09f6567a85b7181a20810be850e0bc1ef32d173a
[linux-2.6.git] / net / ipv4 / netfilter / ip_tables.c
1 /*
2  * Packet matching code.
3  *
4  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
5  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License version 2 as
9  * published by the Free Software Foundation.
10  */
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12 #include <linux/cache.h>
13 #include <linux/capability.h>
14 #include <linux/skbuff.h>
15 #include <linux/kmod.h>
16 #include <linux/vmalloc.h>
17 #include <linux/netdevice.h>
18 #include <linux/module.h>
19 #include <linux/icmp.h>
20 #include <net/ip.h>
21 #include <net/compat.h>
22 #include <asm/uaccess.h>
23 #include <linux/mutex.h>
24 #include <linux/proc_fs.h>
25 #include <linux/err.h>
26 #include <linux/cpumask.h>
27
28 #include <linux/netfilter/x_tables.h>
29 #include <linux/netfilter_ipv4/ip_tables.h>
30 #include <net/netfilter/nf_log.h>
31 #include "../../netfilter/xt_repldata.h"
32
33 MODULE_LICENSE("GPL");
34 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
35 MODULE_DESCRIPTION("IPv4 packet filter");
36
37 /*#define DEBUG_IP_FIREWALL*/
38 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
39 /*#define DEBUG_IP_FIREWALL_USER*/
40
41 #ifdef DEBUG_IP_FIREWALL
42 #define dprintf(format, args...) pr_info(format , ## args)
43 #else
44 #define dprintf(format, args...)
45 #endif
46
47 #ifdef DEBUG_IP_FIREWALL_USER
48 #define duprintf(format, args...) pr_info(format , ## args)
49 #else
50 #define duprintf(format, args...)
51 #endif
52
53 #ifdef CONFIG_NETFILTER_DEBUG
54 #define IP_NF_ASSERT(x)                                         \
55 do {                                                            \
56         if (!(x))                                               \
57                 printk("IP_NF_ASSERT: %s:%s:%u\n",              \
58                        __func__, __FILE__, __LINE__);   \
59 } while(0)
60 #else
61 #define IP_NF_ASSERT(x)
62 #endif
63
64 #if 0
65 /* All the better to debug you with... */
66 #define static
67 #define inline
68 #endif
69
70 void *ipt_alloc_initial_table(const struct xt_table *info)
71 {
72         return xt_alloc_initial_table(ipt, IPT);
73 }
74 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
75
76 /*
77    We keep a set of rules for each CPU, so we can avoid write-locking
78    them in the softirq when updating the counters and therefore
79    only need to read-lock in the softirq; doing a write_lock_bh() in user
80    context stops packets coming through and allows user context to read
81    the counters or update the rules.
82
83    Hence the start of any table is given by get_table() below.  */
84
85 /* Returns whether matches rule or not. */
86 /* Performance critical - called for every packet */
87 static inline bool
88 ip_packet_match(const struct iphdr *ip,
89                 const char *indev,
90                 const char *outdev,
91                 const struct ipt_ip *ipinfo,
92                 int isfrag)
93 {
94         unsigned long ret;
95
96 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
97
98         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
99                   IPT_INV_SRCIP) ||
100             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
101                   IPT_INV_DSTIP)) {
102                 dprintf("Source or dest mismatch.\n");
103
104                 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
105                         &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
106                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
107                 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
108                         &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
109                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
110                 return false;
111         }
112
113         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
114
115         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
116                 dprintf("VIA in mismatch (%s vs %s).%s\n",
117                         indev, ipinfo->iniface,
118                         ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
119                 return false;
120         }
121
122         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
123
124         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
125                 dprintf("VIA out mismatch (%s vs %s).%s\n",
126                         outdev, ipinfo->outiface,
127                         ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
128                 return false;
129         }
130
131         /* Check specific protocol */
132         if (ipinfo->proto &&
133             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
134                 dprintf("Packet protocol %hi does not match %hi.%s\n",
135                         ip->protocol, ipinfo->proto,
136                         ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
137                 return false;
138         }
139
140         /* If we have a fragment rule but the packet is not a fragment
141          * then we return zero */
142         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
143                 dprintf("Fragment rule but not fragment.%s\n",
144                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
145                 return false;
146         }
147
148         return true;
149 }
150
151 static bool
152 ip_checkentry(const struct ipt_ip *ip)
153 {
154         if (ip->flags & ~IPT_F_MASK) {
155                 duprintf("Unknown flag bits set: %08X\n",
156                          ip->flags & ~IPT_F_MASK);
157                 return false;
158         }
159         if (ip->invflags & ~IPT_INV_MASK) {
160                 duprintf("Unknown invflag bits set: %08X\n",
161                          ip->invflags & ~IPT_INV_MASK);
162                 return false;
163         }
164         return true;
165 }
166
167 static unsigned int
168 ipt_error(struct sk_buff *skb, const struct xt_target_param *par)
169 {
170         if (net_ratelimit())
171                 pr_info("error: `%s'\n", (const char *)par->targinfo);
172
173         return NF_DROP;
174 }
175
176 /* Performance critical - called for every packet */
177 static inline bool
178 do_match(const struct ipt_entry_match *m, const struct sk_buff *skb,
179          struct xt_match_param *par)
180 {
181         par->match     = m->u.kernel.match;
182         par->matchinfo = m->data;
183
184         /* Stop iteration if it doesn't match */
185         if (!m->u.kernel.match->match(skb, par))
186                 return true;
187         else
188                 return false;
189 }
190
191 /* Performance critical */
192 static inline struct ipt_entry *
193 get_entry(const void *base, unsigned int offset)
194 {
195         return (struct ipt_entry *)(base + offset);
196 }
197
198 /* All zeroes == unconditional rule. */
199 /* Mildly perf critical (only if packet tracing is on) */
200 static inline bool unconditional(const struct ipt_ip *ip)
201 {
202         static const struct ipt_ip uncond;
203
204         return memcmp(ip, &uncond, sizeof(uncond)) == 0;
205 #undef FWINV
206 }
207
208 /* for const-correctness */
209 static inline const struct ipt_entry_target *
210 ipt_get_target_c(const struct ipt_entry *e)
211 {
212         return ipt_get_target((struct ipt_entry *)e);
213 }
214
215 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
216     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
217 static const char *const hooknames[] = {
218         [NF_INET_PRE_ROUTING]           = "PREROUTING",
219         [NF_INET_LOCAL_IN]              = "INPUT",
220         [NF_INET_FORWARD]               = "FORWARD",
221         [NF_INET_LOCAL_OUT]             = "OUTPUT",
222         [NF_INET_POST_ROUTING]          = "POSTROUTING",
223 };
224
225 enum nf_ip_trace_comments {
226         NF_IP_TRACE_COMMENT_RULE,
227         NF_IP_TRACE_COMMENT_RETURN,
228         NF_IP_TRACE_COMMENT_POLICY,
229 };
230
231 static const char *const comments[] = {
232         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
233         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
234         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
235 };
236
237 static struct nf_loginfo trace_loginfo = {
238         .type = NF_LOG_TYPE_LOG,
239         .u = {
240                 .log = {
241                         .level = 4,
242                         .logflags = NF_LOG_MASK,
243                 },
244         },
245 };
246
247 /* Mildly perf critical (only if packet tracing is on) */
248 static inline int
249 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
250                       const char *hookname, const char **chainname,
251                       const char **comment, unsigned int *rulenum)
252 {
253         const struct ipt_standard_target *t = (void *)ipt_get_target_c(s);
254
255         if (strcmp(t->target.u.kernel.target->name, IPT_ERROR_TARGET) == 0) {
256                 /* Head of user chain: ERROR target with chainname */
257                 *chainname = t->target.data;
258                 (*rulenum) = 0;
259         } else if (s == e) {
260                 (*rulenum)++;
261
262                 if (s->target_offset == sizeof(struct ipt_entry) &&
263                     strcmp(t->target.u.kernel.target->name,
264                            IPT_STANDARD_TARGET) == 0 &&
265                    t->verdict < 0 &&
266                    unconditional(&s->ip)) {
267                         /* Tail of chains: STANDARD target (return/policy) */
268                         *comment = *chainname == hookname
269                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
270                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
271                 }
272                 return 1;
273         } else
274                 (*rulenum)++;
275
276         return 0;
277 }
278
279 static void trace_packet(const struct sk_buff *skb,
280                          unsigned int hook,
281                          const struct net_device *in,
282                          const struct net_device *out,
283                          const char *tablename,
284                          const struct xt_table_info *private,
285                          const struct ipt_entry *e)
286 {
287         const void *table_base;
288         const struct ipt_entry *root;
289         const char *hookname, *chainname, *comment;
290         const struct ipt_entry *iter;
291         unsigned int rulenum = 0;
292
293         table_base = private->entries[smp_processor_id()];
294         root = get_entry(table_base, private->hook_entry[hook]);
295
296         hookname = chainname = hooknames[hook];
297         comment = comments[NF_IP_TRACE_COMMENT_RULE];
298
299         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
300                 if (get_chainname_rulenum(iter, e, hookname,
301                     &chainname, &comment, &rulenum) != 0)
302                         break;
303
304         nf_log_packet(AF_INET, hook, skb, in, out, &trace_loginfo,
305                       "TRACE: %s:%s:%s:%u ",
306                       tablename, chainname, comment, rulenum);
307 }
308 #endif
309
310 static inline __pure
311 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
312 {
313         return (void *)entry + entry->next_offset;
314 }
315
316 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
317 unsigned int
318 ipt_do_table(struct sk_buff *skb,
319              unsigned int hook,
320              const struct net_device *in,
321              const struct net_device *out,
322              struct xt_table *table)
323 {
324 #define tb_comefrom ((struct ipt_entry *)table_base)->comefrom
325
326         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
327         const struct iphdr *ip;
328         bool hotdrop = false;
329         /* Initializing verdict to NF_DROP keeps gcc happy. */
330         unsigned int verdict = NF_DROP;
331         const char *indev, *outdev;
332         const void *table_base;
333         struct ipt_entry *e, *back;
334         const struct xt_table_info *private;
335         struct xt_match_param mtpar;
336         struct xt_target_param tgpar;
337
338         /* Initialization */
339         ip = ip_hdr(skb);
340         indev = in ? in->name : nulldevname;
341         outdev = out ? out->name : nulldevname;
342         /* We handle fragments by dealing with the first fragment as
343          * if it was a normal packet.  All other fragments are treated
344          * normally, except that they will NEVER match rules that ask
345          * things we don't know, ie. tcp syn flag or ports).  If the
346          * rule is also a fragment-specific rule, non-fragments won't
347          * match it. */
348         mtpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
349         mtpar.thoff   = ip_hdrlen(skb);
350         mtpar.hotdrop = &hotdrop;
351         mtpar.in      = tgpar.in  = in;
352         mtpar.out     = tgpar.out = out;
353         mtpar.family  = tgpar.family = NFPROTO_IPV4;
354         mtpar.hooknum = tgpar.hooknum = hook;
355
356         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
357         xt_info_rdlock_bh();
358         private = table->private;
359         table_base = private->entries[smp_processor_id()];
360
361         e = get_entry(table_base, private->hook_entry[hook]);
362
363         /* For return from builtin chain */
364         back = get_entry(table_base, private->underflow[hook]);
365
366         do {
367                 const struct ipt_entry_target *t;
368                 const struct xt_entry_match *ematch;
369
370                 IP_NF_ASSERT(e);
371                 IP_NF_ASSERT(back);
372                 if (!ip_packet_match(ip, indev, outdev,
373                     &e->ip, mtpar.fragoff)) {
374  no_match:
375                         e = ipt_next_entry(e);
376                         continue;
377                 }
378
379                 xt_ematch_foreach(ematch, e)
380                         if (do_match(ematch, skb, &mtpar) != 0)
381                                 goto no_match;
382
383                 ADD_COUNTER(e->counters, ntohs(ip->tot_len), 1);
384
385                 t = ipt_get_target(e);
386                 IP_NF_ASSERT(t->u.kernel.target);
387
388 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
389     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
390                 /* The packet is traced: log it */
391                 if (unlikely(skb->nf_trace))
392                         trace_packet(skb, hook, in, out,
393                                      table->name, private, e);
394 #endif
395                 /* Standard target? */
396                 if (!t->u.kernel.target->target) {
397                         int v;
398
399                         v = ((struct ipt_standard_target *)t)->verdict;
400                         if (v < 0) {
401                                 /* Pop from stack? */
402                                 if (v != IPT_RETURN) {
403                                         verdict = (unsigned)(-v) - 1;
404                                         break;
405                                 }
406                                 e = back;
407                                 back = get_entry(table_base, back->comefrom);
408                                 continue;
409                         }
410                         if (table_base + v != ipt_next_entry(e) &&
411                             !(e->ip.flags & IPT_F_GOTO)) {
412                                 /* Save old back ptr in next entry */
413                                 struct ipt_entry *next = ipt_next_entry(e);
414                                 next->comefrom = (void *)back - table_base;
415                                 /* set back pointer to next entry */
416                                 back = next;
417                         }
418
419                         e = get_entry(table_base, v);
420                         continue;
421                 }
422
423                 /* Targets which reenter must return
424                    abs. verdicts */
425                 tgpar.target   = t->u.kernel.target;
426                 tgpar.targinfo = t->data;
427
428
429 #ifdef CONFIG_NETFILTER_DEBUG
430                 tb_comefrom = 0xeeeeeeec;
431 #endif
432                 verdict = t->u.kernel.target->target(skb, &tgpar);
433 #ifdef CONFIG_NETFILTER_DEBUG
434                 if (tb_comefrom != 0xeeeeeeec && verdict == IPT_CONTINUE) {
435                         printk("Target %s reentered!\n",
436                                t->u.kernel.target->name);
437                         verdict = NF_DROP;
438                 }
439                 tb_comefrom = 0x57acc001;
440 #endif
441                 /* Target might have changed stuff. */
442                 ip = ip_hdr(skb);
443                 if (verdict == IPT_CONTINUE)
444                         e = ipt_next_entry(e);
445                 else
446                         /* Verdict */
447                         break;
448         } while (!hotdrop);
449         xt_info_rdunlock_bh();
450
451 #ifdef DEBUG_ALLOW_ALL
452         return NF_ACCEPT;
453 #else
454         if (hotdrop)
455                 return NF_DROP;
456         else return verdict;
457 #endif
458
459 #undef tb_comefrom
460 }
461
462 /* Figures out from what hook each rule can be called: returns 0 if
463    there are loops.  Puts hook bitmask in comefrom. */
464 static int
465 mark_source_chains(const struct xt_table_info *newinfo,
466                    unsigned int valid_hooks, void *entry0)
467 {
468         unsigned int hook;
469
470         /* No recursion; use packet counter to save back ptrs (reset
471            to 0 as we leave), and comefrom to save source hook bitmask */
472         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
473                 unsigned int pos = newinfo->hook_entry[hook];
474                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
475
476                 if (!(valid_hooks & (1 << hook)))
477                         continue;
478
479                 /* Set initial back pointer. */
480                 e->counters.pcnt = pos;
481
482                 for (;;) {
483                         const struct ipt_standard_target *t
484                                 = (void *)ipt_get_target_c(e);
485                         int visited = e->comefrom & (1 << hook);
486
487                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
488                                 printk("iptables: loop hook %u pos %u %08X.\n",
489                                        hook, pos, e->comefrom);
490                                 return 0;
491                         }
492                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
493
494                         /* Unconditional return/END. */
495                         if ((e->target_offset == sizeof(struct ipt_entry) &&
496                              (strcmp(t->target.u.user.name,
497                                      IPT_STANDARD_TARGET) == 0) &&
498                              t->verdict < 0 && unconditional(&e->ip)) ||
499                             visited) {
500                                 unsigned int oldpos, size;
501
502                                 if ((strcmp(t->target.u.user.name,
503                                             IPT_STANDARD_TARGET) == 0) &&
504                                     t->verdict < -NF_MAX_VERDICT - 1) {
505                                         duprintf("mark_source_chains: bad "
506                                                 "negative verdict (%i)\n",
507                                                                 t->verdict);
508                                         return 0;
509                                 }
510
511                                 /* Return: backtrack through the last
512                                    big jump. */
513                                 do {
514                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
515 #ifdef DEBUG_IP_FIREWALL_USER
516                                         if (e->comefrom
517                                             & (1 << NF_INET_NUMHOOKS)) {
518                                                 duprintf("Back unset "
519                                                          "on hook %u "
520                                                          "rule %u\n",
521                                                          hook, pos);
522                                         }
523 #endif
524                                         oldpos = pos;
525                                         pos = e->counters.pcnt;
526                                         e->counters.pcnt = 0;
527
528                                         /* We're at the start. */
529                                         if (pos == oldpos)
530                                                 goto next;
531
532                                         e = (struct ipt_entry *)
533                                                 (entry0 + pos);
534                                 } while (oldpos == pos + e->next_offset);
535
536                                 /* Move along one */
537                                 size = e->next_offset;
538                                 e = (struct ipt_entry *)
539                                         (entry0 + pos + size);
540                                 e->counters.pcnt = pos;
541                                 pos += size;
542                         } else {
543                                 int newpos = t->verdict;
544
545                                 if (strcmp(t->target.u.user.name,
546                                            IPT_STANDARD_TARGET) == 0 &&
547                                     newpos >= 0) {
548                                         if (newpos > newinfo->size -
549                                                 sizeof(struct ipt_entry)) {
550                                                 duprintf("mark_source_chains: "
551                                                         "bad verdict (%i)\n",
552                                                                 newpos);
553                                                 return 0;
554                                         }
555                                         /* This a jump; chase it. */
556                                         duprintf("Jump rule %u -> %u\n",
557                                                  pos, newpos);
558                                 } else {
559                                         /* ... this is a fallthru */
560                                         newpos = pos + e->next_offset;
561                                 }
562                                 e = (struct ipt_entry *)
563                                         (entry0 + newpos);
564                                 e->counters.pcnt = pos;
565                                 pos = newpos;
566                         }
567                 }
568                 next:
569                 duprintf("Finished chain %u\n", hook);
570         }
571         return 1;
572 }
573
574 static void cleanup_match(struct ipt_entry_match *m, struct net *net)
575 {
576         struct xt_mtdtor_param par;
577
578         par.net       = net;
579         par.match     = m->u.kernel.match;
580         par.matchinfo = m->data;
581         par.family    = NFPROTO_IPV4;
582         if (par.match->destroy != NULL)
583                 par.match->destroy(&par);
584         module_put(par.match->me);
585 }
586
587 static int
588 check_entry(const struct ipt_entry *e, const char *name)
589 {
590         const struct ipt_entry_target *t;
591
592         if (!ip_checkentry(&e->ip)) {
593                 duprintf("ip check failed %p %s.\n", e, name);
594                 return -EINVAL;
595         }
596
597         if (e->target_offset + sizeof(struct ipt_entry_target) >
598             e->next_offset)
599                 return -EINVAL;
600
601         t = ipt_get_target_c(e);
602         if (e->target_offset + t->u.target_size > e->next_offset)
603                 return -EINVAL;
604
605         return 0;
606 }
607
608 static int
609 check_match(struct ipt_entry_match *m, struct xt_mtchk_param *par)
610 {
611         const struct ipt_ip *ip = par->entryinfo;
612         int ret;
613
614         par->match     = m->u.kernel.match;
615         par->matchinfo = m->data;
616
617         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
618               ip->proto, ip->invflags & IPT_INV_PROTO);
619         if (ret < 0) {
620                 duprintf("check failed for `%s'.\n", par.match->name);
621                 return ret;
622         }
623         return 0;
624 }
625
626 static int
627 find_check_match(struct ipt_entry_match *m, struct xt_mtchk_param *par)
628 {
629         struct xt_match *match;
630         int ret;
631
632         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
633                                       m->u.user.revision);
634         if (IS_ERR(match)) {
635                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
636                 return PTR_ERR(match);
637         }
638         m->u.kernel.match = match;
639
640         ret = check_match(m, par);
641         if (ret)
642                 goto err;
643
644         return 0;
645 err:
646         module_put(m->u.kernel.match->me);
647         return ret;
648 }
649
650 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
651 {
652         struct ipt_entry_target *t = ipt_get_target(e);
653         struct xt_tgchk_param par = {
654                 .net       = net,
655                 .table     = name,
656                 .entryinfo = e,
657                 .target    = t->u.kernel.target,
658                 .targinfo  = t->data,
659                 .hook_mask = e->comefrom,
660                 .family    = NFPROTO_IPV4,
661         };
662         int ret;
663
664         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
665               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
666         if (ret < 0) {
667                 duprintf("check failed for `%s'.\n",
668                          t->u.kernel.target->name);
669                 return ret;
670         }
671         return 0;
672 }
673
674 static int
675 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
676                  unsigned int size)
677 {
678         struct ipt_entry_target *t;
679         struct xt_target *target;
680         int ret;
681         unsigned int j;
682         struct xt_mtchk_param mtpar;
683         struct xt_entry_match *ematch;
684
685         ret = check_entry(e, name);
686         if (ret)
687                 return ret;
688
689         j = 0;
690         mtpar.net       = net;
691         mtpar.table     = name;
692         mtpar.entryinfo = &e->ip;
693         mtpar.hook_mask = e->comefrom;
694         mtpar.family    = NFPROTO_IPV4;
695         xt_ematch_foreach(ematch, e) {
696                 ret = find_check_match(ematch, &mtpar);
697                 if (ret != 0)
698                         goto cleanup_matches;
699                 ++j;
700         }
701
702         t = ipt_get_target(e);
703         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
704                                         t->u.user.revision);
705         if (IS_ERR(target)) {
706                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
707                 ret = PTR_ERR(target);
708                 goto cleanup_matches;
709         }
710         t->u.kernel.target = target;
711
712         ret = check_target(e, net, name);
713         if (ret)
714                 goto err;
715         return 0;
716  err:
717         module_put(t->u.kernel.target->me);
718  cleanup_matches:
719         xt_ematch_foreach(ematch, e) {
720                 if (j-- == 0)
721                         break;
722                 cleanup_match(ematch, net);
723         }
724         return ret;
725 }
726
727 static bool check_underflow(const struct ipt_entry *e)
728 {
729         const struct ipt_entry_target *t;
730         unsigned int verdict;
731
732         if (!unconditional(&e->ip))
733                 return false;
734         t = ipt_get_target_c(e);
735         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
736                 return false;
737         verdict = ((struct ipt_standard_target *)t)->verdict;
738         verdict = -verdict - 1;
739         return verdict == NF_DROP || verdict == NF_ACCEPT;
740 }
741
742 static int
743 check_entry_size_and_hooks(struct ipt_entry *e,
744                            struct xt_table_info *newinfo,
745                            const unsigned char *base,
746                            const unsigned char *limit,
747                            const unsigned int *hook_entries,
748                            const unsigned int *underflows,
749                            unsigned int valid_hooks)
750 {
751         unsigned int h;
752
753         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
754             (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
755                 duprintf("Bad offset %p\n", e);
756                 return -EINVAL;
757         }
758
759         if (e->next_offset
760             < sizeof(struct ipt_entry) + sizeof(struct ipt_entry_target)) {
761                 duprintf("checking: element %p size %u\n",
762                          e, e->next_offset);
763                 return -EINVAL;
764         }
765
766         /* Check hooks & underflows */
767         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
768                 if (!(valid_hooks & (1 << h)))
769                         continue;
770                 if ((unsigned char *)e - base == hook_entries[h])
771                         newinfo->hook_entry[h] = hook_entries[h];
772                 if ((unsigned char *)e - base == underflows[h]) {
773                         if (!check_underflow(e)) {
774                                 pr_err("Underflows must be unconditional and "
775                                        "use the STANDARD target with "
776                                        "ACCEPT/DROP\n");
777                                 return -EINVAL;
778                         }
779                         newinfo->underflow[h] = underflows[h];
780                 }
781         }
782
783         /* Clear counters and comefrom */
784         e->counters = ((struct xt_counters) { 0, 0 });
785         e->comefrom = 0;
786         return 0;
787 }
788
789 static void
790 cleanup_entry(struct ipt_entry *e, struct net *net)
791 {
792         struct xt_tgdtor_param par;
793         struct ipt_entry_target *t;
794         struct xt_entry_match *ematch;
795
796         /* Cleanup all matches */
797         xt_ematch_foreach(ematch, e)
798                 cleanup_match(ematch, net);
799         t = ipt_get_target(e);
800
801         par.net      = net;
802         par.target   = t->u.kernel.target;
803         par.targinfo = t->data;
804         par.family   = NFPROTO_IPV4;
805         if (par.target->destroy != NULL)
806                 par.target->destroy(&par);
807         module_put(par.target->me);
808 }
809
810 /* Checks and translates the user-supplied table segment (held in
811    newinfo) */
812 static int
813 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
814                 const struct ipt_replace *repl)
815 {
816         struct ipt_entry *iter;
817         unsigned int i;
818         int ret = 0;
819
820         newinfo->size = repl->size;
821         newinfo->number = repl->num_entries;
822
823         /* Init all hooks to impossible value. */
824         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
825                 newinfo->hook_entry[i] = 0xFFFFFFFF;
826                 newinfo->underflow[i] = 0xFFFFFFFF;
827         }
828
829         duprintf("translate_table: size %u\n", newinfo->size);
830         i = 0;
831         /* Walk through entries, checking offsets. */
832         xt_entry_foreach(iter, entry0, newinfo->size) {
833                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
834                                                  entry0 + repl->size,
835                                                  repl->hook_entry,
836                                                  repl->underflow,
837                                                  repl->valid_hooks);
838                 if (ret != 0)
839                         return ret;
840                 ++i;
841         }
842
843         if (i != repl->num_entries) {
844                 duprintf("translate_table: %u not %u entries\n",
845                          i, repl->num_entries);
846                 return -EINVAL;
847         }
848
849         /* Check hooks all assigned */
850         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
851                 /* Only hooks which are valid */
852                 if (!(repl->valid_hooks & (1 << i)))
853                         continue;
854                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
855                         duprintf("Invalid hook entry %u %u\n",
856                                  i, repl->hook_entry[i]);
857                         return -EINVAL;
858                 }
859                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
860                         duprintf("Invalid underflow %u %u\n",
861                                  i, repl->underflow[i]);
862                         return -EINVAL;
863                 }
864         }
865
866         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
867                 return -ELOOP;
868
869         /* Finally, each sanity check must pass */
870         i = 0;
871         xt_entry_foreach(iter, entry0, newinfo->size) {
872                 ret = find_check_entry(iter, net, repl->name, repl->size);
873                 if (ret != 0)
874                         break;
875                 ++i;
876         }
877
878         if (ret != 0) {
879                 xt_entry_foreach(iter, entry0, newinfo->size) {
880                         if (i-- == 0)
881                                 break;
882                         cleanup_entry(iter, net);
883                 }
884                 return ret;
885         }
886
887         /* And one copy for every other CPU */
888         for_each_possible_cpu(i) {
889                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
890                         memcpy(newinfo->entries[i], entry0, newinfo->size);
891         }
892
893         return ret;
894 }
895
896 static void
897 get_counters(const struct xt_table_info *t,
898              struct xt_counters counters[])
899 {
900         struct ipt_entry *iter;
901         unsigned int cpu;
902         unsigned int i;
903         unsigned int curcpu;
904
905         /* Instead of clearing (by a previous call to memset())
906          * the counters and using adds, we set the counters
907          * with data used by 'current' CPU.
908          *
909          * Bottom half has to be disabled to prevent deadlock
910          * if new softirq were to run and call ipt_do_table
911          */
912         local_bh_disable();
913         curcpu = smp_processor_id();
914
915         i = 0;
916         xt_entry_foreach(iter, t->entries[curcpu], t->size) {
917                 SET_COUNTER(counters[i], iter->counters.bcnt,
918                             iter->counters.pcnt);
919                 ++i;
920         }
921
922         for_each_possible_cpu(cpu) {
923                 if (cpu == curcpu)
924                         continue;
925                 i = 0;
926                 xt_info_wrlock(cpu);
927                 xt_entry_foreach(iter, t->entries[cpu], t->size) {
928                         ADD_COUNTER(counters[i], iter->counters.bcnt,
929                                     iter->counters.pcnt);
930                         ++i; /* macro does multi eval of i */
931                 }
932                 xt_info_wrunlock(cpu);
933         }
934         local_bh_enable();
935 }
936
937 static struct xt_counters *alloc_counters(const struct xt_table *table)
938 {
939         unsigned int countersize;
940         struct xt_counters *counters;
941         const struct xt_table_info *private = table->private;
942
943         /* We need atomic snapshot of counters: rest doesn't change
944            (other than comefrom, which userspace doesn't care
945            about). */
946         countersize = sizeof(struct xt_counters) * private->number;
947         counters = vmalloc_node(countersize, numa_node_id());
948
949         if (counters == NULL)
950                 return ERR_PTR(-ENOMEM);
951
952         get_counters(private, counters);
953
954         return counters;
955 }
956
957 static int
958 copy_entries_to_user(unsigned int total_size,
959                      const struct xt_table *table,
960                      void __user *userptr)
961 {
962         unsigned int off, num;
963         const struct ipt_entry *e;
964         struct xt_counters *counters;
965         const struct xt_table_info *private = table->private;
966         int ret = 0;
967         const void *loc_cpu_entry;
968
969         counters = alloc_counters(table);
970         if (IS_ERR(counters))
971                 return PTR_ERR(counters);
972
973         /* choose the copy that is on our node/cpu, ...
974          * This choice is lazy (because current thread is
975          * allowed to migrate to another cpu)
976          */
977         loc_cpu_entry = private->entries[raw_smp_processor_id()];
978         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
979                 ret = -EFAULT;
980                 goto free_counters;
981         }
982
983         /* FIXME: use iterator macros --RR */
984         /* ... then go back and fix counters and names */
985         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
986                 unsigned int i;
987                 const struct ipt_entry_match *m;
988                 const struct ipt_entry_target *t;
989
990                 e = (struct ipt_entry *)(loc_cpu_entry + off);
991                 if (copy_to_user(userptr + off
992                                  + offsetof(struct ipt_entry, counters),
993                                  &counters[num],
994                                  sizeof(counters[num])) != 0) {
995                         ret = -EFAULT;
996                         goto free_counters;
997                 }
998
999                 for (i = sizeof(struct ipt_entry);
1000                      i < e->target_offset;
1001                      i += m->u.match_size) {
1002                         m = (void *)e + i;
1003
1004                         if (copy_to_user(userptr + off + i
1005                                          + offsetof(struct ipt_entry_match,
1006                                                     u.user.name),
1007                                          m->u.kernel.match->name,
1008                                          strlen(m->u.kernel.match->name)+1)
1009                             != 0) {
1010                                 ret = -EFAULT;
1011                                 goto free_counters;
1012                         }
1013                 }
1014
1015                 t = ipt_get_target_c(e);
1016                 if (copy_to_user(userptr + off + e->target_offset
1017                                  + offsetof(struct ipt_entry_target,
1018                                             u.user.name),
1019                                  t->u.kernel.target->name,
1020                                  strlen(t->u.kernel.target->name)+1) != 0) {
1021                         ret = -EFAULT;
1022                         goto free_counters;
1023                 }
1024         }
1025
1026  free_counters:
1027         vfree(counters);
1028         return ret;
1029 }
1030
1031 #ifdef CONFIG_COMPAT
1032 static void compat_standard_from_user(void *dst, const void *src)
1033 {
1034         int v = *(compat_int_t *)src;
1035
1036         if (v > 0)
1037                 v += xt_compat_calc_jump(AF_INET, v);
1038         memcpy(dst, &v, sizeof(v));
1039 }
1040
1041 static int compat_standard_to_user(void __user *dst, const void *src)
1042 {
1043         compat_int_t cv = *(int *)src;
1044
1045         if (cv > 0)
1046                 cv -= xt_compat_calc_jump(AF_INET, cv);
1047         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1048 }
1049
1050 static int compat_calc_entry(const struct ipt_entry *e,
1051                              const struct xt_table_info *info,
1052                              const void *base, struct xt_table_info *newinfo)
1053 {
1054         const struct xt_entry_match *ematch;
1055         const struct ipt_entry_target *t;
1056         unsigned int entry_offset;
1057         int off, i, ret;
1058
1059         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1060         entry_offset = (void *)e - base;
1061         xt_ematch_foreach(ematch, e)
1062                 off += xt_compat_match_offset(ematch->u.kernel.match);
1063         t = ipt_get_target_c(e);
1064         off += xt_compat_target_offset(t->u.kernel.target);
1065         newinfo->size -= off;
1066         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1067         if (ret)
1068                 return ret;
1069
1070         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1071                 if (info->hook_entry[i] &&
1072                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1073                         newinfo->hook_entry[i] -= off;
1074                 if (info->underflow[i] &&
1075                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1076                         newinfo->underflow[i] -= off;
1077         }
1078         return 0;
1079 }
1080
1081 static int compat_table_info(const struct xt_table_info *info,
1082                              struct xt_table_info *newinfo)
1083 {
1084         struct ipt_entry *iter;
1085         void *loc_cpu_entry;
1086         int ret;
1087
1088         if (!newinfo || !info)
1089                 return -EINVAL;
1090
1091         /* we dont care about newinfo->entries[] */
1092         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1093         newinfo->initial_entries = 0;
1094         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1095         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1096                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1097                 if (ret != 0)
1098                         return ret;
1099         }
1100         return 0;
1101 }
1102 #endif
1103
1104 static int get_info(struct net *net, void __user *user,
1105                     const int *len, int compat)
1106 {
1107         char name[IPT_TABLE_MAXNAMELEN];
1108         struct xt_table *t;
1109         int ret;
1110
1111         if (*len != sizeof(struct ipt_getinfo)) {
1112                 duprintf("length %u != %zu\n", *len,
1113                          sizeof(struct ipt_getinfo));
1114                 return -EINVAL;
1115         }
1116
1117         if (copy_from_user(name, user, sizeof(name)) != 0)
1118                 return -EFAULT;
1119
1120         name[IPT_TABLE_MAXNAMELEN-1] = '\0';
1121 #ifdef CONFIG_COMPAT
1122         if (compat)
1123                 xt_compat_lock(AF_INET);
1124 #endif
1125         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1126                                     "iptable_%s", name);
1127         if (t && !IS_ERR(t)) {
1128                 struct ipt_getinfo info;
1129                 const struct xt_table_info *private = t->private;
1130 #ifdef CONFIG_COMPAT
1131                 struct xt_table_info tmp;
1132
1133                 if (compat) {
1134                         ret = compat_table_info(private, &tmp);
1135                         xt_compat_flush_offsets(AF_INET);
1136                         private = &tmp;
1137                 }
1138 #endif
1139                 info.valid_hooks = t->valid_hooks;
1140                 memcpy(info.hook_entry, private->hook_entry,
1141                        sizeof(info.hook_entry));
1142                 memcpy(info.underflow, private->underflow,
1143                        sizeof(info.underflow));
1144                 info.num_entries = private->number;
1145                 info.size = private->size;
1146                 strcpy(info.name, name);
1147
1148                 if (copy_to_user(user, &info, *len) != 0)
1149                         ret = -EFAULT;
1150                 else
1151                         ret = 0;
1152
1153                 xt_table_unlock(t);
1154                 module_put(t->me);
1155         } else
1156                 ret = t ? PTR_ERR(t) : -ENOENT;
1157 #ifdef CONFIG_COMPAT
1158         if (compat)
1159                 xt_compat_unlock(AF_INET);
1160 #endif
1161         return ret;
1162 }
1163
1164 static int
1165 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1166             const int *len)
1167 {
1168         int ret;
1169         struct ipt_get_entries get;
1170         struct xt_table *t;
1171
1172         if (*len < sizeof(get)) {
1173                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1174                 return -EINVAL;
1175         }
1176         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1177                 return -EFAULT;
1178         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1179                 duprintf("get_entries: %u != %zu\n",
1180                          *len, sizeof(get) + get.size);
1181                 return -EINVAL;
1182         }
1183
1184         t = xt_find_table_lock(net, AF_INET, get.name);
1185         if (t && !IS_ERR(t)) {
1186                 const struct xt_table_info *private = t->private;
1187                 duprintf("t->private->number = %u\n", private->number);
1188                 if (get.size == private->size)
1189                         ret = copy_entries_to_user(private->size,
1190                                                    t, uptr->entrytable);
1191                 else {
1192                         duprintf("get_entries: I've got %u not %u!\n",
1193                                  private->size, get.size);
1194                         ret = -EAGAIN;
1195                 }
1196                 module_put(t->me);
1197                 xt_table_unlock(t);
1198         } else
1199                 ret = t ? PTR_ERR(t) : -ENOENT;
1200
1201         return ret;
1202 }
1203
1204 static int
1205 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1206              struct xt_table_info *newinfo, unsigned int num_counters,
1207              void __user *counters_ptr)
1208 {
1209         int ret;
1210         struct xt_table *t;
1211         struct xt_table_info *oldinfo;
1212         struct xt_counters *counters;
1213         void *loc_cpu_old_entry;
1214         struct ipt_entry *iter;
1215
1216         ret = 0;
1217         counters = vmalloc(num_counters * sizeof(struct xt_counters));
1218         if (!counters) {
1219                 ret = -ENOMEM;
1220                 goto out;
1221         }
1222
1223         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1224                                     "iptable_%s", name);
1225         if (!t || IS_ERR(t)) {
1226                 ret = t ? PTR_ERR(t) : -ENOENT;
1227                 goto free_newinfo_counters_untrans;
1228         }
1229
1230         /* You lied! */
1231         if (valid_hooks != t->valid_hooks) {
1232                 duprintf("Valid hook crap: %08X vs %08X\n",
1233                          valid_hooks, t->valid_hooks);
1234                 ret = -EINVAL;
1235                 goto put_module;
1236         }
1237
1238         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1239         if (!oldinfo)
1240                 goto put_module;
1241
1242         /* Update module usage count based on number of rules */
1243         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1244                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1245         if ((oldinfo->number > oldinfo->initial_entries) ||
1246             (newinfo->number <= oldinfo->initial_entries))
1247                 module_put(t->me);
1248         if ((oldinfo->number > oldinfo->initial_entries) &&
1249             (newinfo->number <= oldinfo->initial_entries))
1250                 module_put(t->me);
1251
1252         /* Get the old counters, and synchronize with replace */
1253         get_counters(oldinfo, counters);
1254
1255         /* Decrease module usage counts and free resource */
1256         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1257         xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1258                 cleanup_entry(iter, net);
1259
1260         xt_free_table_info(oldinfo);
1261         if (copy_to_user(counters_ptr, counters,
1262                          sizeof(struct xt_counters) * num_counters) != 0)
1263                 ret = -EFAULT;
1264         vfree(counters);
1265         xt_table_unlock(t);
1266         return ret;
1267
1268  put_module:
1269         module_put(t->me);
1270         xt_table_unlock(t);
1271  free_newinfo_counters_untrans:
1272         vfree(counters);
1273  out:
1274         return ret;
1275 }
1276
1277 static int
1278 do_replace(struct net *net, const void __user *user, unsigned int len)
1279 {
1280         int ret;
1281         struct ipt_replace tmp;
1282         struct xt_table_info *newinfo;
1283         void *loc_cpu_entry;
1284         struct ipt_entry *iter;
1285
1286         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1287                 return -EFAULT;
1288
1289         /* overflow check */
1290         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1291                 return -ENOMEM;
1292
1293         newinfo = xt_alloc_table_info(tmp.size);
1294         if (!newinfo)
1295                 return -ENOMEM;
1296
1297         /* choose the copy that is on our node/cpu */
1298         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1299         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1300                            tmp.size) != 0) {
1301                 ret = -EFAULT;
1302                 goto free_newinfo;
1303         }
1304
1305         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1306         if (ret != 0)
1307                 goto free_newinfo;
1308
1309         duprintf("Translated table\n");
1310
1311         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1312                            tmp.num_counters, tmp.counters);
1313         if (ret)
1314                 goto free_newinfo_untrans;
1315         return 0;
1316
1317  free_newinfo_untrans:
1318         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1319                 cleanup_entry(iter, net);
1320  free_newinfo:
1321         xt_free_table_info(newinfo);
1322         return ret;
1323 }
1324
1325 static int
1326 do_add_counters(struct net *net, const void __user *user,
1327                 unsigned int len, int compat)
1328 {
1329         unsigned int i, curcpu;
1330         struct xt_counters_info tmp;
1331         struct xt_counters *paddc;
1332         unsigned int num_counters;
1333         const char *name;
1334         int size;
1335         void *ptmp;
1336         struct xt_table *t;
1337         const struct xt_table_info *private;
1338         int ret = 0;
1339         void *loc_cpu_entry;
1340         struct ipt_entry *iter;
1341 #ifdef CONFIG_COMPAT
1342         struct compat_xt_counters_info compat_tmp;
1343
1344         if (compat) {
1345                 ptmp = &compat_tmp;
1346                 size = sizeof(struct compat_xt_counters_info);
1347         } else
1348 #endif
1349         {
1350                 ptmp = &tmp;
1351                 size = sizeof(struct xt_counters_info);
1352         }
1353
1354         if (copy_from_user(ptmp, user, size) != 0)
1355                 return -EFAULT;
1356
1357 #ifdef CONFIG_COMPAT
1358         if (compat) {
1359                 num_counters = compat_tmp.num_counters;
1360                 name = compat_tmp.name;
1361         } else
1362 #endif
1363         {
1364                 num_counters = tmp.num_counters;
1365                 name = tmp.name;
1366         }
1367
1368         if (len != size + num_counters * sizeof(struct xt_counters))
1369                 return -EINVAL;
1370
1371         paddc = vmalloc_node(len - size, numa_node_id());
1372         if (!paddc)
1373                 return -ENOMEM;
1374
1375         if (copy_from_user(paddc, user + size, len - size) != 0) {
1376                 ret = -EFAULT;
1377                 goto free;
1378         }
1379
1380         t = xt_find_table_lock(net, AF_INET, name);
1381         if (!t || IS_ERR(t)) {
1382                 ret = t ? PTR_ERR(t) : -ENOENT;
1383                 goto free;
1384         }
1385
1386         local_bh_disable();
1387         private = t->private;
1388         if (private->number != num_counters) {
1389                 ret = -EINVAL;
1390                 goto unlock_up_free;
1391         }
1392
1393         i = 0;
1394         /* Choose the copy that is on our node */
1395         curcpu = smp_processor_id();
1396         loc_cpu_entry = private->entries[curcpu];
1397         xt_info_wrlock(curcpu);
1398         xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1399                 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1400                 ++i;
1401         }
1402         xt_info_wrunlock(curcpu);
1403  unlock_up_free:
1404         local_bh_enable();
1405         xt_table_unlock(t);
1406         module_put(t->me);
1407  free:
1408         vfree(paddc);
1409
1410         return ret;
1411 }
1412
1413 #ifdef CONFIG_COMPAT
1414 struct compat_ipt_replace {
1415         char                    name[IPT_TABLE_MAXNAMELEN];
1416         u32                     valid_hooks;
1417         u32                     num_entries;
1418         u32                     size;
1419         u32                     hook_entry[NF_INET_NUMHOOKS];
1420         u32                     underflow[NF_INET_NUMHOOKS];
1421         u32                     num_counters;
1422         compat_uptr_t           counters;       /* struct ipt_counters * */
1423         struct compat_ipt_entry entries[0];
1424 };
1425
1426 static int
1427 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1428                           unsigned int *size, struct xt_counters *counters,
1429                           unsigned int i)
1430 {
1431         struct ipt_entry_target *t;
1432         struct compat_ipt_entry __user *ce;
1433         u_int16_t target_offset, next_offset;
1434         compat_uint_t origsize;
1435         const struct xt_entry_match *ematch;
1436         int ret = 0;
1437
1438         origsize = *size;
1439         ce = (struct compat_ipt_entry __user *)*dstptr;
1440         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1441             copy_to_user(&ce->counters, &counters[i],
1442             sizeof(counters[i])) != 0)
1443                 return -EFAULT;
1444
1445         *dstptr += sizeof(struct compat_ipt_entry);
1446         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1447
1448         xt_ematch_foreach(ematch, e) {
1449                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1450                 if (ret != 0)
1451                         return ret;
1452         }
1453         target_offset = e->target_offset - (origsize - *size);
1454         t = ipt_get_target(e);
1455         ret = xt_compat_target_to_user(t, dstptr, size);
1456         if (ret)
1457                 return ret;
1458         next_offset = e->next_offset - (origsize - *size);
1459         if (put_user(target_offset, &ce->target_offset) != 0 ||
1460             put_user(next_offset, &ce->next_offset) != 0)
1461                 return -EFAULT;
1462         return 0;
1463 }
1464
1465 static int
1466 compat_find_calc_match(struct ipt_entry_match *m,
1467                        const char *name,
1468                        const struct ipt_ip *ip,
1469                        unsigned int hookmask,
1470                        int *size)
1471 {
1472         struct xt_match *match;
1473
1474         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1475                                       m->u.user.revision);
1476         if (IS_ERR(match)) {
1477                 duprintf("compat_check_calc_match: `%s' not found\n",
1478                          m->u.user.name);
1479                 return PTR_ERR(match);
1480         }
1481         m->u.kernel.match = match;
1482         *size += xt_compat_match_offset(match);
1483         return 0;
1484 }
1485
1486 static void compat_release_entry(struct compat_ipt_entry *e)
1487 {
1488         struct ipt_entry_target *t;
1489         struct xt_entry_match *ematch;
1490
1491         /* Cleanup all matches */
1492         xt_ematch_foreach(ematch, e)
1493                 module_put(ematch->u.kernel.match->me);
1494         t = compat_ipt_get_target(e);
1495         module_put(t->u.kernel.target->me);
1496 }
1497
1498 static int
1499 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1500                                   struct xt_table_info *newinfo,
1501                                   unsigned int *size,
1502                                   const unsigned char *base,
1503                                   const unsigned char *limit,
1504                                   const unsigned int *hook_entries,
1505                                   const unsigned int *underflows,
1506                                   const char *name)
1507 {
1508         struct xt_entry_match *ematch;
1509         struct ipt_entry_target *t;
1510         struct xt_target *target;
1511         unsigned int entry_offset;
1512         unsigned int j;
1513         int ret, off, h;
1514
1515         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1516         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1517             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1518                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1519                 return -EINVAL;
1520         }
1521
1522         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1523                              sizeof(struct compat_xt_entry_target)) {
1524                 duprintf("checking: element %p size %u\n",
1525                          e, e->next_offset);
1526                 return -EINVAL;
1527         }
1528
1529         /* For purposes of check_entry casting the compat entry is fine */
1530         ret = check_entry((struct ipt_entry *)e, name);
1531         if (ret)
1532                 return ret;
1533
1534         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1535         entry_offset = (void *)e - (void *)base;
1536         j = 0;
1537         xt_ematch_foreach(ematch, e) {
1538                 ret = compat_find_calc_match(ematch, name,
1539                                              &e->ip, e->comefrom, &off);
1540                 if (ret != 0)
1541                         goto release_matches;
1542                 ++j;
1543         }
1544
1545         t = compat_ipt_get_target(e);
1546         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1547                                         t->u.user.revision);
1548         if (IS_ERR(target)) {
1549                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1550                          t->u.user.name);
1551                 ret = PTR_ERR(target);
1552                 goto release_matches;
1553         }
1554         t->u.kernel.target = target;
1555
1556         off += xt_compat_target_offset(target);
1557         *size += off;
1558         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1559         if (ret)
1560                 goto out;
1561
1562         /* Check hooks & underflows */
1563         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1564                 if ((unsigned char *)e - base == hook_entries[h])
1565                         newinfo->hook_entry[h] = hook_entries[h];
1566                 if ((unsigned char *)e - base == underflows[h])
1567                         newinfo->underflow[h] = underflows[h];
1568         }
1569
1570         /* Clear counters and comefrom */
1571         memset(&e->counters, 0, sizeof(e->counters));
1572         e->comefrom = 0;
1573         return 0;
1574
1575 out:
1576         module_put(t->u.kernel.target->me);
1577 release_matches:
1578         xt_ematch_foreach(ematch, e) {
1579                 if (j-- == 0)
1580                         break;
1581                 module_put(ematch->u.kernel.match->me);
1582         }
1583         return ret;
1584 }
1585
1586 static int
1587 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1588                             unsigned int *size, const char *name,
1589                             struct xt_table_info *newinfo, unsigned char *base)
1590 {
1591         struct ipt_entry_target *t;
1592         struct xt_target *target;
1593         struct ipt_entry *de;
1594         unsigned int origsize;
1595         int ret, h;
1596         struct xt_entry_match *ematch;
1597
1598         ret = 0;
1599         origsize = *size;
1600         de = (struct ipt_entry *)*dstptr;
1601         memcpy(de, e, sizeof(struct ipt_entry));
1602         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1603
1604         *dstptr += sizeof(struct ipt_entry);
1605         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1606
1607         xt_ematch_foreach(ematch, e) {
1608                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1609                 if (ret != 0)
1610                         return ret;
1611         }
1612         de->target_offset = e->target_offset - (origsize - *size);
1613         t = compat_ipt_get_target(e);
1614         target = t->u.kernel.target;
1615         xt_compat_target_from_user(t, dstptr, size);
1616
1617         de->next_offset = e->next_offset - (origsize - *size);
1618         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1619                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1620                         newinfo->hook_entry[h] -= origsize - *size;
1621                 if ((unsigned char *)de - base < newinfo->underflow[h])
1622                         newinfo->underflow[h] -= origsize - *size;
1623         }
1624         return ret;
1625 }
1626
1627 static int
1628 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1629 {
1630         struct xt_entry_match *ematch;
1631         struct xt_mtchk_param mtpar;
1632         unsigned int j;
1633         int ret = 0;
1634
1635         j = 0;
1636         mtpar.net       = net;
1637         mtpar.table     = name;
1638         mtpar.entryinfo = &e->ip;
1639         mtpar.hook_mask = e->comefrom;
1640         mtpar.family    = NFPROTO_IPV4;
1641         xt_ematch_foreach(ematch, e) {
1642                 ret = check_match(ematch, &mtpar);
1643                 if (ret != 0)
1644                         goto cleanup_matches;
1645                 ++j;
1646         }
1647
1648         ret = check_target(e, net, name);
1649         if (ret)
1650                 goto cleanup_matches;
1651         return 0;
1652
1653  cleanup_matches:
1654         xt_ematch_foreach(ematch, e) {
1655                 if (j-- == 0)
1656                         break;
1657                 cleanup_match(ematch, net);
1658         }
1659         return ret;
1660 }
1661
1662 static int
1663 translate_compat_table(struct net *net,
1664                        const char *name,
1665                        unsigned int valid_hooks,
1666                        struct xt_table_info **pinfo,
1667                        void **pentry0,
1668                        unsigned int total_size,
1669                        unsigned int number,
1670                        unsigned int *hook_entries,
1671                        unsigned int *underflows)
1672 {
1673         unsigned int i, j;
1674         struct xt_table_info *newinfo, *info;
1675         void *pos, *entry0, *entry1;
1676         struct compat_ipt_entry *iter0;
1677         struct ipt_entry *iter1;
1678         unsigned int size;
1679         int ret;
1680
1681         info = *pinfo;
1682         entry0 = *pentry0;
1683         size = total_size;
1684         info->number = number;
1685
1686         /* Init all hooks to impossible value. */
1687         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1688                 info->hook_entry[i] = 0xFFFFFFFF;
1689                 info->underflow[i] = 0xFFFFFFFF;
1690         }
1691
1692         duprintf("translate_compat_table: size %u\n", info->size);
1693         j = 0;
1694         xt_compat_lock(AF_INET);
1695         /* Walk through entries, checking offsets. */
1696         xt_entry_foreach(iter0, entry0, total_size) {
1697                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1698                                                         entry0,
1699                                                         entry0 + total_size,
1700                                                         hook_entries,
1701                                                         underflows,
1702                                                         name);
1703                 if (ret != 0)
1704                         goto out_unlock;
1705                 ++j;
1706         }
1707
1708         ret = -EINVAL;
1709         if (j != number) {
1710                 duprintf("translate_compat_table: %u not %u entries\n",
1711                          j, number);
1712                 goto out_unlock;
1713         }
1714
1715         /* Check hooks all assigned */
1716         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1717                 /* Only hooks which are valid */
1718                 if (!(valid_hooks & (1 << i)))
1719                         continue;
1720                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1721                         duprintf("Invalid hook entry %u %u\n",
1722                                  i, hook_entries[i]);
1723                         goto out_unlock;
1724                 }
1725                 if (info->underflow[i] == 0xFFFFFFFF) {
1726                         duprintf("Invalid underflow %u %u\n",
1727                                  i, underflows[i]);
1728                         goto out_unlock;
1729                 }
1730         }
1731
1732         ret = -ENOMEM;
1733         newinfo = xt_alloc_table_info(size);
1734         if (!newinfo)
1735                 goto out_unlock;
1736
1737         newinfo->number = number;
1738         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1739                 newinfo->hook_entry[i] = info->hook_entry[i];
1740                 newinfo->underflow[i] = info->underflow[i];
1741         }
1742         entry1 = newinfo->entries[raw_smp_processor_id()];
1743         pos = entry1;
1744         size = total_size;
1745         xt_entry_foreach(iter0, entry0, total_size) {
1746                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1747                                                   name, newinfo, entry1);
1748                 if (ret != 0)
1749                         break;
1750         }
1751         xt_compat_flush_offsets(AF_INET);
1752         xt_compat_unlock(AF_INET);
1753         if (ret)
1754                 goto free_newinfo;
1755
1756         ret = -ELOOP;
1757         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1758                 goto free_newinfo;
1759
1760         i = 0;
1761         xt_entry_foreach(iter1, entry1, newinfo->size) {
1762                 ret = compat_check_entry(iter1, net, name);
1763                 if (ret != 0)
1764                         break;
1765                 ++i;
1766         }
1767         if (ret) {
1768                 /*
1769                  * The first i matches need cleanup_entry (calls ->destroy)
1770                  * because they had called ->check already. The other j-i
1771                  * entries need only release.
1772                  */
1773                 int skip = i;
1774                 j -= i;
1775                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1776                         if (skip-- > 0)
1777                                 continue;
1778                         if (j-- == 0)
1779                                 break;
1780                         compat_release_entry(iter0);
1781                 }
1782                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1783                         if (i-- == 0)
1784                                 break;
1785                         cleanup_entry(iter1, net);
1786                 }
1787                 xt_free_table_info(newinfo);
1788                 return ret;
1789         }
1790
1791         /* And one copy for every other CPU */
1792         for_each_possible_cpu(i)
1793                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1794                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1795
1796         *pinfo = newinfo;
1797         *pentry0 = entry1;
1798         xt_free_table_info(info);
1799         return 0;
1800
1801 free_newinfo:
1802         xt_free_table_info(newinfo);
1803 out:
1804         xt_entry_foreach(iter0, entry0, total_size) {
1805                 if (j-- == 0)
1806                         break;
1807                 compat_release_entry(iter0);
1808         }
1809         return ret;
1810 out_unlock:
1811         xt_compat_flush_offsets(AF_INET);
1812         xt_compat_unlock(AF_INET);
1813         goto out;
1814 }
1815
1816 static int
1817 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1818 {
1819         int ret;
1820         struct compat_ipt_replace tmp;
1821         struct xt_table_info *newinfo;
1822         void *loc_cpu_entry;
1823         struct ipt_entry *iter;
1824
1825         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1826                 return -EFAULT;
1827
1828         /* overflow check */
1829         if (tmp.size >= INT_MAX / num_possible_cpus())
1830                 return -ENOMEM;
1831         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1832                 return -ENOMEM;
1833
1834         newinfo = xt_alloc_table_info(tmp.size);
1835         if (!newinfo)
1836                 return -ENOMEM;
1837
1838         /* choose the copy that is on our node/cpu */
1839         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1840         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1841                            tmp.size) != 0) {
1842                 ret = -EFAULT;
1843                 goto free_newinfo;
1844         }
1845
1846         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1847                                      &newinfo, &loc_cpu_entry, tmp.size,
1848                                      tmp.num_entries, tmp.hook_entry,
1849                                      tmp.underflow);
1850         if (ret != 0)
1851                 goto free_newinfo;
1852
1853         duprintf("compat_do_replace: Translated table\n");
1854
1855         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1856                            tmp.num_counters, compat_ptr(tmp.counters));
1857         if (ret)
1858                 goto free_newinfo_untrans;
1859         return 0;
1860
1861  free_newinfo_untrans:
1862         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1863                 cleanup_entry(iter, net);
1864  free_newinfo:
1865         xt_free_table_info(newinfo);
1866         return ret;
1867 }
1868
1869 static int
1870 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1871                       unsigned int len)
1872 {
1873         int ret;
1874
1875         if (!capable(CAP_NET_ADMIN))
1876                 return -EPERM;
1877
1878         switch (cmd) {
1879         case IPT_SO_SET_REPLACE:
1880                 ret = compat_do_replace(sock_net(sk), user, len);
1881                 break;
1882
1883         case IPT_SO_SET_ADD_COUNTERS:
1884                 ret = do_add_counters(sock_net(sk), user, len, 1);
1885                 break;
1886
1887         default:
1888                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1889                 ret = -EINVAL;
1890         }
1891
1892         return ret;
1893 }
1894
1895 struct compat_ipt_get_entries {
1896         char name[IPT_TABLE_MAXNAMELEN];
1897         compat_uint_t size;
1898         struct compat_ipt_entry entrytable[0];
1899 };
1900
1901 static int
1902 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1903                             void __user *userptr)
1904 {
1905         struct xt_counters *counters;
1906         const struct xt_table_info *private = table->private;
1907         void __user *pos;
1908         unsigned int size;
1909         int ret = 0;
1910         const void *loc_cpu_entry;
1911         unsigned int i = 0;
1912         struct ipt_entry *iter;
1913
1914         counters = alloc_counters(table);
1915         if (IS_ERR(counters))
1916                 return PTR_ERR(counters);
1917
1918         /* choose the copy that is on our node/cpu, ...
1919          * This choice is lazy (because current thread is
1920          * allowed to migrate to another cpu)
1921          */
1922         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1923         pos = userptr;
1924         size = total_size;
1925         xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1926                 ret = compat_copy_entry_to_user(iter, &pos,
1927                                                 &size, counters, i++);
1928                 if (ret != 0)
1929                         break;
1930         }
1931
1932         vfree(counters);
1933         return ret;
1934 }
1935
1936 static int
1937 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1938                    int *len)
1939 {
1940         int ret;
1941         struct compat_ipt_get_entries get;
1942         struct xt_table *t;
1943
1944         if (*len < sizeof(get)) {
1945                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1946                 return -EINVAL;
1947         }
1948
1949         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1950                 return -EFAULT;
1951
1952         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1953                 duprintf("compat_get_entries: %u != %zu\n",
1954                          *len, sizeof(get) + get.size);
1955                 return -EINVAL;
1956         }
1957
1958         xt_compat_lock(AF_INET);
1959         t = xt_find_table_lock(net, AF_INET, get.name);
1960         if (t && !IS_ERR(t)) {
1961                 const struct xt_table_info *private = t->private;
1962                 struct xt_table_info info;
1963                 duprintf("t->private->number = %u\n", private->number);
1964                 ret = compat_table_info(private, &info);
1965                 if (!ret && get.size == info.size) {
1966                         ret = compat_copy_entries_to_user(private->size,
1967                                                           t, uptr->entrytable);
1968                 } else if (!ret) {
1969                         duprintf("compat_get_entries: I've got %u not %u!\n",
1970                                  private->size, get.size);
1971                         ret = -EAGAIN;
1972                 }
1973                 xt_compat_flush_offsets(AF_INET);
1974                 module_put(t->me);
1975                 xt_table_unlock(t);
1976         } else
1977                 ret = t ? PTR_ERR(t) : -ENOENT;
1978
1979         xt_compat_unlock(AF_INET);
1980         return ret;
1981 }
1982
1983 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1984
1985 static int
1986 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1987 {
1988         int ret;
1989
1990         if (!capable(CAP_NET_ADMIN))
1991                 return -EPERM;
1992
1993         switch (cmd) {
1994         case IPT_SO_GET_INFO:
1995                 ret = get_info(sock_net(sk), user, len, 1);
1996                 break;
1997         case IPT_SO_GET_ENTRIES:
1998                 ret = compat_get_entries(sock_net(sk), user, len);
1999                 break;
2000         default:
2001                 ret = do_ipt_get_ctl(sk, cmd, user, len);
2002         }
2003         return ret;
2004 }
2005 #endif
2006
2007 static int
2008 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2009 {
2010         int ret;
2011
2012         if (!capable(CAP_NET_ADMIN))
2013                 return -EPERM;
2014
2015         switch (cmd) {
2016         case IPT_SO_SET_REPLACE:
2017                 ret = do_replace(sock_net(sk), user, len);
2018                 break;
2019
2020         case IPT_SO_SET_ADD_COUNTERS:
2021                 ret = do_add_counters(sock_net(sk), user, len, 0);
2022                 break;
2023
2024         default:
2025                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2026                 ret = -EINVAL;
2027         }
2028
2029         return ret;
2030 }
2031
2032 static int
2033 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2034 {
2035         int ret;
2036
2037         if (!capable(CAP_NET_ADMIN))
2038                 return -EPERM;
2039
2040         switch (cmd) {
2041         case IPT_SO_GET_INFO:
2042                 ret = get_info(sock_net(sk), user, len, 0);
2043                 break;
2044
2045         case IPT_SO_GET_ENTRIES:
2046                 ret = get_entries(sock_net(sk), user, len);
2047                 break;
2048
2049         case IPT_SO_GET_REVISION_MATCH:
2050         case IPT_SO_GET_REVISION_TARGET: {
2051                 struct ipt_get_revision rev;
2052                 int target;
2053
2054                 if (*len != sizeof(rev)) {
2055                         ret = -EINVAL;
2056                         break;
2057                 }
2058                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2059                         ret = -EFAULT;
2060                         break;
2061                 }
2062
2063                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2064                         target = 1;
2065                 else
2066                         target = 0;
2067
2068                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2069                                                          rev.revision,
2070                                                          target, &ret),
2071                                         "ipt_%s", rev.name);
2072                 break;
2073         }
2074
2075         default:
2076                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2077                 ret = -EINVAL;
2078         }
2079
2080         return ret;
2081 }
2082
2083 struct xt_table *ipt_register_table(struct net *net,
2084                                     const struct xt_table *table,
2085                                     const struct ipt_replace *repl)
2086 {
2087         int ret;
2088         struct xt_table_info *newinfo;
2089         struct xt_table_info bootstrap
2090                 = { 0, 0, 0, { 0 }, { 0 }, { } };
2091         void *loc_cpu_entry;
2092         struct xt_table *new_table;
2093
2094         newinfo = xt_alloc_table_info(repl->size);
2095         if (!newinfo) {
2096                 ret = -ENOMEM;
2097                 goto out;
2098         }
2099
2100         /* choose the copy on our node/cpu, but dont care about preemption */
2101         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2102         memcpy(loc_cpu_entry, repl->entries, repl->size);
2103
2104         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2105         if (ret != 0)
2106                 goto out_free;
2107
2108         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2109         if (IS_ERR(new_table)) {
2110                 ret = PTR_ERR(new_table);
2111                 goto out_free;
2112         }
2113
2114         return new_table;
2115
2116 out_free:
2117         xt_free_table_info(newinfo);
2118 out:
2119         return ERR_PTR(ret);
2120 }
2121
2122 void ipt_unregister_table(struct net *net, struct xt_table *table)
2123 {
2124         struct xt_table_info *private;
2125         void *loc_cpu_entry;
2126         struct module *table_owner = table->me;
2127         struct ipt_entry *iter;
2128
2129         private = xt_unregister_table(table);
2130
2131         /* Decrease module usage counts and free resources */
2132         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2133         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2134                 cleanup_entry(iter, net);
2135         if (private->number > private->initial_entries)
2136                 module_put(table_owner);
2137         xt_free_table_info(private);
2138 }
2139
2140 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2141 static inline bool
2142 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2143                      u_int8_t type, u_int8_t code,
2144                      bool invert)
2145 {
2146         return ((test_type == 0xFF) ||
2147                 (type == test_type && code >= min_code && code <= max_code))
2148                 ^ invert;
2149 }
2150
2151 static bool
2152 icmp_match(const struct sk_buff *skb, const struct xt_match_param *par)
2153 {
2154         const struct icmphdr *ic;
2155         struct icmphdr _icmph;
2156         const struct ipt_icmp *icmpinfo = par->matchinfo;
2157
2158         /* Must not be a fragment. */
2159         if (par->fragoff != 0)
2160                 return false;
2161
2162         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2163         if (ic == NULL) {
2164                 /* We've been asked to examine this packet, and we
2165                  * can't.  Hence, no choice but to drop.
2166                  */
2167                 duprintf("Dropping evil ICMP tinygram.\n");
2168                 *par->hotdrop = true;
2169                 return false;
2170         }
2171
2172         return icmp_type_code_match(icmpinfo->type,
2173                                     icmpinfo->code[0],
2174                                     icmpinfo->code[1],
2175                                     ic->type, ic->code,
2176                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2177 }
2178
2179 static bool icmp_checkentry(const struct xt_mtchk_param *par)
2180 {
2181         const struct ipt_icmp *icmpinfo = par->matchinfo;
2182
2183         /* Must specify no unknown invflags */
2184         return !(icmpinfo->invflags & ~IPT_ICMP_INV);
2185 }
2186
2187 /* The built-in targets: standard (NULL) and error. */
2188 static struct xt_target ipt_standard_target __read_mostly = {
2189         .name           = IPT_STANDARD_TARGET,
2190         .targetsize     = sizeof(int),
2191         .family         = NFPROTO_IPV4,
2192 #ifdef CONFIG_COMPAT
2193         .compatsize     = sizeof(compat_int_t),
2194         .compat_from_user = compat_standard_from_user,
2195         .compat_to_user = compat_standard_to_user,
2196 #endif
2197 };
2198
2199 static struct xt_target ipt_error_target __read_mostly = {
2200         .name           = IPT_ERROR_TARGET,
2201         .target         = ipt_error,
2202         .targetsize     = IPT_FUNCTION_MAXNAMELEN,
2203         .family         = NFPROTO_IPV4,
2204 };
2205
2206 static struct nf_sockopt_ops ipt_sockopts = {
2207         .pf             = PF_INET,
2208         .set_optmin     = IPT_BASE_CTL,
2209         .set_optmax     = IPT_SO_SET_MAX+1,
2210         .set            = do_ipt_set_ctl,
2211 #ifdef CONFIG_COMPAT
2212         .compat_set     = compat_do_ipt_set_ctl,
2213 #endif
2214         .get_optmin     = IPT_BASE_CTL,
2215         .get_optmax     = IPT_SO_GET_MAX+1,
2216         .get            = do_ipt_get_ctl,
2217 #ifdef CONFIG_COMPAT
2218         .compat_get     = compat_do_ipt_get_ctl,
2219 #endif
2220         .owner          = THIS_MODULE,
2221 };
2222
2223 static struct xt_match icmp_matchstruct __read_mostly = {
2224         .name           = "icmp",
2225         .match          = icmp_match,
2226         .matchsize      = sizeof(struct ipt_icmp),
2227         .checkentry     = icmp_checkentry,
2228         .proto          = IPPROTO_ICMP,
2229         .family         = NFPROTO_IPV4,
2230 };
2231
2232 static int __net_init ip_tables_net_init(struct net *net)
2233 {
2234         return xt_proto_init(net, NFPROTO_IPV4);
2235 }
2236
2237 static void __net_exit ip_tables_net_exit(struct net *net)
2238 {
2239         xt_proto_fini(net, NFPROTO_IPV4);
2240 }
2241
2242 static struct pernet_operations ip_tables_net_ops = {
2243         .init = ip_tables_net_init,
2244         .exit = ip_tables_net_exit,
2245 };
2246
2247 static int __init ip_tables_init(void)
2248 {
2249         int ret;
2250
2251         ret = register_pernet_subsys(&ip_tables_net_ops);
2252         if (ret < 0)
2253                 goto err1;
2254
2255         /* Noone else will be downing sem now, so we won't sleep */
2256         ret = xt_register_target(&ipt_standard_target);
2257         if (ret < 0)
2258                 goto err2;
2259         ret = xt_register_target(&ipt_error_target);
2260         if (ret < 0)
2261                 goto err3;
2262         ret = xt_register_match(&icmp_matchstruct);
2263         if (ret < 0)
2264                 goto err4;
2265
2266         /* Register setsockopt */
2267         ret = nf_register_sockopt(&ipt_sockopts);
2268         if (ret < 0)
2269                 goto err5;
2270
2271         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2272         return 0;
2273
2274 err5:
2275         xt_unregister_match(&icmp_matchstruct);
2276 err4:
2277         xt_unregister_target(&ipt_error_target);
2278 err3:
2279         xt_unregister_target(&ipt_standard_target);
2280 err2:
2281         unregister_pernet_subsys(&ip_tables_net_ops);
2282 err1:
2283         return ret;
2284 }
2285
2286 static void __exit ip_tables_fini(void)
2287 {
2288         nf_unregister_sockopt(&ipt_sockopts);
2289
2290         xt_unregister_match(&icmp_matchstruct);
2291         xt_unregister_target(&ipt_error_target);
2292         xt_unregister_target(&ipt_standard_target);
2293
2294         unregister_pernet_subsys(&ip_tables_net_ops);
2295 }
2296
2297 EXPORT_SYMBOL(ipt_register_table);
2298 EXPORT_SYMBOL(ipt_unregister_table);
2299 EXPORT_SYMBOL(ipt_do_table);
2300 module_init(ip_tables_init);
2301 module_exit(ip_tables_fini);