526 		goto free_newinfo;
1527 	}
1528 
1529 	ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1530 	if (ret != 0)
1531 		goto free_newinfo;
1532 
1533 	ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1534 			   tmp.num_counters, compat_ptr(tmp.counters));
1535 	if (ret)
1536 		goto free_newinfo_untrans;
1537 	return 0;
1538 
1539  free_newinfo_untrans:
1540 	xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1541 		cleanup_entry(iter, net);
1542  free_newinfo:
1543 	xt_free_table_info(newinfo);
1544 	return ret;
1545 }
1546 
1547 struct compat_ip6t_get_entries {
1548 	char name[XT_TABLE_MAXNAMELEN];
1549 	compat_uint_t size;
1550 	struct compat_ip6t_entry entrytable[];
1551 };
1552 
1553 static int
1554 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1555 			    void __user *userptr)
1556 {
1557 	struct xt_counters *counters;
1558 	const struct xt_table_info *private = table->private;
1559 	void __user *pos;
1560 	unsigned int size;
1561 	int ret = 0;
1562 	unsigned int i = 0;
1563 	struct ip6t_entry *iter;
1564 
1565 	counters = alloc_counters(table);
1566 	if (IS_ERR(counters))
1567 		return PTR_ERR(counters);
1568 
1569 	pos = userptr;
1570 	size = total_size;
1571 	xt_entry_foreach(iter, private->entries, total_size) {
1572 		ret = compat_copy_entry_to_user(iter, &pos,
1573 						&size, counters, i++);
1574 		if (ret != 0)
1575 			break;
1576 	}
1577 
1578 	vfree(counters);
1579 	return ret;
1580 }
1581 
1582 static int
1583 compat_get_entries(struct net *net, struct compat_ip6t_get_entries __user *uptr,
1584 		   int *len)
1585 {
1586 	int ret;
1587 	struct compat_ip6t_get_entries get;
1588 	struct xt_table *t;
1589 
1590 	if (*len < sizeof(get))
1591 		return -EINVAL;
1592 
1593 	if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1594 		return -EFAULT;
1595 
1596 	if (*len != sizeof(struct compat_ip6t_get_entries) + get.size)
1597 		return -EINVAL;
1598 
1599 	get.name[sizeof(get.name) - 1] = '\0';
1600 
1601 	xt_compat_lock(AF_INET6);
1602 	t = xt_find_table_lock(net, AF_INET6, get.name);
1603 	if (!IS_ERR(t)) {
1604 		const struct xt_table_info *private = t->private;
1605 		struct xt_table_info info;
1606 		ret = compat_table_info(private, &info);
1607 		if (!ret && get.size == info.size)
1608 			ret = compat_copy_entries_to_user(private->size,
1609 							  t, uptr->entrytable);
1610 		else if (!ret)
1611 			ret = -EAGAIN;
1612 
1613 		xt_compat_flush_offsets(AF_INET6);
1614 		module_put(t->me);
1615 		xt_table_unlock(t);
1616 	} else
1617 		ret = PTR_ERR(t);
1618 
1619 	xt_compat_unlock(AF_INET6);
1620 	return ret;
1621 }
1622 #endif
1623 
1624 static int
1625 do_ip6t_set_ctl(struct sock *sk, int cmd, sockptr_t arg, unsigned int len)
1626 {
1627 	int ret;
1628 
1629 	if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1630 		return -EPERM;
1631 
1632 	switch (cmd) {
1633 	case IP6T_SO_SET_REPLACE:
1634 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1635 		if (in_compat_syscall())
1636 			ret = compat_do_replace(sock_net(sk), arg, len);
1637 		else
1638 #endif
1639 			ret = do_replace(sock_net(sk), arg, len);
1640 		break;
1641 
1642 	case IP6T_SO_SET_ADD_COUNTERS:
1643 		ret = do_add_counters(sock_net(sk), arg, len);
1644 		break;
1645 
1646 	default:
1647 		ret = -EINVAL;
1648 	}
1649 
1650 	return ret;
1651 }
1652 
1653 static int
1654 do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1655 {
1656 	int ret;
1657 
1658 	if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1659 		return -EPERM;
1660 
1661 	switch (cmd) {
1662 	case IP6T_SO_GET_INFO:
1663 		ret = get_info(sock_net(sk), user, len);
1664 		break;
1665 
1666 	case IP6T_SO_GET_ENTRIES:
1667 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1668 		if (in_compat_syscall())
1669 			ret = compat_get_entries(sock_net(sk), user, len);
1670 		else
1671 #endif
1672 			ret = get_entries(sock_net(sk), user, len);
1673 		break;
1674 
1675 	case IP6T_SO_GET_REVISION_MATCH:
1676 	case IP6T_SO_GET_REVISION_TARGET: {
1677 		struct xt_get_revision rev;
1678 		int target;
1679 
1680 		if (*len != sizeof(rev)) {
1681 			ret = -EINVAL;
1682 			break;
1683 		}
1684 		if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1685 			ret = -EFAULT;
1686 			break;
1687 		}
1688 		rev.name[sizeof(rev.name)-1] = 0;
1689 
1690 		if (cmd == IP6T_SO_GET_REVISION_TARGET)
1691 			target = 1;
1692 		else
1693 			target = 0;
1694 
1695 		try_then_request_module(xt_find_revision(AF_INET6, rev.name,
1696 							 rev.revision,
1697 							 target, &ret),
1698 					"ip6t_%s", rev.name);
1699 		break;
1700 	}
1701 
1702 	default:
1703 		ret = -EINVAL;
1704 	}
1705 
1706 	return ret;
1707 }
1708 
1709 static void __ip6t_unregister_table(struct net *net, struct xt_table *table)
1710 {
1711 	struct xt_table_info *private;
1712 	void *loc_cpu_entry;
1713 	struct module *table_owner = table->me;
1714 	struct ip6t_entry *iter;
1715 
1716 	private = xt_unregister_table(table);
1717 
1718 	/* Decrease module usage counts and free resources */
1719 	loc_cpu_entry = private->entries;
1720 	xt_entry_foreach(iter, loc_cpu_entry, private->size)
1721 		cleanup_entry(iter, net);
1722 	if (private->number > private->initial_entries)
1723 		module_put(table_owner);
1724 	xt_free_table_info(private);
1725 }
1726 
1727 int ip6t_register_table(struct net *net, const struct xt_table *table,
1728 			const struct ip6t_replace *repl,
1729 			const struct nf_hook_ops *template_ops)
1730 {
1731 	struct nf_hook_ops *ops;
1732 	unsigned int num_ops;
1733 	int ret, i;
1734 	struct xt_table_info *newinfo;
1735 	struct xt_table_info bootstrap = {0};
1736 	void *loc_cpu_entry;
1737 	struct xt_table *new_table;
1738 
1739 	newinfo = xt_alloc_table_info(repl->size);
1740 	if (!newinfo)
1741 		return -ENOMEM;
1742 
1743 	loc_cpu_entry = newinfo->entries;
1744 	memcpy(loc_cpu_entry, repl->entries, repl->size);
1745 
1746 	ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1747 	if (ret != 0) {
1748 		xt_free_table_info(newinfo);
1749 		return ret;
1750 	}
1751 
1752 	new_table = xt_register_table(net, table, &bootstrap, newinfo);
1753 	if (IS_ERR(new_table)) {
1754 		xt_free_table_info(newinfo);
1755 		return PTR_ERR(new_table);
1756 	}
1757 
1758 	if (!template_ops)
1759 		return 0;
1760 
1761 	num_ops = hweight32(table->valid_hooks);
1762 	if (num_ops == 0) {
1763 		ret = -EINVAL;
1764 		goto out_free;
1765 	}
1766 
1767 	ops = kmemdup(template_ops, sizeof(*ops) * num_ops, GFP_KERNEL);
1768 	if (!ops) {
1769 		ret = -ENOMEM;
1770 		goto out_free;
1771 	}
1772 
1773 	for (i = 0; i < num_ops; i++)
1774 		ops[i].priv = new_table;
1775 
1776 	new_table->ops = ops;
1777 
1778 	ret = nf_register_net_hooks(net, ops, num_ops);
1779 	if (ret != 0)
1780 		goto out_free;
1781 
1782 	return ret;
1783 
1784 out_free:
1785 	__ip6t_unregister_table(net, new_table);
1786 	return ret;
1787 }
1788 
1789 void ip6t_unregister_table_pre_exit(struct net *net, const char *name)
1790 {
1791 	struct xt_table *table = xt_find_table(net, NFPROTO_IPV6, name);
1792 
1793 	if (table)
1794 		nf_unregister_net_hooks(net, table->ops, hweight32(table->valid_hooks));
1795 }
1796 
1797 void ip6t_unregister_table_exit(struct net *net, const char *name)
1798 {
1799 	struct xt_table *table = xt_find_table(net, NFPROTO_IPV6, name);
1800 
1801 	if (table)
1802 		__ip6t_unregister_table(net, table);
1803 }
1804 
1805 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1806 static inline bool
1807 icmp6_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1808 		     u_int8_t type, u_int8_t code,
1809 		     bool invert)
1810 {
1811 	return (type == test_type && code >= min_code && code <= max_code)
1812 		^ invert;
1813 }
1814 
1815 static bool
1816 icmp6_match(const struct sk_buff *skb, struct xt_action_param *par)
1817 {
1818 	const struct icmp6hdr *ic;
1819 	struct icmp6hdr _icmph;
1820 	const struct ip6t_icmp *icmpinfo = par->matchinfo;
1821 
1822 	/* Must not be a fragment. */
1823 	if (par->fragoff != 0)
1824 		return false;
1825 
1826 	ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1827 	if (ic == NULL) {
1828 		/* We've been asked to examine this packet, and we
1829 		 * can't.  Hence, no choice but to drop.
1830 		 */
1831 		par->hotdrop = true;
1832 		return false;
1833 	}
1834 
1835 	return icmp6_type_code_match(icmpinfo->type,
1836 				     icmpinfo->code[0],
1837 				     icmpinfo->code[1],
1838 				     ic->icmp6_type, ic->icmp6_code,
1839 				     !!(icmpinfo->invflags&IP6T_ICMP_INV));
1840 }
1841 
1842 /* Called when user tries to insert an entry of this type. */
1843 static int icmp6_checkentry(const struct xt_mtchk_param *par)
1844 {
1845 	const struct ip6t_icmp *icmpinfo = par->matchinfo;
1846 
1847 	/* Must specify no unknown invflags */
1848 	return (icmpinfo->invflags & ~IP6T_ICMP_INV) ? -EINVAL : 0;
1849 }
1850 
1851 /* The built-in targets: standard (NULL) and error. */
1852 static struct xt_target ip6t_builtin_tg[] __read_mostly = {
1853 	{
1854 		.name             = XT_STANDARD_TARGET,
1855 		.targetsize       = sizeof(int),
1856 		.family           = NFPROTO_IPV6,
1857 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1858 		.compatsize       = sizeof(compat_int_t),
1859 		.compat_from_user = compat_standard_from_user,
1860 		.compat_to_user   = compat_standard_to_user,
1861 #endif
1862 	},
1863 	{
1864 		.name             = XT_ERROR_TARGET,
1865 		.target           = ip6t_error,
1866 		.targetsize       = XT_FUNCTION_MAXNAMELEN,
1867 		.family           = NFPROTO_IPV6,
1868 	},
1869 };
1870 
1871 static struct nf_sockopt_ops ip6t_sockopts = {
1872 	.pf		= PF_INET6,
1873 	.set_optmin	= IP6T_BASE_CTL,
1874 	.set_optmax	= IP6T_SO_SET_MAX+1,
1875 	.set		= do_ip6t_set_ctl,
1876 	.get_optmin	= IP6T_BASE_CTL,
1877 	.get_optmax	= IP6T_SO_GET_MAX+1,
1878 	.get		= do_ip6t_get_ctl,
1879 	.owner		= THIS_MODULE,
1880 };
1881 
1882 static struct xt_match ip6t_builtin_mt[] __read_mostly = {
1883 	{
1884 		.name       = "icmp6",
1885 		.match      = icmp6_match,
1886 		.matchsize  = sizeof(struct ip6t_icmp),
1887 		.checkentry = icmp6_checkentry,
1888 		.proto      = IPPROTO_ICMPV6,
1889 		.family     = NFPROTO_IPV6,
1890 		.me	    = THIS_MODULE,
1891 	},
1892 };
1893 
1894 static int __net_init ip6_tables_net_init(struct net *net)
1895 {
1896 	return xt_proto_init(net, NFPROTO_IPV6);
1897 }
1898 
1899 static void __net_exit ip6_tables_net_exit(struct net *net)
1900 {
1901 	xt_proto_fini(net, NFPROTO_IPV6);
1902 }
1903 
1904 static struct pernet_operations ip6_tables_net_ops = {
1905 	.init = ip6_tables_net_init,
1906 	.exit = ip6_tables_net_exit,
1907 };
1908 
1909 static int __init ip6_tables_init(void)
1910 {
1911 	int ret;
1912 
1913 	ret = register_pernet_subsys(&ip6_tables_net_ops);
1914 	if (ret < 0)
1915 		goto err1;
1916 
1917 	/* No one else will be downing sem now, so we won't sleep */
1918 	ret = xt_register_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1919 	if (ret < 0)
1920 		goto err2;
1921 	ret = xt_register_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1922 	if (ret < 0)
1923 		goto err4;
1924 
1925 	/* Register setsockopt */
1926 	ret = nf_register_sockopt(&ip6t_sockopts);
1927 	if (ret < 0)
1928 		goto err5;
1929 
1930 	return 0;
1931 
1932 err5:
1933 	xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1934 err4:
1935 	xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1936 err2:
1937 	unregister_pernet_subsys(&ip6_tables_net_ops);
1938 err1:
1939 	return ret;
1940 }
1941 
1942 static void __exit ip6_tables_fini(void)
1943 {
1944 	nf_unregister_sockopt(&ip6t_sockopts);
1945 
1946 	xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1947 	xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1948 	unregister_pernet_subsys(&ip6_tables_net_ops);
1949 }
1950 
1951 EXPORT_SYMBOL(ip6t_register_table);
1952 EXPORT_SYMBOL(ip6t_unregister_table_pre_exit);
1953 EXPORT_SYMBOL(ip6t_unregister_table_exit);
1954 EXPORT_SYMBOL(ip6t_do_table);
1955 
1956 module_init(ip6_tables_init);
1957 module_exit(ip6_tables_fini);
1958