[NETFILTER] x_tables: Abstraction layer for {ip,ip6,arp}_tables
[linux-2.6.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 <asm/semaphore.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 void *targinfo,
212                                void *userinfo)
213 {
214         if (net_ratelimit())
215                 printk("arp_tables: error: '%s'\n", (char *)targinfo);
216
217         return NF_DROP;
218 }
219
220 static inline struct arpt_entry *get_entry(void *base, unsigned int offset)
221 {
222         return (struct arpt_entry *)(base + offset);
223 }
224
225 unsigned int arpt_do_table(struct sk_buff **pskb,
226                            unsigned int hook,
227                            const struct net_device *in,
228                            const struct net_device *out,
229                            struct arpt_table *table,
230                            void *userdata)
231 {
232         static const char nulldevname[IFNAMSIZ];
233         unsigned int verdict = NF_DROP;
234         struct arphdr *arp;
235         int hotdrop = 0;
236         struct arpt_entry *e, *back;
237         const char *indev, *outdev;
238         void *table_base;
239         struct xt_table_info *private = table->private;
240
241         /* ARP header, plus 2 device addresses, plus 2 IP addresses.  */
242         if (!pskb_may_pull((*pskb), (sizeof(struct arphdr) +
243                                      (2 * (*pskb)->dev->addr_len) +
244                                      (2 * sizeof(u32)))))
245                 return NF_DROP;
246
247         indev = in ? in->name : nulldevname;
248         outdev = out ? out->name : nulldevname;
249
250         read_lock_bh(&table->lock);
251         table_base = (void *)private->entries[smp_processor_id()];
252         e = get_entry(table_base, private->hook_entry[hook]);
253         back = get_entry(table_base, private->underflow[hook]);
254
255         arp = (*pskb)->nh.arph;
256         do {
257                 if (arp_packet_match(arp, (*pskb)->dev, indev, outdev, &e->arp)) {
258                         struct arpt_entry_target *t;
259                         int hdr_len;
260
261                         hdr_len = sizeof(*arp) + (2 * sizeof(struct in_addr)) +
262                                 (2 * (*pskb)->dev->addr_len);
263                         ADD_COUNTER(e->counters, hdr_len, 1);
264
265                         t = arpt_get_target(e);
266
267                         /* Standard target? */
268                         if (!t->u.kernel.target->target) {
269                                 int v;
270
271                                 v = ((struct arpt_standard_target *)t)->verdict;
272                                 if (v < 0) {
273                                         /* Pop from stack? */
274                                         if (v != ARPT_RETURN) {
275                                                 verdict = (unsigned)(-v) - 1;
276                                                 break;
277                                         }
278                                         e = back;
279                                         back = get_entry(table_base,
280                                                          back->comefrom);
281                                         continue;
282                                 }
283                                 if (table_base + v
284                                     != (void *)e + e->next_offset) {
285                                         /* Save old back ptr in next entry */
286                                         struct arpt_entry *next
287                                                 = (void *)e + e->next_offset;
288                                         next->comefrom =
289                                                 (void *)back - table_base;
290
291                                         /* set back pointer to next entry */
292                                         back = next;
293                                 }
294
295                                 e = get_entry(table_base, v);
296                         } else {
297                                 /* Targets which reenter must return
298                                  * abs. verdicts
299                                  */
300                                 verdict = t->u.kernel.target->target(pskb,
301                                                                      in, out,
302                                                                      hook,
303                                                                      t->data,
304                                                                      userdata);
305
306                                 /* Target might have changed stuff. */
307                                 arp = (*pskb)->nh.arph;
308
309                                 if (verdict == ARPT_CONTINUE)
310                                         e = (void *)e + e->next_offset;
311                                 else
312                                         /* Verdict */
313                                         break;
314                         }
315                 } else {
316                         e = (void *)e + e->next_offset;
317                 }
318         } while (!hotdrop);
319         read_unlock_bh(&table->lock);
320
321         if (hotdrop)
322                 return NF_DROP;
323         else
324                 return verdict;
325 }
326
327 /* All zeroes == unconditional rule. */
328 static inline int unconditional(const struct arpt_arp *arp)
329 {
330         unsigned int i;
331
332         for (i = 0; i < sizeof(*arp)/sizeof(__u32); i++)
333                 if (((__u32 *)arp)[i])
334                         return 0;
335
336         return 1;
337 }
338
339 /* Figures out from what hook each rule can be called: returns 0 if
340  * there are loops.  Puts hook bitmask in comefrom.
341  */
342 static int mark_source_chains(struct xt_table_info *newinfo,
343                               unsigned int valid_hooks, void *entry0)
344 {
345         unsigned int hook;
346
347         /* No recursion; use packet counter to save back ptrs (reset
348          * to 0 as we leave), and comefrom to save source hook bitmask.
349          */
350         for (hook = 0; hook < NF_ARP_NUMHOOKS; hook++) {
351                 unsigned int pos = newinfo->hook_entry[hook];
352                 struct arpt_entry *e
353                         = (struct arpt_entry *)(entry0 + pos);
354
355                 if (!(valid_hooks & (1 << hook)))
356                         continue;
357
358                 /* Set initial back pointer. */
359                 e->counters.pcnt = pos;
360
361                 for (;;) {
362                         struct arpt_standard_target *t
363                                 = (void *)arpt_get_target(e);
364
365                         if (e->comefrom & (1 << NF_ARP_NUMHOOKS)) {
366                                 printk("arptables: loop hook %u pos %u %08X.\n",
367                                        hook, pos, e->comefrom);
368                                 return 0;
369                         }
370                         e->comefrom
371                                 |= ((1 << hook) | (1 << NF_ARP_NUMHOOKS));
372
373                         /* Unconditional return/END. */
374                         if (e->target_offset == sizeof(struct arpt_entry)
375                             && (strcmp(t->target.u.user.name,
376                                        ARPT_STANDARD_TARGET) == 0)
377                             && t->verdict < 0
378                             && unconditional(&e->arp)) {
379                                 unsigned int oldpos, size;
380
381                                 /* Return: backtrack through the last
382                                  * big jump.
383                                  */
384                                 do {
385                                         e->comefrom ^= (1<<NF_ARP_NUMHOOKS);
386                                         oldpos = pos;
387                                         pos = e->counters.pcnt;
388                                         e->counters.pcnt = 0;
389
390                                         /* We're at the start. */
391                                         if (pos == oldpos)
392                                                 goto next;
393
394                                         e = (struct arpt_entry *)
395                                                 (entry0 + pos);
396                                 } while (oldpos == pos + e->next_offset);
397
398                                 /* Move along one */
399                                 size = e->next_offset;
400                                 e = (struct arpt_entry *)
401                                         (entry0 + pos + size);
402                                 e->counters.pcnt = pos;
403                                 pos += size;
404                         } else {
405                                 int newpos = t->verdict;
406
407                                 if (strcmp(t->target.u.user.name,
408                                            ARPT_STANDARD_TARGET) == 0
409                                     && newpos >= 0) {
410                                         /* This a jump; chase it. */
411                                         duprintf("Jump rule %u -> %u\n",
412                                                  pos, newpos);
413                                 } else {
414                                         /* ... this is a fallthru */
415                                         newpos = pos + e->next_offset;
416                                 }
417                                 e = (struct arpt_entry *)
418                                         (entry0 + newpos);
419                                 e->counters.pcnt = pos;
420                                 pos = newpos;
421                         }
422                 }
423                 next:
424                 duprintf("Finished chain %u\n", hook);
425         }
426         return 1;
427 }
428
429 static inline int standard_check(const struct arpt_entry_target *t,
430                                  unsigned int max_offset)
431 {
432         struct arpt_standard_target *targ = (void *)t;
433
434         /* Check standard info. */
435         if (t->u.target_size
436             != ARPT_ALIGN(sizeof(struct arpt_standard_target))) {
437                 duprintf("arpt_standard_check: target size %u != %Zu\n",
438                          t->u.target_size,
439                          ARPT_ALIGN(sizeof(struct arpt_standard_target)));
440                 return 0;
441         }
442
443         if (targ->verdict >= 0
444             && targ->verdict > max_offset - sizeof(struct arpt_entry)) {
445                 duprintf("arpt_standard_check: bad verdict (%i)\n",
446                          targ->verdict);
447                 return 0;
448         }
449
450         if (targ->verdict < -NF_MAX_VERDICT - 1) {
451                 duprintf("arpt_standard_check: bad negative verdict (%i)\n",
452                          targ->verdict);
453                 return 0;
454         }
455         return 1;
456 }
457
458 static struct arpt_target arpt_standard_target;
459
460 static inline int check_entry(struct arpt_entry *e, const char *name, unsigned int size,
461                               unsigned int *i)
462 {
463         struct arpt_entry_target *t;
464         struct arpt_target *target;
465         int ret;
466
467         if (!arp_checkentry(&e->arp)) {
468                 duprintf("arp_tables: arp check failed %p %s.\n", e, name);
469                 return -EINVAL;
470         }
471
472         t = arpt_get_target(e);
473         target = try_then_request_module(xt_find_target(NF_ARP, t->u.user.name,
474                                                         t->u.user.revision),
475                                          "arpt_%s", t->u.user.name);
476         if (IS_ERR(target) || !target) {
477                 duprintf("check_entry: `%s' not found\n", t->u.user.name);
478                 ret = target ? PTR_ERR(target) : -ENOENT;
479                 goto out;
480         }
481         t->u.kernel.target = target;
482
483         if (t->u.kernel.target == &arpt_standard_target) {
484                 if (!standard_check(t, size)) {
485                         ret = -EINVAL;
486                         goto out;
487                 }
488         } else if (t->u.kernel.target->checkentry
489                    && !t->u.kernel.target->checkentry(name, e, t->data,
490                                                       t->u.target_size
491                                                       - sizeof(*t),
492                                                       e->comefrom)) {
493                 module_put(t->u.kernel.target->me);
494                 duprintf("arp_tables: check failed for `%s'.\n",
495                          t->u.kernel.target->name);
496                 ret = -EINVAL;
497                 goto out;
498         }
499
500         (*i)++;
501         return 0;
502
503 out:
504         return ret;
505 }
506
507 static inline int check_entry_size_and_hooks(struct arpt_entry *e,
508                                              struct xt_table_info *newinfo,
509                                              unsigned char *base,
510                                              unsigned char *limit,
511                                              const unsigned int *hook_entries,
512                                              const unsigned int *underflows,
513                                              unsigned int *i)
514 {
515         unsigned int h;
516
517         if ((unsigned long)e % __alignof__(struct arpt_entry) != 0
518             || (unsigned char *)e + sizeof(struct arpt_entry) >= limit) {
519                 duprintf("Bad offset %p\n", e);
520                 return -EINVAL;
521         }
522
523         if (e->next_offset
524             < sizeof(struct arpt_entry) + sizeof(struct arpt_entry_target)) {
525                 duprintf("checking: element %p size %u\n",
526                          e, e->next_offset);
527                 return -EINVAL;
528         }
529
530         /* Check hooks & underflows */
531         for (h = 0; h < NF_ARP_NUMHOOKS; h++) {
532                 if ((unsigned char *)e - base == hook_entries[h])
533                         newinfo->hook_entry[h] = hook_entries[h];
534                 if ((unsigned char *)e - base == underflows[h])
535                         newinfo->underflow[h] = underflows[h];
536         }
537
538         /* FIXME: underflows must be unconditional, standard verdicts
539            < 0 (not ARPT_RETURN). --RR */
540
541         /* Clear counters and comefrom */
542         e->counters = ((struct xt_counters) { 0, 0 });
543         e->comefrom = 0;
544
545         (*i)++;
546         return 0;
547 }
548
549 static inline int cleanup_entry(struct arpt_entry *e, unsigned int *i)
550 {
551         struct arpt_entry_target *t;
552
553         if (i && (*i)-- == 0)
554                 return 1;
555
556         t = arpt_get_target(e);
557         if (t->u.kernel.target->destroy)
558                 t->u.kernel.target->destroy(t->data,
559                                             t->u.target_size - sizeof(*t));
560         module_put(t->u.kernel.target->me);
561         return 0;
562 }
563
564 /* Checks and translates the user-supplied table segment (held in
565  * newinfo).
566  */
567 static int translate_table(const char *name,
568                            unsigned int valid_hooks,
569                            struct xt_table_info *newinfo,
570                            void *entry0,
571                            unsigned int size,
572                            unsigned int number,
573                            const unsigned int *hook_entries,
574                            const unsigned int *underflows)
575 {
576         unsigned int i;
577         int ret;
578
579         newinfo->size = size;
580         newinfo->number = number;
581
582         /* Init all hooks to impossible value. */
583         for (i = 0; i < NF_ARP_NUMHOOKS; i++) {
584                 newinfo->hook_entry[i] = 0xFFFFFFFF;
585                 newinfo->underflow[i] = 0xFFFFFFFF;
586         }
587
588         duprintf("translate_table: size %u\n", newinfo->size);
589         i = 0;
590
591         /* Walk through entries, checking offsets. */
592         ret = ARPT_ENTRY_ITERATE(entry0, newinfo->size,
593                                  check_entry_size_and_hooks,
594                                  newinfo,
595                                  entry0,
596                                  entry0 + size,
597                                  hook_entries, underflows, &i);
598         duprintf("translate_table: ARPT_ENTRY_ITERATE gives %d\n", ret);
599         if (ret != 0)
600                 return ret;
601
602         if (i != number) {
603                 duprintf("translate_table: %u not %u entries\n",
604                          i, number);
605                 return -EINVAL;
606         }
607
608         /* Check hooks all assigned */
609         for (i = 0; i < NF_ARP_NUMHOOKS; i++) {
610                 /* Only hooks which are valid */
611                 if (!(valid_hooks & (1 << i)))
612                         continue;
613                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
614                         duprintf("Invalid hook entry %u %u\n",
615                                  i, hook_entries[i]);
616                         return -EINVAL;
617                 }
618                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
619                         duprintf("Invalid underflow %u %u\n",
620                                  i, underflows[i]);
621                         return -EINVAL;
622                 }
623         }
624
625         if (!mark_source_chains(newinfo, valid_hooks, entry0)) {
626                 duprintf("Looping hook\n");
627                 return -ELOOP;
628         }
629
630         /* Finally, each sanity check must pass */
631         i = 0;
632         ret = ARPT_ENTRY_ITERATE(entry0, newinfo->size,
633                                  check_entry, name, size, &i);
634
635         if (ret != 0) {
636                 ARPT_ENTRY_ITERATE(entry0, newinfo->size,
637                                    cleanup_entry, &i);
638                 return ret;
639         }
640
641         /* And one copy for every other CPU */
642         for_each_cpu(i) {
643                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
644                         memcpy(newinfo->entries[i], entry0, newinfo->size);
645         }
646
647         return ret;
648 }
649
650 /* Gets counters. */
651 static inline int add_entry_to_counter(const struct arpt_entry *e,
652                                        struct xt_counters total[],
653                                        unsigned int *i)
654 {
655         ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
656
657         (*i)++;
658         return 0;
659 }
660
661 static inline int set_entry_to_counter(const struct arpt_entry *e,
662                                        struct xt_counters total[],
663                                        unsigned int *i)
664 {
665         SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
666
667         (*i)++;
668         return 0;
669 }
670
671 static void get_counters(const struct xt_table_info *t,
672                          struct xt_counters counters[])
673 {
674         unsigned int cpu;
675         unsigned int i;
676         unsigned int curcpu;
677
678         /* Instead of clearing (by a previous call to memset())
679          * the counters and using adds, we set the counters
680          * with data used by 'current' CPU
681          * We dont care about preemption here.
682          */
683         curcpu = raw_smp_processor_id();
684
685         i = 0;
686         ARPT_ENTRY_ITERATE(t->entries[curcpu],
687                            t->size,
688                            set_entry_to_counter,
689                            counters,
690                            &i);
691
692         for_each_cpu(cpu) {
693                 if (cpu == curcpu)
694                         continue;
695                 i = 0;
696                 ARPT_ENTRY_ITERATE(t->entries[cpu],
697                                    t->size,
698                                    add_entry_to_counter,
699                                    counters,
700                                    &i);
701         }
702 }
703
704 static int copy_entries_to_user(unsigned int total_size,
705                                 struct arpt_table *table,
706                                 void __user *userptr)
707 {
708         unsigned int off, num, countersize;
709         struct arpt_entry *e;
710         struct xt_counters *counters;
711         struct xt_table_info *private = table->private;
712         int ret = 0;
713         void *loc_cpu_entry;
714
715         /* We need atomic snapshot of counters: rest doesn't change
716          * (other than comefrom, which userspace doesn't care
717          * about).
718          */
719         countersize = sizeof(struct xt_counters) * private->number;
720         counters = vmalloc_node(countersize, numa_node_id());
721
722         if (counters == NULL)
723                 return -ENOMEM;
724
725         /* First, sum counters... */
726         write_lock_bh(&table->lock);
727         get_counters(private, counters);
728         write_unlock_bh(&table->lock);
729
730         loc_cpu_entry = private->entries[raw_smp_processor_id()];
731         /* ... then copy entire thing ... */
732         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
733                 ret = -EFAULT;
734                 goto free_counters;
735         }
736
737         /* FIXME: use iterator macros --RR */
738         /* ... then go back and fix counters and names */
739         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
740                 struct arpt_entry_target *t;
741
742                 e = (struct arpt_entry *)(loc_cpu_entry + off);
743                 if (copy_to_user(userptr + off
744                                  + offsetof(struct arpt_entry, counters),
745                                  &counters[num],
746                                  sizeof(counters[num])) != 0) {
747                         ret = -EFAULT;
748                         goto free_counters;
749                 }
750
751                 t = arpt_get_target(e);
752                 if (copy_to_user(userptr + off + e->target_offset
753                                  + offsetof(struct arpt_entry_target,
754                                             u.user.name),
755                                  t->u.kernel.target->name,
756                                  strlen(t->u.kernel.target->name)+1) != 0) {
757                         ret = -EFAULT;
758                         goto free_counters;
759                 }
760         }
761
762  free_counters:
763         vfree(counters);
764         return ret;
765 }
766
767 static int get_entries(const struct arpt_get_entries *entries,
768                        struct arpt_get_entries __user *uptr)
769 {
770         int ret;
771         struct arpt_table *t;
772
773         t = xt_find_table_lock(NF_ARP, entries->name);
774         if (t || !IS_ERR(t)) {
775                 struct xt_table_info *private = t->private;
776                 duprintf("t->private->number = %u\n",
777                          private->number);
778                 if (entries->size == private->size)
779                         ret = copy_entries_to_user(private->size,
780                                                    t, uptr->entrytable);
781                 else {
782                         duprintf("get_entries: I've got %u not %u!\n",
783                                  private->size, entries->size);
784                         ret = -EINVAL;
785                 }
786                 module_put(t->me);
787                 xt_table_unlock(t);
788         } else
789                 ret = t ? PTR_ERR(t) : -ENOENT;
790
791         return ret;
792 }
793
794 static int do_replace(void __user *user, unsigned int len)
795 {
796         int ret;
797         struct arpt_replace tmp;
798         struct arpt_table *t;
799         struct xt_table_info *newinfo, *oldinfo;
800         struct xt_counters *counters;
801         void *loc_cpu_entry, *loc_cpu_old_entry;
802
803         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
804                 return -EFAULT;
805
806         /* Hack: Causes ipchains to give correct error msg --RR */
807         if (len != sizeof(tmp) + tmp.size)
808                 return -ENOPROTOOPT;
809
810         newinfo = xt_alloc_table_info(tmp.size);
811         if (!newinfo)
812                 return -ENOMEM;
813
814         /* choose the copy that is on our node/cpu */
815         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
816         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
817                            tmp.size) != 0) {
818                 ret = -EFAULT;
819                 goto free_newinfo;
820         }
821
822         counters = vmalloc(tmp.num_counters * sizeof(struct xt_counters));
823         if (!counters) {
824                 ret = -ENOMEM;
825                 goto free_newinfo;
826         }
827
828         ret = translate_table(tmp.name, tmp.valid_hooks,
829                               newinfo, loc_cpu_entry, tmp.size, tmp.num_entries,
830                               tmp.hook_entry, tmp.underflow);
831         if (ret != 0)
832                 goto free_newinfo_counters;
833
834         duprintf("arp_tables: Translated table\n");
835
836         t = try_then_request_module(xt_find_table_lock(NF_ARP, tmp.name),
837                                     "arptable_%s", tmp.name);
838         if (!t || IS_ERR(t)) {
839                 ret = t ? PTR_ERR(t) : -ENOENT;
840                 goto free_newinfo_counters_untrans;
841         }
842
843         /* You lied! */
844         if (tmp.valid_hooks != t->valid_hooks) {
845                 duprintf("Valid hook crap: %08X vs %08X\n",
846                          tmp.valid_hooks, t->valid_hooks);
847                 ret = -EINVAL;
848                 goto put_module;
849         }
850
851         oldinfo = xt_replace_table(t, tmp.num_counters, newinfo, &ret);
852         if (!oldinfo)
853                 goto put_module;
854
855         /* Update module usage count based on number of rules */
856         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
857                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
858         if ((oldinfo->number > oldinfo->initial_entries) || 
859             (newinfo->number <= oldinfo->initial_entries)) 
860                 module_put(t->me);
861         if ((oldinfo->number > oldinfo->initial_entries) &&
862             (newinfo->number <= oldinfo->initial_entries))
863                 module_put(t->me);
864
865         /* Get the old counters. */
866         get_counters(oldinfo, counters);
867         /* Decrease module usage counts and free resource */
868         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
869         ARPT_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,NULL);
870
871         xt_free_table_info(oldinfo);
872         if (copy_to_user(tmp.counters, counters,
873                          sizeof(struct xt_counters) * tmp.num_counters) != 0)
874                 ret = -EFAULT;
875         vfree(counters);
876         xt_table_unlock(t);
877         return ret;
878
879  put_module:
880         module_put(t->me);
881         xt_table_unlock(t);
882  free_newinfo_counters_untrans:
883         ARPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry, NULL);
884  free_newinfo_counters:
885         vfree(counters);
886  free_newinfo:
887         xt_free_table_info(newinfo);
888         return ret;
889 }
890
891 /* We're lazy, and add to the first CPU; overflow works its fey magic
892  * and everything is OK.
893  */
894 static inline int add_counter_to_entry(struct arpt_entry *e,
895                                        const struct xt_counters addme[],
896                                        unsigned int *i)
897 {
898
899         ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
900
901         (*i)++;
902         return 0;
903 }
904
905 static int do_add_counters(void __user *user, unsigned int len)
906 {
907         unsigned int i;
908         struct xt_counters_info tmp, *paddc;
909         struct arpt_table *t;
910         struct xt_table_info *private;
911         int ret = 0;
912         void *loc_cpu_entry;
913
914         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
915                 return -EFAULT;
916
917         if (len != sizeof(tmp) + tmp.num_counters*sizeof(struct xt_counters))
918                 return -EINVAL;
919
920         paddc = vmalloc(len);
921         if (!paddc)
922                 return -ENOMEM;
923
924         if (copy_from_user(paddc, user, len) != 0) {
925                 ret = -EFAULT;
926                 goto free;
927         }
928
929         t = xt_find_table_lock(NF_ARP, tmp.name);
930         if (!t || IS_ERR(t)) {
931                 ret = t ? PTR_ERR(t) : -ENOENT;
932                 goto free;
933         }
934
935         write_lock_bh(&t->lock);
936         private = t->private;
937         if (private->number != paddc->num_counters) {
938                 ret = -EINVAL;
939                 goto unlock_up_free;
940         }
941
942         i = 0;
943         /* Choose the copy that is on our node */
944         loc_cpu_entry = private->entries[smp_processor_id()];
945         ARPT_ENTRY_ITERATE(loc_cpu_entry,
946                            private->size,
947                            add_counter_to_entry,
948                            paddc->counters,
949                            &i);
950  unlock_up_free:
951         write_unlock_bh(&t->lock);
952         xt_table_unlock(t);
953         module_put(t->me);
954  free:
955         vfree(paddc);
956
957         return ret;
958 }
959
960 static int do_arpt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
961 {
962         int ret;
963
964         if (!capable(CAP_NET_ADMIN))
965                 return -EPERM;
966
967         switch (cmd) {
968         case ARPT_SO_SET_REPLACE:
969                 ret = do_replace(user, len);
970                 break;
971
972         case ARPT_SO_SET_ADD_COUNTERS:
973                 ret = do_add_counters(user, len);
974                 break;
975
976         default:
977                 duprintf("do_arpt_set_ctl:  unknown request %i\n", cmd);
978                 ret = -EINVAL;
979         }
980
981         return ret;
982 }
983
984 static int do_arpt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
985 {
986         int ret;
987
988         if (!capable(CAP_NET_ADMIN))
989                 return -EPERM;
990
991         switch (cmd) {
992         case ARPT_SO_GET_INFO: {
993                 char name[ARPT_TABLE_MAXNAMELEN];
994                 struct arpt_table *t;
995
996                 if (*len != sizeof(struct arpt_getinfo)) {
997                         duprintf("length %u != %Zu\n", *len,
998                                  sizeof(struct arpt_getinfo));
999                         ret = -EINVAL;
1000                         break;
1001                 }
1002
1003                 if (copy_from_user(name, user, sizeof(name)) != 0) {
1004                         ret = -EFAULT;
1005                         break;
1006                 }
1007                 name[ARPT_TABLE_MAXNAMELEN-1] = '\0';
1008
1009                 t = try_then_request_module(xt_find_table_lock(NF_ARP, name),
1010                                             "arptable_%s", name);
1011                 if (t && !IS_ERR(t)) {
1012                         struct arpt_getinfo info;
1013                         struct xt_table_info *private = t->private;
1014
1015                         info.valid_hooks = t->valid_hooks;
1016                         memcpy(info.hook_entry, private->hook_entry,
1017                                sizeof(info.hook_entry));
1018                         memcpy(info.underflow, private->underflow,
1019                                sizeof(info.underflow));
1020                         info.num_entries = private->number;
1021                         info.size = private->size;
1022                         strcpy(info.name, name);
1023
1024                         if (copy_to_user(user, &info, *len) != 0)
1025                                 ret = -EFAULT;
1026                         else
1027                                 ret = 0;
1028                         xt_table_unlock(t);
1029                         module_put(t->me);
1030                 } else
1031                         ret = t ? PTR_ERR(t) : -ENOENT;
1032         }
1033         break;
1034
1035         case ARPT_SO_GET_ENTRIES: {
1036                 struct arpt_get_entries get;
1037
1038                 if (*len < sizeof(get)) {
1039                         duprintf("get_entries: %u < %Zu\n", *len, sizeof(get));
1040                         ret = -EINVAL;
1041                 } else if (copy_from_user(&get, user, sizeof(get)) != 0) {
1042                         ret = -EFAULT;
1043                 } else if (*len != sizeof(struct arpt_get_entries) + get.size) {
1044                         duprintf("get_entries: %u != %Zu\n", *len,
1045                                  sizeof(struct arpt_get_entries) + get.size);
1046                         ret = -EINVAL;
1047                 } else
1048                         ret = get_entries(&get, user);
1049                 break;
1050         }
1051
1052         case ARPT_SO_GET_REVISION_TARGET: {
1053                 struct xt_get_revision rev;
1054
1055                 if (*len != sizeof(rev)) {
1056                         ret = -EINVAL;
1057                         break;
1058                 }
1059                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1060                         ret = -EFAULT;
1061                         break;
1062                 }
1063
1064                 try_then_request_module(xt_find_revision(NF_ARP, rev.name,
1065                                                          rev.revision, 1, &ret),
1066                                         "arpt_%s", rev.name);
1067                 break;
1068         }
1069
1070         default:
1071                 duprintf("do_arpt_get_ctl: unknown request %i\n", cmd);
1072                 ret = -EINVAL;
1073         }
1074
1075         return ret;
1076 }
1077
1078 int arpt_register_table(struct arpt_table *table,
1079                         const struct arpt_replace *repl)
1080 {
1081         int ret;
1082         struct xt_table_info *newinfo;
1083         static struct xt_table_info bootstrap
1084                 = { 0, 0, 0, { 0 }, { 0 }, { } };
1085         void *loc_cpu_entry;
1086
1087         newinfo = xt_alloc_table_info(repl->size);
1088         if (!newinfo) {
1089                 ret = -ENOMEM;
1090                 return ret;
1091         }
1092
1093         /* choose the copy on our node/cpu */
1094         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1095         memcpy(loc_cpu_entry, repl->entries, repl->size);
1096
1097         ret = translate_table(table->name, table->valid_hooks,
1098                               newinfo, loc_cpu_entry, repl->size,
1099                               repl->num_entries,
1100                               repl->hook_entry,
1101                               repl->underflow);
1102
1103         duprintf("arpt_register_table: translate table gives %d\n", ret);
1104         if (ret != 0) {
1105                 xt_free_table_info(newinfo);
1106                 return ret;
1107         }
1108
1109         if (xt_register_table(table, &bootstrap, newinfo) != 0) {
1110                 xt_free_table_info(newinfo);
1111                 return ret;
1112         }
1113
1114         return 0;
1115 }
1116
1117 void arpt_unregister_table(struct arpt_table *table)
1118 {
1119         struct xt_table_info *private;
1120         void *loc_cpu_entry;
1121
1122         private = xt_unregister_table(table);
1123
1124         /* Decrease module usage counts and free resources */
1125         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1126         ARPT_ENTRY_ITERATE(loc_cpu_entry, private->size,
1127                            cleanup_entry, NULL);
1128         xt_free_table_info(private);
1129 }
1130
1131 /* The built-in targets: standard (NULL) and error. */
1132 static struct arpt_target arpt_standard_target = {
1133         .name           = ARPT_STANDARD_TARGET,
1134 };
1135
1136 static struct arpt_target arpt_error_target = {
1137         .name           = ARPT_ERROR_TARGET,
1138         .target         = arpt_error,
1139 };
1140
1141 static struct nf_sockopt_ops arpt_sockopts = {
1142         .pf             = PF_INET,
1143         .set_optmin     = ARPT_BASE_CTL,
1144         .set_optmax     = ARPT_SO_SET_MAX+1,
1145         .set            = do_arpt_set_ctl,
1146         .get_optmin     = ARPT_BASE_CTL,
1147         .get_optmax     = ARPT_SO_GET_MAX+1,
1148         .get            = do_arpt_get_ctl,
1149 };
1150
1151 static int __init init(void)
1152 {
1153         int ret;
1154
1155         xt_proto_init(NF_ARP);
1156
1157         /* Noone else will be downing sem now, so we won't sleep */
1158         xt_register_target(NF_ARP, &arpt_standard_target);
1159         xt_register_target(NF_ARP, &arpt_error_target);
1160
1161         /* Register setsockopt */
1162         ret = nf_register_sockopt(&arpt_sockopts);
1163         if (ret < 0) {
1164                 duprintf("Unable to register sockopts.\n");
1165                 return ret;
1166         }
1167
1168         printk("arp_tables: (C) 2002 David S. Miller\n");
1169         return 0;
1170 }
1171
1172 static void __exit fini(void)
1173 {
1174         nf_unregister_sockopt(&arpt_sockopts);
1175         xt_proto_fini(NF_ARP);
1176 }
1177
1178 EXPORT_SYMBOL(arpt_register_table);
1179 EXPORT_SYMBOL(arpt_unregister_table);
1180 EXPORT_SYMBOL(arpt_do_table);
1181
1182 module_init(init);
1183 module_exit(fini);