/net/ipv4/netfilter/ip_tables.c

   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#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  10#include <linux/cache.h>
  11#include <linux/capability.h>
  12#include <linux/skbuff.h>
  13#include <linux/kmod.h>
  14#include <linux/vmalloc.h>
  15#include <linux/netdevice.h>
  16#include <linux/module.h>
  17#include <linux/icmp.h>
  18#include <net/ip.h>
  19#include <net/compat.h>
  20#include <linux/uaccess.h>
  21#include <linux/mutex.h>
  22#include <linux/proc_fs.h>
  23#include <linux/err.h>
  24#include <linux/cpumask.h>
  25
  26#include <linux/netfilter/x_tables.h>
  27#include <linux/netfilter_ipv4/ip_tables.h>
  28#include <net/netfilter/nf_log.h>
  29#include "../../netfilter/xt_repldata.h"
  30
  31MODULE_LICENSE("GPL");
  32MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  33MODULE_DESCRIPTION("IPv4 packet filter");
  34MODULE_ALIAS("ipt_icmp");
  35
  36void *ipt_alloc_initial_table(const struct xt_table *info)
  37{
  38	return xt_alloc_initial_table(ipt, IPT);
  39}
  40EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  41
  42/* Returns whether matches rule or not. */
  43/* Performance critical - called for every packet */
  44static inline bool
  45ip_packet_match(const struct iphdr *ip,
  46		const char *indev,
  47		const char *outdev,
  48		const struct ipt_ip *ipinfo,
  49		int isfrag)
  50{
  51	unsigned long ret;
  52
  53	if (NF_INVF(ipinfo, IPT_INV_SRCIP,
  54		    (ip->saddr & ipinfo->smsk.s_addr) != ipinfo->src.s_addr) ||
  55	    NF_INVF(ipinfo, IPT_INV_DSTIP,
  56		    (ip->daddr & ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr))
  57		return false;
  58
  59	ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
  60
  61	if (NF_INVF(ipinfo, IPT_INV_VIA_IN, ret != 0))
  62		return false;
  63
  64	ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
  65
  66	if (NF_INVF(ipinfo, IPT_INV_VIA_OUT, ret != 0))
  67		return false;
  68
  69	/* Check specific protocol */
  70	if (ipinfo->proto &&
  71	    NF_INVF(ipinfo, IPT_INV_PROTO, ip->protocol != ipinfo->proto))
  72		return false;
  73
  74	/* If we have a fragment rule but the packet is not a fragment
  75	 * then we return zero */
  76	if (NF_INVF(ipinfo, IPT_INV_FRAG,
  77		    (ipinfo->flags & IPT_F_FRAG) && !isfrag))
  78		return false;
  79
  80	return true;
  81}
  82
  83static bool
  84ip_checkentry(const struct ipt_ip *ip)
  85{
  86	if (ip->flags & ~IPT_F_MASK)
  87		return false;
  88	if (ip->invflags & ~IPT_INV_MASK)
  89		return false;
  90	return true;
  91}
  92
  93static unsigned int
  94ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
  95{
  96	net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
  97
  98	return NF_DROP;
  99}
 100
 101/* Performance critical */
 102static inline struct ipt_entry *
 103get_entry(const void *base, unsigned int offset)
 104{
 105	return (struct ipt_entry *)(base + offset);
 106}
 107
 108/* All zeroes == unconditional rule. */
 109/* Mildly perf critical (only if packet tracing is on) */
 110static inline bool unconditional(const struct ipt_entry *e)
 111{
 112	static const struct ipt_ip uncond;
 113
 114	return e->target_offset == sizeof(struct ipt_entry) &&
 115	       memcmp(&e->ip, &uncond, sizeof(uncond)) == 0;
 116}
 117
 118/* for const-correctness */
 119static inline const struct xt_entry_target *
 120ipt_get_target_c(const struct ipt_entry *e)
 121{
 122	return ipt_get_target((struct ipt_entry *)e);
 123}
 124
 125#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 126static const char *const hooknames[] = {
 127	[NF_INET_PRE_ROUTING]		= "PREROUTING",
 128	[NF_INET_LOCAL_IN]		= "INPUT",
 129	[NF_INET_FORWARD]		= "FORWARD",
 130	[NF_INET_LOCAL_OUT]		= "OUTPUT",
 131	[NF_INET_POST_ROUTING]		= "POSTROUTING",
 132};
 133
 134enum nf_ip_trace_comments {
 135	NF_IP_TRACE_COMMENT_RULE,
 136	NF_IP_TRACE_COMMENT_RETURN,
 137	NF_IP_TRACE_COMMENT_POLICY,
 138};
 139
 140static const char *const comments[] = {
 141	[NF_IP_TRACE_COMMENT_RULE]	= "rule",
 142	[NF_IP_TRACE_COMMENT_RETURN]	= "return",
 143	[NF_IP_TRACE_COMMENT_POLICY]	= "policy",
 144};
 145
 146static const struct nf_loginfo trace_loginfo = {
 147	.type = NF_LOG_TYPE_LOG,
 148	.u = {
 149		.log = {
 150			.level = 4,
 151			.logflags = NF_LOG_DEFAULT_MASK,
 152		},
 153	},
 154};
 155
 156/* Mildly perf critical (only if packet tracing is on) */
 157static inline int
 158get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 159		      const char *hookname, const char **chainname,
 160		      const char **comment, unsigned int *rulenum)
 161{
 162	const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 163
 164	if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 165		/* Head of user chain: ERROR target with chainname */
 166		*chainname = t->target.data;
 167		(*rulenum) = 0;
 168	} else if (s == e) {
 169		(*rulenum)++;
 170
 171		if (unconditional(s) &&
 172		    strcmp(t->target.u.kernel.target->name,
 173			   XT_STANDARD_TARGET) == 0 &&
 174		   t->verdict < 0) {
 175			/* Tail of chains: STANDARD target (return/policy) */
 176			*comment = *chainname == hookname
 177				? comments[NF_IP_TRACE_COMMENT_POLICY]
 178				: comments[NF_IP_TRACE_COMMENT_RETURN];
 179		}
 180		return 1;
 181	} else
 182		(*rulenum)++;
 183
 184	return 0;
 185}
 186
 187static void trace_packet(struct net *net,
 188			 const struct sk_buff *skb,
 189			 unsigned int hook,
 190			 const struct net_device *in,
 191			 const struct net_device *out,
 192			 const char *tablename,
 193			 const struct xt_table_info *private,
 194			 const struct ipt_entry *e)
 195{
 196	const struct ipt_entry *root;
 197	const char *hookname, *chainname, *comment;
 198	const struct ipt_entry *iter;
 199	unsigned int rulenum = 0;
 200
 201	root = get_entry(private->entries, private->hook_entry[hook]);
 202
 203	hookname = chainname = hooknames[hook];
 204	comment = comments[NF_IP_TRACE_COMMENT_RULE];
 205
 206	xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 207		if (get_chainname_rulenum(iter, e, hookname,
 208		    &chainname, &comment, &rulenum) != 0)
 209			break;
 210
 211	nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 212		     "TRACE: %s:%s:%s:%u ",
 213		     tablename, chainname, comment, rulenum);
 214}
 215#endif
 216
 217static inline
 218struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 219{
 220	return (void *)entry + entry->next_offset;
 221}
 222
 223/* Returns one of the generic firewall policies, like NF_ACCEPT. */
 224unsigned int
 225ipt_do_table(struct sk_buff *skb,
 226	     const struct nf_hook_state *state,
 227	     struct xt_table *table)
 228{
 229	unsigned int hook = state->hook;
 230	static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 231	const struct iphdr *ip;
 232	/* Initializing verdict to NF_DROP keeps gcc happy. */
 233	unsigned int verdict = NF_DROP;
 234	const char *indev, *outdev;
 235	const void *table_base;
 236	struct ipt_entry *e, **jumpstack;
 237	unsigned int stackidx, cpu;
 238	const struct xt_table_info *private;
 239	struct xt_action_param acpar;
 240	unsigned int addend;
 241
 242	/* Initialization */
 243	stackidx = 0;
 244	ip = ip_hdr(skb);
 245	indev = state->in ? state->in->name : nulldevname;
 246	outdev = state->out ? state->out->name : nulldevname;
 247	/* We handle fragments by dealing with the first fragment as
 248	 * if it was a normal packet.  All other fragments are treated
 249	 * normally, except that they will NEVER match rules that ask
 250	 * things we don't know, ie. tcp syn flag or ports).  If the
 251	 * rule is also a fragment-specific rule, non-fragments won't
 252	 * match it. */
 253	acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 254	acpar.thoff   = ip_hdrlen(skb);
 255	acpar.hotdrop = false;
 256	acpar.state   = state;
 257
 258	WARN_ON(!(table->valid_hooks & (1 << hook)));
 259	local_bh_disable();
 260	addend = xt_write_recseq_begin();
 261	private = READ_ONCE(table->private); /* Address dependency. */
 262	cpu        = smp_processor_id();
 263	table_base = private->entries;
 264	jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 265
 266	/* Switch to alternate jumpstack if we're being invoked via TEE.
 267	 * TEE issues XT_CONTINUE verdict on original skb so we must not
 268	 * clobber the jumpstack.
 269	 *
 270	 * For recursion via REJECT or SYNPROXY the stack will be clobbered
 271	 * but it is no problem since absolute verdict is issued by these.
 272	 */
 273	if (static_key_false(&xt_tee_enabled))
 274		jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 275
 276	e = get_entry(table_base, private->hook_entry[hook]);
 277
 278	do {
 279		const struct xt_entry_target *t;
 280		const struct xt_entry_match *ematch;
 281		struct xt_counters *counter;
 282
 283		WARN_ON(!e);
 284		if (!ip_packet_match(ip, indev, outdev,
 285		    &e->ip, acpar.fragoff)) {
 286 no_match:
 287			e = ipt_next_entry(e);
 288			continue;
 289		}
 290
 291		xt_ematch_foreach(ematch, e) {
 292			acpar.match     = ematch->u.kernel.match;
 293			acpar.matchinfo = ematch->data;
 294			if (!acpar.match->match(skb, &acpar))
 295				goto no_match;
 296		}
 297
 298		counter = xt_get_this_cpu_counter(&e->counters);
 299		ADD_COUNTER(*counter, skb->len, 1);
 300
 301		t = ipt_get_target_c(e);
 302		WARN_ON(!t->u.kernel.target);
 303
 304#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 305		/* The packet is traced: log it */
 306		if (unlikely(skb->nf_trace))
 307			trace_packet(state->net, skb, hook, state->in,
 308				     state->out, table->name, private, e);
 309#endif
 310		/* Standard target? */
 311		if (!t->u.kernel.target->target) {
 312			int v;
 313
 314			v = ((struct xt_standard_target *)t)->verdict;
 315			if (v < 0) {
 316				/* Pop from stack? */
 317				if (v != XT_RETURN) {
 318					verdict = (unsigned int)(-v) - 1;
 319					break;
 320				}
 321				if (stackidx == 0) {
 322					e = get_entry(table_base,
 323					    private->underflow[hook]);
 324				} else {
 325					e = jumpstack[--stackidx];
 326					e = ipt_next_entry(e);
 327				}
 328				continue;
 329			}
 330			if (table_base + v != ipt_next_entry(e) &&
 331			    !(e->ip.flags & IPT_F_GOTO)) {
 332				if (unlikely(stackidx >= private->stacksize)) {
 333					verdict = NF_DROP;
 334					break;
 335				}
 336				jumpstack[stackidx++] = e;
 337			}
 338
 339			e = get_entry(table_base, v);
 340			continue;
 341		}
 342
 343		acpar.target   = t->u.kernel.target;
 344		acpar.targinfo = t->data;
 345
 346		verdict = t->u.kernel.target->target(skb, &acpar);
 347		if (verdict == XT_CONTINUE) {
 348			/* Target might have changed stuff. */
 349			ip = ip_hdr(skb);
 350			e = ipt_next_entry(e);
 351		} else {
 352			/* Verdict */
 353			break;
 354		}
 355	} while (!acpar.hotdrop);
 356
 357	xt_write_recseq_end(addend);
 358	local_bh_enable();
 359
 360	if (acpar.hotdrop)
 361		return NF_DROP;
 362	else return verdict;
 363}
 364
 365/* Figures out from what hook each rule can be called: returns 0 if
 366   there are loops.  Puts hook bitmask in comefrom. */
 367static int
 368mark_source_chains(const struct xt_table_info *newinfo,
 369		   unsigned int valid_hooks, void *entry0,
 370		   unsigned int *offsets)
 371{
 372	unsigned int hook;
 373
 374	/* No recursion; use packet counter to save back ptrs (reset
 375	   to 0 as we leave), and comefrom to save source hook bitmask */
 376	for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 377		unsigned int pos = newinfo->hook_entry[hook];
 378		struct ipt_entry *e = entry0 + pos;
 379
 380		if (!(valid_hooks & (1 << hook)))
 381			continue;
 382
 383		/* Set initial back pointer. */
 384		e->counters.pcnt = pos;
 385
 386		for (;;) {
 387			const struct xt_standard_target *t
 388				= (void *)ipt_get_target_c(e);
 389			int visited = e->comefrom & (1 << hook);
 390
 391			if (e->comefrom & (1 << NF_INET_NUMHOOKS))
 392				return 0;
 393
 394			e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 395
 396			/* Unconditional return/END. */
 397			if ((unconditional(e) &&
 398			     (strcmp(t->target.u.user.name,
 399				     XT_STANDARD_TARGET) == 0) &&
 400			     t->verdict < 0) || visited) {
 401				unsigned int oldpos, size;
 402
 403				/* Return: backtrack through the last
 404				   big jump. */
 405				do {
 406					e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 407					oldpos = pos;
 408					pos = e->counters.pcnt;
 409					e->counters.pcnt = 0;
 410
 411					/* We're at the start. */
 412					if (pos == oldpos)
 413						goto next;
 414
 415					e = entry0 + pos;
 416				} while (oldpos == pos + e->next_offset);
 417
 418				/* Move along one */
 419				size = e->next_offset;
 420				e = entry0 + pos + size;
 421				if (pos + size >= newinfo->size)
 422					return 0;
 423				e->counters.pcnt = pos;
 424				pos += size;
 425			} else {
 426				int newpos = t->verdict;
 427
 428				if (strcmp(t->target.u.user.name,
 429					   XT_STANDARD_TARGET) == 0 &&
 430				    newpos >= 0) {
 431					/* This a jump; chase it. */
 432					if (!xt_find_jump_offset(offsets, newpos,
 433								 newinfo->number))
 434						return 0;
 435				} else {
 436					/* ... this is a fallthru */
 437					newpos = pos + e->next_offset;
 438					if (newpos >= newinfo->size)
 439						return 0;
 440				}
 441				e = entry0 + newpos;
 442				e->counters.pcnt = pos;
 443				pos = newpos;
 444			}
 445		}
 446next:		;
 447	}
 448	return 1;
 449}
 450
 451static void cleanup_match(struct xt_entry_match *m, struct net *net)
 452{
 453	struct xt_mtdtor_param par;
 454
 455	par.net       = net;
 456	par.match     = m->u.kernel.match;
 457	par.matchinfo = m->data;
 458	par.family    = NFPROTO_IPV4;
 459	if (par.match->destroy != NULL)
 460		par.match->destroy(&par);
 461	module_put(par.match->me);
 462}
 463
 464static int
 465check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 466{
 467	const struct ipt_ip *ip = par->entryinfo;
 468
 469	par->match     = m->u.kernel.match;
 470	par->matchinfo = m->data;
 471
 472	return xt_check_match(par, m->u.match_size - sizeof(*m),
 473			      ip->proto, ip->invflags & IPT_INV_PROTO);
 474}
 475
 476static int
 477find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 478{
 479	struct xt_match *match;
 480	int ret;
 481
 482	match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 483				      m->u.user.revision);
 484	if (IS_ERR(match))
 485		return PTR_ERR(match);
 486	m->u.kernel.match = match;
 487
 488	ret = check_match(m, par);
 489	if (ret)
 490		goto err;
 491
 492	return 0;
 493err:
 494	module_put(m->u.kernel.match->me);
 495	return ret;
 496}
 497
 498static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 499{
 500	struct xt_entry_target *t = ipt_get_target(e);
 501	struct xt_tgchk_param par = {
 502		.net       = net,
 503		.table     = name,
 504		.entryinfo = e,
 505		.target    = t->u.kernel.target,
 506		.targinfo  = t->data,
 507		.hook_mask = e->comefrom,
 508		.family    = NFPROTO_IPV4,
 509	};
 510
 511	return xt_check_target(&par, t->u.target_size - sizeof(*t),
 512			       e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 513}
 514
 515static int
 516find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 517		 unsigned int size,
 518		 struct xt_percpu_counter_alloc_state *alloc_state)
 519{
 520	struct xt_entry_target *t;
 521	struct xt_target *target;
 522	int ret;
 523	unsigned int j;
 524	struct xt_mtchk_param mtpar;
 525	struct xt_entry_match *ematch;
 526
 527	if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 528		return -ENOMEM;
 529
 530	j = 0;
 531	memset(&mtpar, 0, sizeof(mtpar));
 532	mtpar.net	= net;
 533	mtpar.table     = name;
 534	mtpar.entryinfo = &e->ip;
 535	mtpar.hook_mask = e->comefrom;
 536	mtpar.family    = NFPROTO_IPV4;
 537	xt_ematch_foreach(ematch, e) {
 538		ret = find_check_match(ematch, &mtpar);
 539		if (ret != 0)
 540			goto cleanup_matches;
 541		++j;
 542	}
 543
 544	t = ipt_get_target(e);
 545	target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 546					t->u.user.revision);
 547	if (IS_ERR(target)) {
 548		ret = PTR_ERR(target);
 549		goto cleanup_matches;
 550	}
 551	t->u.kernel.target = target;
 552
 553	ret = check_target(e, net, name);
 554	if (ret)
 555		goto err;
 556
 557	return 0;
 558 err:
 559	module_put(t->u.kernel.target->me);
 560 cleanup_matches:
 561	xt_ematch_foreach(ematch, e) {
 562		if (j-- == 0)
 563			break;
 564		cleanup_match(ematch, net);
 565	}
 566
 567	xt_percpu_counter_free(&e->counters);
 568
 569	return ret;
 570}
 571
 572static bool check_underflow(const struct ipt_entry *e)
 573{
 574	const struct xt_entry_target *t;
 575	unsigned int verdict;
 576
 577	if (!unconditional(e))
 578		return false;
 579	t = ipt_get_target_c(e);
 580	if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 581		return false;
 582	verdict = ((struct xt_standard_target *)t)->verdict;
 583	verdict = -verdict - 1;
 584	return verdict == NF_DROP || verdict == NF_ACCEPT;
 585}
 586
 587static int
 588check_entry_size_and_hooks(struct ipt_entry *e,
 589			   struct xt_table_info *newinfo,
 590			   const unsigned char *base,
 591			   const unsigned char *limit,
 592			   const unsigned int *hook_entries,
 593			   const unsigned int *underflows,
 594			   unsigned int valid_hooks)
 595{
 596	unsigned int h;
 597	int err;
 598
 599	if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 600	    (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 601	    (unsigned char *)e + e->next_offset > limit)
 602		return -EINVAL;
 603
 604	if (e->next_offset
 605	    < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target))
 606		return -EINVAL;
 607
 608	if (!ip_checkentry(&e->ip))
 609		return -EINVAL;
 610
 611	err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 612				     e->next_offset);
 613	if (err)
 614		return err;
 615
 616	/* Check hooks & underflows */
 617	for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 618		if (!(valid_hooks & (1 << h)))
 619			continue;
 620		if ((unsigned char *)e - base == hook_entries[h])
 621			newinfo->hook_entry[h] = hook_entries[h];
 622		if ((unsigned char *)e - base == underflows[h]) {
 623			if (!check_underflow(e))
 624				return -EINVAL;
 625
 626			newinfo->underflow[h] = underflows[h];
 627		}
 628	}
 629
 630	/* Clear counters and comefrom */
 631	e->counters = ((struct xt_counters) { 0, 0 });
 632	e->comefrom = 0;
 633	return 0;
 634}
 635
 636static void
 637cleanup_entry(struct ipt_entry *e, struct net *net)
 638{
 639	struct xt_tgdtor_param par;
 640	struct xt_entry_target *t;
 641	struct xt_entry_match *ematch;
 642
 643	/* Cleanup all matches */
 644	xt_ematch_foreach(ematch, e)
 645		cleanup_match(ematch, net);
 646	t = ipt_get_target(e);
 647
 648	par.net      = net;
 649	par.target   = t->u.kernel.target;
 650	par.targinfo = t->data;
 651	par.family   = NFPROTO_IPV4;
 652	if (par.target->destroy != NULL)
 653		par.target->destroy(&par);
 654	module_put(par.target->me);
 655	xt_percpu_counter_free(&e->counters);
 656}
 657
 658/* Checks and translates the user-supplied table segment (held in
 659   newinfo) */
 660static int
 661translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 662		const struct ipt_replace *repl)
 663{
 664	struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 665	struct ipt_entry *iter;
 666	unsigned int *offsets;
 667	unsigned int i;
 668	int ret = 0;
 669
 670	newinfo->size = repl->size;
 671	newinfo->number = repl->num_entries;
 672
 673	/* Init all hooks to impossible value. */
 674	for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 675		newinfo->hook_entry[i] = 0xFFFFFFFF;
 676		newinfo->underflow[i] = 0xFFFFFFFF;
 677	}
 678
 679	offsets = xt_alloc_entry_offsets(newinfo->number);
 680	if (!offsets)
 681		return -ENOMEM;
 682	i = 0;
 683	/* Walk through entries, checking offsets. */
 684	xt_entry_foreach(iter, entry0, newinfo->size) {
 685		ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 686						 entry0 + repl->size,
 687						 repl->hook_entry,
 688						 repl->underflow,
 689						 repl->valid_hooks);
 690		if (ret != 0)
 691			goto out_free;
 692		if (i < repl->num_entries)
 693			offsets[i] = (void *)iter - entry0;
 694		++i;
 695		if (strcmp(ipt_get_target(iter)->u.user.name,
 696		    XT_ERROR_TARGET) == 0)
 697			++newinfo->stacksize;
 698	}
 699
 700	ret = -EINVAL;
 701	if (i != repl->num_entries)
 702		goto out_free;
 703
 704	ret = xt_check_table_hooks(newinfo, repl->valid_hooks);
 705	if (ret)
 706		goto out_free;
 707
 708	if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 709		ret = -ELOOP;
 710		goto out_free;
 711	}
 712	kvfree(offsets);
 713
 714	/* Finally, each sanity check must pass */
 715	i = 0;
 716	xt_entry_foreach(iter, entry0, newinfo->size) {
 717		ret = find_check_entry(iter, net, repl->name, repl->size,
 718				       &alloc_state);
 719		if (ret != 0)
 720			break;
 721		++i;
 722	}
 723
 724	if (ret != 0) {
 725		xt_entry_foreach(iter, entry0, newinfo->size) {
 726			if (i-- == 0)
 727				break;
 728			cleanup_entry(iter, net);
 729		}
 730		return ret;
 731	}
 732
 733	return ret;
 734 out_free:
 735	kvfree(offsets);
 736	return ret;
 737}
 738
 739static void
 740get_counters(const struct xt_table_info *t,
 741	     struct xt_counters counters[])
 742{
 743	struct ipt_entry *iter;
 744	unsigned int cpu;
 745	unsigned int i;
 746
 747	for_each_possible_cpu(cpu) {
 748		seqcount_t *s = &per_cpu(xt_recseq, cpu);
 749
 750		i = 0;
 751		xt_entry_foreach(iter, t->entries, t->size) {
 752			struct xt_counters *tmp;
 753			u64 bcnt, pcnt;
 754			unsigned int start;
 755
 756			tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 757			do {
 758				start = read_seqcount_begin(s);
 759				bcnt = tmp->bcnt;
 760				pcnt = tmp->pcnt;
 761			} while (read_seqcount_retry(s, start));
 762
 763			ADD_COUNTER(counters[i], bcnt, pcnt);
 764			++i; /* macro does multi eval of i */
 765			cond_resched();
 766		}
 767	}
 768}
 769
 770static void get_old_counters(const struct xt_table_info *t,
 771			     struct xt_counters counters[])
 772{
 773	struct ipt_entry *iter;
 774	unsigned int cpu, i;
 775
 776	for_each_possible_cpu(cpu) {
 777		i = 0;
 778		xt_entry_foreach(iter, t->entries, t->size) {
 779			const struct xt_counters *tmp;
 780
 781			tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 782			ADD_COUNTER(counters[i], tmp->bcnt, tmp->pcnt);
 783			++i; /* macro does multi eval of i */
 784		}
 785
 786		cond_resched();
 787	}
 788}
 789
 790static struct xt_counters *alloc_counters(const struct xt_table *table)
 791{
 792	unsigned int countersize;
 793	struct xt_counters *counters;
 794	const struct xt_table_info *private = table->private;
 795
 796	/* We need atomic snapshot of counters: rest doesn't change
 797	   (other than comefrom, which userspace doesn't care
 798	   about). */
 799	countersize = sizeof(struct xt_counters) * private->number;
 800	counters = vzalloc(countersize);
 801
 802	if (counters == NULL)
 803		return ERR_PTR(-ENOMEM);
 804
 805	get_counters(private, counters);
 806
 807	return counters;
 808}
 809
 810static int
 811copy_entries_to_user(unsigned int total_size,
 812		     const struct xt_table *table,
 813		     void __user *userptr)
 814{
 815	unsigned int off, num;
 816	const struct ipt_entry *e;
 817	struct xt_counters *counters;
 818	const struct xt_table_info *private = table->private;
 819	int ret = 0;
 820	const void *loc_cpu_entry;
 821
 822	counters = alloc_counters(table);
 823	if (IS_ERR(counters))
 824		return PTR_ERR(counters);
 825
 826	loc_cpu_entry = private->entries;
 827
 828	/* FIXME: use iterator macros --RR */
 829	/* ... then go back and fix counters and names */
 830	for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 831		unsigned int i;
 832		const struct xt_entry_match *m;
 833		const struct xt_entry_target *t;
 834
 835		e = loc_cpu_entry + off;
 836		if (copy_to_user(userptr + off, e, sizeof(*e))) {
 837			ret = -EFAULT;
 838			goto free_counters;
 839		}
 840		if (copy_to_user(userptr + off
 841				 + offsetof(struct ipt_entry, counters),
 842				 &counters[num],
 843				 sizeof(counters[num])) != 0) {
 844			ret = -EFAULT;
 845			goto free_counters;
 846		}
 847
 848		for (i = sizeof(struct ipt_entry);
 849		     i < e->target_offset;
 850		     i += m->u.match_size) {
 851			m = (void *)e + i;
 852
 853			if (xt_match_to_user(m, userptr + off + i)) {
 854				ret = -EFAULT;
 855				goto free_counters;
 856			}
 857		}
 858
 859		t = ipt_get_target_c(e);
 860		if (xt_target_to_user(t, userptr + off + e->target_offset)) {
 861			ret = -EFAULT;
 862			goto free_counters;
 863		}
 864	}
 865
 866 free_counters:
 867	vfree(counters);
 868	return ret;
 869}
 870
 871#ifdef CONFIG_COMPAT
 872static void compat_standard_from_user(void *dst, const void *src)
 873{
 874	int v = *(compat_int_t *)src;
 875
 876	if (v > 0)
 877		v += xt_compat_calc_jump(AF_INET, v);
 878	memcpy(dst, &v, sizeof(v));
 879}
 880
 881static int compat_standard_to_user(void __user *dst, const void *src)
 882{
 883	compat_int_t cv = *(int *)src;
 884
 885	if (cv > 0)
 886		cv -= xt_compat_calc_jump(AF_INET, cv);
 887	return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 888}
 889
 890static int compat_calc_entry(const struct ipt_entry *e,
 891			     const struct xt_table_info *info,
 892			     const void *base, struct xt_table_info *newinfo)
 893{
 894	const struct xt_entry_match *ematch;
 895	const struct xt_entry_target *t;
 896	unsigned int entry_offset;
 897	int off, i, ret;
 898
 899	off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
 900	entry_offset = (void *)e - base;
 901	xt_ematch_foreach(ematch, e)
 902		off += xt_compat_match_offset(ematch->u.kernel.match);
 903	t = ipt_get_target_c(e);
 904	off += xt_compat_target_offset(t->u.kernel.target);
 905	newinfo->size -= off;
 906	ret = xt_compat_add_offset(AF_INET, entry_offset, off);
 907	if (ret)
 908		return ret;
 909
 910	for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 911		if (info->hook_entry[i] &&
 912		    (e < (struct ipt_entry *)(base + info->hook_entry[i])))
 913			newinfo->hook_entry[i] -= off;
 914		if (info->underflow[i] &&
 915		    (e < (struct ipt_entry *)(base + info->underflow[i])))
 916			newinfo->underflow[i] -= off;
 917	}
 918	return 0;
 919}
 920
 921static int compat_table_info(const struct xt_table_info *info,
 922			     struct xt_table_info *newinfo)
 923{
 924	struct ipt_entry *iter;
 925	const void *loc_cpu_entry;
 926	int ret;
 927
 928	if (!newinfo || !info)
 929		return -EINVAL;
 930
 931	/* we dont care about newinfo->entries */
 932	memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 933	newinfo->initial_entries = 0;
 934	loc_cpu_entry = info->entries;
 935	ret = xt_compat_init_offsets(AF_INET, info->number);
 936	if (ret)
 937		return ret;
 938	xt_entry_foreach(iter, loc_cpu_entry, info->size) {
 939		ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
 940		if (ret != 0)
 941			return ret;
 942	}
 943	return 0;
 944}
 945#endif
 946
 947static int get_info(struct net *net, void __user *user,
 948		    const int *len, int compat)
 949{
 950	char name[XT_TABLE_MAXNAMELEN];
 951	struct xt_table *t;
 952	int ret;
 953
 954	if (*len != sizeof(struct ipt_getinfo))
 955		return -EINVAL;
 956
 957	if (copy_from_user(name, user, sizeof(name)) != 0)
 958		return -EFAULT;
 959
 960	name[XT_TABLE_MAXNAMELEN-1] = '\0';
 961#ifdef CONFIG_COMPAT
 962	if (compat)
 963		xt_compat_lock(AF_INET);
 964#endif
 965	t = xt_request_find_table_lock(net, AF_INET, name);
 966	if (!IS_ERR(t)) {
 967		struct ipt_getinfo info;
 968		const struct xt_table_info *private = t->private;
 969#ifdef CONFIG_COMPAT
 970		struct xt_table_info tmp;
 971
 972		if (compat) {
 973			ret = compat_table_info(private, &tmp);
 974			xt_compat_flush_offsets(AF_INET);
 975			private = &tmp;
 976		}
 977#endif
 978		memset(&info, 0, sizeof(info));
 979		info.valid_hooks = t->valid_hooks;
 980		memcpy(info.hook_entry, private->hook_entry,
 981		       sizeof(info.hook_entry));
 982		memcpy(info.underflow, private->underflow,
 983		       sizeof(info.underflow));
 984		info.num_entries = private->number;
 985		info.size = private->size;
 986		strcpy(info.name, name);
 987
 988		if (copy_to_user(user, &info, *len) != 0)
 989			ret = -EFAULT;
 990		else
 991			ret = 0;
 992
 993		xt_table_unlock(t);
 994		module_put(t->me);
 995	} else
 996		ret = PTR_ERR(t);
 997#ifdef CONFIG_COMPAT
 998	if (compat)
 999		xt_compat_unlock(AF_INET);
1000#endif
1001	return ret;
1002}
1003
1004static int
1005get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1006	    const int *len)
1007{
1008	int ret;
1009	struct ipt_get_entries get;
1010	struct xt_table *t;
1011
1012	if (*len < sizeof(get))
1013		return -EINVAL;
1014	if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1015		return -EFAULT;
1016	if (*len != sizeof(struct ipt_get_entries) + get.size)
1017		return -EINVAL;
1018	get.name[sizeof(get.name) - 1] = '\0';
1019
1020	t = xt_find_table_lock(net, AF_INET, get.name);
1021	if (!IS_ERR(t)) {
1022		const struct xt_table_info *private = t->private;
1023		if (get.size == private->size)
1024			ret = copy_entries_to_user(private->size,
1025						   t, uptr->entrytable);
1026		else
1027			ret = -EAGAIN;
1028
1029		module_put(t->me);
1030		xt_table_unlock(t);
1031	} else
1032		ret = PTR_ERR(t);
1033
1034	return ret;
1035}
1036
1037static int
1038__do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1039	     struct xt_table_info *newinfo, unsigned int num_counters,
1040	     void __user *counters_ptr)
1041{
1042	int ret;
1043	struct xt_table *t;
1044	struct xt_table_info *oldinfo;
1045	struct xt_counters *counters;
1046	struct ipt_entry *iter;
1047
1048	ret = 0;
1049	counters = xt_counters_alloc(num_counters);
1050	if (!counters) {
1051		ret = -ENOMEM;
1052		goto out;
1053	}
1054
1055	t = xt_request_find_table_lock(net, AF_INET, name);
1056	if (IS_ERR(t)) {
1057		ret = PTR_ERR(t);
1058		goto free_newinfo_counters_untrans;
1059	}
1060
1061	/* You lied! */
1062	if (valid_hooks != t->valid_hooks) {
1063		ret = -EINVAL;
1064		goto put_module;
1065	}
1066
1067	oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1068	if (!oldinfo)
1069		goto put_module;
1070
1071	/* Update module usage count based on number of rules */
1072	if ((oldinfo->number > oldinfo->initial_entries) ||
1073	    (newinfo->number <= oldinfo->initial_entries))
1074		module_put(t->me);
1075	if ((oldinfo->number > oldinfo->initial_entries) &&
1076	    (newinfo->number <= oldinfo->initial_entries))
1077		module_put(t->me);
1078
1079	xt_table_unlock(t);
1080
1081	get_old_counters(oldinfo, counters);
1082
1083	/* Decrease module usage counts and free resource */
1084	xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1085		cleanup_entry(iter, net);
1086
1087	xt_free_table_info(oldinfo);
1088	if (copy_to_user(counters_ptr, counters,
1089			 sizeof(struct xt_counters) * num_counters) != 0) {
1090		/* Silent error, can't fail, new table is already in place */
1091		net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1092	}
1093	vfree(counters);
1094	return ret;
1095
1096 put_module:
1097	module_put(t->me);
1098	xt_table_unlock(t);
1099 free_newinfo_counters_untrans:
1100	vfree(counters);
1101 out:
1102	return ret;
1103}
1104
1105static int
1106do_replace(struct net *net, const void __user *user, unsigned int len)
1107{
1108	int ret;
1109	struct ipt_replace tmp;
1110	struct xt_table_info *newinfo;
1111	void *loc_cpu_entry;
1112	struct ipt_entry *iter;
1113
1114	if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1115		return -EFAULT;
1116
1117	/* overflow check */
1118	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1119		return -ENOMEM;
1120	if (tmp.num_counters == 0)
1121		return -EINVAL;
1122
1123	tmp.name[sizeof(tmp.name)-1] = 0;
1124
1125	newinfo = xt_alloc_table_info(tmp.size);
1126	if (!newinfo)
1127		return -ENOMEM;
1128
1129	loc_cpu_entry = newinfo->entries;
1130	if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1131			   tmp.size) != 0) {
1132		ret = -EFAULT;
1133		goto free_newinfo;
1134	}
1135
1136	ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1137	if (ret != 0)
1138		goto free_newinfo;
1139
1140	ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1141			   tmp.num_counters, tmp.counters);
1142	if (ret)
1143		goto free_newinfo_untrans;
1144	return 0;
1145
1146 free_newinfo_untrans:
1147	xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1148		cleanup_entry(iter, net);
1149 free_newinfo:
1150	xt_free_table_info(newinfo);
1151	return ret;
1152}
1153
1154static int
1155do_add_counters(struct net *net, const void __user *user,
1156		unsigned int len, int compat)
1157{
1158	unsigned int i;
1159	struct xt_counters_info tmp;
1160	struct xt_counters *paddc;
1161	struct xt_table *t;
1162	const struct xt_table_info *private;
1163	int ret = 0;
1164	struct ipt_entry *iter;
1165	unsigned int addend;
1166
1167	paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1168	if (IS_ERR(paddc))
1169		return PTR_ERR(paddc);
1170
1171	t = xt_find_table_lock(net, AF_INET, tmp.name);
1172	if (IS_ERR(t)) {
1173		ret = PTR_ERR(t);
1174		goto free;
1175	}
1176
1177	local_bh_disable();
1178	private = t->private;
1179	if (private->number != tmp.num_counters) {
1180		ret = -EINVAL;
1181		goto unlock_up_free;
1182	}
1183
1184	i = 0;
1185	addend = xt_write_recseq_begin();
1186	xt_entry_foreach(iter, private->entries, private->size) {
1187		struct xt_counters *tmp;
1188
1189		tmp = xt_get_this_cpu_counter(&iter->counters);
1190		ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1191		++i;
1192	}
1193	xt_write_recseq_end(addend);
1194 unlock_up_free:
1195	local_bh_enable();
1196	xt_table_unlock(t);
1197	module_put(t->me);
1198 free:
1199	vfree(paddc);
1200
1201	return ret;
1202}
1203
1204#ifdef CONFIG_COMPAT
1205struct compat_ipt_replace {
1206	char			name[XT_TABLE_MAXNAMELEN];
1207	u32			valid_hooks;
1208	u32			num_entries;
1209	u32			size;
1210	u32			hook_entry[NF_INET_NUMHOOKS];
1211	u32			underflow[NF_INET_NUMHOOKS];
1212	u32			num_counters;
1213	compat_uptr_t		counters;	/* struct xt_counters * */
1214	struct compat_ipt_entry	entries[];
1215};
1216
1217static int
1218compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1219			  unsigned int *size, struct xt_counters *counters,
1220			  unsigned int i)
1221{
1222	struct xt_entry_target *t;
1223	struct compat_ipt_entry __user *ce;
1224	u_int16_t target_offset, next_offset;
1225	compat_uint_t origsize;
1226	const struct xt_entry_match *ematch;
1227	int ret = 0;
1228
1229	origsize = *size;
1230	ce = *dstptr;
1231	if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1232	    copy_to_user(&ce->counters, &counters[i],
1233	    sizeof(counters[i])) != 0)
1234		return -EFAULT;
1235
1236	*dstptr += sizeof(struct compat_ipt_entry);
1237	*size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1238
1239	xt_ematch_foreach(ematch, e) {
1240		ret = xt_compat_match_to_user(ematch, dstptr, size);
1241		if (ret != 0)
1242			return ret;
1243	}
1244	target_offset = e->target_offset - (origsize - *size);
1245	t = ipt_get_target(e);
1246	ret = xt_compat_target_to_user(t, dstptr, size);
1247	if (ret)
1248		return ret;
1249	next_offset = e->next_offset - (origsize - *size);
1250	if (put_user(target_offset, &ce->target_offset) != 0 ||
1251	    put_user(next_offset, &ce->next_offset) != 0)
1252		return -EFAULT;
1253	return 0;
1254}
1255
1256static int
1257compat_find_calc_match(struct xt_entry_match *m,
1258		       const struct ipt_ip *ip,
1259		       int *size)
1260{
1261	struct xt_match *match;
1262
1263	match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1264				      m->u.user.revision);
1265	if (IS_ERR(match))
1266		return PTR_ERR(match);
1267
1268	m->u.kernel.match = match;
1269	*size += xt_compat_match_offset(match);
1270	return 0;
1271}
1272
1273static void compat_release_entry(struct compat_ipt_entry *e)
1274{
1275	struct xt_entry_target *t;
1276	struct xt_entry_match *ematch;
1277
1278	/* Cleanup all matches */
1279	xt_ematch_foreach(ematch, e)
1280		module_put(ematch->u.kernel.match->me);
1281	t = compat_ipt_get_target(e);
1282	module_put(t->u.kernel.target->me);
1283}
1284
1285static int
1286check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1287				  struct xt_table_info *newinfo,
1288				  unsigned int *size,
1289				  const unsigned char *base,
1290				  const unsigned char *limit)
1291{
1292	struct xt_entry_match *ematch;
1293	struct xt_entry_target *t;
1294	struct xt_target *target;
1295	unsigned int entry_offset;
1296	unsigned int j;
1297	int ret, off;
1298
1299	if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1300	    (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1301	    (unsigned char *)e + e->next_offset > limit)
1302		return -EINVAL;
1303
1304	if (e->next_offset < sizeof(struct compat_ipt_entry) +
1305			     sizeof(struct compat_xt_entry_target))
1306		return -EINVAL;
1307
1308	if (!ip_checkentry(&e->ip))
1309		return -EINVAL;
1310
1311	ret = xt_compat_check_entry_offsets(e, e->elems,
1312					    e->target_offset, e->next_offset);
1313	if (ret)
1314		return ret;
1315
1316	off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1317	entry_offset = (void *)e - (void *)base;
1318	j = 0;
1319	xt_ematch_foreach(ematch, e) {
1320		ret = compat_find_calc_match(ematch, &e->ip, &off);
1321		if (ret != 0)
1322			goto release_matches;
1323		++j;
1324	}
1325
1326	t = compat_ipt_get_target(e);
1327	target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1328					t->u.user.revision);
1329	if (IS_ERR(target)) {
1330		ret = PTR_ERR(target);
1331		goto release_matches;
1332	}
1333	t->u.kernel.target = target;
1334
1335	off += xt_compat_target_offset(target);
1336	*size += off;
1337	ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1338	if (ret)
1339		goto out;
1340
1341	return 0;
1342
1343out:
1344	module_put(t->u.kernel.target->me);
1345release_matches:
1346	xt_ematch_foreach(ematch, e) {
1347		if (j-- == 0)
1348			break;
1349		module_put(ematch->u.kernel.match->me);
1350	}
1351	return ret;
1352}
1353
1354static void
1355compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1356			    unsigned int *size,
1357			    struct xt_table_info *newinfo, unsigned char *base)
1358{
1359	struct xt_entry_target *t;
1360	struct ipt_entry *de;
1361	unsigned int origsize;
1362	int h;
1363	struct xt_entry_match *ematch;
1364
1365	origsize = *size;
1366	de = *dstptr;
1367	memcpy(de, e, sizeof(struct ipt_entry));
1368	memcpy(&de->counters, &e->counters, sizeof(e->counters));
1369
1370	*dstptr += sizeof(struct ipt_entry);
1371	*size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1372
1373	xt_ematch_foreach(ematch, e)
1374		xt_compat_match_from_user(ematch, dstptr, size);
1375
1376	de->target_offset = e->target_offset - (origsize - *size);
1377	t = compat_ipt_get_target(e);
1378	xt_compat_target_from_user(t, dstptr, size);
1379
1380	de->next_offset = e->next_offset - (origsize - *size);
1381
1382	for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1383		if ((unsigned char *)de - base < newinfo->hook_entry[h])
1384			newinfo->hook_entry[h] -= origsize - *size;
1385		if ((unsigned char *)de - base < newinfo->underflow[h])
1386			newinfo->underflow[h] -= origsize - *size;
1387	}
1388}
1389
1390static int
1391translate_compat_table(struct net *net,
1392		       struct xt_table_info **pinfo,
1393		       void **pentry0,
1394		       const struct compat_ipt_replace *compatr)
1395{
1396	unsigned int i, j;
1397	struct xt_table_info *newinfo, *info;
1398	void *pos, *entry0, *entry1;
1399	struct compat_ipt_entry *iter0;
1400	struct ipt_replace repl;
1401	unsigned int size;
1402	int ret;
1403
1404	info = *pinfo;
1405	entry0 = *pentry0;
1406	size = compatr->size;
1407	info->number = compatr->num_entries;
1408
1409	j = 0;
1410	xt_compat_lock(AF_INET);
1411	ret = xt_compat_init_offsets(AF_INET, compatr->num_entries);
1412	if (ret)
1413		goto out_unlock;
1414	/* Walk through entries, checking offsets. */
1415	xt_entry_foreach(iter0, entry0, compatr->size) {
1416		ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1417							entry0,
1418							entry0 + compatr->size);
1419		if (ret != 0)
1420			goto out_unlock;
1421		++j;
1422	}
1423
1424	ret = -EINVAL;
1425	if (j != compatr->num_entries)
1426		goto out_unlock;
1427
1428	ret = -ENOMEM;
1429	newinfo = xt_alloc_table_info(size);
1430	if (!newinfo)
1431		goto out_unlock;
1432
1433	newinfo->number = compatr->num_entries;
1434	for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1435		newinfo->hook_entry[i] = compatr->hook_entry[i];
1436		newinfo->underflow[i] = compatr->underflow[i];
1437	}
1438	entry1 = newinfo->entries;
1439	pos = entry1;
1440	size = compatr->size;
1441	xt_entry_foreach(iter0, entry0, compatr->size)
1442		compat_copy_entry_from_user(iter0, &pos, &size,
1443					    newinfo, entry1);
1444
1445	/* all module references in entry0 are now gone.
1446	 * entry1/newinfo contains a 64bit ruleset that looks exactly as
1447	 * generated by 64bit userspace.
1448	 *
1449	 * Call standard translate_table() to validate all hook_entrys,
1450	 * underflows, check for loops, etc.
1451	 */
1452	xt_compat_flush_offsets(AF_INET);
1453	xt_compat_unlock(AF_INET);
1454
1455	memcpy(&repl, compatr, sizeof(*compatr));
1456
1457	for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1458		repl.hook_entry[i] = newinfo->hook_entry[i];
1459		repl.underflow[i] = newinfo->underflow[i];
1460	}
1461
1462	repl.num_counters = 0;
1463	repl.counters = NULL;
1464	repl.size = newinfo->size;
1465	ret = translate_table(net, newinfo, entry1, &repl);
1466	if (ret)
1467		goto free_newinfo;
1468
1469	*pinfo = newinfo;
1470	*pentry0 = entry1;
1471	xt_free_table_info(info);
1472	return 0;
1473
1474free_newinfo:
1475	xt_free_table_info(newinfo);
1476	return ret;
1477out_unlock:
1478	xt_compat_flush_offsets(AF_INET);
1479	xt_compat_unlock(AF_INET);
1480	xt_entry_foreach(iter0, entry0, compatr->size) {
1481		if (j-- == 0)
1482			break;
1483		compat_release_entry(iter0);
1484	}
1485	return ret;
1486}
1487
1488static int
1489compat_do_replace(struct net *net, void __user *user, unsigned int len)
1490{
1491	int ret;
1492	struct compat_ipt_replace tmp;
1493	struct xt_table_info *newinfo;
1494	void *loc_cpu_entry;
1495	struct ipt_entry *iter;
1496
1497	if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1498		return -EFAULT;
1499
1500	/* overflow check */
1501	if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1502		return -ENOMEM;
1503	if (tmp.num_counters == 0)
1504		return -EINVAL;
1505
1506	tmp.name[sizeof(tmp.name)-1] = 0;
1507
1508	newinfo = xt_alloc_table_info(tmp.size);
1509	if (!newinfo)
1510		return -ENOMEM;
1511
1512	loc_cpu_entry = newinfo->entries;
1513	if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1514			   tmp.size) != 0) {
1515		ret = -EFAULT;
1516		goto free_newinfo;
1517	}
1518
1519	ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1520	if (ret != 0)
1521		goto free_newinfo;
1522
1523	ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1524			   tmp.num_counters, compat_ptr(tmp.counters));
1525	if (ret)
1526		goto free_newinfo_untrans;
1527	return 0;
1528
1529 free_newinfo_untrans:
1530	xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1531		cleanup_entry(iter, net);
1532 free_newinfo:
1533	xt_free_table_info(newinfo);
1534	return ret;
1535}
1536
1537static int
1538compat_do_ipt_set_ctl(struct sock *sk,	int cmd, void __user *user,
1539		      unsigned int len)
1540{
1541	int ret;
1542
1543	if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1544		return -EPERM;
1545
1546	switch (cmd) {
1547	case IPT_SO_SET_REPLACE:
1548		ret = compat_do_replace(sock_net(sk), user, len);
1549		break;
1550
1551	case IPT_SO_SET_ADD_COUNTERS:
1552		ret = do_add_counters(sock_net(sk), user, len, 1);
1553		break;
1554
1555	default:
1556		ret = -EINVAL;
1557	}
1558
1559	return ret;
1560}
1561
1562struct compat_ipt_get_entries {
1563	char name[XT_TABLE_MAXNAMELEN];
1564	compat_uint_t size;
1565	struct compat_ipt_entry entrytable[];
1566};
1567
1568static int
1569compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1570			    void __user *userptr)
1571{
1572	struct xt_counters *counters;
1573	const struct xt_table_info *private = table->private;
1574	void __user *pos;
1575	unsigned int size;
1576	int ret = 0;
1577	unsigned int i = 0;
1578	struct ipt_entry *iter;
1579
1580	counters = alloc_counters(table);
1581	if (IS_ERR(counters))
1582		return PTR_ERR(counters);
1583
1584	pos = userptr;
1585	size = total_size;
1586	xt_entry_foreach(iter, private->entries, total_size) {
1587		ret = compat_copy_entry_to_user(iter, &pos,
1588						&size, counters, i++);
1589		if (ret != 0)
1590			break;
1591	}
1592
1593	vfree(counters);
1594	return ret;
1595}
1596
1597static int
1598compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1599		   int *len)
1600{
1601	int ret;
1602	struct compat_ipt_get_entries get;
1603	struct xt_table *t;
1604
1605	if (*len < sizeof(get))
1606		return -EINVAL;
1607
1608	if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1609		return -EFAULT;
1610
1611	if (*len != sizeof(struct compat_ipt_get_entries) + get.size)
1612		return -EINVAL;
1613
1614	get.name[sizeof(get.name) - 1] = '\0';
1615
1616	xt_compat_lock(AF_INET);
1617	t = xt_find_table_lock(net, AF_INET, get.name);
1618	if (!IS_ERR(t)) {
1619		const struct xt_table_info *private = t->private;
1620		struct xt_table_info info;
1621		ret = compat_table_info(private, &info);
1622		if (!ret && get.size == info.size)
1623			ret = compat_copy_entries_to_user(private->size,
1624							  t, uptr->entrytable);
1625		else if (!ret)
1626			ret = -EAGAIN;
1627
1628		xt_compat_flush_offsets(AF_INET);
1629		module_put(t->me);
1630		xt_table_unlock(t);
1631	} else
1632		ret = PTR_ERR(t);
1633
1634	xt_compat_unlock(AF_INET);
1635	return ret;
1636}
1637
1638static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1639
1640static int
1641compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1642{
1643	int ret;
1644
1645	if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1646		return -EPERM;
1647
1648	switch (cmd) {
1649	case IPT_SO_GET_INFO:
1650		ret = get_info(sock_net(sk), user, len, 1);
1651		break;
1652	case IPT_SO_GET_ENTRIES:
1653		ret = compat_get_entries(sock_net(sk), user, len);
1654		break;
1655	default:
1656		ret = do_ipt_get_ctl(sk, cmd, user, len);
1657	}
1658	return ret;
1659}
1660#endif
1661
1662static int
1663do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1664{
1665	int ret;
1666
1667	if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1668		return -EPERM;
1669
1670	switch (cmd) {
1671	case IPT_SO_SET_REPLACE:
1672		ret = do_replace(sock_net(sk), user, len);
1673		break;
1674
1675	case IPT_SO_SET_ADD_COUNTERS:
1676		ret = do_add_counters(sock_net(sk), user, len, 0);
1677		break;
1678
1679	default:
1680		ret = -EINVAL;
1681	}
1682
1683	return ret;
1684}
1685
1686static int
1687do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1688{
1689	int ret;
1690
1691	if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1692		return -EPERM;
1693
1694	switch (cmd) {
1695	case IPT_SO_GET_INFO:
1696		ret = get_info(sock_net(sk), user, len, 0);
1697		break;
1698
1699	case IPT_SO_GET_ENTRIES:
1700		ret = get_entries(sock_net(sk), user, len);
1701		break;
1702
1703	case IPT_SO_GET_REVISION_MATCH:
1704	case IPT_SO_GET_REVISION_TARGET: {
1705		struct xt_get_revision rev;
1706		int target;
1707
1708		if (*len != sizeof(rev)) {
1709			ret = -EINVAL;
1710			break;
1711		}
1712		if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1713			ret = -EFAULT;
1714			break;
1715		}
1716		rev.name[sizeof(rev.name)-1] = 0;
1717
1718		if (cmd == IPT_SO_GET_REVISION_TARGET)
1719			target = 1;
1720		else
1721			target = 0;
1722
1723		try_then_request_module(xt_find_revision(AF_INET, rev.name,
1724							 rev.revision,
1725							 target, &ret),
1726					"ipt_%s", rev.name);
1727		break;
1728	}
1729
1730	default:
1731		ret = -EINVAL;
1732	}
1733
1734	return ret;
1735}
1736
1737static void __ipt_unregister_table(struct net *net, struct xt_table *table)
1738{
1739	struct xt_table_info *private;
1740	void *loc_cpu_entry;
1741	struct module *table_owner = table->me;
1742	struct ipt_entry *iter;
1743
1744	private = xt_unregister_table(table);
1745
1746	/* Decrease module usage counts and free resources */
1747	loc_cpu_entry = private->entries;
1748	xt_entry_foreach(iter, loc_cpu_entry, private->size)
1749		cleanup_entry(iter, net);
1750	if (private->number > private->initial_entries)
1751		module_put(table_owner);
1752	xt_free_table_info(private);
1753}
1754
1755int ipt_register_table(struct net *net, const struct xt_table *table,
1756		       const struct ipt_replace *repl,
1757		       const struct nf_hook_ops *ops, struct xt_table **res)
1758{
1759	int ret;
1760	struct xt_table_info *newinfo;
1761	struct xt_table_info bootstrap = {0};
1762	void *loc_cpu_entry;
1763	struct xt_table *new_table;
1764
1765	newinfo = xt_alloc_table_info(repl->size);
1766	if (!newinfo)
1767		return -ENOMEM;
1768
1769	loc_cpu_entry = newinfo->entries;
1770	memcpy(loc_cpu_entry, repl->entries, repl->size);
1771
1772	ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1773	if (ret != 0)
1774		goto out_free;
1775
1776	new_table = xt_register_table(net, table, &bootstrap, newinfo);
1777	if (IS_ERR(new_table)) {
1778		ret = PTR_ERR(new_table);
1779		goto out_free;
1780	}
1781
1782	/* set res now, will see skbs right after nf_register_net_hooks */
1783	WRITE_ONCE(*res, new_table);
1784	if (!ops)
1785		return 0;
1786
1787	ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1788	if (ret != 0) {
1789		__ipt_unregister_table(net, new_table);
1790		*res = NULL;
1791	}
1792
1793	return ret;
1794
1795out_free:
1796	xt_free_table_info(newinfo);
1797	return ret;
1798}
1799
1800void ipt_unregister_table(struct net *net, struct xt_table *table,
1801			  const struct nf_hook_ops *ops)
1802{
1803	if (ops)
1804		nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1805	__ipt_unregister_table(net, table);
1806}
1807
1808/* Returns 1 if the type and code is matched by the range, 0 otherwise */
1809static inline bool
1810icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1811		     u_int8_t type, u_int8_t code,
1812		     bool invert)
1813{
1814	return ((test_type == 0xFF) ||
1815		(type == test_type && code >= min_code && code <= max_code))
1816		^ invert;
1817}
1818
1819static bool
1820icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
1821{
1822	const struct icmphdr *ic;
1823	struct icmphdr _icmph;
1824	const struct ipt_icmp *icmpinfo = par->matchinfo;
1825
1826	/* Must not be a fragment. */
1827	if (par->fragoff != 0)
1828		return false;
1829
1830	ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1831	if (ic == NULL) {
1832		/* We've been asked to examine this packet, and we
1833		 * can't.  Hence, no choice but to drop.
1834		 */
1835		par->hotdrop = true;
1836		return false;
1837	}
1838
1839	return icmp_type_code_match(icmpinfo->type,
1840				    icmpinfo->code[0],
1841				    icmpinfo->code[1],
1842				    ic->type, ic->code,
1843				    !!(icmpinfo->invflags&IPT_ICMP_INV));
1844}
1845
1846static int icmp_checkentry(const struct xt_mtchk_param *par)
1847{
1848	const struct ipt_icmp *icmpinfo = par->matchinfo;
1849
1850	/* Must specify no unknown invflags */
1851	return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
1852}
1853
1854static struct xt_target ipt_builtin_tg[] __read_mostly = {
1855	{
1856		.name             = XT_STANDARD_TARGET,
1857		.targetsize       = sizeof(int),
1858		.family           = NFPROTO_IPV4,
1859#ifdef CONFIG_COMPAT
1860		.compatsize       = sizeof(compat_int_t),
1861		.compat_from_user = compat_standard_from_user,
1862		.compat_to_user   = compat_standard_to_user,
1863#endif
1864	},
1865	{
1866		.name             = XT_ERROR_TARGET,
1867		.target           = ipt_error,
1868		.targetsize       = XT_FUNCTION_MAXNAMELEN,
1869		.family           = NFPROTO_IPV4,
1870	},
1871};
1872
1873static struct nf_sockopt_ops ipt_sockopts = {
1874	.pf		= PF_INET,
1875	.set_optmin	= IPT_BASE_CTL,
1876	.set_optmax	= IPT_SO_SET_MAX+1,
1877	.set		= do_ipt_set_ctl,
1878#ifdef CONFIG_COMPAT
1879	.compat_set	= compat_do_ipt_set_ctl,
1880#endif
1881	.get_optmin	= IPT_BASE_CTL,
1882	.get_optmax	= IPT_SO_GET_MAX+1,
1883	.get		= do_ipt_get_ctl,
1884#ifdef CONFIG_COMPAT
1885	.compat_get	= compat_do_ipt_get_ctl,
1886#endif
1887	.owner		= THIS_MODULE,
1888};
1889
1890static struct xt_match ipt_builtin_mt[] __read_mostly = {
1891	{
1892		.name       = "icmp",
1893		.match      = icmp_match,
1894		.matchsize  = sizeof(struct ipt_icmp),
1895		.checkentry = icmp_checkentry,
1896		.proto      = IPPROTO_ICMP,
1897		.family     = NFPROTO_IPV4,
1898		.me	    = THIS_MODULE,
1899	},
1900};
1901
1902static int __net_init ip_tables_net_init(struct net *net)
1903{
1904	return xt_proto_init(net, NFPROTO_IPV4);
1905}
1906
1907static void __net_exit ip_tables_net_exit(struct net *net)
1908{
1909	xt_proto_fini(net, NFPROTO_IPV4);
1910}
1911
1912static struct pernet_operations ip_tables_net_ops = {
1913	.init = ip_tables_net_init,
1914	.exit = ip_tables_net_exit,
1915};
1916
1917static int __init ip_tables_init(void)
1918{
1919	int ret;
1920
1921	ret = register_pernet_subsys(&ip_tables_net_ops);
1922	if (ret < 0)
1923		goto err1;
1924
1925	/* No one else will be downing sem now, so we won't sleep */
1926	ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1927	if (ret < 0)
1928		goto err2;
1929	ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1930	if (ret < 0)
1931		goto err4;
1932
1933	/* Register setsockopt */
1934	ret = nf_register_sockopt(&ipt_sockopts);
1935	if (ret < 0)
1936		goto err5;
1937
1938	return 0;
1939
1940err5:
1941	xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1942err4:
1943	xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1944err2:
1945	unregister_pernet_subsys(&ip_tables_net_ops);
1946err1:
1947	return ret;
1948}
1949
1950static void __exit ip_tables_fini(void)
1951{
1952	nf_unregister_sockopt(&ipt_sockopts);
1953
1954	xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1955	xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1956	unregister_pernet_subsys(&ip_tables_net_ops);
1957}
1958
1959EXPORT_SYMBOL(ipt_register_table);
1960EXPORT_SYMBOL(ipt_unregister_table);
1961EXPORT_SYMBOL(ipt_do_table);
1962module_init(ip_tables_init);
1963module_exit(ip_tables_fini);