~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ipv4/netfilter/ip_tables.c

Version: ~ [ linux-5.8-rc3 ] ~ [ linux-5.7.5 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.48 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.129 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.185 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.228 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.228 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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  * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net>
  7  *
  8  * This program is free software; you can redistribute it and/or modify
  9  * it under the terms of the GNU General Public License version 2 as
 10  * published by the Free Software Foundation.
 11  */
 12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 13 #include <linux/cache.h>
 14 #include <linux/capability.h>
 15 #include <linux/skbuff.h>
 16 #include <linux/kmod.h>
 17 #include <linux/vmalloc.h>
 18 #include <linux/netdevice.h>
 19 #include <linux/module.h>
 20 #include <linux/icmp.h>
 21 #include <net/ip.h>
 22 #include <net/compat.h>
 23 #include <asm/uaccess.h>
 24 #include <linux/mutex.h>
 25 #include <linux/proc_fs.h>
 26 #include <linux/err.h>
 27 #include <linux/cpumask.h>
 28 
 29 #include <linux/netfilter/x_tables.h>
 30 #include <linux/netfilter_ipv4/ip_tables.h>
 31 #include <net/netfilter/nf_log.h>
 32 #include "../../netfilter/xt_repldata.h"
 33 
 34 MODULE_LICENSE("GPL");
 35 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
 36 MODULE_DESCRIPTION("IPv4 packet filter");
 37 
 38 #ifdef CONFIG_NETFILTER_DEBUG
 39 #define IP_NF_ASSERT(x)         WARN_ON(!(x))
 40 #else
 41 #define IP_NF_ASSERT(x)
 42 #endif
 43 
 44 void *ipt_alloc_initial_table(const struct xt_table *info)
 45 {
 46         return xt_alloc_initial_table(ipt, IPT);
 47 }
 48 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
 49 
 50 /* Returns whether matches rule or not. */
 51 /* Performance critical - called for every packet */
 52 static inline bool
 53 ip_packet_match(const struct iphdr *ip,
 54                 const char *indev,
 55                 const char *outdev,
 56                 const struct ipt_ip *ipinfo,
 57                 int isfrag)
 58 {
 59         unsigned long ret;
 60 
 61 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
 62 
 63         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
 64                   IPT_INV_SRCIP) ||
 65             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
 66                   IPT_INV_DSTIP))
 67                 return false;
 68 
 69         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 70 
 71         if (FWINV(ret != 0, IPT_INV_VIA_IN))
 72                 return false;
 73 
 74         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 75 
 76         if (FWINV(ret != 0, IPT_INV_VIA_OUT))
 77                 return false;
 78 
 79         /* Check specific protocol */
 80         if (ipinfo->proto &&
 81             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO))
 82                 return false;
 83 
 84         /* If we have a fragment rule but the packet is not a fragment
 85          * then we return zero */
 86         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG))
 87                 return false;
 88 
 89         return true;
 90 }
 91 
 92 static bool
 93 ip_checkentry(const struct ipt_ip *ip)
 94 {
 95         if (ip->flags & ~IPT_F_MASK)
 96                 return false;
 97         if (ip->invflags & ~IPT_INV_MASK)
 98                 return false;
 99         return true;
100 }
101 
102 static unsigned int
103 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
104 {
105         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
106 
107         return NF_DROP;
108 }
109 
110 /* Performance critical */
111 static inline struct ipt_entry *
112 get_entry(const void *base, unsigned int offset)
113 {
114         return (struct ipt_entry *)(base + offset);
115 }
116 
117 /* All zeroes == unconditional rule. */
118 /* Mildly perf critical (only if packet tracing is on) */
119 static inline bool unconditional(const struct ipt_entry *e)
120 {
121         static const struct ipt_ip uncond;
122 
123         return e->target_offset == sizeof(struct ipt_entry) &&
124                memcmp(&e->ip, &uncond, sizeof(uncond)) == 0;
125 #undef FWINV
126 }
127 
128 /* for const-correctness */
129 static inline const struct xt_entry_target *
130 ipt_get_target_c(const struct ipt_entry *e)
131 {
132         return ipt_get_target((struct ipt_entry *)e);
133 }
134 
135 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
136 static const char *const hooknames[] = {
137         [NF_INET_PRE_ROUTING]           = "PREROUTING",
138         [NF_INET_LOCAL_IN]              = "INPUT",
139         [NF_INET_FORWARD]               = "FORWARD",
140         [NF_INET_LOCAL_OUT]             = "OUTPUT",
141         [NF_INET_POST_ROUTING]          = "POSTROUTING",
142 };
143 
144 enum nf_ip_trace_comments {
145         NF_IP_TRACE_COMMENT_RULE,
146         NF_IP_TRACE_COMMENT_RETURN,
147         NF_IP_TRACE_COMMENT_POLICY,
148 };
149 
150 static const char *const comments[] = {
151         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
152         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
153         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
154 };
155 
156 static struct nf_loginfo trace_loginfo = {
157         .type = NF_LOG_TYPE_LOG,
158         .u = {
159                 .log = {
160                         .level = 4,
161                         .logflags = NF_LOG_MASK,
162                 },
163         },
164 };
165 
166 /* Mildly perf critical (only if packet tracing is on) */
167 static inline int
168 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
169                       const char *hookname, const char **chainname,
170                       const char **comment, unsigned int *rulenum)
171 {
172         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
173 
174         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
175                 /* Head of user chain: ERROR target with chainname */
176                 *chainname = t->target.data;
177                 (*rulenum) = 0;
178         } else if (s == e) {
179                 (*rulenum)++;
180 
181                 if (unconditional(s) &&
182                     strcmp(t->target.u.kernel.target->name,
183                            XT_STANDARD_TARGET) == 0 &&
184                    t->verdict < 0) {
185                         /* Tail of chains: STANDARD target (return/policy) */
186                         *comment = *chainname == hookname
187                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
188                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
189                 }
190                 return 1;
191         } else
192                 (*rulenum)++;
193 
194         return 0;
195 }
196 
197 static void trace_packet(struct net *net,
198                          const struct sk_buff *skb,
199                          unsigned int hook,
200                          const struct net_device *in,
201                          const struct net_device *out,
202                          const char *tablename,
203                          const struct xt_table_info *private,
204                          const struct ipt_entry *e)
205 {
206         const struct ipt_entry *root;
207         const char *hookname, *chainname, *comment;
208         const struct ipt_entry *iter;
209         unsigned int rulenum = 0;
210 
211         root = get_entry(private->entries, private->hook_entry[hook]);
212 
213         hookname = chainname = hooknames[hook];
214         comment = comments[NF_IP_TRACE_COMMENT_RULE];
215 
216         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
217                 if (get_chainname_rulenum(iter, e, hookname,
218                     &chainname, &comment, &rulenum) != 0)
219                         break;
220 
221         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
222                      "TRACE: %s:%s:%s:%u ",
223                      tablename, chainname, comment, rulenum);
224 }
225 #endif
226 
227 static inline
228 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
229 {
230         return (void *)entry + entry->next_offset;
231 }
232 
233 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
234 unsigned int
235 ipt_do_table(struct sk_buff *skb,
236              const struct nf_hook_state *state,
237              struct xt_table *table)
238 {
239         unsigned int hook = state->hook;
240         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
241         const struct iphdr *ip;
242         /* Initializing verdict to NF_DROP keeps gcc happy. */
243         unsigned int verdict = NF_DROP;
244         const char *indev, *outdev;
245         const void *table_base;
246         struct ipt_entry *e, **jumpstack;
247         unsigned int stackidx, cpu;
248         const struct xt_table_info *private;
249         struct xt_action_param acpar;
250         unsigned int addend;
251 
252         /* Initialization */
253         stackidx = 0;
254         ip = ip_hdr(skb);
255         indev = state->in ? state->in->name : nulldevname;
256         outdev = state->out ? state->out->name : nulldevname;
257         /* We handle fragments by dealing with the first fragment as
258          * if it was a normal packet.  All other fragments are treated
259          * normally, except that they will NEVER match rules that ask
260          * things we don't know, ie. tcp syn flag or ports).  If the
261          * rule is also a fragment-specific rule, non-fragments won't
262          * match it. */
263         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
264         acpar.thoff   = ip_hdrlen(skb);
265         acpar.hotdrop = false;
266         acpar.net     = state->net;
267         acpar.in      = state->in;
268         acpar.out     = state->out;
269         acpar.family  = NFPROTO_IPV4;
270         acpar.hooknum = hook;
271 
272         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
273         local_bh_disable();
274         addend = xt_write_recseq_begin();
275         private = table->private;
276         cpu        = smp_processor_id();
277         /*
278          * Ensure we load private-> members after we've fetched the base
279          * pointer.
280          */
281         smp_read_barrier_depends();
282         table_base = private->entries;
283         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
284 
285         /* Switch to alternate jumpstack if we're being invoked via TEE.
286          * TEE issues XT_CONTINUE verdict on original skb so we must not
287          * clobber the jumpstack.
288          *
289          * For recursion via REJECT or SYNPROXY the stack will be clobbered
290          * but it is no problem since absolute verdict is issued by these.
291          */
292         if (static_key_false(&xt_tee_enabled))
293                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
294 
295         e = get_entry(table_base, private->hook_entry[hook]);
296 
297         do {
298                 const struct xt_entry_target *t;
299                 const struct xt_entry_match *ematch;
300                 struct xt_counters *counter;
301 
302                 IP_NF_ASSERT(e);
303                 if (!ip_packet_match(ip, indev, outdev,
304                     &e->ip, acpar.fragoff)) {
305  no_match:
306                         e = ipt_next_entry(e);
307                         continue;
308                 }
309 
310                 xt_ematch_foreach(ematch, e) {
311                         acpar.match     = ematch->u.kernel.match;
312                         acpar.matchinfo = ematch->data;
313                         if (!acpar.match->match(skb, &acpar))
314                                 goto no_match;
315                 }
316 
317                 counter = xt_get_this_cpu_counter(&e->counters);
318                 ADD_COUNTER(*counter, skb->len, 1);
319 
320                 t = ipt_get_target(e);
321                 IP_NF_ASSERT(t->u.kernel.target);
322 
323 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
324                 /* The packet is traced: log it */
325                 if (unlikely(skb->nf_trace))
326                         trace_packet(state->net, skb, hook, state->in,
327                                      state->out, table->name, private, e);
328 #endif
329                 /* Standard target? */
330                 if (!t->u.kernel.target->target) {
331                         int v;
332 
333                         v = ((struct xt_standard_target *)t)->verdict;
334                         if (v < 0) {
335                                 /* Pop from stack? */
336                                 if (v != XT_RETURN) {
337                                         verdict = (unsigned int)(-v) - 1;
338                                         break;
339                                 }
340                                 if (stackidx == 0) {
341                                         e = get_entry(table_base,
342                                             private->underflow[hook]);
343                                 } else {
344                                         e = jumpstack[--stackidx];
345                                         e = ipt_next_entry(e);
346                                 }
347                                 continue;
348                         }
349                         if (table_base + v != ipt_next_entry(e) &&
350                             !(e->ip.flags & IPT_F_GOTO))
351                                 jumpstack[stackidx++] = e;
352 
353                         e = get_entry(table_base, v);
354                         continue;
355                 }
356 
357                 acpar.target   = t->u.kernel.target;
358                 acpar.targinfo = t->data;
359 
360                 verdict = t->u.kernel.target->target(skb, &acpar);
361                 /* Target might have changed stuff. */
362                 ip = ip_hdr(skb);
363                 if (verdict == XT_CONTINUE)
364                         e = ipt_next_entry(e);
365                 else
366                         /* Verdict */
367                         break;
368         } while (!acpar.hotdrop);
369 
370         xt_write_recseq_end(addend);
371         local_bh_enable();
372 
373         if (acpar.hotdrop)
374                 return NF_DROP;
375         else return verdict;
376 }
377 
378 static bool find_jump_target(const struct xt_table_info *t,
379                              const struct ipt_entry *target)
380 {
381         struct ipt_entry *iter;
382 
383         xt_entry_foreach(iter, t->entries, t->size) {
384                  if (iter == target)
385                         return true;
386         }
387         return false;
388 }
389 
390 /* Figures out from what hook each rule can be called: returns 0 if
391    there are loops.  Puts hook bitmask in comefrom. */
392 static int
393 mark_source_chains(const struct xt_table_info *newinfo,
394                    unsigned int valid_hooks, void *entry0)
395 {
396         unsigned int hook;
397 
398         /* No recursion; use packet counter to save back ptrs (reset
399            to 0 as we leave), and comefrom to save source hook bitmask */
400         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
401                 unsigned int pos = newinfo->hook_entry[hook];
402                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
403 
404                 if (!(valid_hooks & (1 << hook)))
405                         continue;
406 
407                 /* Set initial back pointer. */
408                 e->counters.pcnt = pos;
409 
410                 for (;;) {
411                         const struct xt_standard_target *t
412                                 = (void *)ipt_get_target_c(e);
413                         int visited = e->comefrom & (1 << hook);
414 
415                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
416                                 return 0;
417 
418                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
419 
420                         /* Unconditional return/END. */
421                         if ((unconditional(e) &&
422                              (strcmp(t->target.u.user.name,
423                                      XT_STANDARD_TARGET) == 0) &&
424                              t->verdict < 0) || visited) {
425                                 unsigned int oldpos, size;
426 
427                                 if ((strcmp(t->target.u.user.name,
428                                             XT_STANDARD_TARGET) == 0) &&
429                                     t->verdict < -NF_MAX_VERDICT - 1)
430                                         return 0;
431 
432                                 /* Return: backtrack through the last
433                                    big jump. */
434                                 do {
435                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
436                                         oldpos = pos;
437                                         pos = e->counters.pcnt;
438                                         e->counters.pcnt = 0;
439 
440                                         /* We're at the start. */
441                                         if (pos == oldpos)
442                                                 goto next;
443 
444                                         e = (struct ipt_entry *)
445                                                 (entry0 + pos);
446                                 } while (oldpos == pos + e->next_offset);
447 
448                                 /* Move along one */
449                                 size = e->next_offset;
450                                 e = (struct ipt_entry *)
451                                         (entry0 + pos + size);
452                                 if (pos + size >= newinfo->size)
453                                         return 0;
454                                 e->counters.pcnt = pos;
455                                 pos += size;
456                         } else {
457                                 int newpos = t->verdict;
458 
459                                 if (strcmp(t->target.u.user.name,
460                                            XT_STANDARD_TARGET) == 0 &&
461                                     newpos >= 0) {
462                                         /* This a jump; chase it. */
463                                         e = (struct ipt_entry *)
464                                                 (entry0 + newpos);
465                                         if (!find_jump_target(newinfo, e))
466                                                 return 0;
467                                 } else {
468                                         /* ... this is a fallthru */
469                                         newpos = pos + e->next_offset;
470                                         if (newpos >= newinfo->size)
471                                                 return 0;
472                                 }
473                                 e = (struct ipt_entry *)
474                                         (entry0 + newpos);
475                                 e->counters.pcnt = pos;
476                                 pos = newpos;
477                         }
478                 }
479 next:           ;
480         }
481         return 1;
482 }
483 
484 static void cleanup_match(struct xt_entry_match *m, struct net *net)
485 {
486         struct xt_mtdtor_param par;
487 
488         par.net       = net;
489         par.match     = m->u.kernel.match;
490         par.matchinfo = m->data;
491         par.family    = NFPROTO_IPV4;
492         if (par.match->destroy != NULL)
493                 par.match->destroy(&par);
494         module_put(par.match->me);
495 }
496 
497 static int
498 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
499 {
500         const struct ipt_ip *ip = par->entryinfo;
501 
502         par->match     = m->u.kernel.match;
503         par->matchinfo = m->data;
504 
505         return xt_check_match(par, m->u.match_size - sizeof(*m),
506                               ip->proto, ip->invflags & IPT_INV_PROTO);
507 }
508 
509 static int
510 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
511 {
512         struct xt_match *match;
513         int ret;
514 
515         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
516                                       m->u.user.revision);
517         if (IS_ERR(match))
518                 return PTR_ERR(match);
519         m->u.kernel.match = match;
520 
521         ret = check_match(m, par);
522         if (ret)
523                 goto err;
524 
525         return 0;
526 err:
527         module_put(m->u.kernel.match->me);
528         return ret;
529 }
530 
531 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
532 {
533         struct xt_entry_target *t = ipt_get_target(e);
534         struct xt_tgchk_param par = {
535                 .net       = net,
536                 .table     = name,
537                 .entryinfo = e,
538                 .target    = t->u.kernel.target,
539                 .targinfo  = t->data,
540                 .hook_mask = e->comefrom,
541                 .family    = NFPROTO_IPV4,
542         };
543 
544         return xt_check_target(&par, t->u.target_size - sizeof(*t),
545                                e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
546 }
547 
548 static int
549 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
550                  unsigned int size)
551 {
552         struct xt_entry_target *t;
553         struct xt_target *target;
554         int ret;
555         unsigned int j;
556         struct xt_mtchk_param mtpar;
557         struct xt_entry_match *ematch;
558         unsigned long pcnt;
559 
560         pcnt = xt_percpu_counter_alloc();
561         if (IS_ERR_VALUE(pcnt))
562                 return -ENOMEM;
563         e->counters.pcnt = pcnt;
564 
565         j = 0;
566         mtpar.net       = net;
567         mtpar.table     = name;
568         mtpar.entryinfo = &e->ip;
569         mtpar.hook_mask = e->comefrom;
570         mtpar.family    = NFPROTO_IPV4;
571         xt_ematch_foreach(ematch, e) {
572                 ret = find_check_match(ematch, &mtpar);
573                 if (ret != 0)
574                         goto cleanup_matches;
575                 ++j;
576         }
577 
578         t = ipt_get_target(e);
579         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
580                                         t->u.user.revision);
581         if (IS_ERR(target)) {
582                 ret = PTR_ERR(target);
583                 goto cleanup_matches;
584         }
585         t->u.kernel.target = target;
586 
587         ret = check_target(e, net, name);
588         if (ret)
589                 goto err;
590 
591         return 0;
592  err:
593         module_put(t->u.kernel.target->me);
594  cleanup_matches:
595         xt_ematch_foreach(ematch, e) {
596                 if (j-- == 0)
597                         break;
598                 cleanup_match(ematch, net);
599         }
600 
601         xt_percpu_counter_free(e->counters.pcnt);
602 
603         return ret;
604 }
605 
606 static bool check_underflow(const struct ipt_entry *e)
607 {
608         const struct xt_entry_target *t;
609         unsigned int verdict;
610 
611         if (!unconditional(e))
612                 return false;
613         t = ipt_get_target_c(e);
614         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
615                 return false;
616         verdict = ((struct xt_standard_target *)t)->verdict;
617         verdict = -verdict - 1;
618         return verdict == NF_DROP || verdict == NF_ACCEPT;
619 }
620 
621 static int
622 check_entry_size_and_hooks(struct ipt_entry *e,
623                            struct xt_table_info *newinfo,
624                            const unsigned char *base,
625                            const unsigned char *limit,
626                            const unsigned int *hook_entries,
627                            const unsigned int *underflows,
628                            unsigned int valid_hooks)
629 {
630         unsigned int h;
631         int err;
632 
633         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
634             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
635             (unsigned char *)e + e->next_offset > limit)
636                 return -EINVAL;
637 
638         if (e->next_offset
639             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target))
640                 return -EINVAL;
641 
642         if (!ip_checkentry(&e->ip))
643                 return -EINVAL;
644 
645         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
646                                      e->next_offset);
647         if (err)
648                 return err;
649 
650         /* Check hooks & underflows */
651         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
652                 if (!(valid_hooks & (1 << h)))
653                         continue;
654                 if ((unsigned char *)e - base == hook_entries[h])
655                         newinfo->hook_entry[h] = hook_entries[h];
656                 if ((unsigned char *)e - base == underflows[h]) {
657                         if (!check_underflow(e))
658                                 return -EINVAL;
659 
660                         newinfo->underflow[h] = underflows[h];
661                 }
662         }
663 
664         /* Clear counters and comefrom */
665         e->counters = ((struct xt_counters) { 0, 0 });
666         e->comefrom = 0;
667         return 0;
668 }
669 
670 static void
671 cleanup_entry(struct ipt_entry *e, struct net *net)
672 {
673         struct xt_tgdtor_param par;
674         struct xt_entry_target *t;
675         struct xt_entry_match *ematch;
676 
677         /* Cleanup all matches */
678         xt_ematch_foreach(ematch, e)
679                 cleanup_match(ematch, net);
680         t = ipt_get_target(e);
681 
682         par.net      = net;
683         par.target   = t->u.kernel.target;
684         par.targinfo = t->data;
685         par.family   = NFPROTO_IPV4;
686         if (par.target->destroy != NULL)
687                 par.target->destroy(&par);
688         module_put(par.target->me);
689         xt_percpu_counter_free(e->counters.pcnt);
690 }
691 
692 /* Checks and translates the user-supplied table segment (held in
693    newinfo) */
694 static int
695 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
696                 const struct ipt_replace *repl)
697 {
698         struct ipt_entry *iter;
699         unsigned int i;
700         int ret = 0;
701 
702         newinfo->size = repl->size;
703         newinfo->number = repl->num_entries;
704 
705         /* Init all hooks to impossible value. */
706         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
707                 newinfo->hook_entry[i] = 0xFFFFFFFF;
708                 newinfo->underflow[i] = 0xFFFFFFFF;
709         }
710 
711         i = 0;
712         /* Walk through entries, checking offsets. */
713         xt_entry_foreach(iter, entry0, newinfo->size) {
714                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
715                                                  entry0 + repl->size,
716                                                  repl->hook_entry,
717                                                  repl->underflow,
718                                                  repl->valid_hooks);
719                 if (ret != 0)
720                         return ret;
721                 ++i;
722                 if (strcmp(ipt_get_target(iter)->u.user.name,
723                     XT_ERROR_TARGET) == 0)
724                         ++newinfo->stacksize;
725         }
726 
727         if (i != repl->num_entries)
728                 return -EINVAL;
729 
730         /* Check hooks all assigned */
731         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
732                 /* Only hooks which are valid */
733                 if (!(repl->valid_hooks & (1 << i)))
734                         continue;
735                 if (newinfo->hook_entry[i] == 0xFFFFFFFF)
736                         return -EINVAL;
737                 if (newinfo->underflow[i] == 0xFFFFFFFF)
738                         return -EINVAL;
739         }
740 
741         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
742                 return -ELOOP;
743 
744         /* Finally, each sanity check must pass */
745         i = 0;
746         xt_entry_foreach(iter, entry0, newinfo->size) {
747                 ret = find_check_entry(iter, net, repl->name, repl->size);
748                 if (ret != 0)
749                         break;
750                 ++i;
751         }
752 
753         if (ret != 0) {
754                 xt_entry_foreach(iter, entry0, newinfo->size) {
755                         if (i-- == 0)
756                                 break;
757                         cleanup_entry(iter, net);
758                 }
759                 return ret;
760         }
761 
762         return ret;
763 }
764 
765 static void
766 get_counters(const struct xt_table_info *t,
767              struct xt_counters counters[])
768 {
769         struct ipt_entry *iter;
770         unsigned int cpu;
771         unsigned int i;
772 
773         for_each_possible_cpu(cpu) {
774                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
775 
776                 i = 0;
777                 xt_entry_foreach(iter, t->entries, t->size) {
778                         struct xt_counters *tmp;
779                         u64 bcnt, pcnt;
780                         unsigned int start;
781 
782                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
783                         do {
784                                 start = read_seqcount_begin(s);
785                                 bcnt = tmp->bcnt;
786                                 pcnt = tmp->pcnt;
787                         } while (read_seqcount_retry(s, start));
788 
789                         ADD_COUNTER(counters[i], bcnt, pcnt);
790                         ++i; /* macro does multi eval of i */
791                 }
792         }
793 }
794 
795 static struct xt_counters *alloc_counters(const struct xt_table *table)
796 {
797         unsigned int countersize;
798         struct xt_counters *counters;
799         const struct xt_table_info *private = table->private;
800 
801         /* We need atomic snapshot of counters: rest doesn't change
802            (other than comefrom, which userspace doesn't care
803            about). */
804         countersize = sizeof(struct xt_counters) * private->number;
805         counters = vzalloc(countersize);
806 
807         if (counters == NULL)
808                 return ERR_PTR(-ENOMEM);
809 
810         get_counters(private, counters);
811 
812         return counters;
813 }
814 
815 static int
816 copy_entries_to_user(unsigned int total_size,
817                      const struct xt_table *table,
818                      void __user *userptr)
819 {
820         unsigned int off, num;
821         const struct ipt_entry *e;
822         struct xt_counters *counters;
823         const struct xt_table_info *private = table->private;
824         int ret = 0;
825         const void *loc_cpu_entry;
826 
827         counters = alloc_counters(table);
828         if (IS_ERR(counters))
829                 return PTR_ERR(counters);
830 
831         loc_cpu_entry = private->entries;
832         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
833                 ret = -EFAULT;
834                 goto free_counters;
835         }
836 
837         /* FIXME: use iterator macros --RR */
838         /* ... then go back and fix counters and names */
839         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
840                 unsigned int i;
841                 const struct xt_entry_match *m;
842                 const struct xt_entry_target *t;
843 
844                 e = (struct ipt_entry *)(loc_cpu_entry + off);
845                 if (copy_to_user(userptr + off
846                                  + offsetof(struct ipt_entry, counters),
847                                  &counters[num],
848                                  sizeof(counters[num])) != 0) {
849                         ret = -EFAULT;
850                         goto free_counters;
851                 }
852 
853                 for (i = sizeof(struct ipt_entry);
854                      i < e->target_offset;
855                      i += m->u.match_size) {
856                         m = (void *)e + i;
857 
858                         if (copy_to_user(userptr + off + i
859                                          + offsetof(struct xt_entry_match,
860                                                     u.user.name),
861                                          m->u.kernel.match->name,
862                                          strlen(m->u.kernel.match->name)+1)
863                             != 0) {
864                                 ret = -EFAULT;
865                                 goto free_counters;
866                         }
867                 }
868 
869                 t = ipt_get_target_c(e);
870                 if (copy_to_user(userptr + off + e->target_offset
871                                  + offsetof(struct xt_entry_target,
872                                             u.user.name),
873                                  t->u.kernel.target->name,
874                                  strlen(t->u.kernel.target->name)+1) != 0) {
875                         ret = -EFAULT;
876                         goto free_counters;
877                 }
878         }
879 
880  free_counters:
881         vfree(counters);
882         return ret;
883 }
884 
885 #ifdef CONFIG_COMPAT
886 static void compat_standard_from_user(void *dst, const void *src)
887 {
888         int v = *(compat_int_t *)src;
889 
890         if (v > 0)
891                 v += xt_compat_calc_jump(AF_INET, v);
892         memcpy(dst, &v, sizeof(v));
893 }
894 
895 static int compat_standard_to_user(void __user *dst, const void *src)
896 {
897         compat_int_t cv = *(int *)src;
898 
899         if (cv > 0)
900                 cv -= xt_compat_calc_jump(AF_INET, cv);
901         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
902 }
903 
904 static int compat_calc_entry(const struct ipt_entry *e,
905                              const struct xt_table_info *info,
906                              const void *base, struct xt_table_info *newinfo)
907 {
908         const struct xt_entry_match *ematch;
909         const struct xt_entry_target *t;
910         unsigned int entry_offset;
911         int off, i, ret;
912 
913         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
914         entry_offset = (void *)e - base;
915         xt_ematch_foreach(ematch, e)
916                 off += xt_compat_match_offset(ematch->u.kernel.match);
917         t = ipt_get_target_c(e);
918         off += xt_compat_target_offset(t->u.kernel.target);
919         newinfo->size -= off;
920         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
921         if (ret)
922                 return ret;
923 
924         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
925                 if (info->hook_entry[i] &&
926                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
927                         newinfo->hook_entry[i] -= off;
928                 if (info->underflow[i] &&
929                     (e < (struct ipt_entry *)(base + info->underflow[i])))
930                         newinfo->underflow[i] -= off;
931         }
932         return 0;
933 }
934 
935 static int compat_table_info(const struct xt_table_info *info,
936                              struct xt_table_info *newinfo)
937 {
938         struct ipt_entry *iter;
939         const void *loc_cpu_entry;
940         int ret;
941 
942         if (!newinfo || !info)
943                 return -EINVAL;
944 
945         /* we dont care about newinfo->entries */
946         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
947         newinfo->initial_entries = 0;
948         loc_cpu_entry = info->entries;
949         xt_compat_init_offsets(AF_INET, info->number);
950         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
951                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
952                 if (ret != 0)
953                         return ret;
954         }
955         return 0;
956 }
957 #endif
958 
959 static int get_info(struct net *net, void __user *user,
960                     const int *len, int compat)
961 {
962         char name[XT_TABLE_MAXNAMELEN];
963         struct xt_table *t;
964         int ret;
965 
966         if (*len != sizeof(struct ipt_getinfo))
967                 return -EINVAL;
968 
969         if (copy_from_user(name, user, sizeof(name)) != 0)
970                 return -EFAULT;
971 
972         name[XT_TABLE_MAXNAMELEN-1] = '\0';
973 #ifdef CONFIG_COMPAT
974         if (compat)
975                 xt_compat_lock(AF_INET);
976 #endif
977         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
978                                     "iptable_%s", name);
979         if (!IS_ERR_OR_NULL(t)) {
980                 struct ipt_getinfo info;
981                 const struct xt_table_info *private = t->private;
982 #ifdef CONFIG_COMPAT
983                 struct xt_table_info tmp;
984 
985                 if (compat) {
986                         ret = compat_table_info(private, &tmp);
987                         xt_compat_flush_offsets(AF_INET);
988                         private = &tmp;
989                 }
990 #endif
991                 memset(&info, 0, sizeof(info));
992                 info.valid_hooks = t->valid_hooks;
993                 memcpy(info.hook_entry, private->hook_entry,
994                        sizeof(info.hook_entry));
995                 memcpy(info.underflow, private->underflow,
996                        sizeof(info.underflow));
997                 info.num_entries = private->number;
998                 info.size = private->size;
999                 strcpy(info.name, name);
1000 
1001                 if (copy_to_user(user, &info, *len) != 0)
1002                         ret = -EFAULT;
1003                 else
1004                         ret = 0;
1005 
1006                 xt_table_unlock(t);
1007                 module_put(t->me);
1008         } else
1009                 ret = t ? PTR_ERR(t) : -ENOENT;
1010 #ifdef CONFIG_COMPAT
1011         if (compat)
1012                 xt_compat_unlock(AF_INET);
1013 #endif
1014         return ret;
1015 }
1016 
1017 static int
1018 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1019             const int *len)
1020 {
1021         int ret;
1022         struct ipt_get_entries get;
1023         struct xt_table *t;
1024 
1025         if (*len < sizeof(get))
1026                 return -EINVAL;
1027         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1028                 return -EFAULT;
1029         if (*len != sizeof(struct ipt_get_entries) + get.size)
1030                 return -EINVAL;
1031         get.name[sizeof(get.name) - 1] = '\0';
1032 
1033         t = xt_find_table_lock(net, AF_INET, get.name);
1034         if (!IS_ERR_OR_NULL(t)) {
1035                 const struct xt_table_info *private = t->private;
1036                 if (get.size == private->size)
1037                         ret = copy_entries_to_user(private->size,
1038                                                    t, uptr->entrytable);
1039                 else
1040                         ret = -EAGAIN;
1041 
1042                 module_put(t->me);
1043                 xt_table_unlock(t);
1044         } else
1045                 ret = t ? PTR_ERR(t) : -ENOENT;
1046 
1047         return ret;
1048 }
1049 
1050 static int
1051 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1052              struct xt_table_info *newinfo, unsigned int num_counters,
1053              void __user *counters_ptr)
1054 {
1055         int ret;
1056         struct xt_table *t;
1057         struct xt_table_info *oldinfo;
1058         struct xt_counters *counters;
1059         struct ipt_entry *iter;
1060 
1061         ret = 0;
1062         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1063         if (!counters) {
1064                 ret = -ENOMEM;
1065                 goto out;
1066         }
1067 
1068         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1069                                     "iptable_%s", name);
1070         if (IS_ERR_OR_NULL(t)) {
1071                 ret = t ? PTR_ERR(t) : -ENOENT;
1072                 goto free_newinfo_counters_untrans;
1073         }
1074 
1075         /* You lied! */
1076         if (valid_hooks != t->valid_hooks) {
1077                 ret = -EINVAL;
1078                 goto put_module;
1079         }
1080 
1081         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1082         if (!oldinfo)
1083                 goto put_module;
1084 
1085         /* Update module usage count based on number of rules */
1086         if ((oldinfo->number > oldinfo->initial_entries) ||
1087             (newinfo->number <= oldinfo->initial_entries))
1088                 module_put(t->me);
1089         if ((oldinfo->number > oldinfo->initial_entries) &&
1090             (newinfo->number <= oldinfo->initial_entries))
1091                 module_put(t->me);
1092 
1093         /* Get the old counters, and synchronize with replace */
1094         get_counters(oldinfo, counters);
1095 
1096         /* Decrease module usage counts and free resource */
1097         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1098                 cleanup_entry(iter, net);
1099 
1100         xt_free_table_info(oldinfo);
1101         if (copy_to_user(counters_ptr, counters,
1102                          sizeof(struct xt_counters) * num_counters) != 0) {
1103                 /* Silent error, can't fail, new table is already in place */
1104                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1105         }
1106         vfree(counters);
1107         xt_table_unlock(t);
1108         return ret;
1109 
1110  put_module:
1111         module_put(t->me);
1112         xt_table_unlock(t);
1113  free_newinfo_counters_untrans:
1114         vfree(counters);
1115  out:
1116         return ret;
1117 }
1118 
1119 static int
1120 do_replace(struct net *net, const void __user *user, unsigned int len)
1121 {
1122         int ret;
1123         struct ipt_replace tmp;
1124         struct xt_table_info *newinfo;
1125         void *loc_cpu_entry;
1126         struct ipt_entry *iter;
1127 
1128         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1129                 return -EFAULT;
1130 
1131         /* overflow check */
1132         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1133                 return -ENOMEM;
1134         if (tmp.num_counters == 0)
1135                 return -EINVAL;
1136 
1137         tmp.name[sizeof(tmp.name)-1] = 0;
1138 
1139         newinfo = xt_alloc_table_info(tmp.size);
1140         if (!newinfo)
1141                 return -ENOMEM;
1142 
1143         loc_cpu_entry = newinfo->entries;
1144         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1145                            tmp.size) != 0) {
1146                 ret = -EFAULT;
1147                 goto free_newinfo;
1148         }
1149 
1150         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1151         if (ret != 0)
1152                 goto free_newinfo;
1153 
1154         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1155                            tmp.num_counters, tmp.counters);
1156         if (ret)
1157                 goto free_newinfo_untrans;
1158         return 0;
1159 
1160  free_newinfo_untrans:
1161         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1162                 cleanup_entry(iter, net);
1163  free_newinfo:
1164         xt_free_table_info(newinfo);
1165         return ret;
1166 }
1167 
1168 static int
1169 do_add_counters(struct net *net, const void __user *user,
1170                 unsigned int len, int compat)
1171 {
1172         unsigned int i;
1173         struct xt_counters_info tmp;
1174         struct xt_counters *paddc;
1175         struct xt_table *t;
1176         const struct xt_table_info *private;
1177         int ret = 0;
1178         struct ipt_entry *iter;
1179         unsigned int addend;
1180 
1181         paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1182         if (IS_ERR(paddc))
1183                 return PTR_ERR(paddc);
1184 
1185         t = xt_find_table_lock(net, AF_INET, tmp.name);
1186         if (IS_ERR_OR_NULL(t)) {
1187                 ret = t ? PTR_ERR(t) : -ENOENT;
1188                 goto free;
1189         }
1190 
1191         local_bh_disable();
1192         private = t->private;
1193         if (private->number != tmp.num_counters) {
1194                 ret = -EINVAL;
1195                 goto unlock_up_free;
1196         }
1197 
1198         i = 0;
1199         addend = xt_write_recseq_begin();
1200         xt_entry_foreach(iter, private->entries, private->size) {
1201                 struct xt_counters *tmp;
1202 
1203                 tmp = xt_get_this_cpu_counter(&iter->counters);
1204                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1205                 ++i;
1206         }
1207         xt_write_recseq_end(addend);
1208  unlock_up_free:
1209         local_bh_enable();
1210         xt_table_unlock(t);
1211         module_put(t->me);
1212  free:
1213         vfree(paddc);
1214 
1215         return ret;
1216 }
1217 
1218 #ifdef CONFIG_COMPAT
1219 struct compat_ipt_replace {
1220         char                    name[XT_TABLE_MAXNAMELEN];
1221         u32                     valid_hooks;
1222         u32                     num_entries;
1223         u32                     size;
1224         u32                     hook_entry[NF_INET_NUMHOOKS];
1225         u32                     underflow[NF_INET_NUMHOOKS];
1226         u32                     num_counters;
1227         compat_uptr_t           counters;       /* struct xt_counters * */
1228         struct compat_ipt_entry entries[0];
1229 };
1230 
1231 static int
1232 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1233                           unsigned int *size, struct xt_counters *counters,
1234                           unsigned int i)
1235 {
1236         struct xt_entry_target *t;
1237         struct compat_ipt_entry __user *ce;
1238         u_int16_t target_offset, next_offset;
1239         compat_uint_t origsize;
1240         const struct xt_entry_match *ematch;
1241         int ret = 0;
1242 
1243         origsize = *size;
1244         ce = (struct compat_ipt_entry __user *)*dstptr;
1245         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1246             copy_to_user(&ce->counters, &counters[i],
1247             sizeof(counters[i])) != 0)
1248                 return -EFAULT;
1249 
1250         *dstptr += sizeof(struct compat_ipt_entry);
1251         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1252 
1253         xt_ematch_foreach(ematch, e) {
1254                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1255                 if (ret != 0)
1256                         return ret;
1257         }
1258         target_offset = e->target_offset - (origsize - *size);
1259         t = ipt_get_target(e);
1260         ret = xt_compat_target_to_user(t, dstptr, size);
1261         if (ret)
1262                 return ret;
1263         next_offset = e->next_offset - (origsize - *size);
1264         if (put_user(target_offset, &ce->target_offset) != 0 ||
1265             put_user(next_offset, &ce->next_offset) != 0)
1266                 return -EFAULT;
1267         return 0;
1268 }
1269 
1270 static int
1271 compat_find_calc_match(struct xt_entry_match *m,
1272                        const struct ipt_ip *ip,
1273                        int *size)
1274 {
1275         struct xt_match *match;
1276 
1277         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1278                                       m->u.user.revision);
1279         if (IS_ERR(match))
1280                 return PTR_ERR(match);
1281 
1282         m->u.kernel.match = match;
1283         *size += xt_compat_match_offset(match);
1284         return 0;
1285 }
1286 
1287 static void compat_release_entry(struct compat_ipt_entry *e)
1288 {
1289         struct xt_entry_target *t;
1290         struct xt_entry_match *ematch;
1291 
1292         /* Cleanup all matches */
1293         xt_ematch_foreach(ematch, e)
1294                 module_put(ematch->u.kernel.match->me);
1295         t = compat_ipt_get_target(e);
1296         module_put(t->u.kernel.target->me);
1297 }
1298 
1299 static int
1300 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1301                                   struct xt_table_info *newinfo,
1302                                   unsigned int *size,
1303                                   const unsigned char *base,
1304                                   const unsigned char *limit)
1305 {
1306         struct xt_entry_match *ematch;
1307         struct xt_entry_target *t;
1308         struct xt_target *target;
1309         unsigned int entry_offset;
1310         unsigned int j;
1311         int ret, off;
1312 
1313         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1314             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1315             (unsigned char *)e + e->next_offset > limit)
1316                 return -EINVAL;
1317 
1318         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1319                              sizeof(struct compat_xt_entry_target))
1320                 return -EINVAL;
1321 
1322         if (!ip_checkentry(&e->ip))
1323                 return -EINVAL;
1324 
1325         ret = xt_compat_check_entry_offsets(e, e->elems,
1326                                             e->target_offset, e->next_offset);
1327         if (ret)
1328                 return ret;
1329 
1330         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1331         entry_offset = (void *)e - (void *)base;
1332         j = 0;
1333         xt_ematch_foreach(ematch, e) {
1334                 ret = compat_find_calc_match(ematch, &e->ip, &off);
1335                 if (ret != 0)
1336                         goto release_matches;
1337                 ++j;
1338         }
1339 
1340         t = compat_ipt_get_target(e);
1341         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1342                                         t->u.user.revision);
1343         if (IS_ERR(target)) {
1344                 ret = PTR_ERR(target);
1345                 goto release_matches;
1346         }
1347         t->u.kernel.target = target;
1348 
1349         off += xt_compat_target_offset(target);
1350         *size += off;
1351         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1352         if (ret)
1353                 goto out;
1354 
1355         return 0;
1356 
1357 out:
1358         module_put(t->u.kernel.target->me);
1359 release_matches:
1360         xt_ematch_foreach(ematch, e) {
1361                 if (j-- == 0)
1362                         break;
1363                 module_put(ematch->u.kernel.match->me);
1364         }
1365         return ret;
1366 }
1367 
1368 static void
1369 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1370                             unsigned int *size,
1371                             struct xt_table_info *newinfo, unsigned char *base)
1372 {
1373         struct xt_entry_target *t;
1374         struct xt_target *target;
1375         struct ipt_entry *de;
1376         unsigned int origsize;
1377         int h;
1378         struct xt_entry_match *ematch;
1379 
1380         origsize = *size;
1381         de = (struct ipt_entry *)*dstptr;
1382         memcpy(de, e, sizeof(struct ipt_entry));
1383         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1384 
1385         *dstptr += sizeof(struct ipt_entry);
1386         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1387 
1388         xt_ematch_foreach(ematch, e)
1389                 xt_compat_match_from_user(ematch, dstptr, size);
1390 
1391         de->target_offset = e->target_offset - (origsize - *size);
1392         t = compat_ipt_get_target(e);
1393         target = t->u.kernel.target;
1394         xt_compat_target_from_user(t, dstptr, size);
1395 
1396         de->next_offset = e->next_offset - (origsize - *size);
1397 
1398         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1399                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1400                         newinfo->hook_entry[h] -= origsize - *size;
1401                 if ((unsigned char *)de - base < newinfo->underflow[h])
1402                         newinfo->underflow[h] -= origsize - *size;
1403         }
1404 }
1405 
1406 static int
1407 translate_compat_table(struct net *net,
1408                        struct xt_table_info **pinfo,
1409                        void **pentry0,
1410                        const struct compat_ipt_replace *compatr)
1411 {
1412         unsigned int i, j;
1413         struct xt_table_info *newinfo, *info;
1414         void *pos, *entry0, *entry1;
1415         struct compat_ipt_entry *iter0;
1416         struct ipt_replace repl;
1417         unsigned int size;
1418         int ret;
1419 
1420         info = *pinfo;
1421         entry0 = *pentry0;
1422         size = compatr->size;
1423         info->number = compatr->num_entries;
1424 
1425         j = 0;
1426         xt_compat_lock(AF_INET);
1427         xt_compat_init_offsets(AF_INET, compatr->num_entries);
1428         /* Walk through entries, checking offsets. */
1429         xt_entry_foreach(iter0, entry0, compatr->size) {
1430                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1431                                                         entry0,
1432                                                         entry0 + compatr->size);
1433                 if (ret != 0)
1434                         goto out_unlock;
1435                 ++j;
1436         }
1437 
1438         ret = -EINVAL;
1439         if (j != compatr->num_entries)
1440                 goto out_unlock;
1441 
1442         ret = -ENOMEM;
1443         newinfo = xt_alloc_table_info(size);
1444         if (!newinfo)
1445                 goto out_unlock;
1446 
1447         newinfo->number = compatr->num_entries;
1448         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1449                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1450                 newinfo->underflow[i] = compatr->underflow[i];
1451         }
1452         entry1 = newinfo->entries;
1453         pos = entry1;
1454         size = compatr->size;
1455         xt_entry_foreach(iter0, entry0, compatr->size)
1456                 compat_copy_entry_from_user(iter0, &pos, &size,
1457                                             newinfo, entry1);
1458 
1459         /* all module references in entry0 are now gone.
1460          * entry1/newinfo contains a 64bit ruleset that looks exactly as
1461          * generated by 64bit userspace.
1462          *
1463          * Call standard translate_table() to validate all hook_entrys,
1464          * underflows, check for loops, etc.
1465          */
1466         xt_compat_flush_offsets(AF_INET);
1467         xt_compat_unlock(AF_INET);
1468 
1469         memcpy(&repl, compatr, sizeof(*compatr));
1470 
1471         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1472                 repl.hook_entry[i] = newinfo->hook_entry[i];
1473                 repl.underflow[i] = newinfo->underflow[i];
1474         }
1475 
1476         repl.num_counters = 0;
1477         repl.counters = NULL;
1478         repl.size = newinfo->size;
1479         ret = translate_table(net, newinfo, entry1, &repl);
1480         if (ret)
1481                 goto free_newinfo;
1482 
1483         *pinfo = newinfo;
1484         *pentry0 = entry1;
1485         xt_free_table_info(info);
1486         return 0;
1487 
1488 free_newinfo:
1489         xt_free_table_info(newinfo);
1490         return ret;
1491 out_unlock:
1492         xt_compat_flush_offsets(AF_INET);
1493         xt_compat_unlock(AF_INET);
1494         xt_entry_foreach(iter0, entry0, compatr->size) {
1495                 if (j-- == 0)
1496                         break;
1497                 compat_release_entry(iter0);
1498         }
1499         return ret;
1500 }
1501 
1502 static int
1503 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1504 {
1505         int ret;
1506         struct compat_ipt_replace tmp;
1507         struct xt_table_info *newinfo;
1508         void *loc_cpu_entry;
1509         struct ipt_entry *iter;
1510 
1511         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1512                 return -EFAULT;
1513 
1514         /* overflow check */
1515         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1516                 return -ENOMEM;
1517         if (tmp.num_counters == 0)
1518                 return -EINVAL;
1519 
1520         tmp.name[sizeof(tmp.name)-1] = 0;
1521 
1522         newinfo = xt_alloc_table_info(tmp.size);
1523         if (!newinfo)
1524                 return -ENOMEM;
1525 
1526         loc_cpu_entry = newinfo->entries;
1527         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1528                            tmp.size) != 0) {
1529                 ret = -EFAULT;
1530                 goto free_newinfo;
1531         }
1532 
1533         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1534         if (ret != 0)
1535                 goto free_newinfo;
1536 
1537         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1538                            tmp.num_counters, compat_ptr(tmp.counters));
1539         if (ret)
1540                 goto free_newinfo_untrans;
1541         return 0;
1542 
1543  free_newinfo_untrans:
1544         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1545                 cleanup_entry(iter, net);
1546  free_newinfo:
1547         xt_free_table_info(newinfo);
1548         return ret;
1549 }
1550 
1551 static int
1552 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1553                       unsigned int len)
1554 {
1555         int ret;
1556 
1557         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1558                 return -EPERM;
1559 
1560         switch (cmd) {
1561         case IPT_SO_SET_REPLACE:
1562                 ret = compat_do_replace(sock_net(sk), user, len);
1563                 break;
1564 
1565         case IPT_SO_SET_ADD_COUNTERS:
1566                 ret = do_add_counters(sock_net(sk), user, len, 1);
1567                 break;
1568 
1569         default:
1570                 ret = -EINVAL;
1571         }
1572 
1573         return ret;
1574 }
1575 
1576 struct compat_ipt_get_entries {
1577         char name[XT_TABLE_MAXNAMELEN];
1578         compat_uint_t size;
1579         struct compat_ipt_entry entrytable[0];
1580 };
1581 
1582 static int
1583 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1584                             void __user *userptr)
1585 {
1586         struct xt_counters *counters;
1587         const struct xt_table_info *private = table->private;
1588         void __user *pos;
1589         unsigned int size;
1590         int ret = 0;
1591         unsigned int i = 0;
1592         struct ipt_entry *iter;
1593 
1594         counters = alloc_counters(table);
1595         if (IS_ERR(counters))
1596                 return PTR_ERR(counters);
1597 
1598         pos = userptr;
1599         size = total_size;
1600         xt_entry_foreach(iter, private->entries, total_size) {
1601                 ret = compat_copy_entry_to_user(iter, &pos,
1602                                                 &size, counters, i++);
1603                 if (ret != 0)
1604                         break;
1605         }
1606 
1607         vfree(counters);
1608         return ret;
1609 }
1610 
1611 static int
1612 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1613                    int *len)
1614 {
1615         int ret;
1616         struct compat_ipt_get_entries get;
1617         struct xt_table *t;
1618 
1619         if (*len < sizeof(get))
1620                 return -EINVAL;
1621 
1622         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1623                 return -EFAULT;
1624 
1625         if (*len != sizeof(struct compat_ipt_get_entries) + get.size)
1626                 return -EINVAL;
1627 
1628         get.name[sizeof(get.name) - 1] = '\0';
1629 
1630         xt_compat_lock(AF_INET);
1631         t = xt_find_table_lock(net, AF_INET, get.name);
1632         if (!IS_ERR_OR_NULL(t)) {
1633                 const struct xt_table_info *private = t->private;
1634                 struct xt_table_info info;
1635                 ret = compat_table_info(private, &info);
1636                 if (!ret && get.size == info.size)
1637                         ret = compat_copy_entries_to_user(private->size,
1638                                                           t, uptr->entrytable);
1639                 else if (!ret)
1640                         ret = -EAGAIN;
1641 
1642                 xt_compat_flush_offsets(AF_INET);
1643                 module_put(t->me);
1644                 xt_table_unlock(t);
1645         } else
1646                 ret = t ? PTR_ERR(t) : -ENOENT;
1647 
1648         xt_compat_unlock(AF_INET);
1649         return ret;
1650 }
1651 
1652 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1653 
1654 static int
1655 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1656 {
1657         int ret;
1658 
1659         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1660                 return -EPERM;
1661 
1662         switch (cmd) {
1663         case IPT_SO_GET_INFO:
1664                 ret = get_info(sock_net(sk), user, len, 1);
1665                 break;
1666         case IPT_SO_GET_ENTRIES:
1667                 ret = compat_get_entries(sock_net(sk), user, len);
1668                 break;
1669         default:
1670                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1671         }
1672         return ret;
1673 }
1674 #endif
1675 
1676 static int
1677 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1678 {
1679         int ret;
1680 
1681         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1682                 return -EPERM;
1683 
1684         switch (cmd) {
1685         case IPT_SO_SET_REPLACE:
1686                 ret = do_replace(sock_net(sk), user, len);
1687                 break;
1688 
1689         case IPT_SO_SET_ADD_COUNTERS:
1690                 ret = do_add_counters(sock_net(sk), user, len, 0);
1691                 break;
1692 
1693         default:
1694                 ret = -EINVAL;
1695         }
1696 
1697         return ret;
1698 }
1699 
1700 static int
1701 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1702 {
1703         int ret;
1704 
1705         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1706                 return -EPERM;
1707 
1708         switch (cmd) {
1709         case IPT_SO_GET_INFO:
1710                 ret = get_info(sock_net(sk), user, len, 0);
1711                 break;
1712 
1713         case IPT_SO_GET_ENTRIES:
1714                 ret = get_entries(sock_net(sk), user, len);
1715                 break;
1716 
1717         case IPT_SO_GET_REVISION_MATCH:
1718         case IPT_SO_GET_REVISION_TARGET: {
1719                 struct xt_get_revision rev;
1720                 int target;
1721 
1722                 if (*len != sizeof(rev)) {
1723                         ret = -EINVAL;
1724                         break;
1725                 }
1726                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1727                         ret = -EFAULT;
1728                         break;
1729                 }
1730                 rev.name[sizeof(rev.name)-1] = 0;
1731 
1732                 if (cmd == IPT_SO_GET_REVISION_TARGET)
1733                         target = 1;
1734                 else
1735                         target = 0;
1736 
1737                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
1738                                                          rev.revision,
1739                                                          target, &ret),
1740                                         "ipt_%s", rev.name);
1741                 break;
1742         }
1743 
1744         default:
1745                 ret = -EINVAL;
1746         }
1747 
1748         return ret;
1749 }
1750 
1751 static void __ipt_unregister_table(struct net *net, struct xt_table *table)
1752 {
1753         struct xt_table_info *private;
1754         void *loc_cpu_entry;
1755         struct module *table_owner = table->me;
1756         struct ipt_entry *iter;
1757 
1758         private = xt_unregister_table(table);
1759 
1760         /* Decrease module usage counts and free resources */
1761         loc_cpu_entry = private->entries;
1762         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1763                 cleanup_entry(iter, net);
1764         if (private->number > private->initial_entries)
1765                 module_put(table_owner);
1766         xt_free_table_info(private);
1767 }
1768 
1769 int ipt_register_table(struct net *net, const struct xt_table *table,
1770                        const struct ipt_replace *repl,
1771                        const struct nf_hook_ops *ops, struct xt_table **res)
1772 {
1773         int ret;
1774         struct xt_table_info *newinfo;
1775         struct xt_table_info bootstrap = {0};
1776         void *loc_cpu_entry;
1777         struct xt_table *new_table;
1778 
1779         newinfo = xt_alloc_table_info(repl->size);
1780         if (!newinfo)
1781                 return -ENOMEM;
1782 
1783         loc_cpu_entry = newinfo->entries;
1784         memcpy(loc_cpu_entry, repl->entries, repl->size);
1785 
1786         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1787         if (ret != 0)
1788                 goto out_free;
1789 
1790         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1791         if (IS_ERR(new_table)) {
1792                 ret = PTR_ERR(new_table);
1793                 goto out_free;
1794         }
1795 
1796         /* set res now, will see skbs right after nf_register_net_hooks */
1797         WRITE_ONCE(*res, new_table);
1798 
1799         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1800         if (ret != 0) {
1801                 __ipt_unregister_table(net, new_table);
1802                 *res = NULL;
1803         }
1804 
1805         return ret;
1806 
1807 out_free:
1808         xt_free_table_info(newinfo);
1809         return ret;
1810 }
1811 
1812 void ipt_unregister_table(struct net *net, struct xt_table *table,
1813                           const struct nf_hook_ops *ops)
1814 {
1815         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1816         __ipt_unregister_table(net, table);
1817 }
1818 
1819 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1820 static inline bool
1821 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1822                      u_int8_t type, u_int8_t code,
1823                      bool invert)
1824 {
1825         return ((test_type == 0xFF) ||
1826                 (type == test_type && code >= min_code && code <= max_code))
1827                 ^ invert;
1828 }
1829 
1830 static bool
1831 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
1832 {
1833         const struct icmphdr *ic;
1834         struct icmphdr _icmph;
1835         const struct ipt_icmp *icmpinfo = par->matchinfo;
1836 
1837         /* Must not be a fragment. */
1838         if (par->fragoff != 0)
1839                 return false;
1840 
1841         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1842         if (ic == NULL) {
1843                 /* We've been asked to examine this packet, and we
1844                  * can't.  Hence, no choice but to drop.
1845                  */
1846                 par->hotdrop = true;
1847                 return false;
1848         }
1849 
1850         return icmp_type_code_match(icmpinfo->type,
1851                                     icmpinfo->code[0],
1852                                     icmpinfo->code[1],
1853                                     ic->type, ic->code,
1854                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
1855 }
1856 
1857 static int icmp_checkentry(const struct xt_mtchk_param *par)
1858 {
1859         const struct ipt_icmp *icmpinfo = par->matchinfo;
1860 
1861         /* Must specify no unknown invflags */
1862         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
1863 }
1864 
1865 static struct xt_target ipt_builtin_tg[] __read_mostly = {
1866         {
1867                 .name             = XT_STANDARD_TARGET,
1868                 .targetsize       = sizeof(int),
1869                 .family           = NFPROTO_IPV4,
1870 #ifdef CONFIG_COMPAT
1871                 .compatsize       = sizeof(compat_int_t),
1872                 .compat_from_user = compat_standard_from_user,
1873                 .compat_to_user   = compat_standard_to_user,
1874 #endif
1875         },
1876         {
1877                 .name             = XT_ERROR_TARGET,
1878                 .target           = ipt_error,
1879                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1880                 .family           = NFPROTO_IPV4,
1881         },
1882 };
1883 
1884 static struct nf_sockopt_ops ipt_sockopts = {
1885         .pf             = PF_INET,
1886         .set_optmin     = IPT_BASE_CTL,
1887         .set_optmax     = IPT_SO_SET_MAX+1,
1888         .set            = do_ipt_set_ctl,
1889 #ifdef CONFIG_COMPAT
1890         .compat_set     = compat_do_ipt_set_ctl,
1891 #endif
1892         .get_optmin     = IPT_BASE_CTL,
1893         .get_optmax     = IPT_SO_GET_MAX+1,
1894         .get            = do_ipt_get_ctl,
1895 #ifdef CONFIG_COMPAT
1896         .compat_get     = compat_do_ipt_get_ctl,
1897 #endif
1898         .owner          = THIS_MODULE,
1899 };
1900 
1901 static struct xt_match ipt_builtin_mt[] __read_mostly = {
1902         {
1903                 .name       = "icmp",
1904                 .match      = icmp_match,
1905                 .matchsize  = sizeof(struct ipt_icmp),
1906                 .checkentry = icmp_checkentry,
1907                 .proto      = IPPROTO_ICMP,
1908                 .family     = NFPROTO_IPV4,
1909         },
1910 };
1911 
1912 static int __net_init ip_tables_net_init(struct net *net)
1913 {
1914         return xt_proto_init(net, NFPROTO_IPV4);
1915 }
1916 
1917 static void __net_exit ip_tables_net_exit(struct net *net)
1918 {
1919         xt_proto_fini(net, NFPROTO_IPV4);
1920 }
1921 
1922 static struct pernet_operations ip_tables_net_ops = {
1923         .init = ip_tables_net_init,
1924         .exit = ip_tables_net_exit,
1925 };
1926 
1927 static int __init ip_tables_init(void)
1928 {
1929         int ret;
1930 
1931         ret = register_pernet_subsys(&ip_tables_net_ops);
1932         if (ret < 0)
1933                 goto err1;
1934 
1935         /* No one else will be downing sem now, so we won't sleep */
1936         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1937         if (ret < 0)
1938                 goto err2;
1939         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1940         if (ret < 0)
1941                 goto err4;
1942 
1943         /* Register setsockopt */
1944         ret = nf_register_sockopt(&ipt_sockopts);
1945         if (ret < 0)
1946                 goto err5;
1947 
1948         pr_info("(C) 2000-2006 Netfilter Core Team\n");
1949         return 0;
1950 
1951 err5:
1952         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1953 err4:
1954         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1955 err2:
1956         unregister_pernet_subsys(&ip_tables_net_ops);
1957 err1:
1958         return ret;
1959 }
1960 
1961 static void __exit ip_tables_fini(void)
1962 {
1963         nf_unregister_sockopt(&ipt_sockopts);
1964 
1965         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1966         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1967         unregister_pernet_subsys(&ip_tables_net_ops);
1968 }
1969 
1970 EXPORT_SYMBOL(ipt_register_table);
1971 EXPORT_SYMBOL(ipt_unregister_table);
1972 EXPORT_SYMBOL(ipt_do_table);
1973 module_init(ip_tables_init);
1974 module_exit(ip_tables_fini);
1975 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp