[NETFILTER]: x_tables: add TRACE target
[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 #include <linux/cache.h>
12 #include <linux/capability.h>
13 #include <linux/skbuff.h>
14 #include <linux/kmod.h>
15 #include <linux/vmalloc.h>
16 #include <linux/netdevice.h>
17 #include <linux/module.h>
18 #include <linux/icmp.h>
19 #include <net/ip.h>
20 #include <net/compat.h>
21 #include <asm/uaccess.h>
22 #include <linux/mutex.h>
23 #include <linux/proc_fs.h>
24 #include <linux/err.h>
25 #include <linux/cpumask.h>
26
27 #include <linux/netfilter/x_tables.h>
28 #include <linux/netfilter_ipv4/ip_tables.h>
29
30 MODULE_LICENSE("GPL");
31 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
32 MODULE_DESCRIPTION("IPv4 packet filter");
33
34 /*#define DEBUG_IP_FIREWALL*/
35 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
36 /*#define DEBUG_IP_FIREWALL_USER*/
37
38 #ifdef DEBUG_IP_FIREWALL
39 #define dprintf(format, args...)  printk(format , ## args)
40 #else
41 #define dprintf(format, args...)
42 #endif
43
44 #ifdef DEBUG_IP_FIREWALL_USER
45 #define duprintf(format, args...) printk(format , ## args)
46 #else
47 #define duprintf(format, args...)
48 #endif
49
50 #ifdef CONFIG_NETFILTER_DEBUG
51 #define IP_NF_ASSERT(x)                                         \
52 do {                                                            \
53         if (!(x))                                               \
54                 printk("IP_NF_ASSERT: %s:%s:%u\n",              \
55                        __FUNCTION__, __FILE__, __LINE__);       \
56 } while(0)
57 #else
58 #define IP_NF_ASSERT(x)
59 #endif
60
61 #if 0
62 /* All the better to debug you with... */
63 #define static
64 #define inline
65 #endif
66
67 /*
68    We keep a set of rules for each CPU, so we can avoid write-locking
69    them in the softirq when updating the counters and therefore
70    only need to read-lock in the softirq; doing a write_lock_bh() in user
71    context stops packets coming through and allows user context to read
72    the counters or update the rules.
73
74    Hence the start of any table is given by get_table() below.  */
75
76 /* Returns whether matches rule or not. */
77 static inline int
78 ip_packet_match(const struct iphdr *ip,
79                 const char *indev,
80                 const char *outdev,
81                 const struct ipt_ip *ipinfo,
82                 int isfrag)
83 {
84         size_t i;
85         unsigned long ret;
86
87 #define FWINV(bool,invflg) ((bool) ^ !!(ipinfo->invflags & invflg))
88
89         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
90                   IPT_INV_SRCIP)
91             || FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
92                      IPT_INV_DSTIP)) {
93                 dprintf("Source or dest mismatch.\n");
94
95                 dprintf("SRC: %u.%u.%u.%u. Mask: %u.%u.%u.%u. Target: %u.%u.%u.%u.%s\n",
96                         NIPQUAD(ip->saddr),
97                         NIPQUAD(ipinfo->smsk.s_addr),
98                         NIPQUAD(ipinfo->src.s_addr),
99                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
100                 dprintf("DST: %u.%u.%u.%u Mask: %u.%u.%u.%u Target: %u.%u.%u.%u.%s\n",
101                         NIPQUAD(ip->daddr),
102                         NIPQUAD(ipinfo->dmsk.s_addr),
103                         NIPQUAD(ipinfo->dst.s_addr),
104                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
105                 return 0;
106         }
107
108         /* Look for ifname matches; this should unroll nicely. */
109         for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
110                 ret |= (((const unsigned long *)indev)[i]
111                         ^ ((const unsigned long *)ipinfo->iniface)[i])
112                         & ((const unsigned long *)ipinfo->iniface_mask)[i];
113         }
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 0;
120         }
121
122         for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
123                 ret |= (((const unsigned long *)outdev)[i]
124                         ^ ((const unsigned long *)ipinfo->outiface)[i])
125                         & ((const unsigned long *)ipinfo->outiface_mask)[i];
126         }
127
128         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
129                 dprintf("VIA out mismatch (%s vs %s).%s\n",
130                         outdev, ipinfo->outiface,
131                         ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
132                 return 0;
133         }
134
135         /* Check specific protocol */
136         if (ipinfo->proto
137             && FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
138                 dprintf("Packet protocol %hi does not match %hi.%s\n",
139                         ip->protocol, ipinfo->proto,
140                         ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
141                 return 0;
142         }
143
144         /* If we have a fragment rule but the packet is not a fragment
145          * then we return zero */
146         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
147                 dprintf("Fragment rule but not fragment.%s\n",
148                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
149                 return 0;
150         }
151
152         return 1;
153 }
154
155 static inline bool
156 ip_checkentry(const struct ipt_ip *ip)
157 {
158         if (ip->flags & ~IPT_F_MASK) {
159                 duprintf("Unknown flag bits set: %08X\n",
160                          ip->flags & ~IPT_F_MASK);
161                 return false;
162         }
163         if (ip->invflags & ~IPT_INV_MASK) {
164                 duprintf("Unknown invflag bits set: %08X\n",
165                          ip->invflags & ~IPT_INV_MASK);
166                 return false;
167         }
168         return true;
169 }
170
171 static unsigned int
172 ipt_error(struct sk_buff **pskb,
173           const struct net_device *in,
174           const struct net_device *out,
175           unsigned int hooknum,
176           const struct xt_target *target,
177           const void *targinfo)
178 {
179         if (net_ratelimit())
180                 printk("ip_tables: error: `%s'\n", (char *)targinfo);
181
182         return NF_DROP;
183 }
184
185 static inline
186 bool do_match(struct ipt_entry_match *m,
187               const struct sk_buff *skb,
188               const struct net_device *in,
189               const struct net_device *out,
190               int offset,
191               bool *hotdrop)
192 {
193         /* Stop iteration if it doesn't match */
194         if (!m->u.kernel.match->match(skb, in, out, m->u.kernel.match, m->data,
195                                       offset, ip_hdrlen(skb), hotdrop))
196                 return true;
197         else
198                 return false;
199 }
200
201 static inline struct ipt_entry *
202 get_entry(void *base, unsigned int offset)
203 {
204         return (struct ipt_entry *)(base + offset);
205 }
206
207 /* All zeroes == unconditional rule. */
208 static inline int
209 unconditional(const struct ipt_ip *ip)
210 {
211         unsigned int i;
212
213         for (i = 0; i < sizeof(*ip)/sizeof(__u32); i++)
214                 if (((__u32 *)ip)[i])
215                         return 0;
216
217         return 1;
218 }
219
220 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
221     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
222 static const char *hooknames[] = {
223         [NF_IP_PRE_ROUTING]             = "PREROUTING",
224         [NF_IP_LOCAL_IN]                = "INPUT",
225         [NF_IP_FORWARD]                 = "FORWARD",
226         [NF_IP_LOCAL_OUT]               = "OUTPUT",
227         [NF_IP_POST_ROUTING]            = "POSTROUTING",
228 };
229
230 enum nf_ip_trace_comments {
231         NF_IP_TRACE_COMMENT_RULE,
232         NF_IP_TRACE_COMMENT_RETURN,
233         NF_IP_TRACE_COMMENT_POLICY,
234 };
235
236 static const char *comments[] = {
237         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
238         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
239         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
240 };
241
242 static struct nf_loginfo trace_loginfo = {
243         .type = NF_LOG_TYPE_LOG,
244         .u = {
245                 .log = {
246                         .level = 4,
247                         .logflags = NF_LOG_MASK,
248                 },
249         },
250 };
251
252 static inline int
253 get_chainname_rulenum(struct ipt_entry *s, struct ipt_entry *e,
254                       char *hookname, char **chainname,
255                       char **comment, unsigned int *rulenum)
256 {
257         struct ipt_standard_target *t = (void *)ipt_get_target(s);
258
259         if (strcmp(t->target.u.kernel.target->name, IPT_ERROR_TARGET) == 0) {
260                 /* Head of user chain: ERROR target with chainname */
261                 *chainname = t->target.data;
262                 (*rulenum) = 0;
263         } else if (s == e) {
264                 (*rulenum)++;
265
266                 if (s->target_offset == sizeof(struct ipt_entry)
267                    && strcmp(t->target.u.kernel.target->name,
268                              IPT_STANDARD_TARGET) == 0
269                    && t->verdict < 0
270                    && unconditional(&s->ip)) {
271                         /* Tail of chains: STANDARD target (return/policy) */
272                         *comment = *chainname == hookname
273                                 ? (char *)comments[NF_IP_TRACE_COMMENT_POLICY]
274                                 : (char *)comments[NF_IP_TRACE_COMMENT_RETURN];
275                 }
276                 return 1;
277         } else
278                 (*rulenum)++;
279
280         return 0;
281 }
282
283 static void trace_packet(struct sk_buff *skb,
284                          unsigned int hook,
285                          const struct net_device *in,
286                          const struct net_device *out,
287                          char *tablename,
288                          struct xt_table_info *private,
289                          struct ipt_entry *e)
290 {
291         void *table_base;
292         struct ipt_entry *root;
293         char *hookname, *chainname, *comment;
294         unsigned int rulenum = 0;
295
296         table_base = (void *)private->entries[smp_processor_id()];
297         root = get_entry(table_base, private->hook_entry[hook]);
298
299         hookname = chainname = (char *)hooknames[hook];
300         comment = (char *)comments[NF_IP_TRACE_COMMENT_RULE];
301
302         IPT_ENTRY_ITERATE(root,
303                           private->size - private->hook_entry[hook],
304                           get_chainname_rulenum,
305                           e, hookname, &chainname, &comment, &rulenum);
306
307         nf_log_packet(AF_INET, hook, skb, in, out, &trace_loginfo,
308                       "TRACE: %s:%s:%s:%u ",
309                       tablename, chainname, comment, rulenum);
310 }
311 #endif
312
313 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
314 unsigned int
315 ipt_do_table(struct sk_buff **pskb,
316              unsigned int hook,
317              const struct net_device *in,
318              const struct net_device *out,
319              struct xt_table *table)
320 {
321         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
322         u_int16_t offset;
323         struct iphdr *ip;
324         u_int16_t datalen;
325         bool hotdrop = false;
326         /* Initializing verdict to NF_DROP keeps gcc happy. */
327         unsigned int verdict = NF_DROP;
328         const char *indev, *outdev;
329         void *table_base;
330         struct ipt_entry *e, *back;
331         struct xt_table_info *private;
332
333         /* Initialization */
334         ip = ip_hdr(*pskb);
335         datalen = (*pskb)->len - ip->ihl * 4;
336         indev = in ? in->name : nulldevname;
337         outdev = out ? out->name : nulldevname;
338         /* We handle fragments by dealing with the first fragment as
339          * if it was a normal packet.  All other fragments are treated
340          * normally, except that they will NEVER match rules that ask
341          * things we don't know, ie. tcp syn flag or ports).  If the
342          * rule is also a fragment-specific rule, non-fragments won't
343          * match it. */
344         offset = ntohs(ip->frag_off) & IP_OFFSET;
345
346         read_lock_bh(&table->lock);
347         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
348         private = table->private;
349         table_base = (void *)private->entries[smp_processor_id()];
350         e = get_entry(table_base, private->hook_entry[hook]);
351
352         /* For return from builtin chain */
353         back = get_entry(table_base, private->underflow[hook]);
354
355         do {
356                 IP_NF_ASSERT(e);
357                 IP_NF_ASSERT(back);
358                 if (ip_packet_match(ip, indev, outdev, &e->ip, offset)) {
359                         struct ipt_entry_target *t;
360
361                         if (IPT_MATCH_ITERATE(e, do_match,
362                                               *pskb, in, out,
363                                               offset, &hotdrop) != 0)
364                                 goto no_match;
365
366                         ADD_COUNTER(e->counters, ntohs(ip->tot_len), 1);
367
368                         t = ipt_get_target(e);
369                         IP_NF_ASSERT(t->u.kernel.target);
370
371 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
372     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
373                         /* The packet is traced: log it */
374                         if (unlikely((*pskb)->nf_trace))
375                                 trace_packet(*pskb, hook, in, out,
376                                              table->name, private, e);
377 #endif
378                         /* Standard target? */
379                         if (!t->u.kernel.target->target) {
380                                 int v;
381
382                                 v = ((struct ipt_standard_target *)t)->verdict;
383                                 if (v < 0) {
384                                         /* Pop from stack? */
385                                         if (v != IPT_RETURN) {
386                                                 verdict = (unsigned)(-v) - 1;
387                                                 break;
388                                         }
389                                         e = back;
390                                         back = get_entry(table_base,
391                                                          back->comefrom);
392                                         continue;
393                                 }
394                                 if (table_base + v != (void *)e + e->next_offset
395                                     && !(e->ip.flags & IPT_F_GOTO)) {
396                                         /* Save old back ptr in next entry */
397                                         struct ipt_entry *next
398                                                 = (void *)e + e->next_offset;
399                                         next->comefrom
400                                                 = (void *)back - table_base;
401                                         /* set back pointer to next entry */
402                                         back = next;
403                                 }
404
405                                 e = get_entry(table_base, v);
406                         } else {
407                                 /* Targets which reenter must return
408                                    abs. verdicts */
409 #ifdef CONFIG_NETFILTER_DEBUG
410                                 ((struct ipt_entry *)table_base)->comefrom
411                                         = 0xeeeeeeec;
412 #endif
413                                 verdict = t->u.kernel.target->target(pskb,
414                                                                      in, out,
415                                                                      hook,
416                                                                      t->u.kernel.target,
417                                                                      t->data);
418
419 #ifdef CONFIG_NETFILTER_DEBUG
420                                 if (((struct ipt_entry *)table_base)->comefrom
421                                     != 0xeeeeeeec
422                                     && verdict == IPT_CONTINUE) {
423                                         printk("Target %s reentered!\n",
424                                                t->u.kernel.target->name);
425                                         verdict = NF_DROP;
426                                 }
427                                 ((struct ipt_entry *)table_base)->comefrom
428                                         = 0x57acc001;
429 #endif
430                                 /* Target might have changed stuff. */
431                                 ip = ip_hdr(*pskb);
432                                 datalen = (*pskb)->len - ip->ihl * 4;
433
434                                 if (verdict == IPT_CONTINUE)
435                                         e = (void *)e + e->next_offset;
436                                 else
437                                         /* Verdict */
438                                         break;
439                         }
440                 } else {
441
442                 no_match:
443                         e = (void *)e + e->next_offset;
444                 }
445         } while (!hotdrop);
446
447         read_unlock_bh(&table->lock);
448
449 #ifdef DEBUG_ALLOW_ALL
450         return NF_ACCEPT;
451 #else
452         if (hotdrop)
453                 return NF_DROP;
454         else return verdict;
455 #endif
456 }
457
458 /* Figures out from what hook each rule can be called: returns 0 if
459    there are loops.  Puts hook bitmask in comefrom. */
460 static int
461 mark_source_chains(struct xt_table_info *newinfo,
462                    unsigned int valid_hooks, void *entry0)
463 {
464         unsigned int hook;
465
466         /* No recursion; use packet counter to save back ptrs (reset
467            to 0 as we leave), and comefrom to save source hook bitmask */
468         for (hook = 0; hook < NF_IP_NUMHOOKS; hook++) {
469                 unsigned int pos = newinfo->hook_entry[hook];
470                 struct ipt_entry *e
471                         = (struct ipt_entry *)(entry0 + pos);
472
473                 if (!(valid_hooks & (1 << hook)))
474                         continue;
475
476                 /* Set initial back pointer. */
477                 e->counters.pcnt = pos;
478
479                 for (;;) {
480                         struct ipt_standard_target *t
481                                 = (void *)ipt_get_target(e);
482                         int visited = e->comefrom & (1 << hook);
483
484                         if (e->comefrom & (1 << NF_IP_NUMHOOKS)) {
485                                 printk("iptables: loop hook %u pos %u %08X.\n",
486                                        hook, pos, e->comefrom);
487                                 return 0;
488                         }
489                         e->comefrom
490                                 |= ((1 << hook) | (1 << NF_IP_NUMHOOKS));
491
492                         /* Unconditional return/END. */
493                         if ((e->target_offset == sizeof(struct ipt_entry)
494                             && (strcmp(t->target.u.user.name,
495                                        IPT_STANDARD_TARGET) == 0)
496                             && t->verdict < 0
497                             && unconditional(&e->ip)) || visited) {
498                                 unsigned int oldpos, size;
499
500                                 if (t->verdict < -NF_MAX_VERDICT - 1) {
501                                         duprintf("mark_source_chains: bad "
502                                                 "negative verdict (%i)\n",
503                                                                 t->verdict);
504                                         return 0;
505                                 }
506
507                                 /* Return: backtrack through the last
508                                    big jump. */
509                                 do {
510                                         e->comefrom ^= (1<<NF_IP_NUMHOOKS);
511 #ifdef DEBUG_IP_FIREWALL_USER
512                                         if (e->comefrom
513                                             & (1 << NF_IP_NUMHOOKS)) {
514                                                 duprintf("Back unset "
515                                                          "on hook %u "
516                                                          "rule %u\n",
517                                                          hook, pos);
518                                         }
519 #endif
520                                         oldpos = pos;
521                                         pos = e->counters.pcnt;
522                                         e->counters.pcnt = 0;
523
524                                         /* We're at the start. */
525                                         if (pos == oldpos)
526                                                 goto next;
527
528                                         e = (struct ipt_entry *)
529                                                 (entry0 + pos);
530                                 } while (oldpos == pos + e->next_offset);
531
532                                 /* Move along one */
533                                 size = e->next_offset;
534                                 e = (struct ipt_entry *)
535                                         (entry0 + pos + size);
536                                 e->counters.pcnt = pos;
537                                 pos += size;
538                         } else {
539                                 int newpos = t->verdict;
540
541                                 if (strcmp(t->target.u.user.name,
542                                            IPT_STANDARD_TARGET) == 0
543                                     && newpos >= 0) {
544                                         if (newpos > newinfo->size -
545                                                 sizeof(struct ipt_entry)) {
546                                                 duprintf("mark_source_chains: "
547                                                         "bad verdict (%i)\n",
548                                                                 newpos);
549                                                 return 0;
550                                         }
551                                         /* This a jump; chase it. */
552                                         duprintf("Jump rule %u -> %u\n",
553                                                  pos, newpos);
554                                 } else {
555                                         /* ... this is a fallthru */
556                                         newpos = pos + e->next_offset;
557                                 }
558                                 e = (struct ipt_entry *)
559                                         (entry0 + newpos);
560                                 e->counters.pcnt = pos;
561                                 pos = newpos;
562                         }
563                 }
564                 next:
565                 duprintf("Finished chain %u\n", hook);
566         }
567         return 1;
568 }
569
570 static inline int
571 cleanup_match(struct ipt_entry_match *m, unsigned int *i)
572 {
573         if (i && (*i)-- == 0)
574                 return 1;
575
576         if (m->u.kernel.match->destroy)
577                 m->u.kernel.match->destroy(m->u.kernel.match, m->data);
578         module_put(m->u.kernel.match->me);
579         return 0;
580 }
581
582 static inline int
583 check_entry(struct ipt_entry *e, const char *name)
584 {
585         struct ipt_entry_target *t;
586
587         if (!ip_checkentry(&e->ip)) {
588                 duprintf("ip_tables: ip check failed %p %s.\n", e, name);
589                 return -EINVAL;
590         }
591
592         if (e->target_offset + sizeof(struct ipt_entry_target) > e->next_offset)
593                 return -EINVAL;
594
595         t = ipt_get_target(e);
596         if (e->target_offset + t->u.target_size > e->next_offset)
597                 return -EINVAL;
598
599         return 0;
600 }
601
602 static inline int check_match(struct ipt_entry_match *m, const char *name,
603                                 const struct ipt_ip *ip, unsigned int hookmask,
604                                 unsigned int *i)
605 {
606         struct xt_match *match;
607         int ret;
608
609         match = m->u.kernel.match;
610         ret = xt_check_match(match, AF_INET, m->u.match_size - sizeof(*m),
611                              name, hookmask, ip->proto,
612                              ip->invflags & IPT_INV_PROTO);
613         if (!ret && m->u.kernel.match->checkentry
614             && !m->u.kernel.match->checkentry(name, ip, match, m->data,
615                                               hookmask)) {
616                 duprintf("ip_tables: check failed for `%s'.\n",
617                          m->u.kernel.match->name);
618                 ret = -EINVAL;
619         }
620         if (!ret)
621                 (*i)++;
622         return ret;
623 }
624
625 static inline int
626 find_check_match(struct ipt_entry_match *m,
627             const char *name,
628             const struct ipt_ip *ip,
629             unsigned int hookmask,
630             unsigned int *i)
631 {
632         struct xt_match *match;
633         int ret;
634
635         match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
636                                                    m->u.user.revision),
637                                         "ipt_%s", m->u.user.name);
638         if (IS_ERR(match) || !match) {
639                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
640                 return match ? PTR_ERR(match) : -ENOENT;
641         }
642         m->u.kernel.match = match;
643
644         ret = check_match(m, name, ip, hookmask, i);
645         if (ret)
646                 goto err;
647
648         return 0;
649 err:
650         module_put(m->u.kernel.match->me);
651         return ret;
652 }
653
654 static inline int check_target(struct ipt_entry *e, const char *name)
655 {
656         struct ipt_entry_target *t;
657         struct xt_target *target;
658         int ret;
659
660         t = ipt_get_target(e);
661         target = t->u.kernel.target;
662         ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
663                               name, e->comefrom, e->ip.proto,
664                               e->ip.invflags & IPT_INV_PROTO);
665         if (!ret && t->u.kernel.target->checkentry
666                    && !t->u.kernel.target->checkentry(name, e, target,
667                                                       t->data, e->comefrom)) {
668                 duprintf("ip_tables: check failed for `%s'.\n",
669                          t->u.kernel.target->name);
670                 ret = -EINVAL;
671         }
672         return ret;
673 }
674
675 static inline int
676 find_check_entry(struct ipt_entry *e, const char *name, unsigned int size,
677             unsigned int *i)
678 {
679         struct ipt_entry_target *t;
680         struct xt_target *target;
681         int ret;
682         unsigned int j;
683
684         ret = check_entry(e, name);
685         if (ret)
686                 return ret;
687
688         j = 0;
689         ret = IPT_MATCH_ITERATE(e, find_check_match, name, &e->ip,
690                                                         e->comefrom, &j);
691         if (ret != 0)
692                 goto cleanup_matches;
693
694         t = ipt_get_target(e);
695         target = try_then_request_module(xt_find_target(AF_INET,
696                                                      t->u.user.name,
697                                                      t->u.user.revision),
698                                          "ipt_%s", t->u.user.name);
699         if (IS_ERR(target) || !target) {
700                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
701                 ret = target ? PTR_ERR(target) : -ENOENT;
702                 goto cleanup_matches;
703         }
704         t->u.kernel.target = target;
705
706         ret = check_target(e, name);
707         if (ret)
708                 goto err;
709
710         (*i)++;
711         return 0;
712  err:
713         module_put(t->u.kernel.target->me);
714  cleanup_matches:
715         IPT_MATCH_ITERATE(e, cleanup_match, &j);
716         return ret;
717 }
718
719 static inline int
720 check_entry_size_and_hooks(struct ipt_entry *e,
721                            struct xt_table_info *newinfo,
722                            unsigned char *base,
723                            unsigned char *limit,
724                            const unsigned int *hook_entries,
725                            const unsigned int *underflows,
726                            unsigned int *i)
727 {
728         unsigned int h;
729
730         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0
731             || (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
732                 duprintf("Bad offset %p\n", e);
733                 return -EINVAL;
734         }
735
736         if (e->next_offset
737             < sizeof(struct ipt_entry) + sizeof(struct ipt_entry_target)) {
738                 duprintf("checking: element %p size %u\n",
739                          e, e->next_offset);
740                 return -EINVAL;
741         }
742
743         /* Check hooks & underflows */
744         for (h = 0; h < NF_IP_NUMHOOKS; h++) {
745                 if ((unsigned char *)e - base == hook_entries[h])
746                         newinfo->hook_entry[h] = hook_entries[h];
747                 if ((unsigned char *)e - base == underflows[h])
748                         newinfo->underflow[h] = underflows[h];
749         }
750
751         /* FIXME: underflows must be unconditional, standard verdicts
752            < 0 (not IPT_RETURN). --RR */
753
754         /* Clear counters and comefrom */
755         e->counters = ((struct xt_counters) { 0, 0 });
756         e->comefrom = 0;
757
758         (*i)++;
759         return 0;
760 }
761
762 static inline int
763 cleanup_entry(struct ipt_entry *e, unsigned int *i)
764 {
765         struct ipt_entry_target *t;
766
767         if (i && (*i)-- == 0)
768                 return 1;
769
770         /* Cleanup all matches */
771         IPT_MATCH_ITERATE(e, cleanup_match, NULL);
772         t = ipt_get_target(e);
773         if (t->u.kernel.target->destroy)
774                 t->u.kernel.target->destroy(t->u.kernel.target, t->data);
775         module_put(t->u.kernel.target->me);
776         return 0;
777 }
778
779 /* Checks and translates the user-supplied table segment (held in
780    newinfo) */
781 static int
782 translate_table(const char *name,
783                 unsigned int valid_hooks,
784                 struct xt_table_info *newinfo,
785                 void *entry0,
786                 unsigned int size,
787                 unsigned int number,
788                 const unsigned int *hook_entries,
789                 const unsigned int *underflows)
790 {
791         unsigned int i;
792         int ret;
793
794         newinfo->size = size;
795         newinfo->number = number;
796
797         /* Init all hooks to impossible value. */
798         for (i = 0; i < NF_IP_NUMHOOKS; i++) {
799                 newinfo->hook_entry[i] = 0xFFFFFFFF;
800                 newinfo->underflow[i] = 0xFFFFFFFF;
801         }
802
803         duprintf("translate_table: size %u\n", newinfo->size);
804         i = 0;
805         /* Walk through entries, checking offsets. */
806         ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
807                                 check_entry_size_and_hooks,
808                                 newinfo,
809                                 entry0,
810                                 entry0 + size,
811                                 hook_entries, underflows, &i);
812         if (ret != 0)
813                 return ret;
814
815         if (i != number) {
816                 duprintf("translate_table: %u not %u entries\n",
817                          i, number);
818                 return -EINVAL;
819         }
820
821         /* Check hooks all assigned */
822         for (i = 0; i < NF_IP_NUMHOOKS; i++) {
823                 /* Only hooks which are valid */
824                 if (!(valid_hooks & (1 << i)))
825                         continue;
826                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
827                         duprintf("Invalid hook entry %u %u\n",
828                                  i, hook_entries[i]);
829                         return -EINVAL;
830                 }
831                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
832                         duprintf("Invalid underflow %u %u\n",
833                                  i, underflows[i]);
834                         return -EINVAL;
835                 }
836         }
837
838         if (!mark_source_chains(newinfo, valid_hooks, entry0))
839                 return -ELOOP;
840
841         /* Finally, each sanity check must pass */
842         i = 0;
843         ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
844                                 find_check_entry, name, size, &i);
845
846         if (ret != 0) {
847                 IPT_ENTRY_ITERATE(entry0, newinfo->size,
848                                 cleanup_entry, &i);
849                 return ret;
850         }
851
852         /* And one copy for every other CPU */
853         for_each_possible_cpu(i) {
854                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
855                         memcpy(newinfo->entries[i], entry0, newinfo->size);
856         }
857
858         return ret;
859 }
860
861 /* Gets counters. */
862 static inline int
863 add_entry_to_counter(const struct ipt_entry *e,
864                      struct xt_counters total[],
865                      unsigned int *i)
866 {
867         ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
868
869         (*i)++;
870         return 0;
871 }
872
873 static inline int
874 set_entry_to_counter(const struct ipt_entry *e,
875                      struct ipt_counters total[],
876                      unsigned int *i)
877 {
878         SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
879
880         (*i)++;
881         return 0;
882 }
883
884 static void
885 get_counters(const struct xt_table_info *t,
886              struct xt_counters counters[])
887 {
888         unsigned int cpu;
889         unsigned int i;
890         unsigned int curcpu;
891
892         /* Instead of clearing (by a previous call to memset())
893          * the counters and using adds, we set the counters
894          * with data used by 'current' CPU
895          * We dont care about preemption here.
896          */
897         curcpu = raw_smp_processor_id();
898
899         i = 0;
900         IPT_ENTRY_ITERATE(t->entries[curcpu],
901                           t->size,
902                           set_entry_to_counter,
903                           counters,
904                           &i);
905
906         for_each_possible_cpu(cpu) {
907                 if (cpu == curcpu)
908                         continue;
909                 i = 0;
910                 IPT_ENTRY_ITERATE(t->entries[cpu],
911                                   t->size,
912                                   add_entry_to_counter,
913                                   counters,
914                                   &i);
915         }
916 }
917
918 static inline struct xt_counters * alloc_counters(struct xt_table *table)
919 {
920         unsigned int countersize;
921         struct xt_counters *counters;
922         struct xt_table_info *private = table->private;
923
924         /* We need atomic snapshot of counters: rest doesn't change
925            (other than comefrom, which userspace doesn't care
926            about). */
927         countersize = sizeof(struct xt_counters) * private->number;
928         counters = vmalloc_node(countersize, numa_node_id());
929
930         if (counters == NULL)
931                 return ERR_PTR(-ENOMEM);
932
933         /* First, sum counters... */
934         write_lock_bh(&table->lock);
935         get_counters(private, counters);
936         write_unlock_bh(&table->lock);
937
938         return counters;
939 }
940
941 static int
942 copy_entries_to_user(unsigned int total_size,
943                      struct xt_table *table,
944                      void __user *userptr)
945 {
946         unsigned int off, num;
947         struct ipt_entry *e;
948         struct xt_counters *counters;
949         struct xt_table_info *private = table->private;
950         int ret = 0;
951         void *loc_cpu_entry;
952
953         counters = alloc_counters(table);
954         if (IS_ERR(counters))
955                 return PTR_ERR(counters);
956
957         /* choose the copy that is on our node/cpu, ...
958          * This choice is lazy (because current thread is
959          * allowed to migrate to another cpu)
960          */
961         loc_cpu_entry = private->entries[raw_smp_processor_id()];
962         /* ... then copy entire thing ... */
963         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
964                 ret = -EFAULT;
965                 goto free_counters;
966         }
967
968         /* FIXME: use iterator macros --RR */
969         /* ... then go back and fix counters and names */
970         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
971                 unsigned int i;
972                 struct ipt_entry_match *m;
973                 struct ipt_entry_target *t;
974
975                 e = (struct ipt_entry *)(loc_cpu_entry + off);
976                 if (copy_to_user(userptr + off
977                                  + offsetof(struct ipt_entry, counters),
978                                  &counters[num],
979                                  sizeof(counters[num])) != 0) {
980                         ret = -EFAULT;
981                         goto free_counters;
982                 }
983
984                 for (i = sizeof(struct ipt_entry);
985                      i < e->target_offset;
986                      i += m->u.match_size) {
987                         m = (void *)e + i;
988
989                         if (copy_to_user(userptr + off + i
990                                          + offsetof(struct ipt_entry_match,
991                                                     u.user.name),
992                                          m->u.kernel.match->name,
993                                          strlen(m->u.kernel.match->name)+1)
994                             != 0) {
995                                 ret = -EFAULT;
996                                 goto free_counters;
997                         }
998                 }
999
1000                 t = ipt_get_target(e);
1001                 if (copy_to_user(userptr + off + e->target_offset
1002                                  + offsetof(struct ipt_entry_target,
1003                                             u.user.name),
1004                                  t->u.kernel.target->name,
1005                                  strlen(t->u.kernel.target->name)+1) != 0) {
1006                         ret = -EFAULT;
1007                         goto free_counters;
1008                 }
1009         }
1010
1011  free_counters:
1012         vfree(counters);
1013         return ret;
1014 }
1015
1016 #ifdef CONFIG_COMPAT
1017 struct compat_delta {
1018         struct compat_delta *next;
1019         unsigned int offset;
1020         short delta;
1021 };
1022
1023 static struct compat_delta *compat_offsets = NULL;
1024
1025 static int compat_add_offset(unsigned int offset, short delta)
1026 {
1027         struct compat_delta *tmp;
1028
1029         tmp = kmalloc(sizeof(struct compat_delta), GFP_KERNEL);
1030         if (!tmp)
1031                 return -ENOMEM;
1032         tmp->offset = offset;
1033         tmp->delta = delta;
1034         if (compat_offsets) {
1035                 tmp->next = compat_offsets->next;
1036                 compat_offsets->next = tmp;
1037         } else {
1038                 compat_offsets = tmp;
1039                 tmp->next = NULL;
1040         }
1041         return 0;
1042 }
1043
1044 static void compat_flush_offsets(void)
1045 {
1046         struct compat_delta *tmp, *next;
1047
1048         if (compat_offsets) {
1049                 for(tmp = compat_offsets; tmp; tmp = next) {
1050                         next = tmp->next;
1051                         kfree(tmp);
1052                 }
1053                 compat_offsets = NULL;
1054         }
1055 }
1056
1057 static short compat_calc_jump(unsigned int offset)
1058 {
1059         struct compat_delta *tmp;
1060         short delta;
1061
1062         for(tmp = compat_offsets, delta = 0; tmp; tmp = tmp->next)
1063                 if (tmp->offset < offset)
1064                         delta += tmp->delta;
1065         return delta;
1066 }
1067
1068 static void compat_standard_from_user(void *dst, void *src)
1069 {
1070         int v = *(compat_int_t *)src;
1071
1072         if (v > 0)
1073                 v += compat_calc_jump(v);
1074         memcpy(dst, &v, sizeof(v));
1075 }
1076
1077 static int compat_standard_to_user(void __user *dst, void *src)
1078 {
1079         compat_int_t cv = *(int *)src;
1080
1081         if (cv > 0)
1082                 cv -= compat_calc_jump(cv);
1083         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1084 }
1085
1086 static inline int
1087 compat_calc_match(struct ipt_entry_match *m, int * size)
1088 {
1089         *size += xt_compat_match_offset(m->u.kernel.match);
1090         return 0;
1091 }
1092
1093 static int compat_calc_entry(struct ipt_entry *e, struct xt_table_info *info,
1094                 void *base, struct xt_table_info *newinfo)
1095 {
1096         struct ipt_entry_target *t;
1097         unsigned int entry_offset;
1098         int off, i, ret;
1099
1100         off = 0;
1101         entry_offset = (void *)e - base;
1102         IPT_MATCH_ITERATE(e, compat_calc_match, &off);
1103         t = ipt_get_target(e);
1104         off += xt_compat_target_offset(t->u.kernel.target);
1105         newinfo->size -= off;
1106         ret = compat_add_offset(entry_offset, off);
1107         if (ret)
1108                 return ret;
1109
1110         for (i = 0; i< NF_IP_NUMHOOKS; i++) {
1111                 if (info->hook_entry[i] && (e < (struct ipt_entry *)
1112                                 (base + info->hook_entry[i])))
1113                         newinfo->hook_entry[i] -= off;
1114                 if (info->underflow[i] && (e < (struct ipt_entry *)
1115                                 (base + info->underflow[i])))
1116                         newinfo->underflow[i] -= off;
1117         }
1118         return 0;
1119 }
1120
1121 static int compat_table_info(struct xt_table_info *info,
1122                 struct xt_table_info *newinfo)
1123 {
1124         void *loc_cpu_entry;
1125         int i;
1126
1127         if (!newinfo || !info)
1128                 return -EINVAL;
1129
1130         memset(newinfo, 0, sizeof(struct xt_table_info));
1131         newinfo->size = info->size;
1132         newinfo->number = info->number;
1133         for (i = 0; i < NF_IP_NUMHOOKS; i++) {
1134                 newinfo->hook_entry[i] = info->hook_entry[i];
1135                 newinfo->underflow[i] = info->underflow[i];
1136         }
1137         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1138         return IPT_ENTRY_ITERATE(loc_cpu_entry, info->size,
1139                         compat_calc_entry, info, loc_cpu_entry, newinfo);
1140 }
1141 #endif
1142
1143 static int get_info(void __user *user, int *len, int compat)
1144 {
1145         char name[IPT_TABLE_MAXNAMELEN];
1146         struct xt_table *t;
1147         int ret;
1148
1149         if (*len != sizeof(struct ipt_getinfo)) {
1150                 duprintf("length %u != %u\n", *len,
1151                         (unsigned int)sizeof(struct ipt_getinfo));
1152                 return -EINVAL;
1153         }
1154
1155         if (copy_from_user(name, user, sizeof(name)) != 0)
1156                 return -EFAULT;
1157
1158         name[IPT_TABLE_MAXNAMELEN-1] = '\0';
1159 #ifdef CONFIG_COMPAT
1160         if (compat)
1161                 xt_compat_lock(AF_INET);
1162 #endif
1163         t = try_then_request_module(xt_find_table_lock(AF_INET, name),
1164                         "iptable_%s", name);
1165         if (t && !IS_ERR(t)) {
1166                 struct ipt_getinfo info;
1167                 struct xt_table_info *private = t->private;
1168
1169 #ifdef CONFIG_COMPAT
1170                 if (compat) {
1171                         struct xt_table_info tmp;
1172                         ret = compat_table_info(private, &tmp);
1173                         compat_flush_offsets();
1174                         private =  &tmp;
1175                 }
1176 #endif
1177                 info.valid_hooks = t->valid_hooks;
1178                 memcpy(info.hook_entry, private->hook_entry,
1179                                 sizeof(info.hook_entry));
1180                 memcpy(info.underflow, private->underflow,
1181                                 sizeof(info.underflow));
1182                 info.num_entries = private->number;
1183                 info.size = private->size;
1184                 strcpy(info.name, name);
1185
1186                 if (copy_to_user(user, &info, *len) != 0)
1187                         ret = -EFAULT;
1188                 else
1189                         ret = 0;
1190
1191                 xt_table_unlock(t);
1192                 module_put(t->me);
1193         } else
1194                 ret = t ? PTR_ERR(t) : -ENOENT;
1195 #ifdef CONFIG_COMPAT
1196         if (compat)
1197                 xt_compat_unlock(AF_INET);
1198 #endif
1199         return ret;
1200 }
1201
1202 static int
1203 get_entries(struct ipt_get_entries __user *uptr, int *len)
1204 {
1205         int ret;
1206         struct ipt_get_entries get;
1207         struct xt_table *t;
1208
1209         if (*len < sizeof(get)) {
1210                 duprintf("get_entries: %u < %d\n", *len,
1211                                 (unsigned int)sizeof(get));
1212                 return -EINVAL;
1213         }
1214         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1215                 return -EFAULT;
1216         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1217                 duprintf("get_entries: %u != %u\n", *len,
1218                                 (unsigned int)(sizeof(struct ipt_get_entries) +
1219                                 get.size));
1220                 return -EINVAL;
1221         }
1222
1223         t = xt_find_table_lock(AF_INET, get.name);
1224         if (t && !IS_ERR(t)) {
1225                 struct xt_table_info *private = t->private;
1226                 duprintf("t->private->number = %u\n",
1227                          private->number);
1228                 if (get.size == private->size)
1229                         ret = copy_entries_to_user(private->size,
1230                                                    t, uptr->entrytable);
1231                 else {
1232                         duprintf("get_entries: I've got %u not %u!\n",
1233                                  private->size,
1234                                  get.size);
1235                         ret = -EINVAL;
1236                 }
1237                 module_put(t->me);
1238                 xt_table_unlock(t);
1239         } else
1240                 ret = t ? PTR_ERR(t) : -ENOENT;
1241
1242         return ret;
1243 }
1244
1245 static int
1246 __do_replace(const char *name, unsigned int valid_hooks,
1247                 struct xt_table_info *newinfo, unsigned int num_counters,
1248                 void __user *counters_ptr)
1249 {
1250         int ret;
1251         struct xt_table *t;
1252         struct xt_table_info *oldinfo;
1253         struct xt_counters *counters;
1254         void *loc_cpu_old_entry;
1255
1256         ret = 0;
1257         counters = vmalloc(num_counters * sizeof(struct xt_counters));
1258         if (!counters) {
1259                 ret = -ENOMEM;
1260                 goto out;
1261         }
1262
1263         t = try_then_request_module(xt_find_table_lock(AF_INET, name),
1264                                     "iptable_%s", name);
1265         if (!t || IS_ERR(t)) {
1266                 ret = t ? PTR_ERR(t) : -ENOENT;
1267                 goto free_newinfo_counters_untrans;
1268         }
1269
1270         /* You lied! */
1271         if (valid_hooks != t->valid_hooks) {
1272                 duprintf("Valid hook crap: %08X vs %08X\n",
1273                          valid_hooks, t->valid_hooks);
1274                 ret = -EINVAL;
1275                 goto put_module;
1276         }
1277
1278         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1279         if (!oldinfo)
1280                 goto put_module;
1281
1282         /* Update module usage count based on number of rules */
1283         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1284                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1285         if ((oldinfo->number > oldinfo->initial_entries) ||
1286             (newinfo->number <= oldinfo->initial_entries))
1287                 module_put(t->me);
1288         if ((oldinfo->number > oldinfo->initial_entries) &&
1289             (newinfo->number <= oldinfo->initial_entries))
1290                 module_put(t->me);
1291
1292         /* Get the old counters. */
1293         get_counters(oldinfo, counters);
1294         /* Decrease module usage counts and free resource */
1295         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1296         IPT_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,NULL);
1297         xt_free_table_info(oldinfo);
1298         if (copy_to_user(counters_ptr, counters,
1299                          sizeof(struct xt_counters) * num_counters) != 0)
1300                 ret = -EFAULT;
1301         vfree(counters);
1302         xt_table_unlock(t);
1303         return ret;
1304
1305  put_module:
1306         module_put(t->me);
1307         xt_table_unlock(t);
1308  free_newinfo_counters_untrans:
1309         vfree(counters);
1310  out:
1311         return ret;
1312 }
1313
1314 static int
1315 do_replace(void __user *user, unsigned int len)
1316 {
1317         int ret;
1318         struct ipt_replace tmp;
1319         struct xt_table_info *newinfo;
1320         void *loc_cpu_entry;
1321
1322         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1323                 return -EFAULT;
1324
1325         /* Hack: Causes ipchains to give correct error msg --RR */
1326         if (len != sizeof(tmp) + tmp.size)
1327                 return -ENOPROTOOPT;
1328
1329         /* overflow check */
1330         if (tmp.size >= (INT_MAX - sizeof(struct xt_table_info)) / NR_CPUS -
1331                         SMP_CACHE_BYTES)
1332                 return -ENOMEM;
1333         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1334                 return -ENOMEM;
1335
1336         newinfo = xt_alloc_table_info(tmp.size);
1337         if (!newinfo)
1338                 return -ENOMEM;
1339
1340         /* choose the copy that is our node/cpu */
1341         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1342         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1343                            tmp.size) != 0) {
1344                 ret = -EFAULT;
1345                 goto free_newinfo;
1346         }
1347
1348         ret = translate_table(tmp.name, tmp.valid_hooks,
1349                               newinfo, loc_cpu_entry, tmp.size, tmp.num_entries,
1350                               tmp.hook_entry, tmp.underflow);
1351         if (ret != 0)
1352                 goto free_newinfo;
1353
1354         duprintf("ip_tables: Translated table\n");
1355
1356         ret = __do_replace(tmp.name, tmp.valid_hooks,
1357                               newinfo, tmp.num_counters,
1358                               tmp.counters);
1359         if (ret)
1360                 goto free_newinfo_untrans;
1361         return 0;
1362
1363  free_newinfo_untrans:
1364         IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry,NULL);
1365  free_newinfo:
1366         xt_free_table_info(newinfo);
1367         return ret;
1368 }
1369
1370 /* We're lazy, and add to the first CPU; overflow works its fey magic
1371  * and everything is OK. */
1372 static inline int
1373 add_counter_to_entry(struct ipt_entry *e,
1374                      const struct xt_counters addme[],
1375                      unsigned int *i)
1376 {
1377 #if 0
1378         duprintf("add_counter: Entry %u %lu/%lu + %lu/%lu\n",
1379                  *i,
1380                  (long unsigned int)e->counters.pcnt,
1381                  (long unsigned int)e->counters.bcnt,
1382                  (long unsigned int)addme[*i].pcnt,
1383                  (long unsigned int)addme[*i].bcnt);
1384 #endif
1385
1386         ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
1387
1388         (*i)++;
1389         return 0;
1390 }
1391
1392 static int
1393 do_add_counters(void __user *user, unsigned int len, int compat)
1394 {
1395         unsigned int i;
1396         struct xt_counters_info tmp;
1397         struct xt_counters *paddc;
1398         unsigned int num_counters;
1399         char *name;
1400         int size;
1401         void *ptmp;
1402         struct xt_table *t;
1403         struct xt_table_info *private;
1404         int ret = 0;
1405         void *loc_cpu_entry;
1406 #ifdef CONFIG_COMPAT
1407         struct compat_xt_counters_info compat_tmp;
1408
1409         if (compat) {
1410                 ptmp = &compat_tmp;
1411                 size = sizeof(struct compat_xt_counters_info);
1412         } else
1413 #endif
1414         {
1415                 ptmp = &tmp;
1416                 size = sizeof(struct xt_counters_info);
1417         }
1418
1419         if (copy_from_user(ptmp, user, size) != 0)
1420                 return -EFAULT;
1421
1422 #ifdef CONFIG_COMPAT
1423         if (compat) {
1424                 num_counters = compat_tmp.num_counters;
1425                 name = compat_tmp.name;
1426         } else
1427 #endif
1428         {
1429                 num_counters = tmp.num_counters;
1430                 name = tmp.name;
1431         }
1432
1433         if (len != size + num_counters * sizeof(struct xt_counters))
1434                 return -EINVAL;
1435
1436         paddc = vmalloc_node(len - size, numa_node_id());
1437         if (!paddc)
1438                 return -ENOMEM;
1439
1440         if (copy_from_user(paddc, user + size, len - size) != 0) {
1441                 ret = -EFAULT;
1442                 goto free;
1443         }
1444
1445         t = xt_find_table_lock(AF_INET, name);
1446         if (!t || IS_ERR(t)) {
1447                 ret = t ? PTR_ERR(t) : -ENOENT;
1448                 goto free;
1449         }
1450
1451         write_lock_bh(&t->lock);
1452         private = t->private;
1453         if (private->number != num_counters) {
1454                 ret = -EINVAL;
1455                 goto unlock_up_free;
1456         }
1457
1458         i = 0;
1459         /* Choose the copy that is on our node */
1460         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1461         IPT_ENTRY_ITERATE(loc_cpu_entry,
1462                           private->size,
1463                           add_counter_to_entry,
1464                           paddc,
1465                           &i);
1466  unlock_up_free:
1467         write_unlock_bh(&t->lock);
1468         xt_table_unlock(t);
1469         module_put(t->me);
1470  free:
1471         vfree(paddc);
1472
1473         return ret;
1474 }
1475
1476 #ifdef CONFIG_COMPAT
1477 struct compat_ipt_replace {
1478         char                    name[IPT_TABLE_MAXNAMELEN];
1479         u32                     valid_hooks;
1480         u32                     num_entries;
1481         u32                     size;
1482         u32                     hook_entry[NF_IP_NUMHOOKS];
1483         u32                     underflow[NF_IP_NUMHOOKS];
1484         u32                     num_counters;
1485         compat_uptr_t           counters;       /* struct ipt_counters * */
1486         struct compat_ipt_entry entries[0];
1487 };
1488
1489 static inline int compat_copy_match_to_user(struct ipt_entry_match *m,
1490                 void __user **dstptr, compat_uint_t *size)
1491 {
1492         return xt_compat_match_to_user(m, dstptr, size);
1493 }
1494
1495 static int compat_copy_entry_to_user(struct ipt_entry *e,
1496                 void __user **dstptr, compat_uint_t *size)
1497 {
1498         struct ipt_entry_target *t;
1499         struct compat_ipt_entry __user *ce;
1500         u_int16_t target_offset, next_offset;
1501         compat_uint_t origsize;
1502         int ret;
1503
1504         ret = -EFAULT;
1505         origsize = *size;
1506         ce = (struct compat_ipt_entry __user *)*dstptr;
1507         if (copy_to_user(ce, e, sizeof(struct ipt_entry)))
1508                 goto out;
1509
1510         *dstptr += sizeof(struct compat_ipt_entry);
1511         ret = IPT_MATCH_ITERATE(e, compat_copy_match_to_user, dstptr, size);
1512         target_offset = e->target_offset - (origsize - *size);
1513         if (ret)
1514                 goto out;
1515         t = ipt_get_target(e);
1516         ret = xt_compat_target_to_user(t, dstptr, size);
1517         if (ret)
1518                 goto out;
1519         ret = -EFAULT;
1520         next_offset = e->next_offset - (origsize - *size);
1521         if (put_user(target_offset, &ce->target_offset))
1522                 goto out;
1523         if (put_user(next_offset, &ce->next_offset))
1524                 goto out;
1525         return 0;
1526 out:
1527         return ret;
1528 }
1529
1530 static inline int
1531 compat_find_calc_match(struct ipt_entry_match *m,
1532             const char *name,
1533             const struct ipt_ip *ip,
1534             unsigned int hookmask,
1535             int *size, int *i)
1536 {
1537         struct xt_match *match;
1538
1539         match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
1540                                                    m->u.user.revision),
1541                                         "ipt_%s", m->u.user.name);
1542         if (IS_ERR(match) || !match) {
1543                 duprintf("compat_check_calc_match: `%s' not found\n",
1544                                 m->u.user.name);
1545                 return match ? PTR_ERR(match) : -ENOENT;
1546         }
1547         m->u.kernel.match = match;
1548         *size += xt_compat_match_offset(match);
1549
1550         (*i)++;
1551         return 0;
1552 }
1553
1554 static inline int
1555 compat_release_match(struct ipt_entry_match *m, unsigned int *i)
1556 {
1557         if (i && (*i)-- == 0)
1558                 return 1;
1559
1560         module_put(m->u.kernel.match->me);
1561         return 0;
1562 }
1563
1564 static inline int
1565 compat_release_entry(struct ipt_entry *e, unsigned int *i)
1566 {
1567         struct ipt_entry_target *t;
1568
1569         if (i && (*i)-- == 0)
1570                 return 1;
1571
1572         /* Cleanup all matches */
1573         IPT_MATCH_ITERATE(e, compat_release_match, NULL);
1574         t = ipt_get_target(e);
1575         module_put(t->u.kernel.target->me);
1576         return 0;
1577 }
1578
1579 static inline int
1580 check_compat_entry_size_and_hooks(struct ipt_entry *e,
1581                            struct xt_table_info *newinfo,
1582                            unsigned int *size,
1583                            unsigned char *base,
1584                            unsigned char *limit,
1585                            unsigned int *hook_entries,
1586                            unsigned int *underflows,
1587                            unsigned int *i,
1588                            const char *name)
1589 {
1590         struct ipt_entry_target *t;
1591         struct xt_target *target;
1592         unsigned int entry_offset;
1593         int ret, off, h, j;
1594
1595         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1596         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0
1597             || (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1598                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1599                 return -EINVAL;
1600         }
1601
1602         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1603                         sizeof(struct compat_xt_entry_target)) {
1604                 duprintf("checking: element %p size %u\n",
1605                          e, e->next_offset);
1606                 return -EINVAL;
1607         }
1608
1609         ret = check_entry(e, name);
1610         if (ret)
1611                 return ret;
1612
1613         off = 0;
1614         entry_offset = (void *)e - (void *)base;
1615         j = 0;
1616         ret = IPT_MATCH_ITERATE(e, compat_find_calc_match, name, &e->ip,
1617                         e->comefrom, &off, &j);
1618         if (ret != 0)
1619                 goto release_matches;
1620
1621         t = ipt_get_target(e);
1622         target = try_then_request_module(xt_find_target(AF_INET,
1623                                                      t->u.user.name,
1624                                                      t->u.user.revision),
1625                                          "ipt_%s", t->u.user.name);
1626         if (IS_ERR(target) || !target) {
1627                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1628                                                         t->u.user.name);
1629                 ret = target ? PTR_ERR(target) : -ENOENT;
1630                 goto release_matches;
1631         }
1632         t->u.kernel.target = target;
1633
1634         off += xt_compat_target_offset(target);
1635         *size += off;
1636         ret = compat_add_offset(entry_offset, off);
1637         if (ret)
1638                 goto out;
1639
1640         /* Check hooks & underflows */
1641         for (h = 0; h < NF_IP_NUMHOOKS; h++) {
1642                 if ((unsigned char *)e - base == hook_entries[h])
1643                         newinfo->hook_entry[h] = hook_entries[h];
1644                 if ((unsigned char *)e - base == underflows[h])
1645                         newinfo->underflow[h] = underflows[h];
1646         }
1647
1648         /* Clear counters and comefrom */
1649         e->counters = ((struct ipt_counters) { 0, 0 });
1650         e->comefrom = 0;
1651
1652         (*i)++;
1653         return 0;
1654
1655 out:
1656         module_put(t->u.kernel.target->me);
1657 release_matches:
1658         IPT_MATCH_ITERATE(e, compat_release_match, &j);
1659         return ret;
1660 }
1661
1662 static inline int compat_copy_match_from_user(struct ipt_entry_match *m,
1663         void **dstptr, compat_uint_t *size, const char *name,
1664         const struct ipt_ip *ip, unsigned int hookmask)
1665 {
1666         xt_compat_match_from_user(m, dstptr, size);
1667         return 0;
1668 }
1669
1670 static int compat_copy_entry_from_user(struct ipt_entry *e, void **dstptr,
1671         unsigned int *size, const char *name,
1672         struct xt_table_info *newinfo, unsigned char *base)
1673 {
1674         struct ipt_entry_target *t;
1675         struct xt_target *target;
1676         struct ipt_entry *de;
1677         unsigned int origsize;
1678         int ret, h;
1679
1680         ret = 0;
1681         origsize = *size;
1682         de = (struct ipt_entry *)*dstptr;
1683         memcpy(de, e, sizeof(struct ipt_entry));
1684
1685         *dstptr += sizeof(struct compat_ipt_entry);
1686         ret = IPT_MATCH_ITERATE(e, compat_copy_match_from_user, dstptr, size,
1687                         name, &de->ip, de->comefrom);
1688         if (ret)
1689                 return ret;
1690         de->target_offset = e->target_offset - (origsize - *size);
1691         t = ipt_get_target(e);
1692         target = t->u.kernel.target;
1693         xt_compat_target_from_user(t, dstptr, size);
1694
1695         de->next_offset = e->next_offset - (origsize - *size);
1696         for (h = 0; h < NF_IP_NUMHOOKS; h++) {
1697                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1698                         newinfo->hook_entry[h] -= origsize - *size;
1699                 if ((unsigned char *)de - base < newinfo->underflow[h])
1700                         newinfo->underflow[h] -= origsize - *size;
1701         }
1702         return ret;
1703 }
1704
1705 static inline int compat_check_entry(struct ipt_entry *e, const char *name,
1706                                                 unsigned int *i)
1707 {
1708         int j, ret;
1709
1710         j = 0;
1711         ret = IPT_MATCH_ITERATE(e, check_match, name, &e->ip, e->comefrom, &j);
1712         if (ret)
1713                 goto cleanup_matches;
1714
1715         ret = check_target(e, name);
1716         if (ret)
1717                 goto cleanup_matches;
1718
1719         (*i)++;
1720         return 0;
1721
1722  cleanup_matches:
1723         IPT_MATCH_ITERATE(e, cleanup_match, &j);
1724         return ret;
1725 }
1726
1727 static int
1728 translate_compat_table(const char *name,
1729                 unsigned int valid_hooks,
1730                 struct xt_table_info **pinfo,
1731                 void **pentry0,
1732                 unsigned int total_size,
1733                 unsigned int number,
1734                 unsigned int *hook_entries,
1735                 unsigned int *underflows)
1736 {
1737         unsigned int i, j;
1738         struct xt_table_info *newinfo, *info;
1739         void *pos, *entry0, *entry1;
1740         unsigned int size;
1741         int ret;
1742
1743         info = *pinfo;
1744         entry0 = *pentry0;
1745         size = total_size;
1746         info->number = number;
1747
1748         /* Init all hooks to impossible value. */
1749         for (i = 0; i < NF_IP_NUMHOOKS; i++) {
1750                 info->hook_entry[i] = 0xFFFFFFFF;
1751                 info->underflow[i] = 0xFFFFFFFF;
1752         }
1753
1754         duprintf("translate_compat_table: size %u\n", info->size);
1755         j = 0;
1756         xt_compat_lock(AF_INET);
1757         /* Walk through entries, checking offsets. */
1758         ret = IPT_ENTRY_ITERATE(entry0, total_size,
1759                                 check_compat_entry_size_and_hooks,
1760                                 info, &size, entry0,
1761                                 entry0 + total_size,
1762                                 hook_entries, underflows, &j, name);
1763         if (ret != 0)
1764                 goto out_unlock;
1765
1766         ret = -EINVAL;
1767         if (j != number) {
1768                 duprintf("translate_compat_table: %u not %u entries\n",
1769                          j, number);
1770                 goto out_unlock;
1771         }
1772
1773         /* Check hooks all assigned */
1774         for (i = 0; i < NF_IP_NUMHOOKS; i++) {
1775                 /* Only hooks which are valid */
1776                 if (!(valid_hooks & (1 << i)))
1777                         continue;
1778                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1779                         duprintf("Invalid hook entry %u %u\n",
1780                                  i, hook_entries[i]);
1781                         goto out_unlock;
1782                 }
1783                 if (info->underflow[i] == 0xFFFFFFFF) {
1784                         duprintf("Invalid underflow %u %u\n",
1785                                  i, underflows[i]);
1786                         goto out_unlock;
1787                 }
1788         }
1789
1790         ret = -ENOMEM;
1791         newinfo = xt_alloc_table_info(size);
1792         if (!newinfo)
1793                 goto out_unlock;
1794
1795         newinfo->number = number;
1796         for (i = 0; i < NF_IP_NUMHOOKS; i++) {
1797                 newinfo->hook_entry[i] = info->hook_entry[i];
1798                 newinfo->underflow[i] = info->underflow[i];
1799         }
1800         entry1 = newinfo->entries[raw_smp_processor_id()];
1801         pos = entry1;
1802         size =  total_size;
1803         ret = IPT_ENTRY_ITERATE(entry0, total_size,
1804                         compat_copy_entry_from_user, &pos, &size,
1805                         name, newinfo, entry1);
1806         compat_flush_offsets();
1807         xt_compat_unlock(AF_INET);
1808         if (ret)
1809                 goto free_newinfo;
1810
1811         ret = -ELOOP;
1812         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1813                 goto free_newinfo;
1814
1815         i = 0;
1816         ret = IPT_ENTRY_ITERATE(entry1, newinfo->size, compat_check_entry,
1817                                                                 name, &i);
1818         if (ret) {
1819                 j -= i;
1820                 IPT_ENTRY_ITERATE_CONTINUE(entry1, newinfo->size, i,
1821                                                 compat_release_entry, &j);
1822                 IPT_ENTRY_ITERATE(entry1, newinfo->size, cleanup_entry, &i);
1823                 xt_free_table_info(newinfo);
1824                 return ret;
1825         }
1826
1827         /* And one copy for every other CPU */
1828         for_each_possible_cpu(i)
1829                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1830                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1831
1832         *pinfo = newinfo;
1833         *pentry0 = entry1;
1834         xt_free_table_info(info);
1835         return 0;
1836
1837 free_newinfo:
1838         xt_free_table_info(newinfo);
1839 out:
1840         IPT_ENTRY_ITERATE(entry0, total_size, compat_release_entry, &j);
1841         return ret;
1842 out_unlock:
1843         compat_flush_offsets();
1844         xt_compat_unlock(AF_INET);
1845         goto out;
1846 }
1847
1848 static int
1849 compat_do_replace(void __user *user, unsigned int len)
1850 {
1851         int ret;
1852         struct compat_ipt_replace tmp;
1853         struct xt_table_info *newinfo;
1854         void *loc_cpu_entry;
1855
1856         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1857                 return -EFAULT;
1858
1859         /* Hack: Causes ipchains to give correct error msg --RR */
1860         if (len != sizeof(tmp) + tmp.size)
1861                 return -ENOPROTOOPT;
1862
1863         /* overflow check */
1864         if (tmp.size >= (INT_MAX - sizeof(struct xt_table_info)) / NR_CPUS -
1865                         SMP_CACHE_BYTES)
1866                 return -ENOMEM;
1867         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1868                 return -ENOMEM;
1869
1870         newinfo = xt_alloc_table_info(tmp.size);
1871         if (!newinfo)
1872                 return -ENOMEM;
1873
1874         /* choose the copy that is our node/cpu */
1875         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1876         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1877                            tmp.size) != 0) {
1878                 ret = -EFAULT;
1879                 goto free_newinfo;
1880         }
1881
1882         ret = translate_compat_table(tmp.name, tmp.valid_hooks,
1883                               &newinfo, &loc_cpu_entry, tmp.size,
1884                               tmp.num_entries, tmp.hook_entry, tmp.underflow);
1885         if (ret != 0)
1886                 goto free_newinfo;
1887
1888         duprintf("compat_do_replace: Translated table\n");
1889
1890         ret = __do_replace(tmp.name, tmp.valid_hooks,
1891                               newinfo, tmp.num_counters,
1892                               compat_ptr(tmp.counters));
1893         if (ret)
1894                 goto free_newinfo_untrans;
1895         return 0;
1896
1897  free_newinfo_untrans:
1898         IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry,NULL);
1899  free_newinfo:
1900         xt_free_table_info(newinfo);
1901         return ret;
1902 }
1903
1904 static int
1905 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1906                 unsigned int len)
1907 {
1908         int ret;
1909
1910         if (!capable(CAP_NET_ADMIN))
1911                 return -EPERM;
1912
1913         switch (cmd) {
1914         case IPT_SO_SET_REPLACE:
1915                 ret = compat_do_replace(user, len);
1916                 break;
1917
1918         case IPT_SO_SET_ADD_COUNTERS:
1919                 ret = do_add_counters(user, len, 1);
1920                 break;
1921
1922         default:
1923                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1924                 ret = -EINVAL;
1925         }
1926
1927         return ret;
1928 }
1929
1930 struct compat_ipt_get_entries
1931 {
1932         char name[IPT_TABLE_MAXNAMELEN];
1933         compat_uint_t size;
1934         struct compat_ipt_entry entrytable[0];
1935 };
1936
1937 static int compat_copy_entries_to_user(unsigned int total_size,
1938                      struct xt_table *table, void __user *userptr)
1939 {
1940         unsigned int off, num;
1941         struct compat_ipt_entry e;
1942         struct xt_counters *counters;
1943         struct xt_table_info *private = table->private;
1944         void __user *pos;
1945         unsigned int size;
1946         int ret = 0;
1947         void *loc_cpu_entry;
1948
1949         counters = alloc_counters(table);
1950         if (IS_ERR(counters))
1951                 return PTR_ERR(counters);
1952
1953         /* choose the copy that is on our node/cpu, ...
1954          * This choice is lazy (because current thread is
1955          * allowed to migrate to another cpu)
1956          */
1957         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1958         pos = userptr;
1959         size = total_size;
1960         ret = IPT_ENTRY_ITERATE(loc_cpu_entry, total_size,
1961                         compat_copy_entry_to_user, &pos, &size);
1962         if (ret)
1963                 goto free_counters;
1964
1965         /* ... then go back and fix counters and names */
1966         for (off = 0, num = 0; off < size; off += e.next_offset, num++) {
1967                 unsigned int i;
1968                 struct ipt_entry_match m;
1969                 struct ipt_entry_target t;
1970
1971                 ret = -EFAULT;
1972                 if (copy_from_user(&e, userptr + off,
1973                                         sizeof(struct compat_ipt_entry)))
1974                         goto free_counters;
1975                 if (copy_to_user(userptr + off +
1976                         offsetof(struct compat_ipt_entry, counters),
1977                          &counters[num], sizeof(counters[num])))
1978                         goto free_counters;
1979
1980                 for (i = sizeof(struct compat_ipt_entry);
1981                                 i < e.target_offset; i += m.u.match_size) {
1982                         if (copy_from_user(&m, userptr + off + i,
1983                                         sizeof(struct ipt_entry_match)))
1984                                 goto free_counters;
1985                         if (copy_to_user(userptr + off + i +
1986                                 offsetof(struct ipt_entry_match, u.user.name),
1987                                 m.u.kernel.match->name,
1988                                 strlen(m.u.kernel.match->name) + 1))
1989                                 goto free_counters;
1990                 }
1991
1992                 if (copy_from_user(&t, userptr + off + e.target_offset,
1993                                         sizeof(struct ipt_entry_target)))
1994                         goto free_counters;
1995                 if (copy_to_user(userptr + off + e.target_offset +
1996                         offsetof(struct ipt_entry_target, u.user.name),
1997                         t.u.kernel.target->name,
1998                         strlen(t.u.kernel.target->name) + 1))
1999                         goto free_counters;
2000         }
2001         ret = 0;
2002 free_counters:
2003         vfree(counters);
2004         return ret;
2005 }
2006
2007 static int
2008 compat_get_entries(struct compat_ipt_get_entries __user *uptr, int *len)
2009 {
2010         int ret;
2011         struct compat_ipt_get_entries get;
2012         struct xt_table *t;
2013
2014
2015         if (*len < sizeof(get)) {
2016                 duprintf("compat_get_entries: %u < %u\n",
2017                                 *len, (unsigned int)sizeof(get));
2018                 return -EINVAL;
2019         }
2020
2021         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
2022                 return -EFAULT;
2023
2024         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
2025                 duprintf("compat_get_entries: %u != %u\n", *len,
2026                         (unsigned int)(sizeof(struct compat_ipt_get_entries) +
2027                         get.size));
2028                 return -EINVAL;
2029         }
2030
2031         xt_compat_lock(AF_INET);
2032         t = xt_find_table_lock(AF_INET, get.name);
2033         if (t && !IS_ERR(t)) {
2034                 struct xt_table_info *private = t->private;
2035                 struct xt_table_info info;
2036                 duprintf("t->private->number = %u\n",
2037                          private->number);
2038                 ret = compat_table_info(private, &info);
2039                 if (!ret && get.size == info.size) {
2040                         ret = compat_copy_entries_to_user(private->size,
2041                                                    t, uptr->entrytable);
2042                 } else if (!ret) {
2043                         duprintf("compat_get_entries: I've got %u not %u!\n",
2044                                  private->size,
2045                                  get.size);
2046                         ret = -EINVAL;
2047                 }
2048                 compat_flush_offsets();
2049                 module_put(t->me);
2050                 xt_table_unlock(t);
2051         } else
2052                 ret = t ? PTR_ERR(t) : -ENOENT;
2053
2054         xt_compat_unlock(AF_INET);
2055         return ret;
2056 }
2057
2058 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
2059
2060 static int
2061 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2062 {
2063         int ret;
2064
2065         if (!capable(CAP_NET_ADMIN))
2066                 return -EPERM;
2067
2068         switch (cmd) {
2069         case IPT_SO_GET_INFO:
2070                 ret = get_info(user, len, 1);
2071                 break;
2072         case IPT_SO_GET_ENTRIES:
2073                 ret = compat_get_entries(user, len);
2074                 break;
2075         default:
2076                 ret = do_ipt_get_ctl(sk, cmd, user, len);
2077         }
2078         return ret;
2079 }
2080 #endif
2081
2082 static int
2083 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2084 {
2085         int ret;
2086
2087         if (!capable(CAP_NET_ADMIN))
2088                 return -EPERM;
2089
2090         switch (cmd) {
2091         case IPT_SO_SET_REPLACE:
2092                 ret = do_replace(user, len);
2093                 break;
2094
2095         case IPT_SO_SET_ADD_COUNTERS:
2096                 ret = do_add_counters(user, len, 0);
2097                 break;
2098
2099         default:
2100                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2101                 ret = -EINVAL;
2102         }
2103
2104         return ret;
2105 }
2106
2107 static int
2108 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2109 {
2110         int ret;
2111
2112         if (!capable(CAP_NET_ADMIN))
2113                 return -EPERM;
2114
2115         switch (cmd) {
2116         case IPT_SO_GET_INFO:
2117                 ret = get_info(user, len, 0);
2118                 break;
2119
2120         case IPT_SO_GET_ENTRIES:
2121                 ret = get_entries(user, len);
2122                 break;
2123
2124         case IPT_SO_GET_REVISION_MATCH:
2125         case IPT_SO_GET_REVISION_TARGET: {
2126                 struct ipt_get_revision rev;
2127                 int target;
2128
2129                 if (*len != sizeof(rev)) {
2130                         ret = -EINVAL;
2131                         break;
2132                 }
2133                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2134                         ret = -EFAULT;
2135                         break;
2136                 }
2137
2138                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2139                         target = 1;
2140                 else
2141                         target = 0;
2142
2143                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2144                                                          rev.revision,
2145                                                          target, &ret),
2146                                         "ipt_%s", rev.name);
2147                 break;
2148         }
2149
2150         default:
2151                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2152                 ret = -EINVAL;
2153         }
2154
2155         return ret;
2156 }
2157
2158 int ipt_register_table(struct xt_table *table, const struct ipt_replace *repl)
2159 {
2160         int ret;
2161         struct xt_table_info *newinfo;
2162         static struct xt_table_info bootstrap
2163                 = { 0, 0, 0, { 0 }, { 0 }, { } };
2164         void *loc_cpu_entry;
2165
2166         newinfo = xt_alloc_table_info(repl->size);
2167         if (!newinfo)
2168                 return -ENOMEM;
2169
2170         /* choose the copy on our node/cpu
2171          * but dont care of preemption
2172          */
2173         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2174         memcpy(loc_cpu_entry, repl->entries, repl->size);
2175
2176         ret = translate_table(table->name, table->valid_hooks,
2177                               newinfo, loc_cpu_entry, repl->size,
2178                               repl->num_entries,
2179                               repl->hook_entry,
2180                               repl->underflow);
2181         if (ret != 0) {
2182                 xt_free_table_info(newinfo);
2183                 return ret;
2184         }
2185
2186         ret = xt_register_table(table, &bootstrap, newinfo);
2187         if (ret != 0) {
2188                 xt_free_table_info(newinfo);
2189                 return ret;
2190         }
2191
2192         return 0;
2193 }
2194
2195 void ipt_unregister_table(struct xt_table *table)
2196 {
2197         struct xt_table_info *private;
2198         void *loc_cpu_entry;
2199
2200         private = xt_unregister_table(table);
2201
2202         /* Decrease module usage counts and free resources */
2203         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2204         IPT_ENTRY_ITERATE(loc_cpu_entry, private->size, cleanup_entry, NULL);
2205         xt_free_table_info(private);
2206 }
2207
2208 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2209 static inline bool
2210 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2211                      u_int8_t type, u_int8_t code,
2212                      bool invert)
2213 {
2214         return ((test_type == 0xFF) || (type == test_type && code >= min_code && code <= max_code))
2215                 ^ invert;
2216 }
2217
2218 static bool
2219 icmp_match(const struct sk_buff *skb,
2220            const struct net_device *in,
2221            const struct net_device *out,
2222            const struct xt_match *match,
2223            const void *matchinfo,
2224            int offset,
2225            unsigned int protoff,
2226            bool *hotdrop)
2227 {
2228         struct icmphdr _icmph, *ic;
2229         const struct ipt_icmp *icmpinfo = matchinfo;
2230
2231         /* Must not be a fragment. */
2232         if (offset)
2233                 return false;
2234
2235         ic = skb_header_pointer(skb, protoff, sizeof(_icmph), &_icmph);
2236         if (ic == NULL) {
2237                 /* We've been asked to examine this packet, and we
2238                  * can't.  Hence, no choice but to drop.
2239                  */
2240                 duprintf("Dropping evil ICMP tinygram.\n");
2241                 *hotdrop = true;
2242                 return false;
2243         }
2244
2245         return icmp_type_code_match(icmpinfo->type,
2246                                     icmpinfo->code[0],
2247                                     icmpinfo->code[1],
2248                                     ic->type, ic->code,
2249                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2250 }
2251
2252 /* Called when user tries to insert an entry of this type. */
2253 static bool
2254 icmp_checkentry(const char *tablename,
2255            const void *info,
2256            const struct xt_match *match,
2257            void *matchinfo,
2258            unsigned int hook_mask)
2259 {
2260         const struct ipt_icmp *icmpinfo = matchinfo;
2261
2262         /* Must specify no unknown invflags */
2263         return !(icmpinfo->invflags & ~IPT_ICMP_INV);
2264 }
2265
2266 /* The built-in targets: standard (NULL) and error. */
2267 static struct xt_target ipt_standard_target = {
2268         .name           = IPT_STANDARD_TARGET,
2269         .targetsize     = sizeof(int),
2270         .family         = AF_INET,
2271 #ifdef CONFIG_COMPAT
2272         .compatsize     = sizeof(compat_int_t),
2273         .compat_from_user = compat_standard_from_user,
2274         .compat_to_user = compat_standard_to_user,
2275 #endif
2276 };
2277
2278 static struct xt_target ipt_error_target = {
2279         .name           = IPT_ERROR_TARGET,
2280         .target         = ipt_error,
2281         .targetsize     = IPT_FUNCTION_MAXNAMELEN,
2282         .family         = AF_INET,
2283 };
2284
2285 static struct nf_sockopt_ops ipt_sockopts = {
2286         .pf             = PF_INET,
2287         .set_optmin     = IPT_BASE_CTL,
2288         .set_optmax     = IPT_SO_SET_MAX+1,
2289         .set            = do_ipt_set_ctl,
2290 #ifdef CONFIG_COMPAT
2291         .compat_set     = compat_do_ipt_set_ctl,
2292 #endif
2293         .get_optmin     = IPT_BASE_CTL,
2294         .get_optmax     = IPT_SO_GET_MAX+1,
2295         .get            = do_ipt_get_ctl,
2296 #ifdef CONFIG_COMPAT
2297         .compat_get     = compat_do_ipt_get_ctl,
2298 #endif
2299 };
2300
2301 static struct xt_match icmp_matchstruct = {
2302         .name           = "icmp",
2303         .match          = icmp_match,
2304         .matchsize      = sizeof(struct ipt_icmp),
2305         .proto          = IPPROTO_ICMP,
2306         .family         = AF_INET,
2307         .checkentry     = icmp_checkentry,
2308 };
2309
2310 static int __init ip_tables_init(void)
2311 {
2312         int ret;
2313
2314         ret = xt_proto_init(AF_INET);
2315         if (ret < 0)
2316                 goto err1;
2317
2318         /* Noone else will be downing sem now, so we won't sleep */
2319         ret = xt_register_target(&ipt_standard_target);
2320         if (ret < 0)
2321                 goto err2;
2322         ret = xt_register_target(&ipt_error_target);
2323         if (ret < 0)
2324                 goto err3;
2325         ret = xt_register_match(&icmp_matchstruct);
2326         if (ret < 0)
2327                 goto err4;
2328
2329         /* Register setsockopt */
2330         ret = nf_register_sockopt(&ipt_sockopts);
2331         if (ret < 0)
2332                 goto err5;
2333
2334         printk("ip_tables: (C) 2000-2006 Netfilter Core Team\n");
2335         return 0;
2336
2337 err5:
2338         xt_unregister_match(&icmp_matchstruct);
2339 err4:
2340         xt_unregister_target(&ipt_error_target);
2341 err3:
2342         xt_unregister_target(&ipt_standard_target);
2343 err2:
2344         xt_proto_fini(AF_INET);
2345 err1:
2346         return ret;
2347 }
2348
2349 static void __exit ip_tables_fini(void)
2350 {
2351         nf_unregister_sockopt(&ipt_sockopts);
2352
2353         xt_unregister_match(&icmp_matchstruct);
2354         xt_unregister_target(&ipt_error_target);
2355         xt_unregister_target(&ipt_standard_target);
2356
2357         xt_proto_fini(AF_INET);
2358 }
2359
2360 EXPORT_SYMBOL(ipt_register_table);
2361 EXPORT_SYMBOL(ipt_unregister_table);
2362 EXPORT_SYMBOL(ipt_do_table);
2363 module_init(ip_tables_init);
2364 module_exit(ip_tables_fini);