Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Packet matching code. | |
3 | * | |
4 | * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling | |
6b7d31fc | 5 | * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org> |
1da177e4 LT |
6 | * |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | * | |
11 | * 19 Jan 2002 Harald Welte <laforge@gnumonks.org> | |
12 | * - increase module usage count as soon as we have rules inside | |
13 | * a table | |
14 | * 06 Jun 2002 Andras Kis-Szabo <kisza@sch.bme.hu> | |
15 | * - new extension header parser code | |
16 | */ | |
17 | #include <linux/config.h> | |
18 | #include <linux/skbuff.h> | |
19 | #include <linux/kmod.h> | |
20 | #include <linux/vmalloc.h> | |
21 | #include <linux/netdevice.h> | |
22 | #include <linux/module.h> | |
23 | #include <linux/tcp.h> | |
24 | #include <linux/udp.h> | |
25 | #include <linux/icmpv6.h> | |
1da177e4 LT |
26 | #include <net/ipv6.h> |
27 | #include <asm/uaccess.h> | |
28 | #include <asm/semaphore.h> | |
29 | #include <linux/proc_fs.h> | |
c8923c6b | 30 | #include <linux/cpumask.h> |
1da177e4 LT |
31 | |
32 | #include <linux/netfilter_ipv6/ip6_tables.h> | |
33 | ||
34 | MODULE_LICENSE("GPL"); | |
35 | MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>"); | |
36 | MODULE_DESCRIPTION("IPv6 packet filter"); | |
37 | ||
38 | #define IPV6_HDR_LEN (sizeof(struct ipv6hdr)) | |
39 | #define IPV6_OPTHDR_LEN (sizeof(struct ipv6_opt_hdr)) | |
40 | ||
41 | /*#define DEBUG_IP_FIREWALL*/ | |
42 | /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */ | |
43 | /*#define DEBUG_IP_FIREWALL_USER*/ | |
44 | ||
45 | #ifdef DEBUG_IP_FIREWALL | |
46 | #define dprintf(format, args...) printk(format , ## args) | |
47 | #else | |
48 | #define dprintf(format, args...) | |
49 | #endif | |
50 | ||
51 | #ifdef DEBUG_IP_FIREWALL_USER | |
52 | #define duprintf(format, args...) printk(format , ## args) | |
53 | #else | |
54 | #define duprintf(format, args...) | |
55 | #endif | |
56 | ||
57 | #ifdef CONFIG_NETFILTER_DEBUG | |
58 | #define IP_NF_ASSERT(x) \ | |
59 | do { \ | |
60 | if (!(x)) \ | |
61 | printk("IP_NF_ASSERT: %s:%s:%u\n", \ | |
62 | __FUNCTION__, __FILE__, __LINE__); \ | |
63 | } while(0) | |
64 | #else | |
65 | #define IP_NF_ASSERT(x) | |
66 | #endif | |
67 | #define SMP_ALIGN(x) (((x) + SMP_CACHE_BYTES-1) & ~(SMP_CACHE_BYTES-1)) | |
68 | ||
69 | static DECLARE_MUTEX(ip6t_mutex); | |
70 | ||
71 | /* Must have mutex */ | |
72 | #define ASSERT_READ_LOCK(x) IP_NF_ASSERT(down_trylock(&ip6t_mutex) != 0) | |
73 | #define ASSERT_WRITE_LOCK(x) IP_NF_ASSERT(down_trylock(&ip6t_mutex) != 0) | |
1da177e4 LT |
74 | #include <linux/netfilter_ipv4/listhelp.h> |
75 | ||
76 | #if 0 | |
77 | /* All the better to debug you with... */ | |
78 | #define static | |
79 | #define inline | |
80 | #endif | |
81 | ||
6b7d31fc | 82 | /* |
1da177e4 | 83 | We keep a set of rules for each CPU, so we can avoid write-locking |
6b7d31fc HW |
84 | them in the softirq when updating the counters and therefore |
85 | only need to read-lock in the softirq; doing a write_lock_bh() in user | |
86 | context stops packets coming through and allows user context to read | |
87 | the counters or update the rules. | |
1da177e4 | 88 | |
1da177e4 LT |
89 | Hence the start of any table is given by get_table() below. */ |
90 | ||
91 | /* The table itself */ | |
92 | struct ip6t_table_info | |
93 | { | |
94 | /* Size per table */ | |
95 | unsigned int size; | |
96 | /* Number of entries: FIXME. --RR */ | |
97 | unsigned int number; | |
98 | /* Initial number of entries. Needed for module usage count */ | |
99 | unsigned int initial_entries; | |
100 | ||
101 | /* Entry points and underflows */ | |
102 | unsigned int hook_entry[NF_IP6_NUMHOOKS]; | |
103 | unsigned int underflow[NF_IP6_NUMHOOKS]; | |
104 | ||
105 | /* ip6t_entry tables: one per CPU */ | |
31836064 | 106 | void *entries[NR_CPUS]; |
1da177e4 LT |
107 | }; |
108 | ||
109 | static LIST_HEAD(ip6t_target); | |
110 | static LIST_HEAD(ip6t_match); | |
111 | static LIST_HEAD(ip6t_tables); | |
31836064 | 112 | #define SET_COUNTER(c,b,p) do { (c).bcnt = (b); (c).pcnt = (p); } while(0) |
1da177e4 LT |
113 | #define ADD_COUNTER(c,b,p) do { (c).bcnt += (b); (c).pcnt += (p); } while(0) |
114 | ||
1da177e4 LT |
115 | #if 0 |
116 | #define down(x) do { printk("DOWN:%u:" #x "\n", __LINE__); down(x); } while(0) | |
117 | #define down_interruptible(x) ({ int __r; printk("DOWNi:%u:" #x "\n", __LINE__); __r = down_interruptible(x); if (__r != 0) printk("ABORT-DOWNi:%u\n", __LINE__); __r; }) | |
118 | #define up(x) do { printk("UP:%u:" #x "\n", __LINE__); up(x); } while(0) | |
119 | #endif | |
120 | ||
121 | static int ip6_masked_addrcmp(struct in6_addr addr1, struct in6_addr mask, | |
122 | struct in6_addr addr2) | |
123 | { | |
124 | int i; | |
125 | for( i = 0; i < 16; i++){ | |
126 | if((addr1.s6_addr[i] & mask.s6_addr[i]) != | |
127 | (addr2.s6_addr[i] & mask.s6_addr[i])) | |
128 | return 1; | |
129 | } | |
130 | return 0; | |
131 | } | |
132 | ||
133 | /* Check for an extension */ | |
134 | int | |
135 | ip6t_ext_hdr(u8 nexthdr) | |
136 | { | |
137 | return ( (nexthdr == IPPROTO_HOPOPTS) || | |
138 | (nexthdr == IPPROTO_ROUTING) || | |
139 | (nexthdr == IPPROTO_FRAGMENT) || | |
140 | (nexthdr == IPPROTO_ESP) || | |
141 | (nexthdr == IPPROTO_AH) || | |
142 | (nexthdr == IPPROTO_NONE) || | |
143 | (nexthdr == IPPROTO_DSTOPTS) ); | |
144 | } | |
145 | ||
146 | /* Returns whether matches rule or not. */ | |
147 | static inline int | |
148 | ip6_packet_match(const struct sk_buff *skb, | |
149 | const char *indev, | |
150 | const char *outdev, | |
151 | const struct ip6t_ip6 *ip6info, | |
152 | unsigned int *protoff, | |
153 | int *fragoff) | |
154 | { | |
155 | size_t i; | |
156 | unsigned long ret; | |
157 | const struct ipv6hdr *ipv6 = skb->nh.ipv6h; | |
158 | ||
159 | #define FWINV(bool,invflg) ((bool) ^ !!(ip6info->invflags & invflg)) | |
160 | ||
161 | if (FWINV(ip6_masked_addrcmp(ipv6->saddr,ip6info->smsk,ip6info->src), | |
162 | IP6T_INV_SRCIP) | |
163 | || FWINV(ip6_masked_addrcmp(ipv6->daddr,ip6info->dmsk,ip6info->dst), | |
164 | IP6T_INV_DSTIP)) { | |
165 | dprintf("Source or dest mismatch.\n"); | |
166 | /* | |
167 | dprintf("SRC: %u. Mask: %u. Target: %u.%s\n", ip->saddr, | |
168 | ipinfo->smsk.s_addr, ipinfo->src.s_addr, | |
169 | ipinfo->invflags & IP6T_INV_SRCIP ? " (INV)" : ""); | |
170 | dprintf("DST: %u. Mask: %u. Target: %u.%s\n", ip->daddr, | |
171 | ipinfo->dmsk.s_addr, ipinfo->dst.s_addr, | |
172 | ipinfo->invflags & IP6T_INV_DSTIP ? " (INV)" : "");*/ | |
173 | return 0; | |
174 | } | |
175 | ||
176 | /* Look for ifname matches; this should unroll nicely. */ | |
177 | for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) { | |
178 | ret |= (((const unsigned long *)indev)[i] | |
179 | ^ ((const unsigned long *)ip6info->iniface)[i]) | |
180 | & ((const unsigned long *)ip6info->iniface_mask)[i]; | |
181 | } | |
182 | ||
183 | if (FWINV(ret != 0, IP6T_INV_VIA_IN)) { | |
184 | dprintf("VIA in mismatch (%s vs %s).%s\n", | |
185 | indev, ip6info->iniface, | |
186 | ip6info->invflags&IP6T_INV_VIA_IN ?" (INV)":""); | |
187 | return 0; | |
188 | } | |
189 | ||
190 | for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) { | |
191 | ret |= (((const unsigned long *)outdev)[i] | |
192 | ^ ((const unsigned long *)ip6info->outiface)[i]) | |
193 | & ((const unsigned long *)ip6info->outiface_mask)[i]; | |
194 | } | |
195 | ||
196 | if (FWINV(ret != 0, IP6T_INV_VIA_OUT)) { | |
197 | dprintf("VIA out mismatch (%s vs %s).%s\n", | |
198 | outdev, ip6info->outiface, | |
199 | ip6info->invflags&IP6T_INV_VIA_OUT ?" (INV)":""); | |
200 | return 0; | |
201 | } | |
202 | ||
203 | /* ... might want to do something with class and flowlabel here ... */ | |
204 | ||
205 | /* look for the desired protocol header */ | |
206 | if((ip6info->flags & IP6T_F_PROTO)) { | |
207 | u_int8_t currenthdr = ipv6->nexthdr; | |
208 | struct ipv6_opt_hdr _hdr, *hp; | |
209 | u_int16_t ptr; /* Header offset in skb */ | |
210 | u_int16_t hdrlen; /* Header */ | |
211 | u_int16_t _fragoff = 0, *fp = NULL; | |
212 | ||
213 | ptr = IPV6_HDR_LEN; | |
214 | ||
215 | while (ip6t_ext_hdr(currenthdr)) { | |
216 | /* Is there enough space for the next ext header? */ | |
217 | if (skb->len - ptr < IPV6_OPTHDR_LEN) | |
218 | return 0; | |
219 | ||
220 | /* NONE or ESP: there isn't protocol part */ | |
221 | /* If we want to count these packets in '-p all', | |
222 | * we will change the return 0 to 1*/ | |
223 | if ((currenthdr == IPPROTO_NONE) || | |
224 | (currenthdr == IPPROTO_ESP)) | |
225 | break; | |
226 | ||
227 | hp = skb_header_pointer(skb, ptr, sizeof(_hdr), &_hdr); | |
228 | BUG_ON(hp == NULL); | |
229 | ||
230 | /* Size calculation */ | |
231 | if (currenthdr == IPPROTO_FRAGMENT) { | |
232 | fp = skb_header_pointer(skb, | |
233 | ptr+offsetof(struct frag_hdr, | |
234 | frag_off), | |
235 | sizeof(_fragoff), | |
236 | &_fragoff); | |
237 | if (fp == NULL) | |
238 | return 0; | |
239 | ||
240 | _fragoff = ntohs(*fp) & ~0x7; | |
241 | hdrlen = 8; | |
242 | } else if (currenthdr == IPPROTO_AH) | |
243 | hdrlen = (hp->hdrlen+2)<<2; | |
244 | else | |
245 | hdrlen = ipv6_optlen(hp); | |
246 | ||
247 | currenthdr = hp->nexthdr; | |
248 | ptr += hdrlen; | |
249 | /* ptr is too large */ | |
250 | if ( ptr > skb->len ) | |
251 | return 0; | |
252 | if (_fragoff) { | |
253 | if (ip6t_ext_hdr(currenthdr)) | |
254 | return 0; | |
255 | break; | |
256 | } | |
257 | } | |
258 | ||
259 | *protoff = ptr; | |
260 | *fragoff = _fragoff; | |
261 | ||
262 | /* currenthdr contains the protocol header */ | |
263 | ||
264 | dprintf("Packet protocol %hi ?= %s%hi.\n", | |
265 | currenthdr, | |
266 | ip6info->invflags & IP6T_INV_PROTO ? "!":"", | |
267 | ip6info->proto); | |
268 | ||
269 | if (ip6info->proto == currenthdr) { | |
270 | if(ip6info->invflags & IP6T_INV_PROTO) { | |
271 | return 0; | |
272 | } | |
273 | return 1; | |
274 | } | |
275 | ||
276 | /* We need match for the '-p all', too! */ | |
277 | if ((ip6info->proto != 0) && | |
278 | !(ip6info->invflags & IP6T_INV_PROTO)) | |
279 | return 0; | |
280 | } | |
281 | return 1; | |
282 | } | |
283 | ||
284 | /* should be ip6 safe */ | |
285 | static inline int | |
286 | ip6_checkentry(const struct ip6t_ip6 *ipv6) | |
287 | { | |
288 | if (ipv6->flags & ~IP6T_F_MASK) { | |
289 | duprintf("Unknown flag bits set: %08X\n", | |
290 | ipv6->flags & ~IP6T_F_MASK); | |
291 | return 0; | |
292 | } | |
293 | if (ipv6->invflags & ~IP6T_INV_MASK) { | |
294 | duprintf("Unknown invflag bits set: %08X\n", | |
295 | ipv6->invflags & ~IP6T_INV_MASK); | |
296 | return 0; | |
297 | } | |
298 | return 1; | |
299 | } | |
300 | ||
301 | static unsigned int | |
302 | ip6t_error(struct sk_buff **pskb, | |
303 | const struct net_device *in, | |
304 | const struct net_device *out, | |
305 | unsigned int hooknum, | |
306 | const void *targinfo, | |
307 | void *userinfo) | |
308 | { | |
309 | if (net_ratelimit()) | |
310 | printk("ip6_tables: error: `%s'\n", (char *)targinfo); | |
311 | ||
312 | return NF_DROP; | |
313 | } | |
314 | ||
315 | static inline | |
316 | int do_match(struct ip6t_entry_match *m, | |
317 | const struct sk_buff *skb, | |
318 | const struct net_device *in, | |
319 | const struct net_device *out, | |
320 | int offset, | |
321 | unsigned int protoff, | |
322 | int *hotdrop) | |
323 | { | |
324 | /* Stop iteration if it doesn't match */ | |
325 | if (!m->u.kernel.match->match(skb, in, out, m->data, | |
326 | offset, protoff, hotdrop)) | |
327 | return 1; | |
328 | else | |
329 | return 0; | |
330 | } | |
331 | ||
332 | static inline struct ip6t_entry * | |
333 | get_entry(void *base, unsigned int offset) | |
334 | { | |
335 | return (struct ip6t_entry *)(base + offset); | |
336 | } | |
337 | ||
338 | /* Returns one of the generic firewall policies, like NF_ACCEPT. */ | |
339 | unsigned int | |
340 | ip6t_do_table(struct sk_buff **pskb, | |
341 | unsigned int hook, | |
342 | const struct net_device *in, | |
343 | const struct net_device *out, | |
344 | struct ip6t_table *table, | |
345 | void *userdata) | |
346 | { | |
6b7d31fc | 347 | static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long)))); |
1da177e4 LT |
348 | int offset = 0; |
349 | unsigned int protoff = 0; | |
350 | int hotdrop = 0; | |
351 | /* Initializing verdict to NF_DROP keeps gcc happy. */ | |
352 | unsigned int verdict = NF_DROP; | |
353 | const char *indev, *outdev; | |
354 | void *table_base; | |
355 | struct ip6t_entry *e, *back; | |
356 | ||
357 | /* Initialization */ | |
358 | indev = in ? in->name : nulldevname; | |
359 | outdev = out ? out->name : nulldevname; | |
1da177e4 LT |
360 | /* We handle fragments by dealing with the first fragment as |
361 | * if it was a normal packet. All other fragments are treated | |
362 | * normally, except that they will NEVER match rules that ask | |
363 | * things we don't know, ie. tcp syn flag or ports). If the | |
364 | * rule is also a fragment-specific rule, non-fragments won't | |
365 | * match it. */ | |
366 | ||
367 | read_lock_bh(&table->lock); | |
368 | IP_NF_ASSERT(table->valid_hooks & (1 << hook)); | |
31836064 | 369 | table_base = (void *)table->private->entries[smp_processor_id()]; |
1da177e4 LT |
370 | e = get_entry(table_base, table->private->hook_entry[hook]); |
371 | ||
372 | #ifdef CONFIG_NETFILTER_DEBUG | |
373 | /* Check noone else using our table */ | |
374 | if (((struct ip6t_entry *)table_base)->comefrom != 0xdead57ac | |
375 | && ((struct ip6t_entry *)table_base)->comefrom != 0xeeeeeeec) { | |
376 | printk("ASSERT: CPU #%u, %s comefrom(%p) = %X\n", | |
377 | smp_processor_id(), | |
378 | table->name, | |
379 | &((struct ip6t_entry *)table_base)->comefrom, | |
380 | ((struct ip6t_entry *)table_base)->comefrom); | |
381 | } | |
382 | ((struct ip6t_entry *)table_base)->comefrom = 0x57acc001; | |
383 | #endif | |
384 | ||
385 | /* For return from builtin chain */ | |
386 | back = get_entry(table_base, table->private->underflow[hook]); | |
387 | ||
388 | do { | |
389 | IP_NF_ASSERT(e); | |
390 | IP_NF_ASSERT(back); | |
1da177e4 LT |
391 | if (ip6_packet_match(*pskb, indev, outdev, &e->ipv6, |
392 | &protoff, &offset)) { | |
393 | struct ip6t_entry_target *t; | |
394 | ||
395 | if (IP6T_MATCH_ITERATE(e, do_match, | |
396 | *pskb, in, out, | |
397 | offset, protoff, &hotdrop) != 0) | |
398 | goto no_match; | |
399 | ||
400 | ADD_COUNTER(e->counters, | |
401 | ntohs((*pskb)->nh.ipv6h->payload_len) | |
402 | + IPV6_HDR_LEN, | |
403 | 1); | |
404 | ||
405 | t = ip6t_get_target(e); | |
406 | IP_NF_ASSERT(t->u.kernel.target); | |
407 | /* Standard target? */ | |
408 | if (!t->u.kernel.target->target) { | |
409 | int v; | |
410 | ||
411 | v = ((struct ip6t_standard_target *)t)->verdict; | |
412 | if (v < 0) { | |
413 | /* Pop from stack? */ | |
414 | if (v != IP6T_RETURN) { | |
415 | verdict = (unsigned)(-v) - 1; | |
416 | break; | |
417 | } | |
418 | e = back; | |
419 | back = get_entry(table_base, | |
420 | back->comefrom); | |
421 | continue; | |
422 | } | |
05465343 PM |
423 | if (table_base + v != (void *)e + e->next_offset |
424 | && !(e->ipv6.flags & IP6T_F_GOTO)) { | |
1da177e4 LT |
425 | /* Save old back ptr in next entry */ |
426 | struct ip6t_entry *next | |
427 | = (void *)e + e->next_offset; | |
428 | next->comefrom | |
429 | = (void *)back - table_base; | |
430 | /* set back pointer to next entry */ | |
431 | back = next; | |
432 | } | |
433 | ||
434 | e = get_entry(table_base, v); | |
435 | } else { | |
436 | /* Targets which reenter must return | |
437 | abs. verdicts */ | |
438 | #ifdef CONFIG_NETFILTER_DEBUG | |
439 | ((struct ip6t_entry *)table_base)->comefrom | |
440 | = 0xeeeeeeec; | |
441 | #endif | |
442 | verdict = t->u.kernel.target->target(pskb, | |
443 | in, out, | |
444 | hook, | |
445 | t->data, | |
446 | userdata); | |
447 | ||
448 | #ifdef CONFIG_NETFILTER_DEBUG | |
449 | if (((struct ip6t_entry *)table_base)->comefrom | |
450 | != 0xeeeeeeec | |
451 | && verdict == IP6T_CONTINUE) { | |
452 | printk("Target %s reentered!\n", | |
453 | t->u.kernel.target->name); | |
454 | verdict = NF_DROP; | |
455 | } | |
456 | ((struct ip6t_entry *)table_base)->comefrom | |
457 | = 0x57acc001; | |
458 | #endif | |
459 | if (verdict == IP6T_CONTINUE) | |
460 | e = (void *)e + e->next_offset; | |
461 | else | |
462 | /* Verdict */ | |
463 | break; | |
464 | } | |
465 | } else { | |
466 | ||
467 | no_match: | |
468 | e = (void *)e + e->next_offset; | |
469 | } | |
470 | } while (!hotdrop); | |
471 | ||
472 | #ifdef CONFIG_NETFILTER_DEBUG | |
473 | ((struct ip6t_entry *)table_base)->comefrom = 0xdead57ac; | |
474 | #endif | |
475 | read_unlock_bh(&table->lock); | |
476 | ||
477 | #ifdef DEBUG_ALLOW_ALL | |
478 | return NF_ACCEPT; | |
479 | #else | |
480 | if (hotdrop) | |
481 | return NF_DROP; | |
482 | else return verdict; | |
483 | #endif | |
484 | } | |
485 | ||
6b7d31fc HW |
486 | /* |
487 | * These are weird, but module loading must not be done with mutex | |
488 | * held (since they will register), and we have to have a single | |
489 | * function to use try_then_request_module(). | |
490 | */ | |
491 | ||
492 | /* Find table by name, grabs mutex & ref. Returns ERR_PTR() on error. */ | |
493 | static inline struct ip6t_table *find_table_lock(const char *name) | |
1da177e4 | 494 | { |
6b7d31fc | 495 | struct ip6t_table *t; |
1da177e4 | 496 | |
6b7d31fc HW |
497 | if (down_interruptible(&ip6t_mutex) != 0) |
498 | return ERR_PTR(-EINTR); | |
1da177e4 | 499 | |
6b7d31fc HW |
500 | list_for_each_entry(t, &ip6t_tables, list) |
501 | if (strcmp(t->name, name) == 0 && try_module_get(t->me)) | |
502 | return t; | |
503 | up(&ip6t_mutex); | |
504 | return NULL; | |
505 | } | |
1da177e4 | 506 | |
6b7d31fc HW |
507 | /* Find match, grabs ref. Returns ERR_PTR() on error. */ |
508 | static inline struct ip6t_match *find_match(const char *name, u8 revision) | |
509 | { | |
510 | struct ip6t_match *m; | |
511 | int err = 0; | |
512 | ||
513 | if (down_interruptible(&ip6t_mutex) != 0) | |
514 | return ERR_PTR(-EINTR); | |
515 | ||
516 | list_for_each_entry(m, &ip6t_match, list) { | |
517 | if (strcmp(m->name, name) == 0) { | |
518 | if (m->revision == revision) { | |
519 | if (try_module_get(m->me)) { | |
520 | up(&ip6t_mutex); | |
521 | return m; | |
522 | } | |
523 | } else | |
524 | err = -EPROTOTYPE; /* Found something. */ | |
525 | } | |
1da177e4 | 526 | } |
6b7d31fc HW |
527 | up(&ip6t_mutex); |
528 | return ERR_PTR(err); | |
1da177e4 LT |
529 | } |
530 | ||
6b7d31fc HW |
531 | /* Find target, grabs ref. Returns ERR_PTR() on error. */ |
532 | static inline struct ip6t_target *find_target(const char *name, u8 revision) | |
1da177e4 | 533 | { |
6b7d31fc HW |
534 | struct ip6t_target *t; |
535 | int err = 0; | |
536 | ||
537 | if (down_interruptible(&ip6t_mutex) != 0) | |
538 | return ERR_PTR(-EINTR); | |
539 | ||
540 | list_for_each_entry(t, &ip6t_target, list) { | |
541 | if (strcmp(t->name, name) == 0) { | |
542 | if (t->revision == revision) { | |
543 | if (try_module_get(t->me)) { | |
544 | up(&ip6t_mutex); | |
545 | return t; | |
546 | } | |
547 | } else | |
548 | err = -EPROTOTYPE; /* Found something. */ | |
549 | } | |
1da177e4 | 550 | } |
6b7d31fc HW |
551 | up(&ip6t_mutex); |
552 | return ERR_PTR(err); | |
553 | } | |
1da177e4 | 554 | |
6b7d31fc HW |
555 | struct ip6t_target *ip6t_find_target(const char *name, u8 revision) |
556 | { | |
557 | struct ip6t_target *target; | |
558 | ||
559 | target = try_then_request_module(find_target(name, revision), | |
560 | "ip6t_%s", name); | |
561 | if (IS_ERR(target) || !target) | |
562 | return NULL; | |
563 | return target; | |
1da177e4 | 564 | } |
1da177e4 | 565 | |
6b7d31fc | 566 | static int match_revfn(const char *name, u8 revision, int *bestp) |
1da177e4 | 567 | { |
6b7d31fc HW |
568 | struct ip6t_match *m; |
569 | int have_rev = 0; | |
570 | ||
571 | list_for_each_entry(m, &ip6t_match, list) { | |
572 | if (strcmp(m->name, name) == 0) { | |
573 | if (m->revision > *bestp) | |
574 | *bestp = m->revision; | |
575 | if (m->revision == revision) | |
576 | have_rev = 1; | |
577 | } | |
578 | } | |
579 | return have_rev; | |
1da177e4 LT |
580 | } |
581 | ||
6b7d31fc | 582 | static int target_revfn(const char *name, u8 revision, int *bestp) |
1da177e4 | 583 | { |
6b7d31fc HW |
584 | struct ip6t_target *t; |
585 | int have_rev = 0; | |
586 | ||
587 | list_for_each_entry(t, &ip6t_target, list) { | |
588 | if (strcmp(t->name, name) == 0) { | |
589 | if (t->revision > *bestp) | |
590 | *bestp = t->revision; | |
591 | if (t->revision == revision) | |
592 | have_rev = 1; | |
593 | } | |
594 | } | |
595 | return have_rev; | |
1da177e4 LT |
596 | } |
597 | ||
6b7d31fc HW |
598 | /* Returns true or fals (if no such extension at all) */ |
599 | static inline int find_revision(const char *name, u8 revision, | |
600 | int (*revfn)(const char *, u8, int *), | |
601 | int *err) | |
1da177e4 | 602 | { |
6b7d31fc HW |
603 | int have_rev, best = -1; |
604 | ||
605 | if (down_interruptible(&ip6t_mutex) != 0) { | |
606 | *err = -EINTR; | |
607 | return 1; | |
608 | } | |
609 | have_rev = revfn(name, revision, &best); | |
610 | up(&ip6t_mutex); | |
611 | ||
612 | /* Nothing at all? Return 0 to try loading module. */ | |
613 | if (best == -1) { | |
614 | *err = -ENOENT; | |
615 | return 0; | |
616 | } | |
617 | ||
618 | *err = best; | |
619 | if (!have_rev) | |
620 | *err = -EPROTONOSUPPORT; | |
621 | return 1; | |
1da177e4 LT |
622 | } |
623 | ||
6b7d31fc | 624 | |
1da177e4 LT |
625 | /* All zeroes == unconditional rule. */ |
626 | static inline int | |
627 | unconditional(const struct ip6t_ip6 *ipv6) | |
628 | { | |
629 | unsigned int i; | |
630 | ||
631 | for (i = 0; i < sizeof(*ipv6); i++) | |
632 | if (((char *)ipv6)[i]) | |
633 | break; | |
634 | ||
635 | return (i == sizeof(*ipv6)); | |
636 | } | |
637 | ||
638 | /* Figures out from what hook each rule can be called: returns 0 if | |
639 | there are loops. Puts hook bitmask in comefrom. */ | |
640 | static int | |
31836064 ED |
641 | mark_source_chains(struct ip6t_table_info *newinfo, |
642 | unsigned int valid_hooks, void *entry0) | |
1da177e4 LT |
643 | { |
644 | unsigned int hook; | |
645 | ||
646 | /* No recursion; use packet counter to save back ptrs (reset | |
647 | to 0 as we leave), and comefrom to save source hook bitmask */ | |
648 | for (hook = 0; hook < NF_IP6_NUMHOOKS; hook++) { | |
649 | unsigned int pos = newinfo->hook_entry[hook]; | |
650 | struct ip6t_entry *e | |
31836064 | 651 | = (struct ip6t_entry *)(entry0 + pos); |
1da177e4 LT |
652 | |
653 | if (!(valid_hooks & (1 << hook))) | |
654 | continue; | |
655 | ||
656 | /* Set initial back pointer. */ | |
657 | e->counters.pcnt = pos; | |
658 | ||
659 | for (;;) { | |
660 | struct ip6t_standard_target *t | |
661 | = (void *)ip6t_get_target(e); | |
662 | ||
663 | if (e->comefrom & (1 << NF_IP6_NUMHOOKS)) { | |
664 | printk("iptables: loop hook %u pos %u %08X.\n", | |
665 | hook, pos, e->comefrom); | |
666 | return 0; | |
667 | } | |
668 | e->comefrom | |
669 | |= ((1 << hook) | (1 << NF_IP6_NUMHOOKS)); | |
670 | ||
671 | /* Unconditional return/END. */ | |
672 | if (e->target_offset == sizeof(struct ip6t_entry) | |
673 | && (strcmp(t->target.u.user.name, | |
674 | IP6T_STANDARD_TARGET) == 0) | |
675 | && t->verdict < 0 | |
676 | && unconditional(&e->ipv6)) { | |
677 | unsigned int oldpos, size; | |
678 | ||
679 | /* Return: backtrack through the last | |
680 | big jump. */ | |
681 | do { | |
682 | e->comefrom ^= (1<<NF_IP6_NUMHOOKS); | |
683 | #ifdef DEBUG_IP_FIREWALL_USER | |
684 | if (e->comefrom | |
685 | & (1 << NF_IP6_NUMHOOKS)) { | |
686 | duprintf("Back unset " | |
687 | "on hook %u " | |
688 | "rule %u\n", | |
689 | hook, pos); | |
690 | } | |
691 | #endif | |
692 | oldpos = pos; | |
693 | pos = e->counters.pcnt; | |
694 | e->counters.pcnt = 0; | |
695 | ||
696 | /* We're at the start. */ | |
697 | if (pos == oldpos) | |
698 | goto next; | |
699 | ||
700 | e = (struct ip6t_entry *) | |
31836064 | 701 | (entry0 + pos); |
1da177e4 LT |
702 | } while (oldpos == pos + e->next_offset); |
703 | ||
704 | /* Move along one */ | |
705 | size = e->next_offset; | |
706 | e = (struct ip6t_entry *) | |
31836064 | 707 | (entry0 + pos + size); |
1da177e4 LT |
708 | e->counters.pcnt = pos; |
709 | pos += size; | |
710 | } else { | |
711 | int newpos = t->verdict; | |
712 | ||
713 | if (strcmp(t->target.u.user.name, | |
714 | IP6T_STANDARD_TARGET) == 0 | |
715 | && newpos >= 0) { | |
716 | /* This a jump; chase it. */ | |
717 | duprintf("Jump rule %u -> %u\n", | |
718 | pos, newpos); | |
719 | } else { | |
720 | /* ... this is a fallthru */ | |
721 | newpos = pos + e->next_offset; | |
722 | } | |
723 | e = (struct ip6t_entry *) | |
31836064 | 724 | (entry0 + newpos); |
1da177e4 LT |
725 | e->counters.pcnt = pos; |
726 | pos = newpos; | |
727 | } | |
728 | } | |
729 | next: | |
730 | duprintf("Finished chain %u\n", hook); | |
731 | } | |
732 | return 1; | |
733 | } | |
734 | ||
735 | static inline int | |
736 | cleanup_match(struct ip6t_entry_match *m, unsigned int *i) | |
737 | { | |
738 | if (i && (*i)-- == 0) | |
739 | return 1; | |
740 | ||
741 | if (m->u.kernel.match->destroy) | |
742 | m->u.kernel.match->destroy(m->data, | |
743 | m->u.match_size - sizeof(*m)); | |
744 | module_put(m->u.kernel.match->me); | |
745 | return 0; | |
746 | } | |
747 | ||
748 | static inline int | |
749 | standard_check(const struct ip6t_entry_target *t, | |
750 | unsigned int max_offset) | |
751 | { | |
752 | struct ip6t_standard_target *targ = (void *)t; | |
753 | ||
754 | /* Check standard info. */ | |
755 | if (t->u.target_size | |
756 | != IP6T_ALIGN(sizeof(struct ip6t_standard_target))) { | |
757 | duprintf("standard_check: target size %u != %u\n", | |
758 | t->u.target_size, | |
759 | IP6T_ALIGN(sizeof(struct ip6t_standard_target))); | |
760 | return 0; | |
761 | } | |
762 | ||
763 | if (targ->verdict >= 0 | |
764 | && targ->verdict > max_offset - sizeof(struct ip6t_entry)) { | |
765 | duprintf("ip6t_standard_check: bad verdict (%i)\n", | |
766 | targ->verdict); | |
767 | return 0; | |
768 | } | |
769 | ||
770 | if (targ->verdict < -NF_MAX_VERDICT - 1) { | |
771 | duprintf("ip6t_standard_check: bad negative verdict (%i)\n", | |
772 | targ->verdict); | |
773 | return 0; | |
774 | } | |
775 | return 1; | |
776 | } | |
777 | ||
778 | static inline int | |
779 | check_match(struct ip6t_entry_match *m, | |
780 | const char *name, | |
781 | const struct ip6t_ip6 *ipv6, | |
782 | unsigned int hookmask, | |
783 | unsigned int *i) | |
784 | { | |
1da177e4 LT |
785 | struct ip6t_match *match; |
786 | ||
6b7d31fc HW |
787 | match = try_then_request_module(find_match(m->u.user.name, |
788 | m->u.user.revision), | |
789 | "ip6t_%s", m->u.user.name); | |
790 | if (IS_ERR(match) || !match) { | |
791 | duprintf("check_match: `%s' not found\n", m->u.user.name); | |
792 | return match ? PTR_ERR(match) : -ENOENT; | |
1da177e4 LT |
793 | } |
794 | m->u.kernel.match = match; | |
1da177e4 LT |
795 | |
796 | if (m->u.kernel.match->checkentry | |
797 | && !m->u.kernel.match->checkentry(name, ipv6, m->data, | |
798 | m->u.match_size - sizeof(*m), | |
799 | hookmask)) { | |
800 | module_put(m->u.kernel.match->me); | |
801 | duprintf("ip_tables: check failed for `%s'.\n", | |
802 | m->u.kernel.match->name); | |
803 | return -EINVAL; | |
804 | } | |
805 | ||
806 | (*i)++; | |
807 | return 0; | |
808 | } | |
809 | ||
810 | static struct ip6t_target ip6t_standard_target; | |
811 | ||
812 | static inline int | |
813 | check_entry(struct ip6t_entry *e, const char *name, unsigned int size, | |
814 | unsigned int *i) | |
815 | { | |
816 | struct ip6t_entry_target *t; | |
817 | struct ip6t_target *target; | |
818 | int ret; | |
819 | unsigned int j; | |
820 | ||
821 | if (!ip6_checkentry(&e->ipv6)) { | |
822 | duprintf("ip_tables: ip check failed %p %s.\n", e, name); | |
823 | return -EINVAL; | |
824 | } | |
825 | ||
826 | j = 0; | |
827 | ret = IP6T_MATCH_ITERATE(e, check_match, name, &e->ipv6, e->comefrom, &j); | |
828 | if (ret != 0) | |
829 | goto cleanup_matches; | |
830 | ||
831 | t = ip6t_get_target(e); | |
6b7d31fc HW |
832 | target = try_then_request_module(find_target(t->u.user.name, |
833 | t->u.user.revision), | |
834 | "ip6t_%s", t->u.user.name); | |
835 | if (IS_ERR(target) || !target) { | |
1da177e4 | 836 | duprintf("check_entry: `%s' not found\n", t->u.user.name); |
6b7d31fc | 837 | ret = target ? PTR_ERR(target) : -ENOENT; |
1da177e4 LT |
838 | goto cleanup_matches; |
839 | } | |
840 | t->u.kernel.target = target; | |
6b7d31fc | 841 | |
1da177e4 LT |
842 | if (t->u.kernel.target == &ip6t_standard_target) { |
843 | if (!standard_check(t, size)) { | |
844 | ret = -EINVAL; | |
845 | goto cleanup_matches; | |
846 | } | |
847 | } else if (t->u.kernel.target->checkentry | |
848 | && !t->u.kernel.target->checkentry(name, e, t->data, | |
849 | t->u.target_size | |
850 | - sizeof(*t), | |
851 | e->comefrom)) { | |
852 | module_put(t->u.kernel.target->me); | |
853 | duprintf("ip_tables: check failed for `%s'.\n", | |
854 | t->u.kernel.target->name); | |
855 | ret = -EINVAL; | |
856 | goto cleanup_matches; | |
857 | } | |
858 | ||
859 | (*i)++; | |
860 | return 0; | |
861 | ||
862 | cleanup_matches: | |
863 | IP6T_MATCH_ITERATE(e, cleanup_match, &j); | |
864 | return ret; | |
865 | } | |
866 | ||
867 | static inline int | |
868 | check_entry_size_and_hooks(struct ip6t_entry *e, | |
869 | struct ip6t_table_info *newinfo, | |
870 | unsigned char *base, | |
871 | unsigned char *limit, | |
872 | const unsigned int *hook_entries, | |
873 | const unsigned int *underflows, | |
874 | unsigned int *i) | |
875 | { | |
876 | unsigned int h; | |
877 | ||
878 | if ((unsigned long)e % __alignof__(struct ip6t_entry) != 0 | |
879 | || (unsigned char *)e + sizeof(struct ip6t_entry) >= limit) { | |
880 | duprintf("Bad offset %p\n", e); | |
881 | return -EINVAL; | |
882 | } | |
883 | ||
884 | if (e->next_offset | |
885 | < sizeof(struct ip6t_entry) + sizeof(struct ip6t_entry_target)) { | |
886 | duprintf("checking: element %p size %u\n", | |
887 | e, e->next_offset); | |
888 | return -EINVAL; | |
889 | } | |
890 | ||
891 | /* Check hooks & underflows */ | |
892 | for (h = 0; h < NF_IP6_NUMHOOKS; h++) { | |
893 | if ((unsigned char *)e - base == hook_entries[h]) | |
894 | newinfo->hook_entry[h] = hook_entries[h]; | |
895 | if ((unsigned char *)e - base == underflows[h]) | |
896 | newinfo->underflow[h] = underflows[h]; | |
897 | } | |
898 | ||
899 | /* FIXME: underflows must be unconditional, standard verdicts | |
900 | < 0 (not IP6T_RETURN). --RR */ | |
901 | ||
902 | /* Clear counters and comefrom */ | |
903 | e->counters = ((struct ip6t_counters) { 0, 0 }); | |
904 | e->comefrom = 0; | |
905 | ||
906 | (*i)++; | |
907 | return 0; | |
908 | } | |
909 | ||
910 | static inline int | |
911 | cleanup_entry(struct ip6t_entry *e, unsigned int *i) | |
912 | { | |
913 | struct ip6t_entry_target *t; | |
914 | ||
915 | if (i && (*i)-- == 0) | |
916 | return 1; | |
917 | ||
918 | /* Cleanup all matches */ | |
919 | IP6T_MATCH_ITERATE(e, cleanup_match, NULL); | |
920 | t = ip6t_get_target(e); | |
921 | if (t->u.kernel.target->destroy) | |
922 | t->u.kernel.target->destroy(t->data, | |
923 | t->u.target_size - sizeof(*t)); | |
924 | module_put(t->u.kernel.target->me); | |
925 | return 0; | |
926 | } | |
927 | ||
928 | /* Checks and translates the user-supplied table segment (held in | |
929 | newinfo) */ | |
930 | static int | |
931 | translate_table(const char *name, | |
932 | unsigned int valid_hooks, | |
933 | struct ip6t_table_info *newinfo, | |
31836064 | 934 | void *entry0, |
1da177e4 LT |
935 | unsigned int size, |
936 | unsigned int number, | |
937 | const unsigned int *hook_entries, | |
938 | const unsigned int *underflows) | |
939 | { | |
940 | unsigned int i; | |
941 | int ret; | |
942 | ||
943 | newinfo->size = size; | |
944 | newinfo->number = number; | |
945 | ||
946 | /* Init all hooks to impossible value. */ | |
947 | for (i = 0; i < NF_IP6_NUMHOOKS; i++) { | |
948 | newinfo->hook_entry[i] = 0xFFFFFFFF; | |
949 | newinfo->underflow[i] = 0xFFFFFFFF; | |
950 | } | |
951 | ||
952 | duprintf("translate_table: size %u\n", newinfo->size); | |
953 | i = 0; | |
954 | /* Walk through entries, checking offsets. */ | |
31836064 | 955 | ret = IP6T_ENTRY_ITERATE(entry0, newinfo->size, |
1da177e4 LT |
956 | check_entry_size_and_hooks, |
957 | newinfo, | |
31836064 ED |
958 | entry0, |
959 | entry0 + size, | |
1da177e4 LT |
960 | hook_entries, underflows, &i); |
961 | if (ret != 0) | |
962 | return ret; | |
963 | ||
964 | if (i != number) { | |
965 | duprintf("translate_table: %u not %u entries\n", | |
966 | i, number); | |
967 | return -EINVAL; | |
968 | } | |
969 | ||
970 | /* Check hooks all assigned */ | |
971 | for (i = 0; i < NF_IP6_NUMHOOKS; i++) { | |
972 | /* Only hooks which are valid */ | |
973 | if (!(valid_hooks & (1 << i))) | |
974 | continue; | |
975 | if (newinfo->hook_entry[i] == 0xFFFFFFFF) { | |
976 | duprintf("Invalid hook entry %u %u\n", | |
977 | i, hook_entries[i]); | |
978 | return -EINVAL; | |
979 | } | |
980 | if (newinfo->underflow[i] == 0xFFFFFFFF) { | |
981 | duprintf("Invalid underflow %u %u\n", | |
982 | i, underflows[i]); | |
983 | return -EINVAL; | |
984 | } | |
985 | } | |
986 | ||
31836064 | 987 | if (!mark_source_chains(newinfo, valid_hooks, entry0)) |
1da177e4 LT |
988 | return -ELOOP; |
989 | ||
990 | /* Finally, each sanity check must pass */ | |
991 | i = 0; | |
31836064 | 992 | ret = IP6T_ENTRY_ITERATE(entry0, newinfo->size, |
1da177e4 LT |
993 | check_entry, name, size, &i); |
994 | ||
995 | if (ret != 0) { | |
31836064 | 996 | IP6T_ENTRY_ITERATE(entry0, newinfo->size, |
1da177e4 LT |
997 | cleanup_entry, &i); |
998 | return ret; | |
999 | } | |
1000 | ||
1001 | /* And one copy for every other CPU */ | |
c8923c6b | 1002 | for_each_cpu(i) { |
31836064 ED |
1003 | if (newinfo->entries[i] && newinfo->entries[i] != entry0) |
1004 | memcpy(newinfo->entries[i], entry0, newinfo->size); | |
1da177e4 LT |
1005 | } |
1006 | ||
1007 | return ret; | |
1008 | } | |
1009 | ||
1010 | static struct ip6t_table_info * | |
1011 | replace_table(struct ip6t_table *table, | |
1012 | unsigned int num_counters, | |
1013 | struct ip6t_table_info *newinfo, | |
1014 | int *error) | |
1015 | { | |
1016 | struct ip6t_table_info *oldinfo; | |
1017 | ||
1018 | #ifdef CONFIG_NETFILTER_DEBUG | |
1019 | { | |
31836064 | 1020 | int cpu; |
1da177e4 | 1021 | |
31836064 ED |
1022 | for_each_cpu(cpu) { |
1023 | struct ip6t_entry *table_base = newinfo->entries[cpu]; | |
1024 | if (table_base) | |
1025 | table_base->comefrom = 0xdead57ac; | |
1da177e4 LT |
1026 | } |
1027 | } | |
1028 | #endif | |
1029 | ||
1030 | /* Do the substitution. */ | |
1031 | write_lock_bh(&table->lock); | |
1032 | /* Check inside lock: is the old number correct? */ | |
1033 | if (num_counters != table->private->number) { | |
1034 | duprintf("num_counters != table->private->number (%u/%u)\n", | |
1035 | num_counters, table->private->number); | |
1036 | write_unlock_bh(&table->lock); | |
1037 | *error = -EAGAIN; | |
1038 | return NULL; | |
1039 | } | |
1040 | oldinfo = table->private; | |
1041 | table->private = newinfo; | |
1042 | newinfo->initial_entries = oldinfo->initial_entries; | |
1043 | write_unlock_bh(&table->lock); | |
1044 | ||
1045 | return oldinfo; | |
1046 | } | |
1047 | ||
1048 | /* Gets counters. */ | |
1049 | static inline int | |
1050 | add_entry_to_counter(const struct ip6t_entry *e, | |
1051 | struct ip6t_counters total[], | |
1052 | unsigned int *i) | |
1053 | { | |
1054 | ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt); | |
1055 | ||
1056 | (*i)++; | |
1057 | return 0; | |
1058 | } | |
1059 | ||
31836064 ED |
1060 | static inline int |
1061 | set_entry_to_counter(const struct ip6t_entry *e, | |
1062 | struct ip6t_counters total[], | |
1063 | unsigned int *i) | |
1064 | { | |
1065 | SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt); | |
1066 | ||
1067 | (*i)++; | |
1068 | return 0; | |
1069 | } | |
1070 | ||
1da177e4 LT |
1071 | static void |
1072 | get_counters(const struct ip6t_table_info *t, | |
1073 | struct ip6t_counters counters[]) | |
1074 | { | |
1075 | unsigned int cpu; | |
1076 | unsigned int i; | |
31836064 ED |
1077 | unsigned int curcpu; |
1078 | ||
1079 | /* Instead of clearing (by a previous call to memset()) | |
1080 | * the counters and using adds, we set the counters | |
1081 | * with data used by 'current' CPU | |
1082 | * We dont care about preemption here. | |
1083 | */ | |
1084 | curcpu = raw_smp_processor_id(); | |
1085 | ||
1086 | i = 0; | |
1087 | IP6T_ENTRY_ITERATE(t->entries[curcpu], | |
1088 | t->size, | |
1089 | set_entry_to_counter, | |
1090 | counters, | |
1091 | &i); | |
1da177e4 | 1092 | |
c8923c6b | 1093 | for_each_cpu(cpu) { |
31836064 ED |
1094 | if (cpu == curcpu) |
1095 | continue; | |
1da177e4 | 1096 | i = 0; |
31836064 | 1097 | IP6T_ENTRY_ITERATE(t->entries[cpu], |
1da177e4 LT |
1098 | t->size, |
1099 | add_entry_to_counter, | |
1100 | counters, | |
1101 | &i); | |
1102 | } | |
1103 | } | |
1104 | ||
1105 | static int | |
1106 | copy_entries_to_user(unsigned int total_size, | |
1107 | struct ip6t_table *table, | |
1108 | void __user *userptr) | |
1109 | { | |
1110 | unsigned int off, num, countersize; | |
1111 | struct ip6t_entry *e; | |
1112 | struct ip6t_counters *counters; | |
1113 | int ret = 0; | |
31836064 | 1114 | void *loc_cpu_entry; |
1da177e4 LT |
1115 | |
1116 | /* We need atomic snapshot of counters: rest doesn't change | |
1117 | (other than comefrom, which userspace doesn't care | |
1118 | about). */ | |
1119 | countersize = sizeof(struct ip6t_counters) * table->private->number; | |
1120 | counters = vmalloc(countersize); | |
1121 | ||
1122 | if (counters == NULL) | |
1123 | return -ENOMEM; | |
1124 | ||
1125 | /* First, sum counters... */ | |
1da177e4 LT |
1126 | write_lock_bh(&table->lock); |
1127 | get_counters(table->private, counters); | |
1128 | write_unlock_bh(&table->lock); | |
1129 | ||
31836064 ED |
1130 | /* choose the copy that is on ourc node/cpu */ |
1131 | loc_cpu_entry = table->private->entries[raw_smp_processor_id()]; | |
1132 | if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) { | |
1da177e4 LT |
1133 | ret = -EFAULT; |
1134 | goto free_counters; | |
1135 | } | |
1136 | ||
1137 | /* FIXME: use iterator macros --RR */ | |
1138 | /* ... then go back and fix counters and names */ | |
1139 | for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){ | |
1140 | unsigned int i; | |
1141 | struct ip6t_entry_match *m; | |
1142 | struct ip6t_entry_target *t; | |
1143 | ||
31836064 | 1144 | e = (struct ip6t_entry *)(loc_cpu_entry + off); |
1da177e4 LT |
1145 | if (copy_to_user(userptr + off |
1146 | + offsetof(struct ip6t_entry, counters), | |
1147 | &counters[num], | |
1148 | sizeof(counters[num])) != 0) { | |
1149 | ret = -EFAULT; | |
1150 | goto free_counters; | |
1151 | } | |
1152 | ||
1153 | for (i = sizeof(struct ip6t_entry); | |
1154 | i < e->target_offset; | |
1155 | i += m->u.match_size) { | |
1156 | m = (void *)e + i; | |
1157 | ||
1158 | if (copy_to_user(userptr + off + i | |
1159 | + offsetof(struct ip6t_entry_match, | |
1160 | u.user.name), | |
1161 | m->u.kernel.match->name, | |
1162 | strlen(m->u.kernel.match->name)+1) | |
1163 | != 0) { | |
1164 | ret = -EFAULT; | |
1165 | goto free_counters; | |
1166 | } | |
1167 | } | |
1168 | ||
1169 | t = ip6t_get_target(e); | |
1170 | if (copy_to_user(userptr + off + e->target_offset | |
1171 | + offsetof(struct ip6t_entry_target, | |
1172 | u.user.name), | |
1173 | t->u.kernel.target->name, | |
1174 | strlen(t->u.kernel.target->name)+1) != 0) { | |
1175 | ret = -EFAULT; | |
1176 | goto free_counters; | |
1177 | } | |
1178 | } | |
1179 | ||
1180 | free_counters: | |
1181 | vfree(counters); | |
1182 | return ret; | |
1183 | } | |
1184 | ||
1185 | static int | |
1186 | get_entries(const struct ip6t_get_entries *entries, | |
1187 | struct ip6t_get_entries __user *uptr) | |
1188 | { | |
1189 | int ret; | |
1190 | struct ip6t_table *t; | |
1191 | ||
6b7d31fc HW |
1192 | t = find_table_lock(entries->name); |
1193 | if (t && !IS_ERR(t)) { | |
1da177e4 LT |
1194 | duprintf("t->private->number = %u\n", |
1195 | t->private->number); | |
1196 | if (entries->size == t->private->size) | |
1197 | ret = copy_entries_to_user(t->private->size, | |
1198 | t, uptr->entrytable); | |
1199 | else { | |
1200 | duprintf("get_entries: I've got %u not %u!\n", | |
1201 | t->private->size, | |
1202 | entries->size); | |
1203 | ret = -EINVAL; | |
1204 | } | |
6b7d31fc | 1205 | module_put(t->me); |
1da177e4 LT |
1206 | up(&ip6t_mutex); |
1207 | } else | |
6b7d31fc | 1208 | ret = t ? PTR_ERR(t) : -ENOENT; |
1da177e4 LT |
1209 | |
1210 | return ret; | |
1211 | } | |
1212 | ||
31836064 ED |
1213 | static void free_table_info(struct ip6t_table_info *info) |
1214 | { | |
1215 | int cpu; | |
1216 | for_each_cpu(cpu) { | |
1217 | if (info->size <= PAGE_SIZE) | |
1218 | kfree(info->entries[cpu]); | |
1219 | else | |
1220 | vfree(info->entries[cpu]); | |
1221 | } | |
1222 | kfree(info); | |
1223 | } | |
1224 | ||
1225 | static struct ip6t_table_info *alloc_table_info(unsigned int size) | |
1226 | { | |
1227 | struct ip6t_table_info *newinfo; | |
1228 | int cpu; | |
1229 | ||
1230 | newinfo = kzalloc(sizeof(struct ip6t_table_info), GFP_KERNEL); | |
1231 | if (!newinfo) | |
1232 | return NULL; | |
1233 | ||
1234 | newinfo->size = size; | |
1235 | ||
1236 | for_each_cpu(cpu) { | |
1237 | if (size <= PAGE_SIZE) | |
1238 | newinfo->entries[cpu] = kmalloc_node(size, | |
1239 | GFP_KERNEL, | |
1240 | cpu_to_node(cpu)); | |
1241 | else | |
1242 | newinfo->entries[cpu] = vmalloc_node(size, | |
1243 | cpu_to_node(cpu)); | |
1244 | if (newinfo->entries[cpu] == NULL) { | |
1245 | free_table_info(newinfo); | |
1246 | return NULL; | |
1247 | } | |
1248 | } | |
1249 | ||
1250 | return newinfo; | |
1251 | } | |
1252 | ||
1da177e4 LT |
1253 | static int |
1254 | do_replace(void __user *user, unsigned int len) | |
1255 | { | |
1256 | int ret; | |
1257 | struct ip6t_replace tmp; | |
1258 | struct ip6t_table *t; | |
1259 | struct ip6t_table_info *newinfo, *oldinfo; | |
1260 | struct ip6t_counters *counters; | |
31836064 | 1261 | void *loc_cpu_entry, *loc_cpu_old_entry; |
1da177e4 LT |
1262 | |
1263 | if (copy_from_user(&tmp, user, sizeof(tmp)) != 0) | |
1264 | return -EFAULT; | |
1265 | ||
1266 | /* Pedantry: prevent them from hitting BUG() in vmalloc.c --RR */ | |
1267 | if ((SMP_ALIGN(tmp.size) >> PAGE_SHIFT) + 2 > num_physpages) | |
1268 | return -ENOMEM; | |
1269 | ||
31836064 | 1270 | newinfo = alloc_table_info(tmp.size); |
1da177e4 LT |
1271 | if (!newinfo) |
1272 | return -ENOMEM; | |
1273 | ||
31836064 ED |
1274 | /* choose the copy that is on our node/cpu */ |
1275 | loc_cpu_entry = newinfo->entries[raw_smp_processor_id()]; | |
1276 | if (copy_from_user(loc_cpu_entry, user + sizeof(tmp), | |
1da177e4 LT |
1277 | tmp.size) != 0) { |
1278 | ret = -EFAULT; | |
1279 | goto free_newinfo; | |
1280 | } | |
1281 | ||
1282 | counters = vmalloc(tmp.num_counters * sizeof(struct ip6t_counters)); | |
1283 | if (!counters) { | |
1284 | ret = -ENOMEM; | |
1285 | goto free_newinfo; | |
1286 | } | |
1da177e4 LT |
1287 | |
1288 | ret = translate_table(tmp.name, tmp.valid_hooks, | |
31836064 | 1289 | newinfo, loc_cpu_entry, tmp.size, tmp.num_entries, |
1da177e4 LT |
1290 | tmp.hook_entry, tmp.underflow); |
1291 | if (ret != 0) | |
1292 | goto free_newinfo_counters; | |
1293 | ||
1294 | duprintf("ip_tables: Translated table\n"); | |
1295 | ||
6b7d31fc HW |
1296 | t = try_then_request_module(find_table_lock(tmp.name), |
1297 | "ip6table_%s", tmp.name); | |
1298 | if (!t || IS_ERR(t)) { | |
1299 | ret = t ? PTR_ERR(t) : -ENOENT; | |
1da177e4 | 1300 | goto free_newinfo_counters_untrans; |
6b7d31fc | 1301 | } |
1da177e4 LT |
1302 | |
1303 | /* You lied! */ | |
1304 | if (tmp.valid_hooks != t->valid_hooks) { | |
1305 | duprintf("Valid hook crap: %08X vs %08X\n", | |
1306 | tmp.valid_hooks, t->valid_hooks); | |
1307 | ret = -EINVAL; | |
6b7d31fc | 1308 | goto put_module; |
1da177e4 LT |
1309 | } |
1310 | ||
1311 | oldinfo = replace_table(t, tmp.num_counters, newinfo, &ret); | |
1312 | if (!oldinfo) | |
1313 | goto put_module; | |
1314 | ||
1315 | /* Update module usage count based on number of rules */ | |
1316 | duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n", | |
1317 | oldinfo->number, oldinfo->initial_entries, newinfo->number); | |
1318 | if ((oldinfo->number > oldinfo->initial_entries) || | |
1319 | (newinfo->number <= oldinfo->initial_entries)) | |
1320 | module_put(t->me); | |
1321 | if ((oldinfo->number > oldinfo->initial_entries) && | |
1322 | (newinfo->number <= oldinfo->initial_entries)) | |
1323 | module_put(t->me); | |
1324 | ||
1325 | /* Get the old counters. */ | |
1326 | get_counters(oldinfo, counters); | |
1327 | /* Decrease module usage counts and free resource */ | |
31836064 ED |
1328 | loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()]; |
1329 | IP6T_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,NULL); | |
1330 | free_table_info(oldinfo); | |
1da177e4 LT |
1331 | if (copy_to_user(tmp.counters, counters, |
1332 | sizeof(struct ip6t_counters) * tmp.num_counters) != 0) | |
1333 | ret = -EFAULT; | |
1334 | vfree(counters); | |
1335 | up(&ip6t_mutex); | |
1336 | return ret; | |
1337 | ||
1338 | put_module: | |
1339 | module_put(t->me); | |
1da177e4 LT |
1340 | up(&ip6t_mutex); |
1341 | free_newinfo_counters_untrans: | |
31836064 | 1342 | IP6T_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry,NULL); |
1da177e4 LT |
1343 | free_newinfo_counters: |
1344 | vfree(counters); | |
1345 | free_newinfo: | |
31836064 | 1346 | free_table_info(newinfo); |
1da177e4 LT |
1347 | return ret; |
1348 | } | |
1349 | ||
1350 | /* We're lazy, and add to the first CPU; overflow works its fey magic | |
1351 | * and everything is OK. */ | |
1352 | static inline int | |
1353 | add_counter_to_entry(struct ip6t_entry *e, | |
1354 | const struct ip6t_counters addme[], | |
1355 | unsigned int *i) | |
1356 | { | |
1357 | #if 0 | |
1358 | duprintf("add_counter: Entry %u %lu/%lu + %lu/%lu\n", | |
1359 | *i, | |
1360 | (long unsigned int)e->counters.pcnt, | |
1361 | (long unsigned int)e->counters.bcnt, | |
1362 | (long unsigned int)addme[*i].pcnt, | |
1363 | (long unsigned int)addme[*i].bcnt); | |
1364 | #endif | |
1365 | ||
1366 | ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt); | |
1367 | ||
1368 | (*i)++; | |
1369 | return 0; | |
1370 | } | |
1371 | ||
1372 | static int | |
1373 | do_add_counters(void __user *user, unsigned int len) | |
1374 | { | |
1375 | unsigned int i; | |
1376 | struct ip6t_counters_info tmp, *paddc; | |
1377 | struct ip6t_table *t; | |
6b7d31fc | 1378 | int ret = 0; |
31836064 | 1379 | void *loc_cpu_entry; |
1da177e4 LT |
1380 | |
1381 | if (copy_from_user(&tmp, user, sizeof(tmp)) != 0) | |
1382 | return -EFAULT; | |
1383 | ||
1384 | if (len != sizeof(tmp) + tmp.num_counters*sizeof(struct ip6t_counters)) | |
1385 | return -EINVAL; | |
1386 | ||
1387 | paddc = vmalloc(len); | |
1388 | if (!paddc) | |
1389 | return -ENOMEM; | |
1390 | ||
1391 | if (copy_from_user(paddc, user, len) != 0) { | |
1392 | ret = -EFAULT; | |
1393 | goto free; | |
1394 | } | |
1395 | ||
6b7d31fc HW |
1396 | t = find_table_lock(tmp.name); |
1397 | if (!t || IS_ERR(t)) { | |
1398 | ret = t ? PTR_ERR(t) : -ENOENT; | |
1da177e4 | 1399 | goto free; |
6b7d31fc | 1400 | } |
1da177e4 LT |
1401 | |
1402 | write_lock_bh(&t->lock); | |
1403 | if (t->private->number != paddc->num_counters) { | |
1404 | ret = -EINVAL; | |
1405 | goto unlock_up_free; | |
1406 | } | |
1407 | ||
1408 | i = 0; | |
31836064 ED |
1409 | /* Choose the copy that is on our node */ |
1410 | loc_cpu_entry = t->private->entries[smp_processor_id()]; | |
1411 | IP6T_ENTRY_ITERATE(loc_cpu_entry, | |
1da177e4 LT |
1412 | t->private->size, |
1413 | add_counter_to_entry, | |
1414 | paddc->counters, | |
1415 | &i); | |
1416 | unlock_up_free: | |
1417 | write_unlock_bh(&t->lock); | |
1418 | up(&ip6t_mutex); | |
6b7d31fc | 1419 | module_put(t->me); |
1da177e4 LT |
1420 | free: |
1421 | vfree(paddc); | |
1422 | ||
1423 | return ret; | |
1424 | } | |
1425 | ||
1426 | static int | |
1427 | do_ip6t_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len) | |
1428 | { | |
1429 | int ret; | |
1430 | ||
1431 | if (!capable(CAP_NET_ADMIN)) | |
1432 | return -EPERM; | |
1433 | ||
1434 | switch (cmd) { | |
1435 | case IP6T_SO_SET_REPLACE: | |
1436 | ret = do_replace(user, len); | |
1437 | break; | |
1438 | ||
1439 | case IP6T_SO_SET_ADD_COUNTERS: | |
1440 | ret = do_add_counters(user, len); | |
1441 | break; | |
1442 | ||
1443 | default: | |
1444 | duprintf("do_ip6t_set_ctl: unknown request %i\n", cmd); | |
1445 | ret = -EINVAL; | |
1446 | } | |
1447 | ||
1448 | return ret; | |
1449 | } | |
1450 | ||
1451 | static int | |
1452 | do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len) | |
1453 | { | |
1454 | int ret; | |
1455 | ||
1456 | if (!capable(CAP_NET_ADMIN)) | |
1457 | return -EPERM; | |
1458 | ||
1459 | switch (cmd) { | |
1460 | case IP6T_SO_GET_INFO: { | |
1461 | char name[IP6T_TABLE_MAXNAMELEN]; | |
1462 | struct ip6t_table *t; | |
1463 | ||
1464 | if (*len != sizeof(struct ip6t_getinfo)) { | |
1465 | duprintf("length %u != %u\n", *len, | |
1466 | sizeof(struct ip6t_getinfo)); | |
1467 | ret = -EINVAL; | |
1468 | break; | |
1469 | } | |
1470 | ||
1471 | if (copy_from_user(name, user, sizeof(name)) != 0) { | |
1472 | ret = -EFAULT; | |
1473 | break; | |
1474 | } | |
1475 | name[IP6T_TABLE_MAXNAMELEN-1] = '\0'; | |
6b7d31fc HW |
1476 | |
1477 | t = try_then_request_module(find_table_lock(name), | |
1478 | "ip6table_%s", name); | |
1479 | if (t && !IS_ERR(t)) { | |
1da177e4 LT |
1480 | struct ip6t_getinfo info; |
1481 | ||
1482 | info.valid_hooks = t->valid_hooks; | |
1483 | memcpy(info.hook_entry, t->private->hook_entry, | |
1484 | sizeof(info.hook_entry)); | |
1485 | memcpy(info.underflow, t->private->underflow, | |
1486 | sizeof(info.underflow)); | |
1487 | info.num_entries = t->private->number; | |
1488 | info.size = t->private->size; | |
1489 | memcpy(info.name, name, sizeof(info.name)); | |
1490 | ||
1491 | if (copy_to_user(user, &info, *len) != 0) | |
1492 | ret = -EFAULT; | |
1493 | else | |
1494 | ret = 0; | |
1da177e4 | 1495 | up(&ip6t_mutex); |
6b7d31fc HW |
1496 | module_put(t->me); |
1497 | } else | |
1498 | ret = t ? PTR_ERR(t) : -ENOENT; | |
1da177e4 LT |
1499 | } |
1500 | break; | |
1501 | ||
1502 | case IP6T_SO_GET_ENTRIES: { | |
1503 | struct ip6t_get_entries get; | |
1504 | ||
1505 | if (*len < sizeof(get)) { | |
1506 | duprintf("get_entries: %u < %u\n", *len, sizeof(get)); | |
1507 | ret = -EINVAL; | |
1508 | } else if (copy_from_user(&get, user, sizeof(get)) != 0) { | |
1509 | ret = -EFAULT; | |
1510 | } else if (*len != sizeof(struct ip6t_get_entries) + get.size) { | |
1511 | duprintf("get_entries: %u != %u\n", *len, | |
1512 | sizeof(struct ip6t_get_entries) + get.size); | |
1513 | ret = -EINVAL; | |
1514 | } else | |
1515 | ret = get_entries(&get, user); | |
1516 | break; | |
1517 | } | |
1518 | ||
6b7d31fc HW |
1519 | case IP6T_SO_GET_REVISION_MATCH: |
1520 | case IP6T_SO_GET_REVISION_TARGET: { | |
1521 | struct ip6t_get_revision rev; | |
1522 | int (*revfn)(const char *, u8, int *); | |
1523 | ||
1524 | if (*len != sizeof(rev)) { | |
1525 | ret = -EINVAL; | |
1526 | break; | |
1527 | } | |
1528 | if (copy_from_user(&rev, user, sizeof(rev)) != 0) { | |
1529 | ret = -EFAULT; | |
1530 | break; | |
1531 | } | |
1532 | ||
1533 | if (cmd == IP6T_SO_GET_REVISION_TARGET) | |
1534 | revfn = target_revfn; | |
1535 | else | |
1536 | revfn = match_revfn; | |
1537 | ||
1538 | try_then_request_module(find_revision(rev.name, rev.revision, | |
1539 | revfn, &ret), | |
1540 | "ip6t_%s", rev.name); | |
1541 | break; | |
1542 | } | |
1543 | ||
1da177e4 LT |
1544 | default: |
1545 | duprintf("do_ip6t_get_ctl: unknown request %i\n", cmd); | |
1546 | ret = -EINVAL; | |
1547 | } | |
1548 | ||
1549 | return ret; | |
1550 | } | |
1551 | ||
1552 | /* Registration hooks for targets. */ | |
1553 | int | |
1554 | ip6t_register_target(struct ip6t_target *target) | |
1555 | { | |
1556 | int ret; | |
1557 | ||
1558 | ret = down_interruptible(&ip6t_mutex); | |
1559 | if (ret != 0) | |
1560 | return ret; | |
6b7d31fc | 1561 | list_add(&target->list, &ip6t_target); |
1da177e4 LT |
1562 | up(&ip6t_mutex); |
1563 | return ret; | |
1564 | } | |
1565 | ||
1566 | void | |
1567 | ip6t_unregister_target(struct ip6t_target *target) | |
1568 | { | |
1569 | down(&ip6t_mutex); | |
1570 | LIST_DELETE(&ip6t_target, target); | |
1571 | up(&ip6t_mutex); | |
1572 | } | |
1573 | ||
1574 | int | |
1575 | ip6t_register_match(struct ip6t_match *match) | |
1576 | { | |
1577 | int ret; | |
1578 | ||
1579 | ret = down_interruptible(&ip6t_mutex); | |
1580 | if (ret != 0) | |
1581 | return ret; | |
1582 | ||
6b7d31fc | 1583 | list_add(&match->list, &ip6t_match); |
1da177e4 LT |
1584 | up(&ip6t_mutex); |
1585 | ||
1586 | return ret; | |
1587 | } | |
1588 | ||
1589 | void | |
1590 | ip6t_unregister_match(struct ip6t_match *match) | |
1591 | { | |
1592 | down(&ip6t_mutex); | |
1593 | LIST_DELETE(&ip6t_match, match); | |
1594 | up(&ip6t_mutex); | |
1595 | } | |
1596 | ||
1597 | int ip6t_register_table(struct ip6t_table *table, | |
1598 | const struct ip6t_replace *repl) | |
1599 | { | |
1600 | int ret; | |
1601 | struct ip6t_table_info *newinfo; | |
1602 | static struct ip6t_table_info bootstrap | |
1603 | = { 0, 0, 0, { 0 }, { 0 }, { } }; | |
31836064 | 1604 | void *loc_cpu_entry; |
1da177e4 | 1605 | |
31836064 | 1606 | newinfo = alloc_table_info(repl->size); |
1da177e4 LT |
1607 | if (!newinfo) |
1608 | return -ENOMEM; | |
1609 | ||
31836064 ED |
1610 | /* choose the copy on our node/cpu */ |
1611 | loc_cpu_entry = newinfo->entries[raw_smp_processor_id()]; | |
1612 | memcpy(loc_cpu_entry, repl->entries, repl->size); | |
1da177e4 LT |
1613 | |
1614 | ret = translate_table(table->name, table->valid_hooks, | |
31836064 | 1615 | newinfo, loc_cpu_entry, repl->size, |
1da177e4 LT |
1616 | repl->num_entries, |
1617 | repl->hook_entry, | |
1618 | repl->underflow); | |
1619 | if (ret != 0) { | |
31836064 | 1620 | free_table_info(newinfo); |
1da177e4 LT |
1621 | return ret; |
1622 | } | |
1623 | ||
1624 | ret = down_interruptible(&ip6t_mutex); | |
1625 | if (ret != 0) { | |
31836064 | 1626 | free_table_info(newinfo); |
1da177e4 LT |
1627 | return ret; |
1628 | } | |
1629 | ||
1630 | /* Don't autoload: we'd eat our tail... */ | |
1631 | if (list_named_find(&ip6t_tables, table->name)) { | |
1632 | ret = -EEXIST; | |
1633 | goto free_unlock; | |
1634 | } | |
1635 | ||
1636 | /* Simplifies replace_table code. */ | |
1637 | table->private = &bootstrap; | |
1638 | if (!replace_table(table, 0, newinfo, &ret)) | |
1639 | goto free_unlock; | |
1640 | ||
1641 | duprintf("table->private->number = %u\n", | |
1642 | table->private->number); | |
1643 | ||
1644 | /* save number of initial entries */ | |
1645 | table->private->initial_entries = table->private->number; | |
1646 | ||
1647 | rwlock_init(&table->lock); | |
1648 | list_prepend(&ip6t_tables, table); | |
1649 | ||
1650 | unlock: | |
1651 | up(&ip6t_mutex); | |
1652 | return ret; | |
1653 | ||
1654 | free_unlock: | |
31836064 | 1655 | free_table_info(newinfo); |
1da177e4 LT |
1656 | goto unlock; |
1657 | } | |
1658 | ||
1659 | void ip6t_unregister_table(struct ip6t_table *table) | |
1660 | { | |
31836064 ED |
1661 | void *loc_cpu_entry; |
1662 | ||
1da177e4 LT |
1663 | down(&ip6t_mutex); |
1664 | LIST_DELETE(&ip6t_tables, table); | |
1665 | up(&ip6t_mutex); | |
1666 | ||
1667 | /* Decrease module usage counts and free resources */ | |
31836064 ED |
1668 | loc_cpu_entry = table->private->entries[raw_smp_processor_id()]; |
1669 | IP6T_ENTRY_ITERATE(loc_cpu_entry, table->private->size, | |
1da177e4 | 1670 | cleanup_entry, NULL); |
31836064 | 1671 | free_table_info(table->private); |
1da177e4 LT |
1672 | } |
1673 | ||
1674 | /* Returns 1 if the port is matched by the range, 0 otherwise */ | |
1675 | static inline int | |
1676 | port_match(u_int16_t min, u_int16_t max, u_int16_t port, int invert) | |
1677 | { | |
1678 | int ret; | |
1679 | ||
1680 | ret = (port >= min && port <= max) ^ invert; | |
1681 | return ret; | |
1682 | } | |
1683 | ||
1684 | static int | |
1685 | tcp_find_option(u_int8_t option, | |
1686 | const struct sk_buff *skb, | |
1687 | unsigned int tcpoff, | |
1688 | unsigned int optlen, | |
1689 | int invert, | |
1690 | int *hotdrop) | |
1691 | { | |
1692 | /* tcp.doff is only 4 bits, ie. max 15 * 4 bytes */ | |
1693 | u_int8_t _opt[60 - sizeof(struct tcphdr)], *op; | |
1694 | unsigned int i; | |
1695 | ||
1696 | duprintf("tcp_match: finding option\n"); | |
1697 | if (!optlen) | |
1698 | return invert; | |
1699 | /* If we don't have the whole header, drop packet. */ | |
1700 | op = skb_header_pointer(skb, tcpoff + sizeof(struct tcphdr), optlen, | |
1701 | _opt); | |
1702 | if (op == NULL) { | |
1703 | *hotdrop = 1; | |
1704 | return 0; | |
1705 | } | |
1706 | ||
1707 | for (i = 0; i < optlen; ) { | |
1708 | if (op[i] == option) return !invert; | |
1709 | if (op[i] < 2) i++; | |
1710 | else i += op[i+1]?:1; | |
1711 | } | |
1712 | ||
1713 | return invert; | |
1714 | } | |
1715 | ||
1716 | static int | |
1717 | tcp_match(const struct sk_buff *skb, | |
1718 | const struct net_device *in, | |
1719 | const struct net_device *out, | |
1720 | const void *matchinfo, | |
1721 | int offset, | |
1722 | unsigned int protoff, | |
1723 | int *hotdrop) | |
1724 | { | |
1725 | struct tcphdr _tcph, *th; | |
1726 | const struct ip6t_tcp *tcpinfo = matchinfo; | |
1727 | ||
1728 | if (offset) { | |
1729 | /* To quote Alan: | |
1730 | ||
1731 | Don't allow a fragment of TCP 8 bytes in. Nobody normal | |
1732 | causes this. Its a cracker trying to break in by doing a | |
1733 | flag overwrite to pass the direction checks. | |
1734 | */ | |
1735 | if (offset == 1) { | |
1736 | duprintf("Dropping evil TCP offset=1 frag.\n"); | |
1737 | *hotdrop = 1; | |
1738 | } | |
1739 | /* Must not be a fragment. */ | |
1740 | return 0; | |
1741 | } | |
1742 | ||
1743 | #define FWINVTCP(bool,invflg) ((bool) ^ !!(tcpinfo->invflags & invflg)) | |
1744 | ||
1745 | th = skb_header_pointer(skb, protoff, sizeof(_tcph), &_tcph); | |
1746 | if (th == NULL) { | |
1747 | /* We've been asked to examine this packet, and we | |
1748 | can't. Hence, no choice but to drop. */ | |
1749 | duprintf("Dropping evil TCP offset=0 tinygram.\n"); | |
1750 | *hotdrop = 1; | |
1751 | return 0; | |
1752 | } | |
1753 | ||
1754 | if (!port_match(tcpinfo->spts[0], tcpinfo->spts[1], | |
1755 | ntohs(th->source), | |
1756 | !!(tcpinfo->invflags & IP6T_TCP_INV_SRCPT))) | |
1757 | return 0; | |
1758 | if (!port_match(tcpinfo->dpts[0], tcpinfo->dpts[1], | |
1759 | ntohs(th->dest), | |
1760 | !!(tcpinfo->invflags & IP6T_TCP_INV_DSTPT))) | |
1761 | return 0; | |
1762 | if (!FWINVTCP((((unsigned char *)th)[13] & tcpinfo->flg_mask) | |
1763 | == tcpinfo->flg_cmp, | |
1764 | IP6T_TCP_INV_FLAGS)) | |
1765 | return 0; | |
1766 | if (tcpinfo->option) { | |
1767 | if (th->doff * 4 < sizeof(_tcph)) { | |
1768 | *hotdrop = 1; | |
1769 | return 0; | |
1770 | } | |
1771 | if (!tcp_find_option(tcpinfo->option, skb, protoff, | |
1772 | th->doff*4 - sizeof(*th), | |
1773 | tcpinfo->invflags & IP6T_TCP_INV_OPTION, | |
1774 | hotdrop)) | |
1775 | return 0; | |
1776 | } | |
1777 | return 1; | |
1778 | } | |
1779 | ||
1780 | /* Called when user tries to insert an entry of this type. */ | |
1781 | static int | |
1782 | tcp_checkentry(const char *tablename, | |
1783 | const struct ip6t_ip6 *ipv6, | |
1784 | void *matchinfo, | |
1785 | unsigned int matchsize, | |
1786 | unsigned int hook_mask) | |
1787 | { | |
1788 | const struct ip6t_tcp *tcpinfo = matchinfo; | |
1789 | ||
1790 | /* Must specify proto == TCP, and no unknown invflags */ | |
1791 | return ipv6->proto == IPPROTO_TCP | |
1792 | && !(ipv6->invflags & IP6T_INV_PROTO) | |
1793 | && matchsize == IP6T_ALIGN(sizeof(struct ip6t_tcp)) | |
1794 | && !(tcpinfo->invflags & ~IP6T_TCP_INV_MASK); | |
1795 | } | |
1796 | ||
1797 | static int | |
1798 | udp_match(const struct sk_buff *skb, | |
1799 | const struct net_device *in, | |
1800 | const struct net_device *out, | |
1801 | const void *matchinfo, | |
1802 | int offset, | |
1803 | unsigned int protoff, | |
1804 | int *hotdrop) | |
1805 | { | |
1806 | struct udphdr _udph, *uh; | |
1807 | const struct ip6t_udp *udpinfo = matchinfo; | |
1808 | ||
1809 | /* Must not be a fragment. */ | |
1810 | if (offset) | |
1811 | return 0; | |
1812 | ||
1813 | uh = skb_header_pointer(skb, protoff, sizeof(_udph), &_udph); | |
1814 | if (uh == NULL) { | |
1815 | /* We've been asked to examine this packet, and we | |
1816 | can't. Hence, no choice but to drop. */ | |
1817 | duprintf("Dropping evil UDP tinygram.\n"); | |
1818 | *hotdrop = 1; | |
1819 | return 0; | |
1820 | } | |
1821 | ||
1822 | return port_match(udpinfo->spts[0], udpinfo->spts[1], | |
1823 | ntohs(uh->source), | |
1824 | !!(udpinfo->invflags & IP6T_UDP_INV_SRCPT)) | |
1825 | && port_match(udpinfo->dpts[0], udpinfo->dpts[1], | |
1826 | ntohs(uh->dest), | |
1827 | !!(udpinfo->invflags & IP6T_UDP_INV_DSTPT)); | |
1828 | } | |
1829 | ||
1830 | /* Called when user tries to insert an entry of this type. */ | |
1831 | static int | |
1832 | udp_checkentry(const char *tablename, | |
1833 | const struct ip6t_ip6 *ipv6, | |
1834 | void *matchinfo, | |
1835 | unsigned int matchinfosize, | |
1836 | unsigned int hook_mask) | |
1837 | { | |
1838 | const struct ip6t_udp *udpinfo = matchinfo; | |
1839 | ||
1840 | /* Must specify proto == UDP, and no unknown invflags */ | |
1841 | if (ipv6->proto != IPPROTO_UDP || (ipv6->invflags & IP6T_INV_PROTO)) { | |
1842 | duprintf("ip6t_udp: Protocol %u != %u\n", ipv6->proto, | |
1843 | IPPROTO_UDP); | |
1844 | return 0; | |
1845 | } | |
1846 | if (matchinfosize != IP6T_ALIGN(sizeof(struct ip6t_udp))) { | |
1847 | duprintf("ip6t_udp: matchsize %u != %u\n", | |
1848 | matchinfosize, IP6T_ALIGN(sizeof(struct ip6t_udp))); | |
1849 | return 0; | |
1850 | } | |
1851 | if (udpinfo->invflags & ~IP6T_UDP_INV_MASK) { | |
1852 | duprintf("ip6t_udp: unknown flags %X\n", | |
1853 | udpinfo->invflags); | |
1854 | return 0; | |
1855 | } | |
1856 | ||
1857 | return 1; | |
1858 | } | |
1859 | ||
1860 | /* Returns 1 if the type and code is matched by the range, 0 otherwise */ | |
1861 | static inline int | |
1862 | icmp6_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code, | |
1863 | u_int8_t type, u_int8_t code, | |
1864 | int invert) | |
1865 | { | |
1866 | return (type == test_type && code >= min_code && code <= max_code) | |
1867 | ^ invert; | |
1868 | } | |
1869 | ||
1870 | static int | |
1871 | icmp6_match(const struct sk_buff *skb, | |
1872 | const struct net_device *in, | |
1873 | const struct net_device *out, | |
1874 | const void *matchinfo, | |
1875 | int offset, | |
1876 | unsigned int protoff, | |
1877 | int *hotdrop) | |
1878 | { | |
1879 | struct icmp6hdr _icmp, *ic; | |
1880 | const struct ip6t_icmp *icmpinfo = matchinfo; | |
1881 | ||
1882 | /* Must not be a fragment. */ | |
1883 | if (offset) | |
1884 | return 0; | |
1885 | ||
1886 | ic = skb_header_pointer(skb, protoff, sizeof(_icmp), &_icmp); | |
1887 | if (ic == NULL) { | |
1888 | /* We've been asked to examine this packet, and we | |
1889 | can't. Hence, no choice but to drop. */ | |
1890 | duprintf("Dropping evil ICMP tinygram.\n"); | |
1891 | *hotdrop = 1; | |
1892 | return 0; | |
1893 | } | |
1894 | ||
1895 | return icmp6_type_code_match(icmpinfo->type, | |
1896 | icmpinfo->code[0], | |
1897 | icmpinfo->code[1], | |
1898 | ic->icmp6_type, ic->icmp6_code, | |
1899 | !!(icmpinfo->invflags&IP6T_ICMP_INV)); | |
1900 | } | |
1901 | ||
1902 | /* Called when user tries to insert an entry of this type. */ | |
1903 | static int | |
1904 | icmp6_checkentry(const char *tablename, | |
1905 | const struct ip6t_ip6 *ipv6, | |
1906 | void *matchinfo, | |
1907 | unsigned int matchsize, | |
1908 | unsigned int hook_mask) | |
1909 | { | |
1910 | const struct ip6t_icmp *icmpinfo = matchinfo; | |
1911 | ||
1912 | /* Must specify proto == ICMP, and no unknown invflags */ | |
1913 | return ipv6->proto == IPPROTO_ICMPV6 | |
1914 | && !(ipv6->invflags & IP6T_INV_PROTO) | |
1915 | && matchsize == IP6T_ALIGN(sizeof(struct ip6t_icmp)) | |
1916 | && !(icmpinfo->invflags & ~IP6T_ICMP_INV); | |
1917 | } | |
1918 | ||
1919 | /* The built-in targets: standard (NULL) and error. */ | |
1920 | static struct ip6t_target ip6t_standard_target = { | |
1921 | .name = IP6T_STANDARD_TARGET, | |
1922 | }; | |
1923 | ||
1924 | static struct ip6t_target ip6t_error_target = { | |
1925 | .name = IP6T_ERROR_TARGET, | |
1926 | .target = ip6t_error, | |
1927 | }; | |
1928 | ||
1929 | static struct nf_sockopt_ops ip6t_sockopts = { | |
1930 | .pf = PF_INET6, | |
1931 | .set_optmin = IP6T_BASE_CTL, | |
1932 | .set_optmax = IP6T_SO_SET_MAX+1, | |
1933 | .set = do_ip6t_set_ctl, | |
1934 | .get_optmin = IP6T_BASE_CTL, | |
1935 | .get_optmax = IP6T_SO_GET_MAX+1, | |
1936 | .get = do_ip6t_get_ctl, | |
1937 | }; | |
1938 | ||
1939 | static struct ip6t_match tcp_matchstruct = { | |
1940 | .name = "tcp", | |
1941 | .match = &tcp_match, | |
1942 | .checkentry = &tcp_checkentry, | |
1943 | }; | |
1944 | ||
1945 | static struct ip6t_match udp_matchstruct = { | |
1946 | .name = "udp", | |
1947 | .match = &udp_match, | |
1948 | .checkentry = &udp_checkentry, | |
1949 | }; | |
1950 | ||
1951 | static struct ip6t_match icmp6_matchstruct = { | |
1952 | .name = "icmp6", | |
1953 | .match = &icmp6_match, | |
1954 | .checkentry = &icmp6_checkentry, | |
1955 | }; | |
1956 | ||
1957 | #ifdef CONFIG_PROC_FS | |
1958 | static inline int print_name(const char *i, | |
1959 | off_t start_offset, char *buffer, int length, | |
1960 | off_t *pos, unsigned int *count) | |
1961 | { | |
1962 | if ((*count)++ >= start_offset) { | |
1963 | unsigned int namelen; | |
1964 | ||
1965 | namelen = sprintf(buffer + *pos, "%s\n", | |
1966 | i + sizeof(struct list_head)); | |
1967 | if (*pos + namelen > length) { | |
1968 | /* Stop iterating */ | |
1969 | return 1; | |
1970 | } | |
1971 | *pos += namelen; | |
1972 | } | |
1973 | return 0; | |
1974 | } | |
1975 | ||
1976 | static inline int print_target(const struct ip6t_target *t, | |
1977 | off_t start_offset, char *buffer, int length, | |
1978 | off_t *pos, unsigned int *count) | |
1979 | { | |
1980 | if (t == &ip6t_standard_target || t == &ip6t_error_target) | |
1981 | return 0; | |
1982 | return print_name((char *)t, start_offset, buffer, length, pos, count); | |
1983 | } | |
1984 | ||
1985 | static int ip6t_get_tables(char *buffer, char **start, off_t offset, int length) | |
1986 | { | |
1987 | off_t pos = 0; | |
1988 | unsigned int count = 0; | |
1989 | ||
1990 | if (down_interruptible(&ip6t_mutex) != 0) | |
1991 | return 0; | |
1992 | ||
1993 | LIST_FIND(&ip6t_tables, print_name, char *, | |
1994 | offset, buffer, length, &pos, &count); | |
1995 | ||
1996 | up(&ip6t_mutex); | |
1997 | ||
1998 | /* `start' hack - see fs/proc/generic.c line ~105 */ | |
1999 | *start=(char *)((unsigned long)count-offset); | |
2000 | return pos; | |
2001 | } | |
2002 | ||
2003 | static int ip6t_get_targets(char *buffer, char **start, off_t offset, int length) | |
2004 | { | |
2005 | off_t pos = 0; | |
2006 | unsigned int count = 0; | |
2007 | ||
2008 | if (down_interruptible(&ip6t_mutex) != 0) | |
2009 | return 0; | |
2010 | ||
2011 | LIST_FIND(&ip6t_target, print_target, struct ip6t_target *, | |
2012 | offset, buffer, length, &pos, &count); | |
2013 | ||
2014 | up(&ip6t_mutex); | |
2015 | ||
2016 | *start = (char *)((unsigned long)count - offset); | |
2017 | return pos; | |
2018 | } | |
2019 | ||
2020 | static int ip6t_get_matches(char *buffer, char **start, off_t offset, int length) | |
2021 | { | |
2022 | off_t pos = 0; | |
2023 | unsigned int count = 0; | |
2024 | ||
2025 | if (down_interruptible(&ip6t_mutex) != 0) | |
2026 | return 0; | |
2027 | ||
2028 | LIST_FIND(&ip6t_match, print_name, char *, | |
2029 | offset, buffer, length, &pos, &count); | |
2030 | ||
2031 | up(&ip6t_mutex); | |
2032 | ||
2033 | *start = (char *)((unsigned long)count - offset); | |
2034 | return pos; | |
2035 | } | |
2036 | ||
9b5b5cff | 2037 | static const struct { char *name; get_info_t *get_info; } ip6t_proc_entry[] = |
1da177e4 LT |
2038 | { { "ip6_tables_names", ip6t_get_tables }, |
2039 | { "ip6_tables_targets", ip6t_get_targets }, | |
2040 | { "ip6_tables_matches", ip6t_get_matches }, | |
2041 | { NULL, NULL} }; | |
2042 | #endif /*CONFIG_PROC_FS*/ | |
2043 | ||
2044 | static int __init init(void) | |
2045 | { | |
2046 | int ret; | |
2047 | ||
2048 | /* Noone else will be downing sem now, so we won't sleep */ | |
2049 | down(&ip6t_mutex); | |
2050 | list_append(&ip6t_target, &ip6t_standard_target); | |
2051 | list_append(&ip6t_target, &ip6t_error_target); | |
2052 | list_append(&ip6t_match, &tcp_matchstruct); | |
2053 | list_append(&ip6t_match, &udp_matchstruct); | |
2054 | list_append(&ip6t_match, &icmp6_matchstruct); | |
2055 | up(&ip6t_mutex); | |
2056 | ||
2057 | /* Register setsockopt */ | |
2058 | ret = nf_register_sockopt(&ip6t_sockopts); | |
2059 | if (ret < 0) { | |
2060 | duprintf("Unable to register sockopts.\n"); | |
2061 | return ret; | |
2062 | } | |
2063 | ||
2064 | #ifdef CONFIG_PROC_FS | |
2065 | { | |
2066 | struct proc_dir_entry *proc; | |
2067 | int i; | |
2068 | ||
2069 | for (i = 0; ip6t_proc_entry[i].name; i++) { | |
2070 | proc = proc_net_create(ip6t_proc_entry[i].name, 0, | |
2071 | ip6t_proc_entry[i].get_info); | |
2072 | if (!proc) { | |
2073 | while (--i >= 0) | |
2074 | proc_net_remove(ip6t_proc_entry[i].name); | |
2075 | nf_unregister_sockopt(&ip6t_sockopts); | |
2076 | return -ENOMEM; | |
2077 | } | |
2078 | proc->owner = THIS_MODULE; | |
2079 | } | |
2080 | } | |
2081 | #endif | |
2082 | ||
2083 | printk("ip6_tables: (C) 2000-2002 Netfilter core team\n"); | |
2084 | return 0; | |
2085 | } | |
2086 | ||
2087 | static void __exit fini(void) | |
2088 | { | |
2089 | nf_unregister_sockopt(&ip6t_sockopts); | |
2090 | #ifdef CONFIG_PROC_FS | |
2091 | { | |
2092 | int i; | |
2093 | for (i = 0; ip6t_proc_entry[i].name; i++) | |
2094 | proc_net_remove(ip6t_proc_entry[i].name); | |
2095 | } | |
2096 | #endif | |
2097 | } | |
2098 | ||
e674d0f3 YK |
2099 | /* |
2100 | * find specified header up to transport protocol header. | |
2101 | * If found target header, the offset to the header is set to *offset | |
2102 | * and return 0. otherwise, return -1. | |
2103 | * | |
2104 | * Notes: - non-1st Fragment Header isn't skipped. | |
2105 | * - ESP header isn't skipped. | |
2106 | * - The target header may be trancated. | |
2107 | */ | |
2108 | int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, u8 target) | |
2109 | { | |
2110 | unsigned int start = (u8*)(skb->nh.ipv6h + 1) - skb->data; | |
2111 | u8 nexthdr = skb->nh.ipv6h->nexthdr; | |
2112 | unsigned int len = skb->len - start; | |
2113 | ||
2114 | while (nexthdr != target) { | |
2115 | struct ipv6_opt_hdr _hdr, *hp; | |
2116 | unsigned int hdrlen; | |
2117 | ||
2118 | if ((!ipv6_ext_hdr(nexthdr)) || nexthdr == NEXTHDR_NONE) | |
2119 | return -1; | |
2120 | hp = skb_header_pointer(skb, start, sizeof(_hdr), &_hdr); | |
2121 | if (hp == NULL) | |
2122 | return -1; | |
2123 | if (nexthdr == NEXTHDR_FRAGMENT) { | |
2124 | unsigned short _frag_off, *fp; | |
2125 | fp = skb_header_pointer(skb, | |
2126 | start+offsetof(struct frag_hdr, | |
2127 | frag_off), | |
2128 | sizeof(_frag_off), | |
2129 | &_frag_off); | |
2130 | if (fp == NULL) | |
2131 | return -1; | |
2132 | ||
2133 | if (ntohs(*fp) & ~0x7) | |
2134 | return -1; | |
2135 | hdrlen = 8; | |
2136 | } else if (nexthdr == NEXTHDR_AUTH) | |
2137 | hdrlen = (hp->hdrlen + 2) << 2; | |
2138 | else | |
2139 | hdrlen = ipv6_optlen(hp); | |
2140 | ||
2141 | nexthdr = hp->nexthdr; | |
2142 | len -= hdrlen; | |
2143 | start += hdrlen; | |
2144 | } | |
2145 | ||
2146 | *offset = start; | |
2147 | return 0; | |
2148 | } | |
2149 | ||
1da177e4 LT |
2150 | EXPORT_SYMBOL(ip6t_register_table); |
2151 | EXPORT_SYMBOL(ip6t_unregister_table); | |
2152 | EXPORT_SYMBOL(ip6t_do_table); | |
2153 | EXPORT_SYMBOL(ip6t_register_match); | |
2154 | EXPORT_SYMBOL(ip6t_unregister_match); | |
2155 | EXPORT_SYMBOL(ip6t_register_target); | |
2156 | EXPORT_SYMBOL(ip6t_unregister_target); | |
2157 | EXPORT_SYMBOL(ip6t_ext_hdr); | |
e674d0f3 | 2158 | EXPORT_SYMBOL(ipv6_find_hdr); |
1da177e4 LT |
2159 | |
2160 | module_init(init); | |
2161 | module_exit(fini); |