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

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

Version: ~ [ linux-5.12-rc7 ] ~ [ linux-5.11.13 ] ~ [ linux-5.10.29 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.111 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.186 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.230 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.266 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.266 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ 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 <linux/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         if (NF_INVF(ipinfo, IPT_INV_SRCIP,
 62                     (ip->saddr & ipinfo->smsk.s_addr) != ipinfo->src.s_addr) ||
 63             NF_INVF(ipinfo, IPT_INV_DSTIP,
 64                     (ip->daddr & ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr))
 65                 return false;
 66 
 67         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 68 
 69         if (NF_INVF(ipinfo, IPT_INV_VIA_IN, ret != 0))
 70                 return false;
 71 
 72         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 73 
 74         if (NF_INVF(ipinfo, IPT_INV_VIA_OUT, ret != 0))
 75                 return false;
 76 
 77         /* Check specific protocol */
 78         if (ipinfo->proto &&
 79             NF_INVF(ipinfo, IPT_INV_PROTO, ip->protocol != ipinfo->proto))
 80                 return false;
 81 
 82         /* If we have a fragment rule but the packet is not a fragment
 83          * then we return zero */
 84         if (NF_INVF(ipinfo, IPT_INV_FRAG,
 85                     (ipinfo->flags & IPT_F_FRAG) && !isfrag))
 86                 return false;
 87 
 88         return true;
 89 }
 90 
 91 static bool
 92 ip_checkentry(const struct ipt_ip *ip)
 93 {
 94         if (ip->flags & ~IPT_F_MASK)
 95                 return false;
 96         if (ip->invflags & ~IPT_INV_MASK)
 97                 return false;
 98         return true;
 99 }
100 
101 static unsigned int
102 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
103 {
104         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
105 
106         return NF_DROP;
107 }
108 
109 /* Performance critical */
110 static inline struct ipt_entry *
111 get_entry(const void *base, unsigned int offset)
112 {
113         return (struct ipt_entry *)(base + offset);
114 }
115 
116 /* All zeroes == unconditional rule. */
117 /* Mildly perf critical (only if packet tracing is on) */
118 static inline bool unconditional(const struct ipt_entry *e)
119 {
120         static const struct ipt_ip uncond;
121 
122         return e->target_offset == sizeof(struct ipt_entry) &&
123                memcmp(&e->ip, &uncond, sizeof(uncond)) == 0;
124 }
125 
126 /* for const-correctness */
127 static inline const struct xt_entry_target *
128 ipt_get_target_c(const struct ipt_entry *e)
129 {
130         return ipt_get_target((struct ipt_entry *)e);
131 }
132 
133 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
134 static const char *const hooknames[] = {
135         [NF_INET_PRE_ROUTING]           = "PREROUTING",
136         [NF_INET_LOCAL_IN]              = "INPUT",
137         [NF_INET_FORWARD]               = "FORWARD",
138         [NF_INET_LOCAL_OUT]             = "OUTPUT",
139         [NF_INET_POST_ROUTING]          = "POSTROUTING",
140 };
141 
142 enum nf_ip_trace_comments {
143         NF_IP_TRACE_COMMENT_RULE,
144         NF_IP_TRACE_COMMENT_RETURN,
145         NF_IP_TRACE_COMMENT_POLICY,
146 };
147 
148 static const char *const comments[] = {
149         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
150         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
151         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
152 };
153 
154 static struct nf_loginfo trace_loginfo = {
155         .type = NF_LOG_TYPE_LOG,
156         .u = {
157                 .log = {
158                         .level = 4,
159                         .logflags = NF_LOG_DEFAULT_MASK,
160                 },
161         },
162 };
163 
164 /* Mildly perf critical (only if packet tracing is on) */
165 static inline int
166 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
167                       const char *hookname, const char **chainname,
168                       const char **comment, unsigned int *rulenum)
169 {
170         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
171 
172         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
173                 /* Head of user chain: ERROR target with chainname */
174                 *chainname = t->target.data;
175                 (*rulenum) = 0;
176         } else if (s == e) {
177                 (*rulenum)++;
178 
179                 if (unconditional(s) &&
180                     strcmp(t->target.u.kernel.target->name,
181                            XT_STANDARD_TARGET) == 0 &&
182                    t->verdict < 0) {
183                         /* Tail of chains: STANDARD target (return/policy) */
184                         *comment = *chainname == hookname
185                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
186                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
187                 }
188                 return 1;
189         } else
190                 (*rulenum)++;
191 
192         return 0;
193 }
194 
195 static void trace_packet(struct net *net,
196                          const struct sk_buff *skb,
197                          unsigned int hook,
198                          const struct net_device *in,
199                          const struct net_device *out,
200                          const char *tablename,
201                          const struct xt_table_info *private,
202                          const struct ipt_entry *e)
203 {
204         const struct ipt_entry *root;
205         const char *hookname, *chainname, *comment;
206         const struct ipt_entry *iter;
207         unsigned int rulenum = 0;
208 
209         root = get_entry(private->entries, private->hook_entry[hook]);
210 
211         hookname = chainname = hooknames[hook];
212         comment = comments[NF_IP_TRACE_COMMENT_RULE];
213 
214         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
215                 if (get_chainname_rulenum(iter, e, hookname,
216                     &chainname, &comment, &rulenum) != 0)
217                         break;
218 
219         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
220                      "TRACE: %s:%s:%s:%u ",
221                      tablename, chainname, comment, rulenum);
222 }
223 #endif
224 
225 static inline
226 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
227 {
228         return (void *)entry + entry->next_offset;
229 }
230 
231 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
232 unsigned int
233 ipt_do_table(struct sk_buff *skb,
234              const struct nf_hook_state *state,
235              struct xt_table *table)
236 {
237         unsigned int hook = state->hook;
238         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
239         const struct iphdr *ip;
240         /* Initializing verdict to NF_DROP keeps gcc happy. */
241         unsigned int verdict = NF_DROP;
242         const char *indev, *outdev;
243         const void *table_base;
244         struct ipt_entry *e, **jumpstack;
245         unsigned int stackidx, cpu;
246         const struct xt_table_info *private;
247         struct xt_action_param acpar;
248         unsigned int addend;
249 
250         /* Initialization */
251         stackidx = 0;
252         ip = ip_hdr(skb);
253         indev = state->in ? state->in->name : nulldevname;
254         outdev = state->out ? state->out->name : nulldevname;
255         /* We handle fragments by dealing with the first fragment as
256          * if it was a normal packet.  All other fragments are treated
257          * normally, except that they will NEVER match rules that ask
258          * things we don't know, ie. tcp syn flag or ports).  If the
259          * rule is also a fragment-specific rule, non-fragments won't
260          * match it. */
261         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
262         acpar.thoff   = ip_hdrlen(skb);
263         acpar.hotdrop = false;
264         acpar.state   = state;
265 
266         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
267         local_bh_disable();
268         addend = xt_write_recseq_begin();
269         private = table->private;
270         cpu        = smp_processor_id();
271         /*
272          * Ensure we load private-> members after we've fetched the base
273          * pointer.
274          */
275         smp_read_barrier_depends();
276         table_base = private->entries;
277         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
278 
279         /* Switch to alternate jumpstack if we're being invoked via TEE.
280          * TEE issues XT_CONTINUE verdict on original skb so we must not
281          * clobber the jumpstack.
282          *
283          * For recursion via REJECT or SYNPROXY the stack will be clobbered
284          * but it is no problem since absolute verdict is issued by these.
285          */
286         if (static_key_false(&xt_tee_enabled))
287                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
288 
289         e = get_entry(table_base, private->hook_entry[hook]);
290 
291         do {
292                 const struct xt_entry_target *t;
293                 const struct xt_entry_match *ematch;
294                 struct xt_counters *counter;
295 
296                 IP_NF_ASSERT(e);
297                 if (!ip_packet_match(ip, indev, outdev,
298                     &e->ip, acpar.fragoff)) {
299  no_match:
300                         e = ipt_next_entry(e);
301                         continue;
302                 }
303 
304                 xt_ematch_foreach(ematch, e) {
305                         acpar.match     = ematch->u.kernel.match;
306                         acpar.matchinfo = ematch->data;
307                         if (!acpar.match->match(skb, &acpar))
308                                 goto no_match;
309                 }
310 
311                 counter = xt_get_this_cpu_counter(&e->counters);
312                 ADD_COUNTER(*counter, skb->len, 1);
313 
314                 t = ipt_get_target(e);
315                 IP_NF_ASSERT(t->u.kernel.target);
316 
317 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
318                 /* The packet is traced: log it */
319                 if (unlikely(skb->nf_trace))
320                         trace_packet(state->net, skb, hook, state->in,
321                                      state->out, table->name, private, e);
322 #endif
323                 /* Standard target? */
324                 if (!t->u.kernel.target->target) {
325                         int v;
326 
327                         v = ((struct xt_standard_target *)t)->verdict;
328                         if (v < 0) {
329                                 /* Pop from stack? */
330                                 if (v != XT_RETURN) {
331                                         verdict = (unsigned int)(-v) - 1;
332                                         break;
333                                 }
334                                 if (stackidx == 0) {
335                                         e = get_entry(table_base,
336                                             private->underflow[hook]);
337                                 } else {
338                                         e = jumpstack[--stackidx];
339                                         e = ipt_next_entry(e);
340                                 }
341                                 continue;
342                         }
343                         if (table_base + v != ipt_next_entry(e) &&
344                             !(e->ip.flags & IPT_F_GOTO))
345                                 jumpstack[stackidx++] = e;
346 
347                         e = get_entry(table_base, v);
348                         continue;
349                 }
350 
351                 acpar.target   = t->u.kernel.target;
352                 acpar.targinfo = t->data;
353 
354                 verdict = t->u.kernel.target->target(skb, &acpar);
355                 /* Target might have changed stuff. */
356                 ip = ip_hdr(skb);
357                 if (verdict == XT_CONTINUE)
358                         e = ipt_next_entry(e);
359                 else
360                         /* Verdict */
361                         break;
362         } while (!acpar.hotdrop);
363 
364         xt_write_recseq_end(addend);
365         local_bh_enable();
366 
367         if (acpar.hotdrop)
368                 return NF_DROP;
369         else return verdict;
370 }
371 
372 /* Figures out from what hook each rule can be called: returns 0 if
373    there are loops.  Puts hook bitmask in comefrom. */
374 static int
375 mark_source_chains(const struct xt_table_info *newinfo,
376                    unsigned int valid_hooks, void *entry0,
377                    unsigned int *offsets)
378 {
379         unsigned int hook;
380 
381         /* No recursion; use packet counter to save back ptrs (reset
382            to 0 as we leave), and comefrom to save source hook bitmask */
383         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
384                 unsigned int pos = newinfo->hook_entry[hook];
385                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
386 
387                 if (!(valid_hooks & (1 << hook)))
388                         continue;
389 
390                 /* Set initial back pointer. */
391                 e->counters.pcnt = pos;
392 
393                 for (;;) {
394                         const struct xt_standard_target *t
395                                 = (void *)ipt_get_target_c(e);
396                         int visited = e->comefrom & (1 << hook);
397 
398                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
399                                 return 0;
400 
401                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
402 
403                         /* Unconditional return/END. */
404                         if ((unconditional(e) &&
405                              (strcmp(t->target.u.user.name,
406                                      XT_STANDARD_TARGET) == 0) &&
407                              t->verdict < 0) || visited) {
408                                 unsigned int oldpos, size;
409 
410                                 if ((strcmp(t->target.u.user.name,
411                                             XT_STANDARD_TARGET) == 0) &&
412                                     t->verdict < -NF_MAX_VERDICT - 1)
413                                         return 0;
414 
415                                 /* Return: backtrack through the last
416                                    big jump. */
417                                 do {
418                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
419                                         oldpos = pos;
420                                         pos = e->counters.pcnt;
421                                         e->counters.pcnt = 0;
422 
423                                         /* We're at the start. */
424                                         if (pos == oldpos)
425                                                 goto next;
426 
427                                         e = (struct ipt_entry *)
428                                                 (entry0 + pos);
429                                 } while (oldpos == pos + e->next_offset);
430 
431                                 /* Move along one */
432                                 size = e->next_offset;
433                                 e = (struct ipt_entry *)
434                                         (entry0 + pos + size);
435                                 if (pos + size >= newinfo->size)
436                                         return 0;
437                                 e->counters.pcnt = pos;
438                                 pos += size;
439                         } else {
440                                 int newpos = t->verdict;
441 
442                                 if (strcmp(t->target.u.user.name,
443                                            XT_STANDARD_TARGET) == 0 &&
444                                     newpos >= 0) {
445                                         /* This a jump; chase it. */
446                                         if (!xt_find_jump_offset(offsets, newpos,
447                                                                  newinfo->number))
448                                                 return 0;
449                                         e = (struct ipt_entry *)
450                                                 (entry0 + newpos);
451                                 } else {
452                                         /* ... this is a fallthru */
453                                         newpos = pos + e->next_offset;
454                                         if (newpos >= newinfo->size)
455                                                 return 0;
456                                 }
457                                 e = (struct ipt_entry *)
458                                         (entry0 + newpos);
459                                 e->counters.pcnt = pos;
460                                 pos = newpos;
461                         }
462                 }
463 next:           ;
464         }
465         return 1;
466 }
467 
468 static void cleanup_match(struct xt_entry_match *m, struct net *net)
469 {
470         struct xt_mtdtor_param par;
471 
472         par.net       = net;
473         par.match     = m->u.kernel.match;
474         par.matchinfo = m->data;
475         par.family    = NFPROTO_IPV4;
476         if (par.match->destroy != NULL)
477                 par.match->destroy(&par);
478         module_put(par.match->me);
479 }
480 
481 static int
482 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
483 {
484         const struct ipt_ip *ip = par->entryinfo;
485 
486         par->match     = m->u.kernel.match;
487         par->matchinfo = m->data;
488 
489         return xt_check_match(par, m->u.match_size - sizeof(*m),
490                               ip->proto, ip->invflags & IPT_INV_PROTO);
491 }
492 
493 static int
494 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
495 {
496         struct xt_match *match;
497         int ret;
498 
499         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
500                                       m->u.user.revision);
501         if (IS_ERR(match))
502                 return PTR_ERR(match);
503         m->u.kernel.match = match;
504 
505         ret = check_match(m, par);
506         if (ret)
507                 goto err;
508 
509         return 0;
510 err:
511         module_put(m->u.kernel.match->me);
512         return ret;
513 }
514 
515 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
516 {
517         struct xt_entry_target *t = ipt_get_target(e);
518         struct xt_tgchk_param par = {
519                 .net       = net,
520                 .table     = name,
521                 .entryinfo = e,
522                 .target    = t->u.kernel.target,
523                 .targinfo  = t->data,
524                 .hook_mask = e->comefrom,
525                 .family    = NFPROTO_IPV4,
526         };
527 
528         return xt_check_target(&par, t->u.target_size - sizeof(*t),
529                                e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
530 }
531 
532 static int
533 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
534                  unsigned int size,
535                  struct xt_percpu_counter_alloc_state *alloc_state)
536 {
537         struct xt_entry_target *t;
538         struct xt_target *target;
539         int ret;
540         unsigned int j;
541         struct xt_mtchk_param mtpar;
542         struct xt_entry_match *ematch;
543 
544         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
545                 return -ENOMEM;
546 
547         j = 0;
548         mtpar.net       = net;
549         mtpar.table     = name;
550         mtpar.entryinfo = &e->ip;
551         mtpar.hook_mask = e->comefrom;
552         mtpar.family    = NFPROTO_IPV4;
553         xt_ematch_foreach(ematch, e) {
554                 ret = find_check_match(ematch, &mtpar);
555                 if (ret != 0)
556                         goto cleanup_matches;
557                 ++j;
558         }
559 
560         t = ipt_get_target(e);
561         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
562                                         t->u.user.revision);
563         if (IS_ERR(target)) {
564                 ret = PTR_ERR(target);
565                 goto cleanup_matches;
566         }
567         t->u.kernel.target = target;
568 
569         ret = check_target(e, net, name);
570         if (ret)
571                 goto err;
572 
573         return 0;
574  err:
575         module_put(t->u.kernel.target->me);
576  cleanup_matches:
577         xt_ematch_foreach(ematch, e) {
578                 if (j-- == 0)
579                         break;
580                 cleanup_match(ematch, net);
581         }
582 
583         xt_percpu_counter_free(&e->counters);
584 
585         return ret;
586 }
587 
588 static bool check_underflow(const struct ipt_entry *e)
589 {
590         const struct xt_entry_target *t;
591         unsigned int verdict;
592 
593         if (!unconditional(e))
594                 return false;
595         t = ipt_get_target_c(e);
596         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
597                 return false;
598         verdict = ((struct xt_standard_target *)t)->verdict;
599         verdict = -verdict - 1;
600         return verdict == NF_DROP || verdict == NF_ACCEPT;
601 }
602 
603 static int
604 check_entry_size_and_hooks(struct ipt_entry *e,
605                            struct xt_table_info *newinfo,
606                            const unsigned char *base,
607                            const unsigned char *limit,
608                            const unsigned int *hook_entries,
609                            const unsigned int *underflows,
610                            unsigned int valid_hooks)
611 {
612         unsigned int h;
613         int err;
614 
615         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
616             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
617             (unsigned char *)e + e->next_offset > limit)
618                 return -EINVAL;
619 
620         if (e->next_offset
621             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target))
622                 return -EINVAL;
623 
624         if (!ip_checkentry(&e->ip))
625                 return -EINVAL;
626 
627         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
628                                      e->next_offset);
629         if (err)
630                 return err;
631 
632         /* Check hooks & underflows */
633         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
634                 if (!(valid_hooks & (1 << h)))
635                         continue;
636                 if ((unsigned char *)e - base == hook_entries[h])
637                         newinfo->hook_entry[h] = hook_entries[h];
638                 if ((unsigned char *)e - base == underflows[h]) {
639                         if (!check_underflow(e))
640                                 return -EINVAL;
641 
642                         newinfo->underflow[h] = underflows[h];
643                 }
644         }
645 
646         /* Clear counters and comefrom */
647         e->counters = ((struct xt_counters) { 0, 0 });
648         e->comefrom = 0;
649         return 0;
650 }
651 
652 static void
653 cleanup_entry(struct ipt_entry *e, struct net *net)
654 {
655         struct xt_tgdtor_param par;
656         struct xt_entry_target *t;
657         struct xt_entry_match *ematch;
658 
659         /* Cleanup all matches */
660         xt_ematch_foreach(ematch, e)
661                 cleanup_match(ematch, net);
662         t = ipt_get_target(e);
663 
664         par.net      = net;
665         par.target   = t->u.kernel.target;
666         par.targinfo = t->data;
667         par.family   = NFPROTO_IPV4;
668         if (par.target->destroy != NULL)
669                 par.target->destroy(&par);
670         module_put(par.target->me);
671         xt_percpu_counter_free(&e->counters);
672 }
673 
674 /* Checks and translates the user-supplied table segment (held in
675    newinfo) */
676 static int
677 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
678                 const struct ipt_replace *repl)
679 {
680         struct xt_percpu_counter_alloc_state alloc_state = { 0 };
681         struct ipt_entry *iter;
682         unsigned int *offsets;
683         unsigned int i;
684         int ret = 0;
685 
686         newinfo->size = repl->size;
687         newinfo->number = repl->num_entries;
688 
689         /* Init all hooks to impossible value. */
690         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
691                 newinfo->hook_entry[i] = 0xFFFFFFFF;
692                 newinfo->underflow[i] = 0xFFFFFFFF;
693         }
694 
695         offsets = xt_alloc_entry_offsets(newinfo->number);
696         if (!offsets)
697                 return -ENOMEM;
698         i = 0;
699         /* Walk through entries, checking offsets. */
700         xt_entry_foreach(iter, entry0, newinfo->size) {
701                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
702                                                  entry0 + repl->size,
703                                                  repl->hook_entry,
704                                                  repl->underflow,
705                                                  repl->valid_hooks);
706                 if (ret != 0)
707                         goto out_free;
708                 if (i < repl->num_entries)
709                         offsets[i] = (void *)iter - entry0;
710                 ++i;
711                 if (strcmp(ipt_get_target(iter)->u.user.name,
712                     XT_ERROR_TARGET) == 0)
713                         ++newinfo->stacksize;
714         }
715 
716         ret = -EINVAL;
717         if (i != repl->num_entries)
718                 goto out_free;
719 
720         /* Check hooks all assigned */
721         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
722                 /* Only hooks which are valid */
723                 if (!(repl->valid_hooks & (1 << i)))
724                         continue;
725                 if (newinfo->hook_entry[i] == 0xFFFFFFFF)
726                         goto out_free;
727                 if (newinfo->underflow[i] == 0xFFFFFFFF)
728                         goto out_free;
729         }
730 
731         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
732                 ret = -ELOOP;
733                 goto out_free;
734         }
735         kvfree(offsets);
736 
737         /* Finally, each sanity check must pass */
738         i = 0;
739         xt_entry_foreach(iter, entry0, newinfo->size) {
740                 ret = find_check_entry(iter, net, repl->name, repl->size,
741                                        &alloc_state);
742                 if (ret != 0)
743                         break;
744                 ++i;
745         }
746 
747         if (ret != 0) {
748                 xt_entry_foreach(iter, entry0, newinfo->size) {
749                         if (i-- == 0)
750                                 break;
751                         cleanup_entry(iter, net);
752                 }
753                 return ret;
754         }
755 
756         return ret;
757  out_free:
758         kvfree(offsets);
759         return ret;
760 }
761 
762 static void
763 get_counters(const struct xt_table_info *t,
764              struct xt_counters counters[])
765 {
766         struct ipt_entry *iter;
767         unsigned int cpu;
768         unsigned int i;
769 
770         for_each_possible_cpu(cpu) {
771                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
772 
773                 i = 0;
774                 xt_entry_foreach(iter, t->entries, t->size) {
775                         struct xt_counters *tmp;
776                         u64 bcnt, pcnt;
777                         unsigned int start;
778 
779                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
780                         do {
781                                 start = read_seqcount_begin(s);
782                                 bcnt = tmp->bcnt;
783                                 pcnt = tmp->pcnt;
784                         } while (read_seqcount_retry(s, start));
785 
786                         ADD_COUNTER(counters[i], bcnt, pcnt);
787                         ++i; /* macro does multi eval of i */
788                 }
789         }
790 }
791 
792 static struct xt_counters *alloc_counters(const struct xt_table *table)
793 {
794         unsigned int countersize;
795         struct xt_counters *counters;
796         const struct xt_table_info *private = table->private;
797 
798         /* We need atomic snapshot of counters: rest doesn't change
799            (other than comefrom, which userspace doesn't care
800            about). */
801         countersize = sizeof(struct xt_counters) * private->number;
802         counters = vzalloc(countersize);
803 
804         if (counters == NULL)
805                 return ERR_PTR(-ENOMEM);
806 
807         get_counters(private, counters);
808 
809         return counters;
810 }
811 
812 static int
813 copy_entries_to_user(unsigned int total_size,
814                      const struct xt_table *table,
815                      void __user *userptr)
816 {
817         unsigned int off, num;
818         const struct ipt_entry *e;
819         struct xt_counters *counters;
820         const struct xt_table_info *private = table->private;
821         int ret = 0;
822         const void *loc_cpu_entry;
823 
824         counters = alloc_counters(table);
825         if (IS_ERR(counters))
826                 return PTR_ERR(counters);
827 
828         loc_cpu_entry = private->entries;
829 
830         /* FIXME: use iterator macros --RR */
831         /* ... then go back and fix counters and names */
832         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
833                 unsigned int i;
834                 const struct xt_entry_match *m;
835                 const struct xt_entry_target *t;
836 
837                 e = (struct ipt_entry *)(loc_cpu_entry + off);
838                 if (copy_to_user(userptr + off, e, sizeof(*e))) {
839                         ret = -EFAULT;
840                         goto free_counters;
841                 }
842                 if (copy_to_user(userptr + off
843                                  + offsetof(struct ipt_entry, counters),
844                                  &counters[num],
845                                  sizeof(counters[num])) != 0) {
846                         ret = -EFAULT;
847                         goto free_counters;
848                 }
849 
850                 for (i = sizeof(struct ipt_entry);
851                      i < e->target_offset;
852                      i += m->u.match_size) {
853                         m = (void *)e + i;
854 
855                         if (xt_match_to_user(m, userptr + off + i)) {
856                                 ret = -EFAULT;
857                                 goto free_counters;
858                         }
859                 }
860 
861                 t = ipt_get_target_c(e);
862                 if (xt_target_to_user(t, userptr + off + e->target_offset)) {
863                         ret = -EFAULT;
864                         goto free_counters;
865                 }
866         }
867 
868  free_counters:
869         vfree(counters);
870         return ret;
871 }
872 
873 #ifdef CONFIG_COMPAT
874 static void compat_standard_from_user(void *dst, const void *src)
875 {
876         int v = *(compat_int_t *)src;
877 
878         if (v > 0)
879                 v += xt_compat_calc_jump(AF_INET, v);
880         memcpy(dst, &v, sizeof(v));
881 }
882 
883 static int compat_standard_to_user(void __user *dst, const void *src)
884 {
885         compat_int_t cv = *(int *)src;
886 
887         if (cv > 0)
888                 cv -= xt_compat_calc_jump(AF_INET, cv);
889         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
890 }
891 
892 static int compat_calc_entry(const struct ipt_entry *e,
893                              const struct xt_table_info *info,
894                              const void *base, struct xt_table_info *newinfo)
895 {
896         const struct xt_entry_match *ematch;
897         const struct xt_entry_target *t;
898         unsigned int entry_offset;
899         int off, i, ret;
900 
901         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
902         entry_offset = (void *)e - base;
903         xt_ematch_foreach(ematch, e)
904                 off += xt_compat_match_offset(ematch->u.kernel.match);
905         t = ipt_get_target_c(e);
906         off += xt_compat_target_offset(t->u.kernel.target);
907         newinfo->size -= off;
908         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
909         if (ret)
910                 return ret;
911 
912         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
913                 if (info->hook_entry[i] &&
914                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
915                         newinfo->hook_entry[i] -= off;
916                 if (info->underflow[i] &&
917                     (e < (struct ipt_entry *)(base + info->underflow[i])))
918                         newinfo->underflow[i] -= off;
919         }
920         return 0;
921 }
922 
923 static int compat_table_info(const struct xt_table_info *info,
924                              struct xt_table_info *newinfo)
925 {
926         struct ipt_entry *iter;
927         const void *loc_cpu_entry;
928         int ret;
929 
930         if (!newinfo || !info)
931                 return -EINVAL;
932 
933         /* we dont care about newinfo->entries */
934         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
935         newinfo->initial_entries = 0;
936         loc_cpu_entry = info->entries;
937         xt_compat_init_offsets(AF_INET, info->number);
938         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
939                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
940                 if (ret != 0)
941                         return ret;
942         }
943         return 0;
944 }
945 #endif
946 
947 static int get_info(struct net *net, void __user *user,
948                     const int *len, int compat)
949 {
950         char name[XT_TABLE_MAXNAMELEN];
951         struct xt_table *t;
952         int ret;
953 
954         if (*len != sizeof(struct ipt_getinfo))
955                 return -EINVAL;
956 
957         if (copy_from_user(name, user, sizeof(name)) != 0)
958                 return -EFAULT;
959 
960         name[XT_TABLE_MAXNAMELEN-1] = '\0';
961 #ifdef CONFIG_COMPAT
962         if (compat)
963                 xt_compat_lock(AF_INET);
964 #endif
965         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
966                                     "iptable_%s", name);
967         if (t) {
968                 struct ipt_getinfo info;
969                 const struct xt_table_info *private = t->private;
970 #ifdef CONFIG_COMPAT
971                 struct xt_table_info tmp;
972 
973                 if (compat) {
974                         ret = compat_table_info(private, &tmp);
975                         xt_compat_flush_offsets(AF_INET);
976                         private = &tmp;
977                 }
978 #endif
979                 memset(&info, 0, sizeof(info));
980                 info.valid_hooks = t->valid_hooks;
981                 memcpy(info.hook_entry, private->hook_entry,
982                        sizeof(info.hook_entry));
983                 memcpy(info.underflow, private->underflow,
984                        sizeof(info.underflow));
985                 info.num_entries = private->number;
986                 info.size = private->size;
987                 strcpy(info.name, name);
988 
989                 if (copy_to_user(user, &info, *len) != 0)
990                         ret = -EFAULT;
991                 else
992                         ret = 0;
993 
994                 xt_table_unlock(t);
995                 module_put(t->me);
996         } else
997                 ret = -ENOENT;
998 #ifdef CONFIG_COMPAT
999         if (compat)
1000                 xt_compat_unlock(AF_INET);
1001 #endif
1002         return ret;
1003 }
1004 
1005 static int
1006 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1007             const int *len)
1008 {
1009         int ret;
1010         struct ipt_get_entries get;
1011         struct xt_table *t;
1012 
1013         if (*len < sizeof(get))
1014                 return -EINVAL;
1015         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1016                 return -EFAULT;
1017         if (*len != sizeof(struct ipt_get_entries) + get.size)
1018                 return -EINVAL;
1019         get.name[sizeof(get.name) - 1] = '\0';
1020 
1021         t = xt_find_table_lock(net, AF_INET, get.name);
1022         if (t) {
1023                 const struct xt_table_info *private = t->private;
1024                 if (get.size == private->size)
1025                         ret = copy_entries_to_user(private->size,
1026                                                    t, uptr->entrytable);
1027                 else
1028                         ret = -EAGAIN;
1029 
1030                 module_put(t->me);
1031                 xt_table_unlock(t);
1032         } else
1033                 ret = -ENOENT;
1034 
1035         return ret;
1036 }
1037 
1038 static int
1039 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1040              struct xt_table_info *newinfo, unsigned int num_counters,
1041              void __user *counters_ptr)
1042 {
1043         int ret;
1044         struct xt_table *t;
1045         struct xt_table_info *oldinfo;
1046         struct xt_counters *counters;
1047         struct ipt_entry *iter;
1048 
1049         ret = 0;
1050         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1051         if (!counters) {
1052                 ret = -ENOMEM;
1053                 goto out;
1054         }
1055 
1056         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1057                                     "iptable_%s", name);
1058         if (!t) {
1059                 ret = -ENOENT;
1060                 goto free_newinfo_counters_untrans;
1061         }
1062 
1063         /* You lied! */
1064         if (valid_hooks != t->valid_hooks) {
1065                 ret = -EINVAL;
1066                 goto put_module;
1067         }
1068 
1069         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1070         if (!oldinfo)
1071                 goto put_module;
1072 
1073         /* Update module usage count based on number of rules */
1074         if ((oldinfo->number > oldinfo->initial_entries) ||
1075             (newinfo->number <= oldinfo->initial_entries))
1076                 module_put(t->me);
1077         if ((oldinfo->number > oldinfo->initial_entries) &&
1078             (newinfo->number <= oldinfo->initial_entries))
1079                 module_put(t->me);
1080 
1081         /* Get the old counters, and synchronize with replace */
1082         get_counters(oldinfo, counters);
1083 
1084         /* Decrease module usage counts and free resource */
1085         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1086                 cleanup_entry(iter, net);
1087 
1088         xt_free_table_info(oldinfo);
1089         if (copy_to_user(counters_ptr, counters,
1090                          sizeof(struct xt_counters) * num_counters) != 0) {
1091                 /* Silent error, can't fail, new table is already in place */
1092                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1093         }
1094         vfree(counters);
1095         xt_table_unlock(t);
1096         return ret;
1097 
1098  put_module:
1099         module_put(t->me);
1100         xt_table_unlock(t);
1101  free_newinfo_counters_untrans:
1102         vfree(counters);
1103  out:
1104         return ret;
1105 }
1106 
1107 static int
1108 do_replace(struct net *net, const void __user *user, unsigned int len)
1109 {
1110         int ret;
1111         struct ipt_replace tmp;
1112         struct xt_table_info *newinfo;
1113         void *loc_cpu_entry;
1114         struct ipt_entry *iter;
1115 
1116         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1117                 return -EFAULT;
1118 
1119         /* overflow check */
1120         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1121                 return -ENOMEM;
1122         if (tmp.num_counters == 0)
1123                 return -EINVAL;
1124 
1125         tmp.name[sizeof(tmp.name)-1] = 0;
1126 
1127         newinfo = xt_alloc_table_info(tmp.size);
1128         if (!newinfo)
1129                 return -ENOMEM;
1130 
1131         loc_cpu_entry = newinfo->entries;
1132         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1133                            tmp.size) != 0) {
1134                 ret = -EFAULT;
1135                 goto free_newinfo;
1136         }
1137 
1138         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1139         if (ret != 0)
1140                 goto free_newinfo;
1141 
1142         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1143                            tmp.num_counters, tmp.counters);
1144         if (ret)
1145                 goto free_newinfo_untrans;
1146         return 0;
1147 
1148  free_newinfo_untrans:
1149         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1150                 cleanup_entry(iter, net);
1151  free_newinfo:
1152         xt_free_table_info(newinfo);
1153         return ret;
1154 }
1155 
1156 static int
1157 do_add_counters(struct net *net, const void __user *user,
1158                 unsigned int len, int compat)
1159 {
1160         unsigned int i;
1161         struct xt_counters_info tmp;
1162         struct xt_counters *paddc;
1163         struct xt_table *t;
1164         const struct xt_table_info *private;
1165         int ret = 0;
1166         struct ipt_entry *iter;
1167         unsigned int addend;
1168 
1169         paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1170         if (IS_ERR(paddc))
1171                 return PTR_ERR(paddc);
1172 
1173         t = xt_find_table_lock(net, AF_INET, tmp.name);
1174         if (!t) {
1175                 ret = -ENOENT;
1176                 goto free;
1177         }
1178 
1179         local_bh_disable();
1180         private = t->private;
1181         if (private->number != tmp.num_counters) {
1182                 ret = -EINVAL;
1183                 goto unlock_up_free;
1184         }
1185 
1186         i = 0;
1187         addend = xt_write_recseq_begin();
1188         xt_entry_foreach(iter, private->entries, private->size) {
1189                 struct xt_counters *tmp;
1190 
1191                 tmp = xt_get_this_cpu_counter(&iter->counters);
1192                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1193                 ++i;
1194         }
1195         xt_write_recseq_end(addend);
1196  unlock_up_free:
1197         local_bh_enable();
1198         xt_table_unlock(t);
1199         module_put(t->me);
1200  free:
1201         vfree(paddc);
1202 
1203         return ret;
1204 }
1205 
1206 #ifdef CONFIG_COMPAT
1207 struct compat_ipt_replace {
1208         char                    name[XT_TABLE_MAXNAMELEN];
1209         u32                     valid_hooks;
1210         u32                     num_entries;
1211         u32                     size;
1212         u32                     hook_entry[NF_INET_NUMHOOKS];
1213         u32                     underflow[NF_INET_NUMHOOKS];
1214         u32                     num_counters;
1215         compat_uptr_t           counters;       /* struct xt_counters * */
1216         struct compat_ipt_entry entries[0];
1217 };
1218 
1219 static int
1220 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1221                           unsigned int *size, struct xt_counters *counters,
1222                           unsigned int i)
1223 {
1224         struct xt_entry_target *t;
1225         struct compat_ipt_entry __user *ce;
1226         u_int16_t target_offset, next_offset;
1227         compat_uint_t origsize;
1228         const struct xt_entry_match *ematch;
1229         int ret = 0;
1230 
1231         origsize = *size;
1232         ce = (struct compat_ipt_entry __user *)*dstptr;
1233         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1234             copy_to_user(&ce->counters, &counters[i],
1235             sizeof(counters[i])) != 0)
1236                 return -EFAULT;
1237 
1238         *dstptr += sizeof(struct compat_ipt_entry);
1239         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1240 
1241         xt_ematch_foreach(ematch, e) {
1242                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1243                 if (ret != 0)
1244                         return ret;
1245         }
1246         target_offset = e->target_offset - (origsize - *size);
1247         t = ipt_get_target(e);
1248         ret = xt_compat_target_to_user(t, dstptr, size);
1249         if (ret)
1250                 return ret;
1251         next_offset = e->next_offset - (origsize - *size);
1252         if (put_user(target_offset, &ce->target_offset) != 0 ||
1253             put_user(next_offset, &ce->next_offset) != 0)
1254                 return -EFAULT;
1255         return 0;
1256 }
1257 
1258 static int
1259 compat_find_calc_match(struct xt_entry_match *m,
1260                        const struct ipt_ip *ip,
1261                        int *size)
1262 {
1263         struct xt_match *match;
1264 
1265         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1266                                       m->u.user.revision);
1267         if (IS_ERR(match))
1268                 return PTR_ERR(match);
1269 
1270         m->u.kernel.match = match;
1271         *size += xt_compat_match_offset(match);
1272         return 0;
1273 }
1274 
1275 static void compat_release_entry(struct compat_ipt_entry *e)
1276 {
1277         struct xt_entry_target *t;
1278         struct xt_entry_match *ematch;
1279 
1280         /* Cleanup all matches */
1281         xt_ematch_foreach(ematch, e)
1282                 module_put(ematch->u.kernel.match->me);
1283         t = compat_ipt_get_target(e);
1284         module_put(t->u.kernel.target->me);
1285 }
1286 
1287 static int
1288 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1289                                   struct xt_table_info *newinfo,
1290                                   unsigned int *size,
1291                                   const unsigned char *base,
1292                                   const unsigned char *limit)
1293 {
1294         struct xt_entry_match *ematch;
1295         struct xt_entry_target *t;
1296         struct xt_target *target;
1297         unsigned int entry_offset;
1298         unsigned int j;
1299         int ret, off;
1300 
1301         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1302             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1303             (unsigned char *)e + e->next_offset > limit)
1304                 return -EINVAL;
1305 
1306         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1307                              sizeof(struct compat_xt_entry_target))
1308                 return -EINVAL;
1309 
1310         if (!ip_checkentry(&e->ip))
1311                 return -EINVAL;
1312 
1313         ret = xt_compat_check_entry_offsets(e, e->elems,
1314                                             e->target_offset, e->next_offset);
1315         if (ret)
1316                 return ret;
1317 
1318         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1319         entry_offset = (void *)e - (void *)base;
1320         j = 0;
1321         xt_ematch_foreach(ematch, e) {
1322                 ret = compat_find_calc_match(ematch, &e->ip, &off);
1323                 if (ret != 0)
1324                         goto release_matches;
1325                 ++j;
1326         }
1327 
1328         t = compat_ipt_get_target(e);
1329         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1330                                         t->u.user.revision);
1331         if (IS_ERR(target)) {
1332                 ret = PTR_ERR(target);
1333                 goto release_matches;
1334         }
1335         t->u.kernel.target = target;
1336 
1337         off += xt_compat_target_offset(target);
1338         *size += off;
1339         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1340         if (ret)
1341                 goto out;
1342 
1343         return 0;
1344 
1345 out:
1346         module_put(t->u.kernel.target->me);
1347 release_matches:
1348         xt_ematch_foreach(ematch, e) {
1349                 if (j-- == 0)
1350                         break;
1351                 module_put(ematch->u.kernel.match->me);
1352         }
1353         return ret;
1354 }
1355 
1356 static void
1357 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1358                             unsigned int *size,
1359                             struct xt_table_info *newinfo, unsigned char *base)
1360 {
1361         struct xt_entry_target *t;
1362         struct xt_target *target;
1363         struct ipt_entry *de;
1364         unsigned int origsize;
1365         int h;
1366         struct xt_entry_match *ematch;
1367 
1368         origsize = *size;
1369         de = (struct ipt_entry *)*dstptr;
1370         memcpy(de, e, sizeof(struct ipt_entry));
1371         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1372 
1373         *dstptr += sizeof(struct ipt_entry);
1374         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1375 
1376         xt_ematch_foreach(ematch, e)
1377                 xt_compat_match_from_user(ematch, dstptr, size);
1378 
1379         de->target_offset = e->target_offset - (origsize - *size);
1380         t = compat_ipt_get_target(e);
1381         target = t->u.kernel.target;
1382         xt_compat_target_from_user(t, dstptr, size);
1383 
1384         de->next_offset = e->next_offset - (origsize - *size);
1385 
1386         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1387                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1388                         newinfo->hook_entry[h] -= origsize - *size;
1389                 if ((unsigned char *)de - base < newinfo->underflow[h])
1390                         newinfo->underflow[h] -= origsize - *size;
1391         }
1392 }
1393 
1394 static int
1395 translate_compat_table(struct net *net,
1396                        struct xt_table_info **pinfo,
1397                        void **pentry0,
1398                        const struct compat_ipt_replace *compatr)
1399 {
1400         unsigned int i, j;
1401         struct xt_table_info *newinfo, *info;
1402         void *pos, *entry0, *entry1;
1403         struct compat_ipt_entry *iter0;
1404         struct ipt_replace repl;
1405         unsigned int size;
1406         int ret;
1407 
1408         info = *pinfo;
1409         entry0 = *pentry0;
1410         size = compatr->size;
1411         info->number = compatr->num_entries;
1412 
1413         j = 0;
1414         xt_compat_lock(AF_INET);
1415         xt_compat_init_offsets(AF_INET, compatr->num_entries);
1416         /* Walk through entries, checking offsets. */
1417         xt_entry_foreach(iter0, entry0, compatr->size) {
1418                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1419                                                         entry0,
1420                                                         entry0 + compatr->size);
1421                 if (ret != 0)
1422                         goto out_unlock;
1423                 ++j;
1424         }
1425 
1426         ret = -EINVAL;
1427         if (j != compatr->num_entries)
1428                 goto out_unlock;
1429 
1430         ret = -ENOMEM;
1431         newinfo = xt_alloc_table_info(size);
1432         if (!newinfo)
1433                 goto out_unlock;
1434 
1435         newinfo->number = compatr->num_entries;
1436         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1437                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1438                 newinfo->underflow[i] = compatr->underflow[i];
1439         }
1440         entry1 = newinfo->entries;
1441         pos = entry1;
1442         size = compatr->size;
1443         xt_entry_foreach(iter0, entry0, compatr->size)
1444                 compat_copy_entry_from_user(iter0, &pos, &size,
1445                                             newinfo, entry1);
1446 
1447         /* all module references in entry0 are now gone.
1448          * entry1/newinfo contains a 64bit ruleset that looks exactly as
1449          * generated by 64bit userspace.
1450          *
1451          * Call standard translate_table() to validate all hook_entrys,
1452          * underflows, check for loops, etc.
1453          */
1454         xt_compat_flush_offsets(AF_INET);
1455         xt_compat_unlock(AF_INET);
1456 
1457         memcpy(&repl, compatr, sizeof(*compatr));
1458 
1459         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1460                 repl.hook_entry[i] = newinfo->hook_entry[i];
1461                 repl.underflow[i] = newinfo->underflow[i];
1462         }
1463 
1464         repl.num_counters = 0;
1465         repl.counters = NULL;
1466         repl.size = newinfo->size;
1467         ret = translate_table(net, newinfo, entry1, &repl);
1468         if (ret)
1469                 goto free_newinfo;
1470 
1471         *pinfo = newinfo;
1472         *pentry0 = entry1;
1473         xt_free_table_info(info);
1474         return 0;
1475 
1476 free_newinfo:
1477         xt_free_table_info(newinfo);
1478         return ret;
1479 out_unlock:
1480         xt_compat_flush_offsets(AF_INET);
1481         xt_compat_unlock(AF_INET);
1482         xt_entry_foreach(iter0, entry0, compatr->size) {
1483                 if (j-- == 0)
1484                         break;
1485                 compat_release_entry(iter0);
1486         }
1487         return ret;
1488 }
1489 
1490 static int
1491 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1492 {
1493         int ret;
1494         struct compat_ipt_replace tmp;
1495         struct xt_table_info *newinfo;
1496         void *loc_cpu_entry;
1497         struct ipt_entry *iter;
1498 
1499         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1500                 return -EFAULT;
1501 
1502         /* overflow check */
1503         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1504                 return -ENOMEM;
1505         if (tmp.num_counters == 0)
1506                 return -EINVAL;
1507 
1508         tmp.name[sizeof(tmp.name)-1] = 0;
1509 
1510         newinfo = xt_alloc_table_info(tmp.size);
1511         if (!newinfo)
1512                 return -ENOMEM;
1513 
1514         loc_cpu_entry = newinfo->entries;
1515         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1516                            tmp.size) != 0) {
1517                 ret = -EFAULT;
1518                 goto free_newinfo;
1519         }
1520 
1521         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1522         if (ret != 0)
1523                 goto free_newinfo;
1524 
1525         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1526                            tmp.num_counters, compat_ptr(tmp.counters));
1527         if (ret)
1528                 goto free_newinfo_untrans;
1529         return 0;
1530 
1531  free_newinfo_untrans:
1532         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1533                 cleanup_entry(iter, net);
1534  free_newinfo:
1535         xt_free_table_info(newinfo);
1536         return ret;
1537 }
1538 
1539 static int
1540 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1541                       unsigned int len)
1542 {
1543         int ret;
1544 
1545         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1546                 return -EPERM;
1547 
1548         switch (cmd) {
1549         case IPT_SO_SET_REPLACE:
1550                 ret = compat_do_replace(sock_net(sk), user, len);
1551                 break;
1552 
1553         case IPT_SO_SET_ADD_COUNTERS:
1554                 ret = do_add_counters(sock_net(sk), user, len, 1);
1555                 break;
1556 
1557         default:
1558                 ret = -EINVAL;
1559         }
1560 
1561         return ret;
1562 }
1563 
1564 struct compat_ipt_get_entries {
1565         char name[XT_TABLE_MAXNAMELEN];
1566         compat_uint_t size;
1567         struct compat_ipt_entry entrytable[0];
1568 };
1569 
1570 static int
1571 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1572                             void __user *userptr)
1573 {
1574         struct xt_counters *counters;
1575         const struct xt_table_info *private = table->private;
1576         void __user *pos;
1577         unsigned int size;
1578         int ret = 0;
1579         unsigned int i = 0;
1580         struct ipt_entry *iter;
1581 
1582         counters = alloc_counters(table);
1583         if (IS_ERR(counters))
1584                 return PTR_ERR(counters);
1585 
1586         pos = userptr;
1587         size = total_size;
1588         xt_entry_foreach(iter, private->entries, total_size) {
1589                 ret = compat_copy_entry_to_user(iter, &pos,
1590                                                 &size, counters, i++);
1591                 if (ret != 0)
1592                         break;
1593         }
1594 
1595         vfree(counters);
1596         return ret;
1597 }
1598 
1599 static int
1600 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1601                    int *len)
1602 {
1603         int ret;
1604         struct compat_ipt_get_entries get;
1605         struct xt_table *t;
1606 
1607         if (*len < sizeof(get))
1608                 return -EINVAL;
1609 
1610         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1611                 return -EFAULT;
1612 
1613         if (*len != sizeof(struct compat_ipt_get_entries) + get.size)
1614                 return -EINVAL;
1615 
1616         get.name[sizeof(get.name) - 1] = '\0';
1617 
1618         xt_compat_lock(AF_INET);
1619         t = xt_find_table_lock(net, AF_INET, get.name);
1620         if (t) {
1621                 const struct xt_table_info *private = t->private;
1622                 struct xt_table_info info;
1623                 ret = compat_table_info(private, &info);
1624                 if (!ret && get.size == info.size)
1625                         ret = compat_copy_entries_to_user(private->size,
1626                                                           t, uptr->entrytable);
1627                 else if (!ret)
1628                         ret = -EAGAIN;
1629 
1630                 xt_compat_flush_offsets(AF_INET);
1631                 module_put(t->me);
1632                 xt_table_unlock(t);
1633         } else
1634                 ret = -ENOENT;
1635 
1636         xt_compat_unlock(AF_INET);
1637         return ret;
1638 }
1639 
1640 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1641 
1642 static int
1643 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1644 {
1645         int ret;
1646 
1647         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1648                 return -EPERM;
1649 
1650         switch (cmd) {
1651         case IPT_SO_GET_INFO:
1652                 ret = get_info(sock_net(sk), user, len, 1);
1653                 break;
1654         case IPT_SO_GET_ENTRIES:
1655                 ret = compat_get_entries(sock_net(sk), user, len);
1656                 break;
1657         default:
1658                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1659         }
1660         return ret;
1661 }
1662 #endif
1663 
1664 static int
1665 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1666 {
1667         int ret;
1668 
1669         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1670                 return -EPERM;
1671 
1672         switch (cmd) {
1673         case IPT_SO_SET_REPLACE:
1674                 ret = do_replace(sock_net(sk), user, len);
1675                 break;
1676 
1677         case IPT_SO_SET_ADD_COUNTERS:
1678                 ret = do_add_counters(sock_net(sk), user, len, 0);
1679                 break;
1680 
1681         default:
1682                 ret = -EINVAL;
1683         }
1684 
1685         return ret;
1686 }
1687 
1688 static int
1689 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1690 {
1691         int ret;
1692 
1693         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1694                 return -EPERM;
1695 
1696         switch (cmd) {
1697         case IPT_SO_GET_INFO:
1698                 ret = get_info(sock_net(sk), user, len, 0);
1699                 break;
1700 
1701         case IPT_SO_GET_ENTRIES:
1702                 ret = get_entries(sock_net(sk), user, len);
1703                 break;
1704 
1705         case IPT_SO_GET_REVISION_MATCH:
1706         case IPT_SO_GET_REVISION_TARGET: {
1707                 struct xt_get_revision rev;
1708                 int target;
1709 
1710                 if (*len != sizeof(rev)) {
1711                         ret = -EINVAL;
1712                         break;
1713                 }
1714                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1715                         ret = -EFAULT;
1716                         break;
1717                 }
1718                 rev.name[sizeof(rev.name)-1] = 0;
1719 
1720                 if (cmd == IPT_SO_GET_REVISION_TARGET)
1721                         target = 1;
1722                 else
1723                         target = 0;
1724 
1725                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
1726                                                          rev.revision,
1727                                                          target, &ret),
1728                                         "ipt_%s", rev.name);
1729                 break;
1730         }
1731 
1732         default:
1733                 ret = -EINVAL;
1734         }
1735 
1736         return ret;
1737 }
1738 
1739 static void __ipt_unregister_table(struct net *net, struct xt_table *table)
1740 {
1741         struct xt_table_info *private;
1742         void *loc_cpu_entry;
1743         struct module *table_owner = table->me;
1744         struct ipt_entry *iter;
1745 
1746         private = xt_unregister_table(table);
1747 
1748         /* Decrease module usage counts and free resources */
1749         loc_cpu_entry = private->entries;
1750         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1751                 cleanup_entry(iter, net);
1752         if (private->number > private->initial_entries)
1753                 module_put(table_owner);
1754         xt_free_table_info(private);
1755 }
1756 
1757 int ipt_register_table(struct net *net, const struct xt_table *table,
1758                        const struct ipt_replace *repl,
1759                        const struct nf_hook_ops *ops, struct xt_table **res)
1760 {
1761         int ret;
1762         struct xt_table_info *newinfo;
1763         struct xt_table_info bootstrap = {0};
1764         void *loc_cpu_entry;
1765         struct xt_table *new_table;
1766 
1767         newinfo = xt_alloc_table_info(repl->size);
1768         if (!newinfo)
1769                 return -ENOMEM;
1770 
1771         loc_cpu_entry = newinfo->entries;
1772         memcpy(loc_cpu_entry, repl->entries, repl->size);
1773 
1774         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1775         if (ret != 0)
1776                 goto out_free;
1777 
1778         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1779         if (IS_ERR(new_table)) {
1780                 ret = PTR_ERR(new_table);
1781                 goto out_free;
1782         }
1783 
1784         /* set res now, will see skbs right after nf_register_net_hooks */
1785         WRITE_ONCE(*res, new_table);
1786 
1787         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1788         if (ret != 0) {
1789                 __ipt_unregister_table(net, new_table);
1790                 *res = NULL;
1791         }
1792 
1793         return ret;
1794 
1795 out_free:
1796         xt_free_table_info(newinfo);
1797         return ret;
1798 }
1799 
1800 void ipt_unregister_table(struct net *net, struct xt_table *table,
1801                           const struct nf_hook_ops *ops)
1802 {
1803         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1804         __ipt_unregister_table(net, table);
1805 }
1806 
1807 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1808 static inline bool
1809 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1810                      u_int8_t type, u_int8_t code,
1811                      bool invert)
1812 {
1813         return ((test_type == 0xFF) ||
1814                 (type == test_type && code >= min_code && code <= max_code))
1815                 ^ invert;
1816 }
1817 
1818 static bool
1819 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
1820 {
1821         const struct icmphdr *ic;
1822         struct icmphdr _icmph;
1823         const struct ipt_icmp *icmpinfo = par->matchinfo;
1824 
1825         /* Must not be a fragment. */
1826         if (par->fragoff != 0)
1827                 return false;
1828 
1829         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1830         if (ic == NULL) {
1831                 /* We've been asked to examine this packet, and we
1832                  * can't.  Hence, no choice but to drop.
1833                  */
1834                 par->hotdrop = true;
1835                 return false;
1836         }
1837 
1838         return icmp_type_code_match(icmpinfo->type,
1839                                     icmpinfo->code[0],
1840                                     icmpinfo->code[1],
1841                                     ic->type, ic->code,
1842                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
1843 }
1844 
1845 static int icmp_checkentry(const struct xt_mtchk_param *par)
1846 {
1847         const struct ipt_icmp *icmpinfo = par->matchinfo;
1848 
1849         /* Must specify no unknown invflags */
1850         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
1851 }
1852 
1853 static struct xt_target ipt_builtin_tg[] __read_mostly = {
1854         {
1855                 .name             = XT_STANDARD_TARGET,
1856                 .targetsize       = sizeof(int),
1857                 .family           = NFPROTO_IPV4,
1858 #ifdef CONFIG_COMPAT
1859                 .compatsize       = sizeof(compat_int_t),
1860                 .compat_from_user = compat_standard_from_user,
1861                 .compat_to_user   = compat_standard_to_user,
1862 #endif
1863         },
1864         {
1865                 .name             = XT_ERROR_TARGET,
1866                 .target           = ipt_error,
1867                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1868                 .family           = NFPROTO_IPV4,
1869         },
1870 };
1871 
1872 static struct nf_sockopt_ops ipt_sockopts = {
1873         .pf             = PF_INET,
1874         .set_optmin     = IPT_BASE_CTL,
1875         .set_optmax     = IPT_SO_SET_MAX+1,
1876         .set            = do_ipt_set_ctl,
1877 #ifdef CONFIG_COMPAT
1878         .compat_set     = compat_do_ipt_set_ctl,
1879 #endif
1880         .get_optmin     = IPT_BASE_CTL,
1881         .get_optmax     = IPT_SO_GET_MAX+1,
1882         .get            = do_ipt_get_ctl,
1883 #ifdef CONFIG_COMPAT
1884         .compat_get     = compat_do_ipt_get_ctl,
1885 #endif
1886         .owner          = THIS_MODULE,
1887 };
1888 
1889 static struct xt_match ipt_builtin_mt[] __read_mostly = {
1890         {
1891                 .name       = "icmp",
1892                 .match      = icmp_match,
1893                 .matchsize  = sizeof(struct ipt_icmp),
1894                 .checkentry = icmp_checkentry,
1895                 .proto      = IPPROTO_ICMP,
1896                 .family     = NFPROTO_IPV4,
1897         },
1898 };
1899 
1900 static int __net_init ip_tables_net_init(struct net *net)
1901 {
1902         return xt_proto_init(net, NFPROTO_IPV4);
1903 }
1904 
1905 static void __net_exit ip_tables_net_exit(struct net *net)
1906 {
1907         xt_proto_fini(net, NFPROTO_IPV4);
1908 }
1909 
1910 static struct pernet_operations ip_tables_net_ops = {
1911         .init = ip_tables_net_init,
1912         .exit = ip_tables_net_exit,
1913 };
1914 
1915 static int __init ip_tables_init(void)
1916 {
1917         int ret;
1918 
1919         ret = register_pernet_subsys(&ip_tables_net_ops);
1920         if (ret < 0)
1921                 goto err1;
1922 
1923         /* No one else will be downing sem now, so we won't sleep */
1924         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1925         if (ret < 0)
1926                 goto err2;
1927         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1928         if (ret < 0)
1929                 goto err4;
1930 
1931         /* Register setsockopt */
1932         ret = nf_register_sockopt(&ipt_sockopts);
1933         if (ret < 0)
1934                 goto err5;
1935 
1936         pr_info("(C) 2000-2006 Netfilter Core Team\n");
1937         return 0;
1938 
1939 err5:
1940         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1941 err4:
1942         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1943 err2:
1944         unregister_pernet_subsys(&ip_tables_net_ops);
1945 err1:
1946         return ret;
1947 }
1948 
1949 static void __exit ip_tables_fini(void)
1950 {
1951         nf_unregister_sockopt(&ipt_sockopts);
1952 
1953         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1954         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1955         unregister_pernet_subsys(&ip_tables_net_ops);
1956 }
1957 
1958 EXPORT_SYMBOL(ipt_register_table);
1959 EXPORT_SYMBOL(ipt_unregister_table);
1960 EXPORT_SYMBOL(ipt_do_table);
1961 module_init(ip_tables_init);
1962 module_exit(ip_tables_fini);
1963 

~ [ 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