d0d19192026d9eb2327cf68a47a4ba1b23dc7fe5
[linux-3.10.git] / net / ipv4 / netfilter / arp_tables.c
1 /*
2  * Packet matching code for ARP packets.
3  *
4  * Based heavily, if not almost entirely, upon ip_tables.c framework.
5  *
6  * Some ARP specific bits are:
7  *
8  * Copyright (C) 2002 David S. Miller (davem@redhat.com)
9  *
10  */
11
12 #include <linux/config.h>
13 #include <linux/kernel.h>
14 #include <linux/skbuff.h>
15 #include <linux/netdevice.h>
16 #include <linux/capability.h>
17 #include <linux/if_arp.h>
18 #include <linux/kmod.h>
19 #include <linux/vmalloc.h>
20 #include <linux/proc_fs.h>
21 #include <linux/module.h>
22 #include <linux/init.h>
23
24 #include <asm/uaccess.h>
25 #include <linux/mutex.h>
26
27 #include <linux/netfilter/x_tables.h>
28 #include <linux/netfilter_arp/arp_tables.h>
29
30 MODULE_LICENSE("GPL");
31 MODULE_AUTHOR("David S. Miller <davem@redhat.com>");
32 MODULE_DESCRIPTION("arptables core");
33
34 /*#define DEBUG_ARP_TABLES*/
35 /*#define DEBUG_ARP_TABLES_USER*/
36
37 #ifdef DEBUG_ARP_TABLES
38 #define dprintf(format, args...)  printk(format , ## args)
39 #else
40 #define dprintf(format, args...)
41 #endif
42
43 #ifdef DEBUG_ARP_TABLES_USER
44 #define duprintf(format, args...) printk(format , ## args)
45 #else
46 #define duprintf(format, args...)
47 #endif
48
49 #ifdef CONFIG_NETFILTER_DEBUG
50 #define ARP_NF_ASSERT(x)                                        \
51 do {                                                            \
52         if (!(x))                                               \
53                 printk("ARP_NF_ASSERT: %s:%s:%u\n",             \
54                        __FUNCTION__, __FILE__, __LINE__);       \
55 } while(0)
56 #else
57 #define ARP_NF_ASSERT(x)
58 #endif
59
60 #include <linux/netfilter_ipv4/listhelp.h>
61
62 static inline int arp_devaddr_compare(const struct arpt_devaddr_info *ap,
63                                       char *hdr_addr, int len)
64 {
65         int i, ret;
66
67         if (len > ARPT_DEV_ADDR_LEN_MAX)
68                 len = ARPT_DEV_ADDR_LEN_MAX;
69
70         ret = 0;
71         for (i = 0; i < len; i++)
72                 ret |= (hdr_addr[i] ^ ap->addr[i]) & ap->mask[i];
73
74         return (ret != 0);
75 }
76
77 /* Returns whether packet matches rule or not. */
78 static inline int arp_packet_match(const struct arphdr *arphdr,
79                                    struct net_device *dev,
80                                    const char *indev,
81                                    const char *outdev,
82                                    const struct arpt_arp *arpinfo)
83 {
84         char *arpptr = (char *)(arphdr + 1);
85         char *src_devaddr, *tgt_devaddr;
86         u32 src_ipaddr, tgt_ipaddr;
87         int i, ret;
88
89 #define FWINV(bool,invflg) ((bool) ^ !!(arpinfo->invflags & invflg))
90
91         if (FWINV((arphdr->ar_op & arpinfo->arpop_mask) != arpinfo->arpop,
92                   ARPT_INV_ARPOP)) {
93                 dprintf("ARP operation field mismatch.\n");
94                 dprintf("ar_op: %04x info->arpop: %04x info->arpop_mask: %04x\n",
95                         arphdr->ar_op, arpinfo->arpop, arpinfo->arpop_mask);
96                 return 0;
97         }
98
99         if (FWINV((arphdr->ar_hrd & arpinfo->arhrd_mask) != arpinfo->arhrd,
100                   ARPT_INV_ARPHRD)) {
101                 dprintf("ARP hardware address format mismatch.\n");
102                 dprintf("ar_hrd: %04x info->arhrd: %04x info->arhrd_mask: %04x\n",
103                         arphdr->ar_hrd, arpinfo->arhrd, arpinfo->arhrd_mask);
104                 return 0;
105         }
106
107         if (FWINV((arphdr->ar_pro & arpinfo->arpro_mask) != arpinfo->arpro,
108                   ARPT_INV_ARPPRO)) {
109                 dprintf("ARP protocol address format mismatch.\n");
110                 dprintf("ar_pro: %04x info->arpro: %04x info->arpro_mask: %04x\n",
111                         arphdr->ar_pro, arpinfo->arpro, arpinfo->arpro_mask);
112                 return 0;
113         }
114
115         if (FWINV((arphdr->ar_hln & arpinfo->arhln_mask) != arpinfo->arhln,
116                   ARPT_INV_ARPHLN)) {
117                 dprintf("ARP hardware address length mismatch.\n");
118                 dprintf("ar_hln: %02x info->arhln: %02x info->arhln_mask: %02x\n",
119                         arphdr->ar_hln, arpinfo->arhln, arpinfo->arhln_mask);
120                 return 0;
121         }
122
123         src_devaddr = arpptr;
124         arpptr += dev->addr_len;
125         memcpy(&src_ipaddr, arpptr, sizeof(u32));
126         arpptr += sizeof(u32);
127         tgt_devaddr = arpptr;
128         arpptr += dev->addr_len;
129         memcpy(&tgt_ipaddr, arpptr, sizeof(u32));
130
131         if (FWINV(arp_devaddr_compare(&arpinfo->src_devaddr, src_devaddr, dev->addr_len),
132                   ARPT_INV_SRCDEVADDR) ||
133             FWINV(arp_devaddr_compare(&arpinfo->tgt_devaddr, tgt_devaddr, dev->addr_len),
134                   ARPT_INV_TGTDEVADDR)) {
135                 dprintf("Source or target device address mismatch.\n");
136
137                 return 0;
138         }
139
140         if (FWINV((src_ipaddr & arpinfo->smsk.s_addr) != arpinfo->src.s_addr,
141                   ARPT_INV_SRCIP) ||
142             FWINV(((tgt_ipaddr & arpinfo->tmsk.s_addr) != arpinfo->tgt.s_addr),
143                   ARPT_INV_TGTIP)) {
144                 dprintf("Source or target IP address mismatch.\n");
145
146                 dprintf("SRC: %u.%u.%u.%u. Mask: %u.%u.%u.%u. Target: %u.%u.%u.%u.%s\n",
147                         NIPQUAD(src_ipaddr),
148                         NIPQUAD(arpinfo->smsk.s_addr),
149                         NIPQUAD(arpinfo->src.s_addr),
150                         arpinfo->invflags & ARPT_INV_SRCIP ? " (INV)" : "");
151                 dprintf("TGT: %u.%u.%u.%u Mask: %u.%u.%u.%u Target: %u.%u.%u.%u.%s\n",
152                         NIPQUAD(tgt_ipaddr),
153                         NIPQUAD(arpinfo->tmsk.s_addr),
154                         NIPQUAD(arpinfo->tgt.s_addr),
155                         arpinfo->invflags & ARPT_INV_TGTIP ? " (INV)" : "");
156                 return 0;
157         }
158
159         /* Look for ifname matches.  */
160         for (i = 0, ret = 0; i < IFNAMSIZ; i++) {
161                 ret |= (indev[i] ^ arpinfo->iniface[i])
162                         & arpinfo->iniface_mask[i];
163         }
164
165         if (FWINV(ret != 0, ARPT_INV_VIA_IN)) {
166                 dprintf("VIA in mismatch (%s vs %s).%s\n",
167                         indev, arpinfo->iniface,
168                         arpinfo->invflags&ARPT_INV_VIA_IN ?" (INV)":"");
169                 return 0;
170         }
171
172         for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
173                 unsigned long odev;
174                 memcpy(&odev, outdev + i*sizeof(unsigned long),
175                        sizeof(unsigned long));
176                 ret |= (odev
177                         ^ ((const unsigned long *)arpinfo->outiface)[i])
178                         & ((const unsigned long *)arpinfo->outiface_mask)[i];
179         }
180
181         if (FWINV(ret != 0, ARPT_INV_VIA_OUT)) {
182                 dprintf("VIA out mismatch (%s vs %s).%s\n",
183                         outdev, arpinfo->outiface,
184                         arpinfo->invflags&ARPT_INV_VIA_OUT ?" (INV)":"");
185                 return 0;
186         }
187
188         return 1;
189 }
190
191 static inline int arp_checkentry(const struct arpt_arp *arp)
192 {
193         if (arp->flags & ~ARPT_F_MASK) {
194                 duprintf("Unknown flag bits set: %08X\n",
195                          arp->flags & ~ARPT_F_MASK);
196                 return 0;
197         }
198         if (arp->invflags & ~ARPT_INV_MASK) {
199                 duprintf("Unknown invflag bits set: %08X\n",
200                          arp->invflags & ~ARPT_INV_MASK);
201                 return 0;
202         }
203
204         return 1;
205 }
206
207 static unsigned int arpt_error(struct sk_buff **pskb,
208                                const struct net_device *in,
209                                const struct net_device *out,
210                                unsigned int hooknum,
211                                const struct xt_target *target,
212                                const void *targinfo,
213                                void *userinfo)
214 {
215         if (net_ratelimit())
216                 printk("arp_tables: error: '%s'\n", (char *)targinfo);
217
218         return NF_DROP;
219 }
220
221 static inline struct arpt_entry *get_entry(void *base, unsigned int offset)
222 {
223         return (struct arpt_entry *)(base + offset);
224 }
225
226 unsigned int arpt_do_table(struct sk_buff **pskb,
227                            unsigned int hook,
228                            const struct net_device *in,
229                            const struct net_device *out,
230                            struct arpt_table *table,
231                            void *userdata)
232 {
233         static const char nulldevname[IFNAMSIZ];
234         unsigned int verdict = NF_DROP;
235         struct arphdr *arp;
236         int hotdrop = 0;
237         struct arpt_entry *e, *back;
238         const char *indev, *outdev;
239         void *table_base;
240         struct xt_table_info *private = table->private;
241
242         /* ARP header, plus 2 device addresses, plus 2 IP addresses.  */
243         if (!pskb_may_pull((*pskb), (sizeof(struct arphdr) +
244                                      (2 * (*pskb)->dev->addr_len) +
245                                      (2 * sizeof(u32)))))
246                 return NF_DROP;
247
248         indev = in ? in->name : nulldevname;
249         outdev = out ? out->name : nulldevname;
250
251         read_lock_bh(&table->lock);
252         table_base = (void *)private->entries[smp_processor_id()];
253         e = get_entry(table_base, private->hook_entry[hook]);
254         back = get_entry(table_base, private->underflow[hook]);
255
256         arp = (*pskb)->nh.arph;
257         do {
258                 if (arp_packet_match(arp, (*pskb)->dev, indev, outdev, &e->arp)) {
259                         struct arpt_entry_target *t;
260                         int hdr_len;
261
262                         hdr_len = sizeof(*arp) + (2 * sizeof(struct in_addr)) +
263                                 (2 * (*pskb)->dev->addr_len);
264                         ADD_COUNTER(e->counters, hdr_len, 1);
265
266                         t = arpt_get_target(e);
267
268                         /* Standard target? */
269                         if (!t->u.kernel.target->target) {
270                                 int v;
271
272                                 v = ((struct arpt_standard_target *)t)->verdict;
273                                 if (v < 0) {
274                                         /* Pop from stack? */
275                                         if (v != ARPT_RETURN) {
276                                                 verdict = (unsigned)(-v) - 1;
277                                                 break;
278                                         }
279                                         e = back;
280                                         back = get_entry(table_base,
281                                                          back->comefrom);
282                                         continue;
283                                 }
284                                 if (table_base + v
285                                     != (void *)e + e->next_offset) {
286                                         /* Save old back ptr in next entry */
287                                         struct arpt_entry *next
288                                                 = (void *)e + e->next_offset;
289                                         next->comefrom =
290                                                 (void *)back - table_base;
291
292                                         /* set back pointer to next entry */
293                                         back = next;
294                                 }
295
296                                 e = get_entry(table_base, v);
297                         } else {
298                                 /* Targets which reenter must return
299                                  * abs. verdicts
300                                  */
301                                 verdict = t->u.kernel.target->target(pskb,
302                                                                      in, out,
303                                                                      hook,
304                                                                      t->u.kernel.target,
305                                                                      t->data,
306                                                                      userdata);
307
308                                 /* Target might have changed stuff. */
309                                 arp = (*pskb)->nh.arph;
310
311                                 if (verdict == ARPT_CONTINUE)
312                                         e = (void *)e + e->next_offset;
313                                 else
314                                         /* Verdict */
315                                         break;
316                         }
317                 } else {
318                         e = (void *)e + e->next_offset;
319                 }
320         } while (!hotdrop);
321         read_unlock_bh(&table->lock);
322
323         if (hotdrop)
324                 return NF_DROP;
325         else
326                 return verdict;
327 }
328
329 /* All zeroes == unconditional rule. */
330 static inline int unconditional(const struct arpt_arp *arp)
331 {
332         unsigned int i;
333
334         for (i = 0; i < sizeof(*arp)/sizeof(__u32); i++)
335                 if (((__u32 *)arp)[i])
336                         return 0;
337
338         return 1;
339 }
340
341 /* Figures out from what hook each rule can be called: returns 0 if
342  * there are loops.  Puts hook bitmask in comefrom.
343  */
344 static int mark_source_chains(struct xt_table_info *newinfo,
345                               unsigned int valid_hooks, void *entry0)
346 {
347         unsigned int hook;
348
349         /* No recursion; use packet counter to save back ptrs (reset
350          * to 0 as we leave), and comefrom to save source hook bitmask.
351          */
352         for (hook = 0; hook < NF_ARP_NUMHOOKS; hook++) {
353                 unsigned int pos = newinfo->hook_entry[hook];
354                 struct arpt_entry *e
355                         = (struct arpt_entry *)(entry0 + pos);
356
357                 if (!(valid_hooks & (1 << hook)))
358                         continue;
359
360                 /* Set initial back pointer. */
361                 e->counters.pcnt = pos;
362
363                 for (;;) {
364                         struct arpt_standard_target *t
365                                 = (void *)arpt_get_target(e);
366
367                         if (e->comefrom & (1 << NF_ARP_NUMHOOKS)) {
368                                 printk("arptables: loop hook %u pos %u %08X.\n",
369                                        hook, pos, e->comefrom);
370                                 return 0;
371                         }
372                         e->comefrom
373                                 |= ((1 << hook) | (1 << NF_ARP_NUMHOOKS));
374
375                         /* Unconditional return/END. */
376                         if (e->target_offset == sizeof(struct arpt_entry)
377                             && (strcmp(t->target.u.user.name,
378                                        ARPT_STANDARD_TARGET) == 0)
379                             && t->verdict < 0
380                             && unconditional(&e->arp)) {
381                                 unsigned int oldpos, size;
382
383                                 /* Return: backtrack through the last
384                                  * big jump.
385                                  */
386                                 do {
387                                         e->comefrom ^= (1<<NF_ARP_NUMHOOKS);
388                                         oldpos = pos;
389                                         pos = e->counters.pcnt;
390                                         e->counters.pcnt = 0;
391
392                                         /* We're at the start. */
393                                         if (pos == oldpos)
394                                                 goto next;
395
396                                         e = (struct arpt_entry *)
397                                                 (entry0 + pos);
398                                 } while (oldpos == pos + e->next_offset);
399
400                                 /* Move along one */
401                                 size = e->next_offset;
402                                 e = (struct arpt_entry *)
403                                         (entry0 + pos + size);
404                                 e->counters.pcnt = pos;
405                                 pos += size;
406                         } else {
407                                 int newpos = t->verdict;
408
409                                 if (strcmp(t->target.u.user.name,
410                                            ARPT_STANDARD_TARGET) == 0
411                                     && newpos >= 0) {
412                                         /* This a jump; chase it. */
413                                         duprintf("Jump rule %u -> %u\n",
414                                                  pos, newpos);
415                                 } else {
416                                         /* ... this is a fallthru */
417                                         newpos = pos + e->next_offset;
418                                 }
419                                 e = (struct arpt_entry *)
420                                         (entry0 + newpos);
421                                 e->counters.pcnt = pos;
422                                 pos = newpos;
423                         }
424                 }
425                 next:
426                 duprintf("Finished chain %u\n", hook);
427         }
428         return 1;
429 }
430
431 static inline int standard_check(const struct arpt_entry_target *t,
432                                  unsigned int max_offset)
433 {
434         struct arpt_standard_target *targ = (void *)t;
435
436         /* Check standard info. */
437         if (t->u.target_size
438             != ARPT_ALIGN(sizeof(struct arpt_standard_target))) {
439                 duprintf("arpt_standard_check: target size %u != %Zu\n",
440                          t->u.target_size,
441                          ARPT_ALIGN(sizeof(struct arpt_standard_target)));
442                 return 0;
443         }
444
445         if (targ->verdict >= 0
446             && targ->verdict > max_offset - sizeof(struct arpt_entry)) {
447                 duprintf("arpt_standard_check: bad verdict (%i)\n",
448                          targ->verdict);
449                 return 0;
450         }
451
452         if (targ->verdict < -NF_MAX_VERDICT - 1) {
453                 duprintf("arpt_standard_check: bad negative verdict (%i)\n",
454                          targ->verdict);
455                 return 0;
456         }
457         return 1;
458 }
459
460 static struct arpt_target arpt_standard_target;
461
462 static inline int check_entry(struct arpt_entry *e, const char *name, unsigned int size,
463                               unsigned int *i)
464 {
465         struct arpt_entry_target *t;
466         struct arpt_target *target;
467         int ret;
468
469         if (!arp_checkentry(&e->arp)) {
470                 duprintf("arp_tables: arp check failed %p %s.\n", e, name);
471                 return -EINVAL;
472         }
473
474         t = arpt_get_target(e);
475         target = try_then_request_module(xt_find_target(NF_ARP, t->u.user.name,
476                                                         t->u.user.revision),
477                                          "arpt_%s", t->u.user.name);
478         if (IS_ERR(target) || !target) {
479                 duprintf("check_entry: `%s' not found\n", t->u.user.name);
480                 ret = target ? PTR_ERR(target) : -ENOENT;
481                 goto out;
482         }
483         t->u.kernel.target = target;
484
485         ret = xt_check_target(target, NF_ARP, t->u.target_size - sizeof(*t),
486                               name, e->comefrom, 0, 0);
487         if (ret)
488                 goto err;
489
490         if (t->u.kernel.target == &arpt_standard_target) {
491                 if (!standard_check(t, size)) {
492                         ret = -EINVAL;
493                         goto out;
494                 }
495         } else if (t->u.kernel.target->checkentry
496                    && !t->u.kernel.target->checkentry(name, e, target, t->data,
497                                                       t->u.target_size
498                                                       - sizeof(*t),
499                                                       e->comefrom)) {
500                 duprintf("arp_tables: check failed for `%s'.\n",
501                          t->u.kernel.target->name);
502                 ret = -EINVAL;
503                 goto err;
504         }
505
506         (*i)++;
507         return 0;
508 err:
509         module_put(t->u.kernel.target->me);
510 out:
511         return ret;
512 }
513
514 static inline int check_entry_size_and_hooks(struct arpt_entry *e,
515                                              struct xt_table_info *newinfo,
516                                              unsigned char *base,
517                                              unsigned char *limit,
518                                              const unsigned int *hook_entries,
519                                              const unsigned int *underflows,
520                                              unsigned int *i)
521 {
522         unsigned int h;
523
524         if ((unsigned long)e % __alignof__(struct arpt_entry) != 0
525             || (unsigned char *)e + sizeof(struct arpt_entry) >= limit) {
526                 duprintf("Bad offset %p\n", e);
527                 return -EINVAL;
528         }
529
530         if (e->next_offset
531             < sizeof(struct arpt_entry) + sizeof(struct arpt_entry_target)) {
532                 duprintf("checking: element %p size %u\n",
533                          e, e->next_offset);
534                 return -EINVAL;
535         }
536
537         /* Check hooks & underflows */
538         for (h = 0; h < NF_ARP_NUMHOOKS; h++) {
539                 if ((unsigned char *)e - base == hook_entries[h])
540                         newinfo->hook_entry[h] = hook_entries[h];
541                 if ((unsigned char *)e - base == underflows[h])
542                         newinfo->underflow[h] = underflows[h];
543         }
544
545         /* FIXME: underflows must be unconditional, standard verdicts
546            < 0 (not ARPT_RETURN). --RR */
547
548         /* Clear counters and comefrom */
549         e->counters = ((struct xt_counters) { 0, 0 });
550         e->comefrom = 0;
551
552         (*i)++;
553         return 0;
554 }
555
556 static inline int cleanup_entry(struct arpt_entry *e, unsigned int *i)
557 {
558         struct arpt_entry_target *t;
559
560         if (i && (*i)-- == 0)
561                 return 1;
562
563         t = arpt_get_target(e);
564         if (t->u.kernel.target->destroy)
565                 t->u.kernel.target->destroy(t->u.kernel.target, t->data,
566                                             t->u.target_size - sizeof(*t));
567         module_put(t->u.kernel.target->me);
568         return 0;
569 }
570
571 /* Checks and translates the user-supplied table segment (held in
572  * newinfo).
573  */
574 static int translate_table(const char *name,
575                            unsigned int valid_hooks,
576                            struct xt_table_info *newinfo,
577                            void *entry0,
578                            unsigned int size,
579                            unsigned int number,
580                            const unsigned int *hook_entries,
581                            const unsigned int *underflows)
582 {
583         unsigned int i;
584         int ret;
585
586         newinfo->size = size;
587         newinfo->number = number;
588
589         /* Init all hooks to impossible value. */
590         for (i = 0; i < NF_ARP_NUMHOOKS; i++) {
591                 newinfo->hook_entry[i] = 0xFFFFFFFF;
592                 newinfo->underflow[i] = 0xFFFFFFFF;
593         }
594
595         duprintf("translate_table: size %u\n", newinfo->size);
596         i = 0;
597
598         /* Walk through entries, checking offsets. */
599         ret = ARPT_ENTRY_ITERATE(entry0, newinfo->size,
600                                  check_entry_size_and_hooks,
601                                  newinfo,
602                                  entry0,
603                                  entry0 + size,
604                                  hook_entries, underflows, &i);
605         duprintf("translate_table: ARPT_ENTRY_ITERATE gives %d\n", ret);
606         if (ret != 0)
607                 return ret;
608
609         if (i != number) {
610                 duprintf("translate_table: %u not %u entries\n",
611                          i, number);
612                 return -EINVAL;
613         }
614
615         /* Check hooks all assigned */
616         for (i = 0; i < NF_ARP_NUMHOOKS; i++) {
617                 /* Only hooks which are valid */
618                 if (!(valid_hooks & (1 << i)))
619                         continue;
620                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
621                         duprintf("Invalid hook entry %u %u\n",
622                                  i, hook_entries[i]);
623                         return -EINVAL;
624                 }
625                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
626                         duprintf("Invalid underflow %u %u\n",
627                                  i, underflows[i]);
628                         return -EINVAL;
629                 }
630         }
631
632         if (!mark_source_chains(newinfo, valid_hooks, entry0)) {
633                 duprintf("Looping hook\n");
634                 return -ELOOP;
635         }
636
637         /* Finally, each sanity check must pass */
638         i = 0;
639         ret = ARPT_ENTRY_ITERATE(entry0, newinfo->size,
640                                  check_entry, name, size, &i);
641
642         if (ret != 0) {
643                 ARPT_ENTRY_ITERATE(entry0, newinfo->size,
644                                    cleanup_entry, &i);
645                 return ret;
646         }
647
648         /* And one copy for every other CPU */
649         for_each_possible_cpu(i) {
650                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
651                         memcpy(newinfo->entries[i], entry0, newinfo->size);
652         }
653
654         return ret;
655 }
656
657 /* Gets counters. */
658 static inline int add_entry_to_counter(const struct arpt_entry *e,
659                                        struct xt_counters total[],
660                                        unsigned int *i)
661 {
662         ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
663
664         (*i)++;
665         return 0;
666 }
667
668 static inline int set_entry_to_counter(const struct arpt_entry *e,
669                                        struct xt_counters total[],
670                                        unsigned int *i)
671 {
672         SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
673
674         (*i)++;
675         return 0;
676 }
677
678 static void get_counters(const struct xt_table_info *t,
679                          struct xt_counters counters[])
680 {
681         unsigned int cpu;
682         unsigned int i;
683         unsigned int curcpu;
684
685         /* Instead of clearing (by a previous call to memset())
686          * the counters and using adds, we set the counters
687          * with data used by 'current' CPU
688          * We dont care about preemption here.
689          */
690         curcpu = raw_smp_processor_id();
691
692         i = 0;
693         ARPT_ENTRY_ITERATE(t->entries[curcpu],
694                            t->size,
695                            set_entry_to_counter,
696                            counters,
697                            &i);
698
699         for_each_possible_cpu(cpu) {
700                 if (cpu == curcpu)
701                         continue;
702                 i = 0;
703                 ARPT_ENTRY_ITERATE(t->entries[cpu],
704                                    t->size,
705                                    add_entry_to_counter,
706                                    counters,
707                                    &i);
708         }
709 }
710
711 static int copy_entries_to_user(unsigned int total_size,
712                                 struct arpt_table *table,
713                                 void __user *userptr)
714 {
715         unsigned int off, num, countersize;
716         struct arpt_entry *e;
717         struct xt_counters *counters;
718         struct xt_table_info *private = table->private;
719         int ret = 0;
720         void *loc_cpu_entry;
721
722         /* We need atomic snapshot of counters: rest doesn't change
723          * (other than comefrom, which userspace doesn't care
724          * about).
725          */
726         countersize = sizeof(struct xt_counters) * private->number;
727         counters = vmalloc_node(countersize, numa_node_id());
728
729         if (counters == NULL)
730                 return -ENOMEM;
731
732         /* First, sum counters... */
733         write_lock_bh(&table->lock);
734         get_counters(private, counters);
735         write_unlock_bh(&table->lock);
736
737         loc_cpu_entry = private->entries[raw_smp_processor_id()];
738         /* ... then copy entire thing ... */
739         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
740                 ret = -EFAULT;
741                 goto free_counters;
742         }
743
744         /* FIXME: use iterator macros --RR */
745         /* ... then go back and fix counters and names */
746         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
747                 struct arpt_entry_target *t;
748
749                 e = (struct arpt_entry *)(loc_cpu_entry + off);
750                 if (copy_to_user(userptr + off
751                                  + offsetof(struct arpt_entry, counters),
752                                  &counters[num],
753                                  sizeof(counters[num])) != 0) {
754                         ret = -EFAULT;
755                         goto free_counters;
756                 }
757
758                 t = arpt_get_target(e);
759                 if (copy_to_user(userptr + off + e->target_offset
760                                  + offsetof(struct arpt_entry_target,
761                                             u.user.name),
762                                  t->u.kernel.target->name,
763                                  strlen(t->u.kernel.target->name)+1) != 0) {
764                         ret = -EFAULT;
765                         goto free_counters;
766                 }
767         }
768
769  free_counters:
770         vfree(counters);
771         return ret;
772 }
773
774 static int get_entries(const struct arpt_get_entries *entries,
775                        struct arpt_get_entries __user *uptr)
776 {
777         int ret;
778         struct arpt_table *t;
779
780         t = xt_find_table_lock(NF_ARP, entries->name);
781         if (t && !IS_ERR(t)) {
782                 struct xt_table_info *private = t->private;
783                 duprintf("t->private->number = %u\n",
784                          private->number);
785                 if (entries->size == private->size)
786                         ret = copy_entries_to_user(private->size,
787                                                    t, uptr->entrytable);
788                 else {
789                         duprintf("get_entries: I've got %u not %u!\n",
790                                  private->size, entries->size);
791                         ret = -EINVAL;
792                 }
793                 module_put(t->me);
794                 xt_table_unlock(t);
795         } else
796                 ret = t ? PTR_ERR(t) : -ENOENT;
797
798         return ret;
799 }
800
801 static int do_replace(void __user *user, unsigned int len)
802 {
803         int ret;
804         struct arpt_replace tmp;
805         struct arpt_table *t;
806         struct xt_table_info *newinfo, *oldinfo;
807         struct xt_counters *counters;
808         void *loc_cpu_entry, *loc_cpu_old_entry;
809
810         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
811                 return -EFAULT;
812
813         /* Hack: Causes ipchains to give correct error msg --RR */
814         if (len != sizeof(tmp) + tmp.size)
815                 return -ENOPROTOOPT;
816
817         /* overflow check */
818         if (tmp.size >= (INT_MAX - sizeof(struct xt_table_info)) / NR_CPUS -
819                         SMP_CACHE_BYTES)
820                 return -ENOMEM;
821         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
822                 return -ENOMEM;
823
824         newinfo = xt_alloc_table_info(tmp.size);
825         if (!newinfo)
826                 return -ENOMEM;
827
828         /* choose the copy that is on our node/cpu */
829         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
830         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
831                            tmp.size) != 0) {
832                 ret = -EFAULT;
833                 goto free_newinfo;
834         }
835
836         counters = vmalloc(tmp.num_counters * sizeof(struct xt_counters));
837         if (!counters) {
838                 ret = -ENOMEM;
839                 goto free_newinfo;
840         }
841
842         ret = translate_table(tmp.name, tmp.valid_hooks,
843                               newinfo, loc_cpu_entry, tmp.size, tmp.num_entries,
844                               tmp.hook_entry, tmp.underflow);
845         if (ret != 0)
846                 goto free_newinfo_counters;
847
848         duprintf("arp_tables: Translated table\n");
849
850         t = try_then_request_module(xt_find_table_lock(NF_ARP, tmp.name),
851                                     "arptable_%s", tmp.name);
852         if (!t || IS_ERR(t)) {
853                 ret = t ? PTR_ERR(t) : -ENOENT;
854                 goto free_newinfo_counters_untrans;
855         }
856
857         /* You lied! */
858         if (tmp.valid_hooks != t->valid_hooks) {
859                 duprintf("Valid hook crap: %08X vs %08X\n",
860                          tmp.valid_hooks, t->valid_hooks);
861                 ret = -EINVAL;
862                 goto put_module;
863         }
864
865         oldinfo = xt_replace_table(t, tmp.num_counters, newinfo, &ret);
866         if (!oldinfo)
867                 goto put_module;
868
869         /* Update module usage count based on number of rules */
870         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
871                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
872         if ((oldinfo->number > oldinfo->initial_entries) || 
873             (newinfo->number <= oldinfo->initial_entries)) 
874                 module_put(t->me);
875         if ((oldinfo->number > oldinfo->initial_entries) &&
876             (newinfo->number <= oldinfo->initial_entries))
877                 module_put(t->me);
878
879         /* Get the old counters. */
880         get_counters(oldinfo, counters);
881         /* Decrease module usage counts and free resource */
882         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
883         ARPT_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,NULL);
884
885         xt_free_table_info(oldinfo);
886         if (copy_to_user(tmp.counters, counters,
887                          sizeof(struct xt_counters) * tmp.num_counters) != 0)
888                 ret = -EFAULT;
889         vfree(counters);
890         xt_table_unlock(t);
891         return ret;
892
893  put_module:
894         module_put(t->me);
895         xt_table_unlock(t);
896  free_newinfo_counters_untrans:
897         ARPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry, NULL);
898  free_newinfo_counters:
899         vfree(counters);
900  free_newinfo:
901         xt_free_table_info(newinfo);
902         return ret;
903 }
904
905 /* We're lazy, and add to the first CPU; overflow works its fey magic
906  * and everything is OK.
907  */
908 static inline int add_counter_to_entry(struct arpt_entry *e,
909                                        const struct xt_counters addme[],
910                                        unsigned int *i)
911 {
912
913         ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
914
915         (*i)++;
916         return 0;
917 }
918
919 static int do_add_counters(void __user *user, unsigned int len)
920 {
921         unsigned int i;
922         struct xt_counters_info tmp, *paddc;
923         struct arpt_table *t;
924         struct xt_table_info *private;
925         int ret = 0;
926         void *loc_cpu_entry;
927
928         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
929                 return -EFAULT;
930
931         if (len != sizeof(tmp) + tmp.num_counters*sizeof(struct xt_counters))
932                 return -EINVAL;
933
934         paddc = vmalloc(len);
935         if (!paddc)
936                 return -ENOMEM;
937
938         if (copy_from_user(paddc, user, len) != 0) {
939                 ret = -EFAULT;
940                 goto free;
941         }
942
943         t = xt_find_table_lock(NF_ARP, tmp.name);
944         if (!t || IS_ERR(t)) {
945                 ret = t ? PTR_ERR(t) : -ENOENT;
946                 goto free;
947         }
948
949         write_lock_bh(&t->lock);
950         private = t->private;
951         if (private->number != tmp.num_counters) {
952                 ret = -EINVAL;
953                 goto unlock_up_free;
954         }
955
956         i = 0;
957         /* Choose the copy that is on our node */
958         loc_cpu_entry = private->entries[smp_processor_id()];
959         ARPT_ENTRY_ITERATE(loc_cpu_entry,
960                            private->size,
961                            add_counter_to_entry,
962                            paddc->counters,
963                            &i);
964  unlock_up_free:
965         write_unlock_bh(&t->lock);
966         xt_table_unlock(t);
967         module_put(t->me);
968  free:
969         vfree(paddc);
970
971         return ret;
972 }
973
974 static int do_arpt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
975 {
976         int ret;
977
978         if (!capable(CAP_NET_ADMIN))
979                 return -EPERM;
980
981         switch (cmd) {
982         case ARPT_SO_SET_REPLACE:
983                 ret = do_replace(user, len);
984                 break;
985
986         case ARPT_SO_SET_ADD_COUNTERS:
987                 ret = do_add_counters(user, len);
988                 break;
989
990         default:
991                 duprintf("do_arpt_set_ctl:  unknown request %i\n", cmd);
992                 ret = -EINVAL;
993         }
994
995         return ret;
996 }
997
998 static int do_arpt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
999 {
1000         int ret;
1001
1002         if (!capable(CAP_NET_ADMIN))
1003                 return -EPERM;
1004
1005         switch (cmd) {
1006         case ARPT_SO_GET_INFO: {
1007                 char name[ARPT_TABLE_MAXNAMELEN];
1008                 struct arpt_table *t;
1009
1010                 if (*len != sizeof(struct arpt_getinfo)) {
1011                         duprintf("length %u != %Zu\n", *len,
1012                                  sizeof(struct arpt_getinfo));
1013                         ret = -EINVAL;
1014                         break;
1015                 }
1016
1017                 if (copy_from_user(name, user, sizeof(name)) != 0) {
1018                         ret = -EFAULT;
1019                         break;
1020                 }
1021                 name[ARPT_TABLE_MAXNAMELEN-1] = '\0';
1022
1023                 t = try_then_request_module(xt_find_table_lock(NF_ARP, name),
1024                                             "arptable_%s", name);
1025                 if (t && !IS_ERR(t)) {
1026                         struct arpt_getinfo info;
1027                         struct xt_table_info *private = t->private;
1028
1029                         info.valid_hooks = t->valid_hooks;
1030                         memcpy(info.hook_entry, private->hook_entry,
1031                                sizeof(info.hook_entry));
1032                         memcpy(info.underflow, private->underflow,
1033                                sizeof(info.underflow));
1034                         info.num_entries = private->number;
1035                         info.size = private->size;
1036                         strcpy(info.name, name);
1037
1038                         if (copy_to_user(user, &info, *len) != 0)
1039                                 ret = -EFAULT;
1040                         else
1041                                 ret = 0;
1042                         xt_table_unlock(t);
1043                         module_put(t->me);
1044                 } else
1045                         ret = t ? PTR_ERR(t) : -ENOENT;
1046         }
1047         break;
1048
1049         case ARPT_SO_GET_ENTRIES: {
1050                 struct arpt_get_entries get;
1051
1052                 if (*len < sizeof(get)) {
1053                         duprintf("get_entries: %u < %Zu\n", *len, sizeof(get));
1054                         ret = -EINVAL;
1055                 } else if (copy_from_user(&get, user, sizeof(get)) != 0) {
1056                         ret = -EFAULT;
1057                 } else if (*len != sizeof(struct arpt_get_entries) + get.size) {
1058                         duprintf("get_entries: %u != %Zu\n", *len,
1059                                  sizeof(struct arpt_get_entries) + get.size);
1060                         ret = -EINVAL;
1061                 } else
1062                         ret = get_entries(&get, user);
1063                 break;
1064         }
1065
1066         case ARPT_SO_GET_REVISION_TARGET: {
1067                 struct xt_get_revision rev;
1068
1069                 if (*len != sizeof(rev)) {
1070                         ret = -EINVAL;
1071                         break;
1072                 }
1073                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1074                         ret = -EFAULT;
1075                         break;
1076                 }
1077
1078                 try_then_request_module(xt_find_revision(NF_ARP, rev.name,
1079                                                          rev.revision, 1, &ret),
1080                                         "arpt_%s", rev.name);
1081                 break;
1082         }
1083
1084         default:
1085                 duprintf("do_arpt_get_ctl: unknown request %i\n", cmd);
1086                 ret = -EINVAL;
1087         }
1088
1089         return ret;
1090 }
1091
1092 int arpt_register_table(struct arpt_table *table,
1093                         const struct arpt_replace *repl)
1094 {
1095         int ret;
1096         struct xt_table_info *newinfo;
1097         static struct xt_table_info bootstrap
1098                 = { 0, 0, 0, { 0 }, { 0 }, { } };
1099         void *loc_cpu_entry;
1100
1101         newinfo = xt_alloc_table_info(repl->size);
1102         if (!newinfo) {
1103                 ret = -ENOMEM;
1104                 return ret;
1105         }
1106
1107         /* choose the copy on our node/cpu */
1108         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1109         memcpy(loc_cpu_entry, repl->entries, repl->size);
1110
1111         ret = translate_table(table->name, table->valid_hooks,
1112                               newinfo, loc_cpu_entry, repl->size,
1113                               repl->num_entries,
1114                               repl->hook_entry,
1115                               repl->underflow);
1116
1117         duprintf("arpt_register_table: translate table gives %d\n", ret);
1118         if (ret != 0) {
1119                 xt_free_table_info(newinfo);
1120                 return ret;
1121         }
1122
1123         if (xt_register_table(table, &bootstrap, newinfo) != 0) {
1124                 xt_free_table_info(newinfo);
1125                 return ret;
1126         }
1127
1128         return 0;
1129 }
1130
1131 void arpt_unregister_table(struct arpt_table *table)
1132 {
1133         struct xt_table_info *private;
1134         void *loc_cpu_entry;
1135
1136         private = xt_unregister_table(table);
1137
1138         /* Decrease module usage counts and free resources */
1139         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1140         ARPT_ENTRY_ITERATE(loc_cpu_entry, private->size,
1141                            cleanup_entry, NULL);
1142         xt_free_table_info(private);
1143 }
1144
1145 /* The built-in targets: standard (NULL) and error. */
1146 static struct arpt_target arpt_standard_target = {
1147         .name           = ARPT_STANDARD_TARGET,
1148         .targetsize     = sizeof(int),
1149         .family         = NF_ARP,
1150 };
1151
1152 static struct arpt_target arpt_error_target = {
1153         .name           = ARPT_ERROR_TARGET,
1154         .target         = arpt_error,
1155         .targetsize     = ARPT_FUNCTION_MAXNAMELEN,
1156         .family         = NF_ARP,
1157 };
1158
1159 static struct nf_sockopt_ops arpt_sockopts = {
1160         .pf             = PF_INET,
1161         .set_optmin     = ARPT_BASE_CTL,
1162         .set_optmax     = ARPT_SO_SET_MAX+1,
1163         .set            = do_arpt_set_ctl,
1164         .get_optmin     = ARPT_BASE_CTL,
1165         .get_optmax     = ARPT_SO_GET_MAX+1,
1166         .get            = do_arpt_get_ctl,
1167 };
1168
1169 static int __init arp_tables_init(void)
1170 {
1171         int ret;
1172
1173         xt_proto_init(NF_ARP);
1174
1175         /* Noone else will be downing sem now, so we won't sleep */
1176         xt_register_target(&arpt_standard_target);
1177         xt_register_target(&arpt_error_target);
1178
1179         /* Register setsockopt */
1180         ret = nf_register_sockopt(&arpt_sockopts);
1181         if (ret < 0) {
1182                 duprintf("Unable to register sockopts.\n");
1183                 return ret;
1184         }
1185
1186         printk("arp_tables: (C) 2002 David S. Miller\n");
1187         return 0;
1188 }
1189
1190 static void __exit arp_tables_fini(void)
1191 {
1192         nf_unregister_sockopt(&arpt_sockopts);
1193         xt_proto_fini(NF_ARP);
1194 }
1195
1196 EXPORT_SYMBOL(arpt_register_table);
1197 EXPORT_SYMBOL(arpt_unregister_table);
1198 EXPORT_SYMBOL(arpt_do_table);
1199
1200 module_init(arp_tables_init);
1201 module_exit(arp_tables_fini);