Bart De Schuymer | d86fb4e | 2002-08-24 09:23:41 +0000 | [diff] [blame] | 1 | /* |
| 2 | * Packet matching code. |
| 3 | * |
| 4 | * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling |
| 5 | * Copyright (C) 2009-2002 Netfilter core team <coreteam@netfilter.org> |
| 6 | * |
| 7 | * 19 Jan 2002 Harald Welte <laforge@gnumonks.org> |
| 8 | * - increase module usage count as soon as we have rules inside |
| 9 | * a table |
| 10 | */ |
| 11 | #include <linux/config.h> |
| 12 | #include <linux/cache.h> |
| 13 | #include <linux/skbuff.h> |
| 14 | #include <linux/kmod.h> |
| 15 | #include <linux/vmalloc.h> |
| 16 | #include <linux/netdevice.h> |
| 17 | #include <linux/module.h> |
| 18 | #include <linux/tcp.h> |
| 19 | #include <linux/udp.h> |
| 20 | #include <linux/icmp.h> |
| 21 | #include <net/ip.h> |
| 22 | #include <asm/uaccess.h> |
| 23 | #include <asm/semaphore.h> |
| 24 | #include <linux/proc_fs.h> |
| 25 | |
| 26 | #include <linux/netfilter_ipv4/ip_tables.h> |
| 27 | |
| 28 | /*#define DEBUG_IP_FIREWALL*/ |
| 29 | /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */ |
| 30 | /*#define DEBUG_IP_FIREWALL_USER*/ |
| 31 | |
| 32 | #ifdef DEBUG_IP_FIREWALL |
| 33 | #define dprintf(format, args...) printk(format , ## args) |
| 34 | #else |
| 35 | #define dprintf(format, args...) |
| 36 | #endif |
| 37 | |
| 38 | #ifdef DEBUG_IP_FIREWALL_USER |
| 39 | #define duprintf(format, args...) printk(format , ## args) |
| 40 | #else |
| 41 | #define duprintf(format, args...) |
| 42 | #endif |
| 43 | |
| 44 | #ifdef CONFIG_NETFILTER_DEBUG |
| 45 | #define IP_NF_ASSERT(x) \ |
| 46 | do { \ |
| 47 | if (!(x)) \ |
| 48 | printk("IP_NF_ASSERT: %s:%s:%u\n", \ |
| 49 | __FUNCTION__, __FILE__, __LINE__); \ |
| 50 | } while(0) |
| 51 | #else |
| 52 | #define IP_NF_ASSERT(x) |
| 53 | #endif |
| 54 | #define SMP_ALIGN(x) (((x) + SMP_CACHE_BYTES-1) & ~(SMP_CACHE_BYTES-1)) |
| 55 | |
| 56 | /* Mutex protects lists (only traversed in user context). */ |
| 57 | static DECLARE_MUTEX(ipt_mutex); |
| 58 | |
| 59 | /* Must have mutex */ |
| 60 | #define ASSERT_READ_LOCK(x) IP_NF_ASSERT(down_trylock(&ipt_mutex) != 0) |
| 61 | #define ASSERT_WRITE_LOCK(x) IP_NF_ASSERT(down_trylock(&ipt_mutex) != 0) |
| 62 | #include <linux/netfilter_ipv4/lockhelp.h> |
| 63 | #include <linux/netfilter_ipv4/listhelp.h> |
| 64 | |
| 65 | #if 0 |
| 66 | /* All the better to debug you with... */ |
| 67 | #define static |
| 68 | #define inline |
| 69 | #endif |
| 70 | |
Bart De Schuymer | a07c59d | 2002-09-18 20:40:54 +0000 | [diff] [blame^] | 71 | /* Locking is simple: we assume at worst case there will be one packet |
| 72 | in user context and one from bottom halves (or soft irq if Alexey's |
| 73 | softnet patch was applied). |
| 74 | |
Bart De Schuymer | d86fb4e | 2002-08-24 09:23:41 +0000 | [diff] [blame] | 75 | We keep a set of rules for each CPU, so we can avoid write-locking |
Bart De Schuymer | 0434b96 | 2002-09-18 20:39:33 +0000 | [diff] [blame] | 76 | them in the softirq when updating the counters and therefore |
| 77 | only need to read-lock in the softirq; doing a write_lock_bh() in user |
| 78 | context stops packets coming through and allows user context to read |
| 79 | the counters or update the rules. |
Bart De Schuymer | d86fb4e | 2002-08-24 09:23:41 +0000 | [diff] [blame] | 80 | |
| 81 | To be cache friendly on SMP, we arrange them like so: |
| 82 | [ n-entries ] |
| 83 | ... cache-align padding ... |
| 84 | [ n-entries ] |
| 85 | |
| 86 | Hence the start of any table is given by get_table() below. */ |
| 87 | |
| 88 | /* The table itself */ |
| 89 | struct ipt_table_info |
| 90 | { |
| 91 | /* Size per table */ |
| 92 | unsigned int size; |
| 93 | /* Number of entries: FIXME. --RR */ |
| 94 | unsigned int number; |
| 95 | /* Initial number of entries. Needed for module usage count */ |
| 96 | unsigned int initial_entries; |
| 97 | |
| 98 | /* Entry points and underflows */ |
| 99 | unsigned int hook_entry[NF_IP_NUMHOOKS]; |
| 100 | unsigned int underflow[NF_IP_NUMHOOKS]; |
| 101 | |
| 102 | /* ipt_entry tables: one per CPU */ |
| 103 | char entries[0] ____cacheline_aligned; |
| 104 | }; |
| 105 | |
| 106 | static LIST_HEAD(ipt_target); |
| 107 | static LIST_HEAD(ipt_match); |
| 108 | static LIST_HEAD(ipt_tables); |
| 109 | #define ADD_COUNTER(c,b,p) do { (c).bcnt += (b); (c).pcnt += (p); } while(0) |
| 110 | |
| 111 | #ifdef CONFIG_SMP |
| 112 | #define TABLE_OFFSET(t,p) (SMP_ALIGN((t)->size)*(p)) |
| 113 | #else |
| 114 | #define TABLE_OFFSET(t,p) 0 |
| 115 | #endif |
| 116 | |
| 117 | #if 0 |
| 118 | #define down(x) do { printk("DOWN:%u:" #x "\n", __LINE__); down(x); } while(0) |
| 119 | #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; }) |
| 120 | #define up(x) do { printk("UP:%u:" #x "\n", __LINE__); up(x); } while(0) |
| 121 | #endif |
| 122 | |
| 123 | /* Returns whether matches rule or not. */ |
| 124 | static inline int |
| 125 | ip_packet_match(const struct iphdr *ip, |
| 126 | const char *indev, |
| 127 | const char *physindev, |
| 128 | const char *outdev, |
| 129 | const char *physoutdev, |
| 130 | const struct ipt_ip *ipinfo, |
| 131 | int isfrag) |
| 132 | { |
| 133 | size_t i; |
Bart De Schuymer | 4142649 | 2002-09-18 18:18:25 +0000 | [diff] [blame] | 134 | unsigned long ret, ret2; |
Bart De Schuymer | d86fb4e | 2002-08-24 09:23:41 +0000 | [diff] [blame] | 135 | |
| 136 | #define FWINV(bool,invflg) ((bool) ^ !!(ipinfo->invflags & invflg)) |
| 137 | |
| 138 | if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr, |
| 139 | IPT_INV_SRCIP) |
| 140 | || FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr, |
| 141 | IPT_INV_DSTIP)) { |
| 142 | dprintf("Source or dest mismatch.\n"); |
| 143 | |
| 144 | dprintf("SRC: %u.%u.%u.%u. Mask: %u.%u.%u.%u. Target: %u.%u.%u.%u.%s\n", |
| 145 | NIPQUAD(ip->saddr), |
| 146 | NIPQUAD(ipinfo->smsk.s_addr), |
| 147 | NIPQUAD(ipinfo->src.s_addr), |
| 148 | ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : ""); |
| 149 | dprintf("DST: %u.%u.%u.%u Mask: %u.%u.%u.%u Target: %u.%u.%u.%u.%s\n", |
| 150 | NIPQUAD(ip->daddr), |
| 151 | NIPQUAD(ipinfo->dmsk.s_addr), |
| 152 | NIPQUAD(ipinfo->dst.s_addr), |
| 153 | ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : ""); |
| 154 | return 0; |
| 155 | } |
| 156 | |
| 157 | /* Look for ifname matches; this should unroll nicely. */ |
| 158 | for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) { |
| 159 | ret |= (((const unsigned long *)indev)[i] |
| 160 | ^ ((const unsigned long *)ipinfo->iniface)[i]) |
| 161 | & ((const unsigned long *)ipinfo->iniface_mask)[i]; |
| 162 | } |
| 163 | |
| 164 | for (i = 0, ret2 = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) { |
| 165 | ret2 |= (((const unsigned long *)physindev)[i] |
| 166 | ^ ((const unsigned long *)ipinfo->iniface)[i]) |
| 167 | & ((const unsigned long *)ipinfo->iniface_mask)[i]; |
| 168 | } |
| 169 | |
| 170 | if (FWINV(ret != 0 && ret2 != 0, IPT_INV_VIA_IN)) { |
| 171 | dprintf("VIA in mismatch (%s vs %s).%s\n", |
| 172 | indev, ipinfo->iniface, |
| 173 | ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":""); |
| 174 | return 0; |
| 175 | } |
| 176 | |
| 177 | for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) { |
| 178 | ret |= (((const unsigned long *)outdev)[i] |
| 179 | ^ ((const unsigned long *)ipinfo->outiface)[i]) |
| 180 | & ((const unsigned long *)ipinfo->outiface_mask)[i]; |
| 181 | } |
| 182 | |
| 183 | for (i = 0, ret2 = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) { |
| 184 | ret2 |= (((const unsigned long *)physoutdev)[i] |
| 185 | ^ ((const unsigned long *)ipinfo->outiface)[i]) |
| 186 | & ((const unsigned long *)ipinfo->outiface_mask)[i]; |
| 187 | } |
| 188 | |
| 189 | if (FWINV(ret != 0 && ret2 != 0, IPT_INV_VIA_OUT)) { |
| 190 | dprintf("VIA out mismatch (%s vs %s).%s\n", |
| 191 | outdev, ipinfo->outiface, |
| 192 | ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":""); |
| 193 | return 0; |
| 194 | } |
| 195 | |
| 196 | /* Check specific protocol */ |
| 197 | if (ipinfo->proto |
| 198 | && FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) { |
| 199 | dprintf("Packet protocol %hi does not match %hi.%s\n", |
| 200 | ip->protocol, ipinfo->proto, |
| 201 | ipinfo->invflags&IPT_INV_PROTO ? " (INV)":""); |
| 202 | return 0; |
| 203 | } |
| 204 | |
| 205 | /* If we have a fragment rule but the packet is not a fragment |
| 206 | * then we return zero */ |
| 207 | if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) { |
| 208 | dprintf("Fragment rule but not fragment.%s\n", |
| 209 | ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : ""); |
| 210 | return 0; |
| 211 | } |
| 212 | |
| 213 | return 1; |
| 214 | } |
| 215 | |
| 216 | static inline int |
| 217 | ip_checkentry(const struct ipt_ip *ip) |
| 218 | { |
| 219 | if (ip->flags & ~IPT_F_MASK) { |
| 220 | duprintf("Unknown flag bits set: %08X\n", |
| 221 | ip->flags & ~IPT_F_MASK); |
| 222 | return 0; |
| 223 | } |
| 224 | if (ip->invflags & ~IPT_INV_MASK) { |
| 225 | duprintf("Unknown invflag bits set: %08X\n", |
| 226 | ip->invflags & ~IPT_INV_MASK); |
| 227 | return 0; |
| 228 | } |
| 229 | return 1; |
| 230 | } |
| 231 | |
| 232 | static unsigned int |
| 233 | ipt_error(struct sk_buff **pskb, |
| 234 | unsigned int hooknum, |
| 235 | const struct net_device *in, |
| 236 | const struct net_device *out, |
| 237 | const void *targinfo, |
| 238 | void *userinfo) |
| 239 | { |
| 240 | if (net_ratelimit()) |
| 241 | printk("ip_tables: error: `%s'\n", (char *)targinfo); |
| 242 | |
| 243 | return NF_DROP; |
| 244 | } |
| 245 | |
| 246 | static inline |
| 247 | int do_match(struct ipt_entry_match *m, |
| 248 | const struct sk_buff *skb, |
| 249 | const struct net_device *in, |
| 250 | const struct net_device *out, |
| 251 | int offset, |
| 252 | const void *hdr, |
| 253 | u_int16_t datalen, |
| 254 | int *hotdrop) |
| 255 | { |
| 256 | /* Stop iteration if it doesn't match */ |
| 257 | if (!m->u.kernel.match->match(skb, in, out, m->data, |
| 258 | offset, hdr, datalen, hotdrop)) |
| 259 | return 1; |
| 260 | else |
| 261 | return 0; |
| 262 | } |
| 263 | |
| 264 | static inline struct ipt_entry * |
| 265 | get_entry(void *base, unsigned int offset) |
| 266 | { |
| 267 | return (struct ipt_entry *)(base + offset); |
| 268 | } |
| 269 | |
| 270 | /* Returns one of the generic firewall policies, like NF_ACCEPT. */ |
| 271 | unsigned int |
| 272 | ipt_do_table(struct sk_buff **pskb, |
| 273 | unsigned int hook, |
| 274 | const struct net_device *in, |
| 275 | const struct net_device *out, |
| 276 | struct ipt_table *table, |
| 277 | void *userdata) |
| 278 | { |
| 279 | static const char nulldevname[IFNAMSIZ] = { 0 }; |
| 280 | u_int16_t offset; |
| 281 | struct iphdr *ip; |
| 282 | void *protohdr; |
| 283 | u_int16_t datalen; |
| 284 | int hotdrop = 0; |
| 285 | /* Initializing verdict to NF_DROP keeps gcc happy. */ |
| 286 | unsigned int verdict = NF_DROP; |
| 287 | const char *indev, *outdev; |
| 288 | const char *physindev, *physoutdev; |
| 289 | void *table_base; |
| 290 | struct ipt_entry *e, *back; |
| 291 | |
| 292 | /* Initialization */ |
| 293 | ip = (*pskb)->nh.iph; |
| 294 | protohdr = (u_int32_t *)ip + ip->ihl; |
| 295 | datalen = (*pskb)->len - ip->ihl * 4; |
| 296 | indev = in ? in->name : nulldevname; |
| 297 | outdev = out ? out->name : nulldevname; |
| 298 | physindev = (*pskb)->physindev ? (*pskb)->physindev->name : nulldevname; |
| 299 | physoutdev = (*pskb)->physoutdev ? (*pskb)->physoutdev->name : nulldevname; |
| 300 | |
| 301 | /* We handle fragments by dealing with the first fragment as |
| 302 | * if it was a normal packet. All other fragments are treated |
| 303 | * normally, except that they will NEVER match rules that ask |
| 304 | * things we don't know, ie. tcp syn flag or ports). If the |
| 305 | * rule is also a fragment-specific rule, non-fragments won't |
| 306 | * match it. */ |
| 307 | offset = ntohs(ip->frag_off) & IP_OFFSET; |
| 308 | |
| 309 | read_lock_bh(&table->lock); |
| 310 | IP_NF_ASSERT(table->valid_hooks & (1 << hook)); |
| 311 | table_base = (void *)table->private->entries |
| 312 | + TABLE_OFFSET(table->private, smp_processor_id()); |
| 313 | e = get_entry(table_base, table->private->hook_entry[hook]); |
| 314 | |
| 315 | #ifdef CONFIG_NETFILTER_DEBUG |
| 316 | /* Check noone else using our table */ |
| 317 | if (((struct ipt_entry *)table_base)->comefrom != 0xdead57ac |
| 318 | && ((struct ipt_entry *)table_base)->comefrom != 0xeeeeeeec) { |
| 319 | printk("ASSERT: CPU #%u, %s comefrom(%p) = %X\n", |
| 320 | smp_processor_id(), |
| 321 | table->name, |
| 322 | &((struct ipt_entry *)table_base)->comefrom, |
| 323 | ((struct ipt_entry *)table_base)->comefrom); |
| 324 | } |
| 325 | ((struct ipt_entry *)table_base)->comefrom = 0x57acc001; |
| 326 | #endif |
| 327 | |
| 328 | /* For return from builtin chain */ |
| 329 | back = get_entry(table_base, table->private->underflow[hook]); |
| 330 | |
| 331 | do { |
| 332 | IP_NF_ASSERT(e); |
| 333 | IP_NF_ASSERT(back); |
| 334 | (*pskb)->nfcache |= e->nfcache; |
Bart De Schuymer | 4142649 | 2002-09-18 18:18:25 +0000 | [diff] [blame] | 335 | if (ip_packet_match(ip, indev, physindev, outdev, physoutdev, |
Bart De Schuymer | 2717b28 | 2002-08-24 13:41:45 +0000 | [diff] [blame] | 336 | &e->ip, offset)) { |
Bart De Schuymer | d86fb4e | 2002-08-24 09:23:41 +0000 | [diff] [blame] | 337 | struct ipt_entry_target *t; |
| 338 | |
| 339 | if (IPT_MATCH_ITERATE(e, do_match, |
| 340 | *pskb, in, out, |
| 341 | offset, protohdr, |
| 342 | datalen, &hotdrop) != 0) |
| 343 | goto no_match; |
| 344 | |
| 345 | ADD_COUNTER(e->counters, ntohs(ip->tot_len), 1); |
| 346 | |
| 347 | t = ipt_get_target(e); |
| 348 | IP_NF_ASSERT(t->u.kernel.target); |
| 349 | /* Standard target? */ |
| 350 | if (!t->u.kernel.target->target) { |
| 351 | int v; |
| 352 | |
| 353 | v = ((struct ipt_standard_target *)t)->verdict; |
| 354 | if (v < 0) { |
| 355 | /* Pop from stack? */ |
| 356 | if (v != IPT_RETURN) { |
| 357 | verdict = (unsigned)(-v) - 1; |
| 358 | break; |
| 359 | } |
| 360 | e = back; |
| 361 | back = get_entry(table_base, |
| 362 | back->comefrom); |
| 363 | continue; |
| 364 | } |
| 365 | if (table_base + v |
| 366 | != (void *)e + e->next_offset) { |
| 367 | /* Save old back ptr in next entry */ |
| 368 | struct ipt_entry *next |
| 369 | = (void *)e + e->next_offset; |
| 370 | next->comefrom |
| 371 | = (void *)back - table_base; |
| 372 | /* set back pointer to next entry */ |
| 373 | back = next; |
| 374 | } |
| 375 | |
| 376 | e = get_entry(table_base, v); |
| 377 | } else { |
| 378 | /* Targets which reenter must return |
| 379 | abs. verdicts */ |
| 380 | #ifdef CONFIG_NETFILTER_DEBUG |
| 381 | ((struct ipt_entry *)table_base)->comefrom |
| 382 | = 0xeeeeeeec; |
| 383 | #endif |
| 384 | verdict = t->u.kernel.target->target(pskb, |
| 385 | hook, |
| 386 | in, out, |
| 387 | t->data, |
| 388 | userdata); |
| 389 | |
| 390 | #ifdef CONFIG_NETFILTER_DEBUG |
| 391 | if (((struct ipt_entry *)table_base)->comefrom |
| 392 | != 0xeeeeeeec |
| 393 | && verdict == IPT_CONTINUE) { |
| 394 | printk("Target %s reentered!\n", |
| 395 | t->u.kernel.target->name); |
| 396 | verdict = NF_DROP; |
| 397 | } |
| 398 | ((struct ipt_entry *)table_base)->comefrom |
| 399 | = 0x57acc001; |
| 400 | #endif |
| 401 | /* Target might have changed stuff. */ |
| 402 | ip = (*pskb)->nh.iph; |
| 403 | protohdr = (u_int32_t *)ip + ip->ihl; |
| 404 | datalen = (*pskb)->len - ip->ihl * 4; |
| 405 | |
| 406 | if (verdict == IPT_CONTINUE) |
| 407 | e = (void *)e + e->next_offset; |
| 408 | else |
| 409 | /* Verdict */ |
| 410 | break; |
| 411 | } |
| 412 | } else { |
| 413 | |
| 414 | no_match: |
| 415 | e = (void *)e + e->next_offset; |
| 416 | } |
| 417 | } while (!hotdrop); |
| 418 | |
| 419 | #ifdef CONFIG_NETFILTER_DEBUG |
| 420 | ((struct ipt_entry *)table_base)->comefrom = 0xdead57ac; |
| 421 | #endif |
| 422 | read_unlock_bh(&table->lock); |
| 423 | |
| 424 | #ifdef DEBUG_ALLOW_ALL |
| 425 | return NF_ACCEPT; |
| 426 | #else |
| 427 | if (hotdrop) |
| 428 | return NF_DROP; |
| 429 | else return verdict; |
| 430 | #endif |
| 431 | } |
| 432 | |
| 433 | /* If it succeeds, returns element and locks mutex */ |
| 434 | static inline void * |
| 435 | find_inlist_lock_noload(struct list_head *head, |
| 436 | const char *name, |
| 437 | int *error, |
| 438 | struct semaphore *mutex) |
| 439 | { |
| 440 | void *ret; |
| 441 | |
| 442 | #if 0 |
| 443 | duprintf("find_inlist: searching for `%s' in %s.\n", |
| 444 | name, head == &ipt_target ? "ipt_target" |
| 445 | : head == &ipt_match ? "ipt_match" |
| 446 | : head == &ipt_tables ? "ipt_tables" : "UNKNOWN"); |
| 447 | #endif |
| 448 | |
| 449 | *error = down_interruptible(mutex); |
| 450 | if (*error != 0) |
| 451 | return NULL; |
| 452 | |
| 453 | ret = list_named_find(head, name); |
| 454 | if (!ret) { |
| 455 | *error = -ENOENT; |
| 456 | up(mutex); |
| 457 | } |
| 458 | return ret; |
| 459 | } |
| 460 | |
| 461 | #ifndef CONFIG_KMOD |
| 462 | #define find_inlist_lock(h,n,p,e,m) find_inlist_lock_noload((h),(n),(e),(m)) |
| 463 | #else |
| 464 | static void * |
| 465 | find_inlist_lock(struct list_head *head, |
| 466 | const char *name, |
| 467 | const char *prefix, |
| 468 | int *error, |
| 469 | struct semaphore *mutex) |
| 470 | { |
| 471 | void *ret; |
| 472 | |
| 473 | ret = find_inlist_lock_noload(head, name, error, mutex); |
| 474 | if (!ret) { |
| 475 | char modulename[IPT_FUNCTION_MAXNAMELEN + strlen(prefix) + 1]; |
| 476 | strcpy(modulename, prefix); |
| 477 | strcat(modulename, name); |
| 478 | duprintf("find_inlist: loading `%s'.\n", modulename); |
| 479 | request_module(modulename); |
| 480 | ret = find_inlist_lock_noload(head, name, error, mutex); |
| 481 | } |
| 482 | |
| 483 | return ret; |
| 484 | } |
| 485 | #endif |
| 486 | |
| 487 | static inline struct ipt_table * |
| 488 | find_table_lock(const char *name, int *error, struct semaphore *mutex) |
| 489 | { |
| 490 | return find_inlist_lock(&ipt_tables, name, "iptable_", error, mutex); |
| 491 | } |
| 492 | |
| 493 | static inline struct ipt_match * |
| 494 | find_match_lock(const char *name, int *error, struct semaphore *mutex) |
| 495 | { |
| 496 | return find_inlist_lock(&ipt_match, name, "ipt_", error, mutex); |
| 497 | } |
| 498 | |
| 499 | static inline struct ipt_target * |
| 500 | find_target_lock(const char *name, int *error, struct semaphore *mutex) |
| 501 | { |
| 502 | return find_inlist_lock(&ipt_target, name, "ipt_", error, mutex); |
| 503 | } |
| 504 | |
| 505 | /* All zeroes == unconditional rule. */ |
| 506 | static inline int |
| 507 | unconditional(const struct ipt_ip *ip) |
| 508 | { |
| 509 | unsigned int i; |
| 510 | |
| 511 | for (i = 0; i < sizeof(*ip)/sizeof(__u32); i++) |
| 512 | if (((__u32 *)ip)[i]) |
| 513 | return 0; |
| 514 | |
| 515 | return 1; |
| 516 | } |
| 517 | |
| 518 | /* Figures out from what hook each rule can be called: returns 0 if |
| 519 | there are loops. Puts hook bitmask in comefrom. */ |
| 520 | static int |
| 521 | mark_source_chains(struct ipt_table_info *newinfo, unsigned int valid_hooks) |
| 522 | { |
| 523 | unsigned int hook; |
| 524 | |
| 525 | /* No recursion; use packet counter to save back ptrs (reset |
| 526 | to 0 as we leave), and comefrom to save source hook bitmask */ |
| 527 | for (hook = 0; hook < NF_IP_NUMHOOKS; hook++) { |
| 528 | unsigned int pos = newinfo->hook_entry[hook]; |
| 529 | struct ipt_entry *e |
| 530 | = (struct ipt_entry *)(newinfo->entries + pos); |
| 531 | |
| 532 | if (!(valid_hooks & (1 << hook))) |
| 533 | continue; |
| 534 | |
| 535 | /* Set initial back pointer. */ |
| 536 | e->counters.pcnt = pos; |
| 537 | |
| 538 | for (;;) { |
| 539 | struct ipt_standard_target *t |
| 540 | = (void *)ipt_get_target(e); |
| 541 | |
| 542 | if (e->comefrom & (1 << NF_IP_NUMHOOKS)) { |
| 543 | printk("iptables: loop hook %u pos %u %08X.\n", |
| 544 | hook, pos, e->comefrom); |
| 545 | return 0; |
| 546 | } |
| 547 | e->comefrom |
| 548 | |= ((1 << hook) | (1 << NF_IP_NUMHOOKS)); |
| 549 | |
| 550 | /* Unconditional return/END. */ |
| 551 | if (e->target_offset == sizeof(struct ipt_entry) |
| 552 | && (strcmp(t->target.u.user.name, |
| 553 | IPT_STANDARD_TARGET) == 0) |
| 554 | && t->verdict < 0 |
| 555 | && unconditional(&e->ip)) { |
| 556 | unsigned int oldpos, size; |
| 557 | |
| 558 | /* Return: backtrack through the last |
| 559 | big jump. */ |
| 560 | do { |
| 561 | e->comefrom ^= (1<<NF_IP_NUMHOOKS); |
| 562 | #ifdef DEBUG_IP_FIREWALL_USER |
| 563 | if (e->comefrom |
| 564 | & (1 << NF_IP_NUMHOOKS)) { |
| 565 | duprintf("Back unset " |
| 566 | "on hook %u " |
| 567 | "rule %u\n", |
| 568 | hook, pos); |
| 569 | } |
| 570 | #endif |
| 571 | oldpos = pos; |
| 572 | pos = e->counters.pcnt; |
| 573 | e->counters.pcnt = 0; |
| 574 | |
| 575 | /* We're at the start. */ |
| 576 | if (pos == oldpos) |
| 577 | goto next; |
| 578 | |
| 579 | e = (struct ipt_entry *) |
| 580 | (newinfo->entries + pos); |
| 581 | } while (oldpos == pos + e->next_offset); |
| 582 | |
| 583 | /* Move along one */ |
| 584 | size = e->next_offset; |
| 585 | e = (struct ipt_entry *) |
| 586 | (newinfo->entries + pos + size); |
| 587 | e->counters.pcnt = pos; |
| 588 | pos += size; |
| 589 | } else { |
| 590 | int newpos = t->verdict; |
| 591 | |
| 592 | if (strcmp(t->target.u.user.name, |
| 593 | IPT_STANDARD_TARGET) == 0 |
| 594 | && newpos >= 0) { |
| 595 | /* This a jump; chase it. */ |
| 596 | duprintf("Jump rule %u -> %u\n", |
| 597 | pos, newpos); |
| 598 | } else { |
| 599 | /* ... this is a fallthru */ |
| 600 | newpos = pos + e->next_offset; |
| 601 | } |
| 602 | e = (struct ipt_entry *) |
| 603 | (newinfo->entries + newpos); |
| 604 | e->counters.pcnt = pos; |
| 605 | pos = newpos; |
| 606 | } |
| 607 | } |
| 608 | next: |
| 609 | duprintf("Finished chain %u\n", hook); |
| 610 | } |
| 611 | return 1; |
| 612 | } |
| 613 | |
| 614 | static inline int |
| 615 | cleanup_match(struct ipt_entry_match *m, unsigned int *i) |
| 616 | { |
| 617 | if (i && (*i)-- == 0) |
| 618 | return 1; |
| 619 | |
| 620 | if (m->u.kernel.match->destroy) |
| 621 | m->u.kernel.match->destroy(m->data, |
| 622 | m->u.match_size - sizeof(*m)); |
| 623 | |
| 624 | if (m->u.kernel.match->me) |
| 625 | __MOD_DEC_USE_COUNT(m->u.kernel.match->me); |
| 626 | |
| 627 | return 0; |
| 628 | } |
| 629 | |
| 630 | static inline int |
| 631 | standard_check(const struct ipt_entry_target *t, |
| 632 | unsigned int max_offset) |
| 633 | { |
| 634 | struct ipt_standard_target *targ = (void *)t; |
| 635 | |
| 636 | /* Check standard info. */ |
| 637 | if (t->u.target_size |
| 638 | != IPT_ALIGN(sizeof(struct ipt_standard_target))) { |
| 639 | duprintf("standard_check: target size %u != %u\n", |
| 640 | t->u.target_size, |
| 641 | IPT_ALIGN(sizeof(struct ipt_standard_target))); |
| 642 | return 0; |
| 643 | } |
| 644 | |
| 645 | if (targ->verdict >= 0 |
| 646 | && targ->verdict > max_offset - sizeof(struct ipt_entry)) { |
| 647 | duprintf("ipt_standard_check: bad verdict (%i)\n", |
| 648 | targ->verdict); |
| 649 | return 0; |
| 650 | } |
| 651 | |
| 652 | if (targ->verdict < -NF_MAX_VERDICT - 1) { |
| 653 | duprintf("ipt_standard_check: bad negative verdict (%i)\n", |
| 654 | targ->verdict); |
| 655 | return 0; |
| 656 | } |
| 657 | return 1; |
| 658 | } |
| 659 | |
| 660 | static inline int |
| 661 | check_match(struct ipt_entry_match *m, |
| 662 | const char *name, |
| 663 | const struct ipt_ip *ip, |
| 664 | unsigned int hookmask, |
| 665 | unsigned int *i) |
| 666 | { |
| 667 | int ret; |
| 668 | struct ipt_match *match; |
| 669 | |
| 670 | match = find_match_lock(m->u.user.name, &ret, &ipt_mutex); |
| 671 | if (!match) { |
| 672 | duprintf("check_match: `%s' not found\n", m->u.user.name); |
| 673 | return ret; |
| 674 | } |
| 675 | if (match->me) |
| 676 | __MOD_INC_USE_COUNT(match->me); |
| 677 | m->u.kernel.match = match; |
| 678 | up(&ipt_mutex); |
| 679 | |
| 680 | if (m->u.kernel.match->checkentry |
| 681 | && !m->u.kernel.match->checkentry(name, ip, m->data, |
| 682 | m->u.match_size - sizeof(*m), |
| 683 | hookmask)) { |
| 684 | if (m->u.kernel.match->me) |
| 685 | __MOD_DEC_USE_COUNT(m->u.kernel.match->me); |
| 686 | duprintf("ip_tables: check failed for `%s'.\n", |
| 687 | m->u.kernel.match->name); |
| 688 | return -EINVAL; |
| 689 | } |
| 690 | |
| 691 | (*i)++; |
| 692 | return 0; |
| 693 | } |
| 694 | |
| 695 | static struct ipt_target ipt_standard_target; |
| 696 | |
| 697 | static inline int |
| 698 | check_entry(struct ipt_entry *e, const char *name, unsigned int size, |
| 699 | unsigned int *i) |
| 700 | { |
| 701 | struct ipt_entry_target *t; |
| 702 | struct ipt_target *target; |
| 703 | int ret; |
| 704 | unsigned int j; |
| 705 | |
| 706 | if (!ip_checkentry(&e->ip)) { |
| 707 | duprintf("ip_tables: ip check failed %p %s.\n", e, name); |
| 708 | return -EINVAL; |
| 709 | } |
| 710 | |
| 711 | j = 0; |
| 712 | ret = IPT_MATCH_ITERATE(e, check_match, name, &e->ip, e->comefrom, &j); |
| 713 | if (ret != 0) |
| 714 | goto cleanup_matches; |
| 715 | |
| 716 | t = ipt_get_target(e); |
| 717 | target = find_target_lock(t->u.user.name, &ret, &ipt_mutex); |
| 718 | if (!target) { |
| 719 | duprintf("check_entry: `%s' not found\n", t->u.user.name); |
| 720 | goto cleanup_matches; |
| 721 | } |
| 722 | if (target->me) |
| 723 | __MOD_INC_USE_COUNT(target->me); |
| 724 | t->u.kernel.target = target; |
| 725 | up(&ipt_mutex); |
| 726 | |
| 727 | if (t->u.kernel.target == &ipt_standard_target) { |
| 728 | if (!standard_check(t, size)) { |
| 729 | ret = -EINVAL; |
| 730 | goto cleanup_matches; |
| 731 | } |
| 732 | } else if (t->u.kernel.target->checkentry |
| 733 | && !t->u.kernel.target->checkentry(name, e, t->data, |
| 734 | t->u.target_size |
| 735 | - sizeof(*t), |
| 736 | e->comefrom)) { |
| 737 | if (t->u.kernel.target->me) |
| 738 | __MOD_DEC_USE_COUNT(t->u.kernel.target->me); |
| 739 | duprintf("ip_tables: check failed for `%s'.\n", |
| 740 | t->u.kernel.target->name); |
| 741 | ret = -EINVAL; |
| 742 | goto cleanup_matches; |
| 743 | } |
| 744 | |
| 745 | (*i)++; |
| 746 | return 0; |
| 747 | |
| 748 | cleanup_matches: |
| 749 | IPT_MATCH_ITERATE(e, cleanup_match, &j); |
| 750 | return ret; |
| 751 | } |
| 752 | |
| 753 | static inline int |
| 754 | check_entry_size_and_hooks(struct ipt_entry *e, |
| 755 | struct ipt_table_info *newinfo, |
| 756 | unsigned char *base, |
| 757 | unsigned char *limit, |
| 758 | const unsigned int *hook_entries, |
| 759 | const unsigned int *underflows, |
| 760 | unsigned int *i) |
| 761 | { |
| 762 | unsigned int h; |
| 763 | |
| 764 | if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 |
| 765 | || (unsigned char *)e + sizeof(struct ipt_entry) >= limit) { |
| 766 | duprintf("Bad offset %p\n", e); |
| 767 | return -EINVAL; |
| 768 | } |
| 769 | |
| 770 | if (e->next_offset |
| 771 | < sizeof(struct ipt_entry) + sizeof(struct ipt_entry_target)) { |
| 772 | duprintf("checking: element %p size %u\n", |
| 773 | e, e->next_offset); |
| 774 | return -EINVAL; |
| 775 | } |
| 776 | |
| 777 | /* Check hooks & underflows */ |
| 778 | for (h = 0; h < NF_IP_NUMHOOKS; h++) { |
| 779 | if ((unsigned char *)e - base == hook_entries[h]) |
| 780 | newinfo->hook_entry[h] = hook_entries[h]; |
| 781 | if ((unsigned char *)e - base == underflows[h]) |
| 782 | newinfo->underflow[h] = underflows[h]; |
| 783 | } |
| 784 | |
| 785 | /* FIXME: underflows must be unconditional, standard verdicts |
| 786 | < 0 (not IPT_RETURN). --RR */ |
| 787 | |
| 788 | /* Clear counters and comefrom */ |
| 789 | e->counters = ((struct ipt_counters) { 0, 0 }); |
| 790 | e->comefrom = 0; |
| 791 | |
| 792 | (*i)++; |
| 793 | return 0; |
| 794 | } |
| 795 | |
| 796 | static inline int |
| 797 | cleanup_entry(struct ipt_entry *e, unsigned int *i) |
| 798 | { |
| 799 | struct ipt_entry_target *t; |
| 800 | |
| 801 | if (i && (*i)-- == 0) |
| 802 | return 1; |
| 803 | |
| 804 | /* Cleanup all matches */ |
| 805 | IPT_MATCH_ITERATE(e, cleanup_match, NULL); |
| 806 | t = ipt_get_target(e); |
| 807 | if (t->u.kernel.target->destroy) |
| 808 | t->u.kernel.target->destroy(t->data, |
| 809 | t->u.target_size - sizeof(*t)); |
| 810 | if (t->u.kernel.target->me) |
| 811 | __MOD_DEC_USE_COUNT(t->u.kernel.target->me); |
| 812 | |
| 813 | return 0; |
| 814 | } |
| 815 | |
| 816 | /* Checks and translates the user-supplied table segment (held in |
| 817 | newinfo) */ |
| 818 | static int |
| 819 | translate_table(const char *name, |
| 820 | unsigned int valid_hooks, |
| 821 | struct ipt_table_info *newinfo, |
| 822 | unsigned int size, |
| 823 | unsigned int number, |
| 824 | const unsigned int *hook_entries, |
| 825 | const unsigned int *underflows) |
| 826 | { |
| 827 | unsigned int i; |
| 828 | int ret; |
| 829 | |
| 830 | newinfo->size = size; |
| 831 | newinfo->number = number; |
| 832 | |
| 833 | /* Init all hooks to impossible value. */ |
| 834 | for (i = 0; i < NF_IP_NUMHOOKS; i++) { |
| 835 | newinfo->hook_entry[i] = 0xFFFFFFFF; |
| 836 | newinfo->underflow[i] = 0xFFFFFFFF; |
| 837 | } |
| 838 | |
| 839 | duprintf("translate_table: size %u\n", newinfo->size); |
| 840 | i = 0; |
| 841 | /* Walk through entries, checking offsets. */ |
| 842 | ret = IPT_ENTRY_ITERATE(newinfo->entries, newinfo->size, |
| 843 | check_entry_size_and_hooks, |
| 844 | newinfo, |
| 845 | newinfo->entries, |
| 846 | newinfo->entries + size, |
| 847 | hook_entries, underflows, &i); |
| 848 | if (ret != 0) |
| 849 | return ret; |
| 850 | |
| 851 | if (i != number) { |
| 852 | duprintf("translate_table: %u not %u entries\n", |
| 853 | i, number); |
| 854 | return -EINVAL; |
| 855 | } |
| 856 | |
| 857 | /* Check hooks all assigned */ |
| 858 | for (i = 0; i < NF_IP_NUMHOOKS; i++) { |
| 859 | /* Only hooks which are valid */ |
| 860 | if (!(valid_hooks & (1 << i))) |
| 861 | continue; |
| 862 | if (newinfo->hook_entry[i] == 0xFFFFFFFF) { |
| 863 | duprintf("Invalid hook entry %u %u\n", |
| 864 | i, hook_entries[i]); |
| 865 | return -EINVAL; |
| 866 | } |
| 867 | if (newinfo->underflow[i] == 0xFFFFFFFF) { |
| 868 | duprintf("Invalid underflow %u %u\n", |
| 869 | i, underflows[i]); |
| 870 | return -EINVAL; |
| 871 | } |
| 872 | } |
| 873 | |
| 874 | if (!mark_source_chains(newinfo, valid_hooks)) |
| 875 | return -ELOOP; |
| 876 | |
| 877 | /* Finally, each sanity check must pass */ |
| 878 | i = 0; |
| 879 | ret = IPT_ENTRY_ITERATE(newinfo->entries, newinfo->size, |
| 880 | check_entry, name, size, &i); |
| 881 | |
| 882 | if (ret != 0) { |
| 883 | IPT_ENTRY_ITERATE(newinfo->entries, newinfo->size, |
| 884 | cleanup_entry, &i); |
| 885 | return ret; |
| 886 | } |
| 887 | |
| 888 | /* And one copy for every other CPU */ |
| 889 | for (i = 1; i < NR_CPUS; i++) { |
| 890 | memcpy(newinfo->entries + SMP_ALIGN(newinfo->size)*i, |
| 891 | newinfo->entries, |
| 892 | SMP_ALIGN(newinfo->size)); |
| 893 | } |
| 894 | |
| 895 | return ret; |
| 896 | } |
| 897 | |
| 898 | static struct ipt_table_info * |
| 899 | replace_table(struct ipt_table *table, |
| 900 | unsigned int num_counters, |
| 901 | struct ipt_table_info *newinfo, |
| 902 | int *error) |
| 903 | { |
| 904 | struct ipt_table_info *oldinfo; |
| 905 | |
| 906 | #ifdef CONFIG_NETFILTER_DEBUG |
| 907 | { |
| 908 | struct ipt_entry *table_base; |
| 909 | unsigned int i; |
| 910 | |
| 911 | for (i = 0; i < NR_CPUS; i++) { |
| 912 | table_base = |
| 913 | (void *)newinfo->entries |
| 914 | + TABLE_OFFSET(newinfo, i); |
| 915 | |
| 916 | table_base->comefrom = 0xdead57ac; |
| 917 | } |
| 918 | } |
| 919 | #endif |
| 920 | |
| 921 | /* Do the substitution. */ |
| 922 | write_lock_bh(&table->lock); |
| 923 | /* Check inside lock: is the old number correct? */ |
| 924 | if (num_counters != table->private->number) { |
| 925 | duprintf("num_counters != table->private->number (%u/%u)\n", |
| 926 | num_counters, table->private->number); |
| 927 | write_unlock_bh(&table->lock); |
| 928 | *error = -EAGAIN; |
| 929 | return NULL; |
| 930 | } |
| 931 | oldinfo = table->private; |
| 932 | table->private = newinfo; |
| 933 | newinfo->initial_entries = oldinfo->initial_entries; |
| 934 | write_unlock_bh(&table->lock); |
| 935 | |
| 936 | return oldinfo; |
| 937 | } |
| 938 | |
| 939 | /* Gets counters. */ |
| 940 | static inline int |
| 941 | add_entry_to_counter(const struct ipt_entry *e, |
| 942 | struct ipt_counters total[], |
| 943 | unsigned int *i) |
| 944 | { |
| 945 | ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt); |
| 946 | |
| 947 | (*i)++; |
| 948 | return 0; |
| 949 | } |
| 950 | |
| 951 | static void |
| 952 | get_counters(const struct ipt_table_info *t, |
| 953 | struct ipt_counters counters[]) |
| 954 | { |
| 955 | unsigned int cpu; |
| 956 | unsigned int i; |
| 957 | |
| 958 | for (cpu = 0; cpu < NR_CPUS; cpu++) { |
| 959 | i = 0; |
| 960 | IPT_ENTRY_ITERATE(t->entries + TABLE_OFFSET(t, cpu), |
| 961 | t->size, |
| 962 | add_entry_to_counter, |
| 963 | counters, |
| 964 | &i); |
| 965 | } |
| 966 | } |
| 967 | |
| 968 | static int |
| 969 | copy_entries_to_user(unsigned int total_size, |
| 970 | struct ipt_table *table, |
| 971 | void *userptr) |
| 972 | { |
| 973 | unsigned int off, num, countersize; |
| 974 | struct ipt_entry *e; |
| 975 | struct ipt_counters *counters; |
| 976 | int ret = 0; |
| 977 | |
| 978 | /* We need atomic snapshot of counters: rest doesn't change |
| 979 | (other than comefrom, which userspace doesn't care |
| 980 | about). */ |
| 981 | countersize = sizeof(struct ipt_counters) * table->private->number; |
| 982 | counters = vmalloc(countersize); |
| 983 | |
| 984 | if (counters == NULL) |
| 985 | return -ENOMEM; |
| 986 | |
| 987 | /* First, sum counters... */ |
| 988 | memset(counters, 0, countersize); |
| 989 | write_lock_bh(&table->lock); |
| 990 | get_counters(table->private, counters); |
| 991 | write_unlock_bh(&table->lock); |
| 992 | |
| 993 | /* ... then copy entire thing from CPU 0... */ |
| 994 | if (copy_to_user(userptr, table->private->entries, total_size) != 0) { |
| 995 | ret = -EFAULT; |
| 996 | goto free_counters; |
| 997 | } |
| 998 | |
| 999 | /* FIXME: use iterator macros --RR */ |
| 1000 | /* ... then go back and fix counters and names */ |
| 1001 | for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){ |
| 1002 | unsigned int i; |
| 1003 | struct ipt_entry_match *m; |
| 1004 | struct ipt_entry_target *t; |
| 1005 | |
| 1006 | e = (struct ipt_entry *)(table->private->entries + off); |
| 1007 | if (copy_to_user(userptr + off |
| 1008 | + offsetof(struct ipt_entry, counters), |
| 1009 | &counters[num], |
| 1010 | sizeof(counters[num])) != 0) { |
| 1011 | ret = -EFAULT; |
| 1012 | goto free_counters; |
| 1013 | } |
| 1014 | |
| 1015 | for (i = sizeof(struct ipt_entry); |
| 1016 | i < e->target_offset; |
| 1017 | i += m->u.match_size) { |
| 1018 | m = (void *)e + i; |
| 1019 | |
| 1020 | if (copy_to_user(userptr + off + i |
| 1021 | + offsetof(struct ipt_entry_match, |
| 1022 | u.user.name), |
| 1023 | m->u.kernel.match->name, |
| 1024 | strlen(m->u.kernel.match->name)+1) |
| 1025 | != 0) { |
| 1026 | ret = -EFAULT; |
| 1027 | goto free_counters; |
| 1028 | } |
| 1029 | } |
| 1030 | |
| 1031 | t = ipt_get_target(e); |
| 1032 | if (copy_to_user(userptr + off + e->target_offset |
| 1033 | + offsetof(struct ipt_entry_target, |
| 1034 | u.user.name), |
| 1035 | t->u.kernel.target->name, |
| 1036 | strlen(t->u.kernel.target->name)+1) != 0) { |
| 1037 | ret = -EFAULT; |
| 1038 | goto free_counters; |
| 1039 | } |
| 1040 | } |
| 1041 | |
| 1042 | free_counters: |
| 1043 | vfree(counters); |
| 1044 | return ret; |
| 1045 | } |
| 1046 | |
| 1047 | static int |
| 1048 | get_entries(const struct ipt_get_entries *entries, |
| 1049 | struct ipt_get_entries *uptr) |
| 1050 | { |
| 1051 | int ret; |
| 1052 | struct ipt_table *t; |
| 1053 | |
| 1054 | t = find_table_lock(entries->name, &ret, &ipt_mutex); |
| 1055 | if (t) { |
| 1056 | duprintf("t->private->number = %u\n", |
| 1057 | t->private->number); |
| 1058 | if (entries->size == t->private->size) |
| 1059 | ret = copy_entries_to_user(t->private->size, |
| 1060 | t, uptr->entrytable); |
| 1061 | else { |
| 1062 | duprintf("get_entries: I've got %u not %u!\n", |
| 1063 | t->private->size, |
| 1064 | entries->size); |
| 1065 | ret = -EINVAL; |
| 1066 | } |
| 1067 | up(&ipt_mutex); |
| 1068 | } else |
| 1069 | duprintf("get_entries: Can't find %s!\n", |
| 1070 | entries->name); |
| 1071 | |
| 1072 | return ret; |
| 1073 | } |
| 1074 | |
| 1075 | static int |
| 1076 | do_replace(void *user, unsigned int len) |
| 1077 | { |
| 1078 | int ret; |
| 1079 | struct ipt_replace tmp; |
| 1080 | struct ipt_table *t; |
| 1081 | struct ipt_table_info *newinfo, *oldinfo; |
| 1082 | struct ipt_counters *counters; |
| 1083 | |
| 1084 | if (copy_from_user(&tmp, user, sizeof(tmp)) != 0) |
| 1085 | return -EFAULT; |
| 1086 | |
| 1087 | /* Hack: Causes ipchains to give correct error msg --RR */ |
| 1088 | if (len != sizeof(tmp) + tmp.size) |
| 1089 | return -ENOPROTOOPT; |
| 1090 | |
| 1091 | /* Pedantry: prevent them from hitting BUG() in vmalloc.c --RR */ |
| 1092 | if ((SMP_ALIGN(tmp.size) >> PAGE_SHIFT) + 2 > num_physpages) |
| 1093 | return -ENOMEM; |
| 1094 | |
| 1095 | newinfo = vmalloc(sizeof(struct ipt_table_info) |
| 1096 | + SMP_ALIGN(tmp.size) * NR_CPUS); |
| 1097 | if (!newinfo) |
| 1098 | return -ENOMEM; |
| 1099 | |
| 1100 | if (copy_from_user(newinfo->entries, user + sizeof(tmp), |
| 1101 | tmp.size) != 0) { |
| 1102 | ret = -EFAULT; |
| 1103 | goto free_newinfo; |
| 1104 | } |
| 1105 | |
| 1106 | counters = vmalloc(tmp.num_counters * sizeof(struct ipt_counters)); |
| 1107 | if (!counters) { |
| 1108 | ret = -ENOMEM; |
| 1109 | goto free_newinfo; |
| 1110 | } |
| 1111 | memset(counters, 0, tmp.num_counters * sizeof(struct ipt_counters)); |
| 1112 | |
| 1113 | ret = translate_table(tmp.name, tmp.valid_hooks, |
| 1114 | newinfo, tmp.size, tmp.num_entries, |
| 1115 | tmp.hook_entry, tmp.underflow); |
| 1116 | if (ret != 0) |
| 1117 | goto free_newinfo_counters; |
| 1118 | |
| 1119 | duprintf("ip_tables: Translated table\n"); |
| 1120 | |
| 1121 | t = find_table_lock(tmp.name, &ret, &ipt_mutex); |
| 1122 | if (!t) |
| 1123 | goto free_newinfo_counters_untrans; |
| 1124 | |
| 1125 | /* You lied! */ |
| 1126 | if (tmp.valid_hooks != t->valid_hooks) { |
| 1127 | duprintf("Valid hook crap: %08X vs %08X\n", |
| 1128 | tmp.valid_hooks, t->valid_hooks); |
| 1129 | ret = -EINVAL; |
| 1130 | goto free_newinfo_counters_untrans_unlock; |
| 1131 | } |
| 1132 | |
| 1133 | oldinfo = replace_table(t, tmp.num_counters, newinfo, &ret); |
| 1134 | if (!oldinfo) |
| 1135 | goto free_newinfo_counters_untrans_unlock; |
| 1136 | |
| 1137 | /* Update module usage count based on number of rules */ |
| 1138 | duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n", |
| 1139 | oldinfo->number, oldinfo->initial_entries, newinfo->number); |
| 1140 | if (t->me && (oldinfo->number <= oldinfo->initial_entries) && |
| 1141 | (newinfo->number > oldinfo->initial_entries)) |
| 1142 | __MOD_INC_USE_COUNT(t->me); |
| 1143 | else if (t->me && (oldinfo->number > oldinfo->initial_entries) && |
| 1144 | (newinfo->number <= oldinfo->initial_entries)) |
| 1145 | __MOD_DEC_USE_COUNT(t->me); |
| 1146 | |
| 1147 | /* Get the old counters. */ |
| 1148 | get_counters(oldinfo, counters); |
| 1149 | /* Decrease module usage counts and free resource */ |
| 1150 | IPT_ENTRY_ITERATE(oldinfo->entries, oldinfo->size, cleanup_entry,NULL); |
| 1151 | vfree(oldinfo); |
| 1152 | /* Silent error: too late now. */ |
| 1153 | copy_to_user(tmp.counters, counters, |
| 1154 | sizeof(struct ipt_counters) * tmp.num_counters); |
| 1155 | vfree(counters); |
| 1156 | up(&ipt_mutex); |
| 1157 | return 0; |
| 1158 | |
| 1159 | free_newinfo_counters_untrans_unlock: |
| 1160 | up(&ipt_mutex); |
| 1161 | free_newinfo_counters_untrans: |
| 1162 | IPT_ENTRY_ITERATE(newinfo->entries, newinfo->size, cleanup_entry,NULL); |
| 1163 | free_newinfo_counters: |
| 1164 | vfree(counters); |
| 1165 | free_newinfo: |
| 1166 | vfree(newinfo); |
| 1167 | return ret; |
| 1168 | } |
| 1169 | |
| 1170 | /* We're lazy, and add to the first CPU; overflow works its fey magic |
| 1171 | * and everything is OK. */ |
| 1172 | static inline int |
| 1173 | add_counter_to_entry(struct ipt_entry *e, |
| 1174 | const struct ipt_counters addme[], |
| 1175 | unsigned int *i) |
| 1176 | { |
| 1177 | #if 0 |
| 1178 | duprintf("add_counter: Entry %u %lu/%lu + %lu/%lu\n", |
| 1179 | *i, |
| 1180 | (long unsigned int)e->counters.pcnt, |
| 1181 | (long unsigned int)e->counters.bcnt, |
| 1182 | (long unsigned int)addme[*i].pcnt, |
| 1183 | (long unsigned int)addme[*i].bcnt); |
| 1184 | #endif |
| 1185 | |
| 1186 | ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt); |
| 1187 | |
| 1188 | (*i)++; |
| 1189 | return 0; |
| 1190 | } |
| 1191 | |
| 1192 | static int |
| 1193 | do_add_counters(void *user, unsigned int len) |
| 1194 | { |
| 1195 | unsigned int i; |
| 1196 | struct ipt_counters_info tmp, *paddc; |
| 1197 | struct ipt_table *t; |
| 1198 | int ret; |
| 1199 | |
| 1200 | if (copy_from_user(&tmp, user, sizeof(tmp)) != 0) |
| 1201 | return -EFAULT; |
| 1202 | |
| 1203 | if (len != sizeof(tmp) + tmp.num_counters*sizeof(struct ipt_counters)) |
| 1204 | return -EINVAL; |
| 1205 | |
| 1206 | paddc = vmalloc(len); |
| 1207 | if (!paddc) |
| 1208 | return -ENOMEM; |
| 1209 | |
| 1210 | if (copy_from_user(paddc, user, len) != 0) { |
| 1211 | ret = -EFAULT; |
| 1212 | goto free; |
| 1213 | } |
| 1214 | |
| 1215 | t = find_table_lock(tmp.name, &ret, &ipt_mutex); |
| 1216 | if (!t) |
| 1217 | goto free; |
| 1218 | |
| 1219 | write_lock_bh(&t->lock); |
| 1220 | if (t->private->number != paddc->num_counters) { |
| 1221 | ret = -EINVAL; |
| 1222 | goto unlock_up_free; |
| 1223 | } |
| 1224 | |
| 1225 | i = 0; |
| 1226 | IPT_ENTRY_ITERATE(t->private->entries, |
| 1227 | t->private->size, |
| 1228 | add_counter_to_entry, |
| 1229 | paddc->counters, |
| 1230 | &i); |
| 1231 | unlock_up_free: |
| 1232 | write_unlock_bh(&t->lock); |
| 1233 | up(&ipt_mutex); |
| 1234 | free: |
| 1235 | vfree(paddc); |
| 1236 | |
| 1237 | return ret; |
| 1238 | } |
| 1239 | |
| 1240 | static int |
| 1241 | do_ipt_set_ctl(struct sock *sk, int cmd, void *user, unsigned int len) |
| 1242 | { |
| 1243 | int ret; |
| 1244 | |
| 1245 | if (!capable(CAP_NET_ADMIN)) |
| 1246 | return -EPERM; |
| 1247 | |
| 1248 | switch (cmd) { |
| 1249 | case IPT_SO_SET_REPLACE: |
| 1250 | ret = do_replace(user, len); |
| 1251 | break; |
| 1252 | |
| 1253 | case IPT_SO_SET_ADD_COUNTERS: |
| 1254 | ret = do_add_counters(user, len); |
| 1255 | break; |
| 1256 | |
| 1257 | default: |
| 1258 | duprintf("do_ipt_set_ctl: unknown request %i\n", cmd); |
| 1259 | ret = -EINVAL; |
| 1260 | } |
| 1261 | |
| 1262 | return ret; |
| 1263 | } |
| 1264 | |
| 1265 | static int |
| 1266 | do_ipt_get_ctl(struct sock *sk, int cmd, void *user, int *len) |
| 1267 | { |
| 1268 | int ret; |
| 1269 | |
| 1270 | if (!capable(CAP_NET_ADMIN)) |
| 1271 | return -EPERM; |
| 1272 | |
| 1273 | switch (cmd) { |
| 1274 | case IPT_SO_GET_INFO: { |
| 1275 | char name[IPT_TABLE_MAXNAMELEN]; |
| 1276 | struct ipt_table *t; |
| 1277 | |
| 1278 | if (*len != sizeof(struct ipt_getinfo)) { |
| 1279 | duprintf("length %u != %u\n", *len, |
| 1280 | sizeof(struct ipt_getinfo)); |
| 1281 | ret = -EINVAL; |
| 1282 | break; |
| 1283 | } |
| 1284 | |
| 1285 | if (copy_from_user(name, user, sizeof(name)) != 0) { |
| 1286 | ret = -EFAULT; |
| 1287 | break; |
| 1288 | } |
| 1289 | name[IPT_TABLE_MAXNAMELEN-1] = '\0'; |
| 1290 | t = find_table_lock(name, &ret, &ipt_mutex); |
| 1291 | if (t) { |
| 1292 | struct ipt_getinfo info; |
| 1293 | |
| 1294 | info.valid_hooks = t->valid_hooks; |
| 1295 | memcpy(info.hook_entry, t->private->hook_entry, |
| 1296 | sizeof(info.hook_entry)); |
| 1297 | memcpy(info.underflow, t->private->underflow, |
| 1298 | sizeof(info.underflow)); |
| 1299 | info.num_entries = t->private->number; |
| 1300 | info.size = t->private->size; |
| 1301 | strcpy(info.name, name); |
| 1302 | |
| 1303 | if (copy_to_user(user, &info, *len) != 0) |
| 1304 | ret = -EFAULT; |
| 1305 | else |
| 1306 | ret = 0; |
| 1307 | |
| 1308 | up(&ipt_mutex); |
| 1309 | } |
| 1310 | } |
| 1311 | break; |
| 1312 | |
| 1313 | case IPT_SO_GET_ENTRIES: { |
| 1314 | struct ipt_get_entries get; |
| 1315 | |
| 1316 | if (*len < sizeof(get)) { |
| 1317 | duprintf("get_entries: %u < %u\n", *len, sizeof(get)); |
| 1318 | ret = -EINVAL; |
| 1319 | } else if (copy_from_user(&get, user, sizeof(get)) != 0) { |
| 1320 | ret = -EFAULT; |
| 1321 | } else if (*len != sizeof(struct ipt_get_entries) + get.size) { |
| 1322 | duprintf("get_entries: %u != %u\n", *len, |
| 1323 | sizeof(struct ipt_get_entries) + get.size); |
| 1324 | ret = -EINVAL; |
| 1325 | } else |
| 1326 | ret = get_entries(&get, user); |
| 1327 | break; |
| 1328 | } |
| 1329 | |
| 1330 | default: |
| 1331 | duprintf("do_ipt_get_ctl: unknown request %i\n", cmd); |
| 1332 | ret = -EINVAL; |
| 1333 | } |
| 1334 | |
| 1335 | return ret; |
| 1336 | } |
| 1337 | |
| 1338 | /* Registration hooks for targets. */ |
| 1339 | int |
| 1340 | ipt_register_target(struct ipt_target *target) |
| 1341 | { |
| 1342 | int ret; |
| 1343 | |
| 1344 | MOD_INC_USE_COUNT; |
| 1345 | ret = down_interruptible(&ipt_mutex); |
| 1346 | if (ret != 0) { |
| 1347 | MOD_DEC_USE_COUNT; |
| 1348 | return ret; |
| 1349 | } |
| 1350 | if (!list_named_insert(&ipt_target, target)) { |
| 1351 | duprintf("ipt_register_target: `%s' already in list!\n", |
| 1352 | target->name); |
| 1353 | ret = -EINVAL; |
| 1354 | MOD_DEC_USE_COUNT; |
| 1355 | } |
| 1356 | up(&ipt_mutex); |
| 1357 | return ret; |
| 1358 | } |
| 1359 | |
| 1360 | void |
| 1361 | ipt_unregister_target(struct ipt_target *target) |
| 1362 | { |
| 1363 | down(&ipt_mutex); |
| 1364 | LIST_DELETE(&ipt_target, target); |
| 1365 | up(&ipt_mutex); |
| 1366 | MOD_DEC_USE_COUNT; |
| 1367 | } |
| 1368 | |
| 1369 | int |
| 1370 | ipt_register_match(struct ipt_match *match) |
| 1371 | { |
| 1372 | int ret; |
| 1373 | |
| 1374 | MOD_INC_USE_COUNT; |
| 1375 | ret = down_interruptible(&ipt_mutex); |
| 1376 | if (ret != 0) { |
| 1377 | MOD_DEC_USE_COUNT; |
| 1378 | return ret; |
| 1379 | } |
| 1380 | if (!list_named_insert(&ipt_match, match)) { |
| 1381 | duprintf("ipt_register_match: `%s' already in list!\n", |
| 1382 | match->name); |
| 1383 | MOD_DEC_USE_COUNT; |
| 1384 | ret = -EINVAL; |
| 1385 | } |
| 1386 | up(&ipt_mutex); |
| 1387 | |
| 1388 | return ret; |
| 1389 | } |
| 1390 | |
| 1391 | void |
| 1392 | ipt_unregister_match(struct ipt_match *match) |
| 1393 | { |
| 1394 | down(&ipt_mutex); |
| 1395 | LIST_DELETE(&ipt_match, match); |
| 1396 | up(&ipt_mutex); |
| 1397 | MOD_DEC_USE_COUNT; |
| 1398 | } |
| 1399 | |
| 1400 | int ipt_register_table(struct ipt_table *table) |
| 1401 | { |
| 1402 | int ret; |
| 1403 | struct ipt_table_info *newinfo; |
| 1404 | static struct ipt_table_info bootstrap |
| 1405 | = { 0, 0, 0, { 0 }, { 0 }, { } }; |
| 1406 | |
| 1407 | MOD_INC_USE_COUNT; |
| 1408 | newinfo = vmalloc(sizeof(struct ipt_table_info) |
| 1409 | + SMP_ALIGN(table->table->size) * NR_CPUS); |
| 1410 | if (!newinfo) { |
| 1411 | ret = -ENOMEM; |
| 1412 | MOD_DEC_USE_COUNT; |
| 1413 | return ret; |
| 1414 | } |
| 1415 | memcpy(newinfo->entries, table->table->entries, table->table->size); |
| 1416 | |
| 1417 | ret = translate_table(table->name, table->valid_hooks, |
| 1418 | newinfo, table->table->size, |
| 1419 | table->table->num_entries, |
| 1420 | table->table->hook_entry, |
| 1421 | table->table->underflow); |
| 1422 | if (ret != 0) { |
| 1423 | vfree(newinfo); |
| 1424 | MOD_DEC_USE_COUNT; |
| 1425 | return ret; |
| 1426 | } |
| 1427 | |
| 1428 | ret = down_interruptible(&ipt_mutex); |
| 1429 | if (ret != 0) { |
| 1430 | vfree(newinfo); |
| 1431 | MOD_DEC_USE_COUNT; |
| 1432 | return ret; |
| 1433 | } |
| 1434 | |
| 1435 | /* Don't autoload: we'd eat our tail... */ |
| 1436 | if (list_named_find(&ipt_tables, table->name)) { |
| 1437 | ret = -EEXIST; |
| 1438 | goto free_unlock; |
| 1439 | } |
| 1440 | |
| 1441 | /* Simplifies replace_table code. */ |
| 1442 | table->private = &bootstrap; |
| 1443 | if (!replace_table(table, 0, newinfo, &ret)) |
| 1444 | goto free_unlock; |
| 1445 | |
| 1446 | duprintf("table->private->number = %u\n", |
| 1447 | table->private->number); |
| 1448 | |
| 1449 | /* save number of initial entries */ |
| 1450 | table->private->initial_entries = table->private->number; |
| 1451 | |
| 1452 | table->lock = RW_LOCK_UNLOCKED; |
| 1453 | list_prepend(&ipt_tables, table); |
| 1454 | |
| 1455 | unlock: |
| 1456 | up(&ipt_mutex); |
| 1457 | return ret; |
| 1458 | |
| 1459 | free_unlock: |
| 1460 | vfree(newinfo); |
| 1461 | MOD_DEC_USE_COUNT; |
| 1462 | goto unlock; |
| 1463 | } |
| 1464 | |
| 1465 | void ipt_unregister_table(struct ipt_table *table) |
| 1466 | { |
| 1467 | down(&ipt_mutex); |
| 1468 | LIST_DELETE(&ipt_tables, table); |
| 1469 | up(&ipt_mutex); |
| 1470 | |
| 1471 | /* Decrease module usage counts and free resources */ |
| 1472 | IPT_ENTRY_ITERATE(table->private->entries, table->private->size, |
| 1473 | cleanup_entry, NULL); |
| 1474 | vfree(table->private); |
| 1475 | MOD_DEC_USE_COUNT; |
| 1476 | } |
| 1477 | |
| 1478 | /* Returns 1 if the port is matched by the range, 0 otherwise */ |
| 1479 | static inline int |
| 1480 | port_match(u_int16_t min, u_int16_t max, u_int16_t port, int invert) |
| 1481 | { |
| 1482 | int ret; |
| 1483 | |
| 1484 | ret = (port >= min && port <= max) ^ invert; |
| 1485 | return ret; |
| 1486 | } |
| 1487 | |
| 1488 | static int |
| 1489 | tcp_find_option(u_int8_t option, |
| 1490 | const struct tcphdr *tcp, |
| 1491 | u_int16_t datalen, |
| 1492 | int invert, |
| 1493 | int *hotdrop) |
| 1494 | { |
| 1495 | unsigned int i = sizeof(struct tcphdr); |
| 1496 | const u_int8_t *opt = (u_int8_t *)tcp; |
| 1497 | |
| 1498 | duprintf("tcp_match: finding option\n"); |
| 1499 | /* If we don't have the whole header, drop packet. */ |
| 1500 | if (tcp->doff * 4 > datalen) { |
| 1501 | *hotdrop = 1; |
| 1502 | return 0; |
| 1503 | } |
| 1504 | |
| 1505 | while (i < tcp->doff * 4) { |
| 1506 | if (opt[i] == option) return !invert; |
| 1507 | if (opt[i] < 2) i++; |
| 1508 | else i += opt[i+1]?:1; |
| 1509 | } |
| 1510 | |
| 1511 | return invert; |
| 1512 | } |
| 1513 | |
| 1514 | static int |
| 1515 | tcp_match(const struct sk_buff *skb, |
| 1516 | const struct net_device *in, |
| 1517 | const struct net_device *out, |
| 1518 | const void *matchinfo, |
| 1519 | int offset, |
| 1520 | const void *hdr, |
| 1521 | u_int16_t datalen, |
| 1522 | int *hotdrop) |
| 1523 | { |
| 1524 | const struct tcphdr *tcp = hdr; |
| 1525 | const struct ipt_tcp *tcpinfo = matchinfo; |
| 1526 | |
| 1527 | /* To quote Alan: |
| 1528 | |
| 1529 | Don't allow a fragment of TCP 8 bytes in. Nobody normal |
| 1530 | causes this. Its a cracker trying to break in by doing a |
| 1531 | flag overwrite to pass the direction checks. |
| 1532 | */ |
| 1533 | |
| 1534 | if (offset == 1) { |
| 1535 | duprintf("Dropping evil TCP offset=1 frag.\n"); |
| 1536 | *hotdrop = 1; |
| 1537 | return 0; |
| 1538 | } else if (offset == 0 && datalen < sizeof(struct tcphdr)) { |
| 1539 | /* We've been asked to examine this packet, and we |
| 1540 | can't. Hence, no choice but to drop. */ |
| 1541 | duprintf("Dropping evil TCP offset=0 tinygram.\n"); |
| 1542 | *hotdrop = 1; |
| 1543 | return 0; |
| 1544 | } |
| 1545 | |
| 1546 | /* FIXME: Try tcp doff >> packet len against various stacks --RR */ |
| 1547 | |
| 1548 | #define FWINVTCP(bool,invflg) ((bool) ^ !!(tcpinfo->invflags & invflg)) |
| 1549 | |
| 1550 | /* Must not be a fragment. */ |
| 1551 | return !offset |
| 1552 | && port_match(tcpinfo->spts[0], tcpinfo->spts[1], |
| 1553 | ntohs(tcp->source), |
| 1554 | !!(tcpinfo->invflags & IPT_TCP_INV_SRCPT)) |
| 1555 | && port_match(tcpinfo->dpts[0], tcpinfo->dpts[1], |
| 1556 | ntohs(tcp->dest), |
| 1557 | !!(tcpinfo->invflags & IPT_TCP_INV_DSTPT)) |
| 1558 | && FWINVTCP((((unsigned char *)tcp)[13] |
| 1559 | & tcpinfo->flg_mask) |
| 1560 | == tcpinfo->flg_cmp, |
| 1561 | IPT_TCP_INV_FLAGS) |
| 1562 | && (!tcpinfo->option |
| 1563 | || tcp_find_option(tcpinfo->option, tcp, datalen, |
| 1564 | tcpinfo->invflags |
| 1565 | & IPT_TCP_INV_OPTION, |
| 1566 | hotdrop)); |
| 1567 | } |
| 1568 | |
| 1569 | /* Called when user tries to insert an entry of this type. */ |
| 1570 | static int |
| 1571 | tcp_checkentry(const char *tablename, |
| 1572 | const struct ipt_ip *ip, |
| 1573 | void *matchinfo, |
| 1574 | unsigned int matchsize, |
| 1575 | unsigned int hook_mask) |
| 1576 | { |
| 1577 | const struct ipt_tcp *tcpinfo = matchinfo; |
| 1578 | |
| 1579 | /* Must specify proto == TCP, and no unknown invflags */ |
| 1580 | return ip->proto == IPPROTO_TCP |
| 1581 | && !(ip->invflags & IPT_INV_PROTO) |
| 1582 | && matchsize == IPT_ALIGN(sizeof(struct ipt_tcp)) |
| 1583 | && !(tcpinfo->invflags & ~IPT_TCP_INV_MASK); |
| 1584 | } |
| 1585 | |
| 1586 | static int |
| 1587 | udp_match(const struct sk_buff *skb, |
| 1588 | const struct net_device *in, |
| 1589 | const struct net_device *out, |
| 1590 | const void *matchinfo, |
| 1591 | int offset, |
| 1592 | const void *hdr, |
| 1593 | u_int16_t datalen, |
| 1594 | int *hotdrop) |
| 1595 | { |
| 1596 | const struct udphdr *udp = hdr; |
| 1597 | const struct ipt_udp *udpinfo = matchinfo; |
| 1598 | |
| 1599 | if (offset == 0 && datalen < sizeof(struct udphdr)) { |
| 1600 | /* We've been asked to examine this packet, and we |
| 1601 | can't. Hence, no choice but to drop. */ |
| 1602 | duprintf("Dropping evil UDP tinygram.\n"); |
| 1603 | *hotdrop = 1; |
| 1604 | return 0; |
| 1605 | } |
| 1606 | |
| 1607 | /* Must not be a fragment. */ |
| 1608 | return !offset |
| 1609 | && port_match(udpinfo->spts[0], udpinfo->spts[1], |
| 1610 | ntohs(udp->source), |
| 1611 | !!(udpinfo->invflags & IPT_UDP_INV_SRCPT)) |
| 1612 | && port_match(udpinfo->dpts[0], udpinfo->dpts[1], |
| 1613 | ntohs(udp->dest), |
| 1614 | !!(udpinfo->invflags & IPT_UDP_INV_DSTPT)); |
| 1615 | } |
| 1616 | |
| 1617 | /* Called when user tries to insert an entry of this type. */ |
| 1618 | static int |
| 1619 | udp_checkentry(const char *tablename, |
| 1620 | const struct ipt_ip *ip, |
| 1621 | void *matchinfo, |
| 1622 | unsigned int matchinfosize, |
| 1623 | unsigned int hook_mask) |
| 1624 | { |
| 1625 | const struct ipt_udp *udpinfo = matchinfo; |
| 1626 | |
| 1627 | /* Must specify proto == UDP, and no unknown invflags */ |
| 1628 | if (ip->proto != IPPROTO_UDP || (ip->invflags & IPT_INV_PROTO)) { |
| 1629 | duprintf("ipt_udp: Protocol %u != %u\n", ip->proto, |
| 1630 | IPPROTO_UDP); |
| 1631 | return 0; |
| 1632 | } |
| 1633 | if (matchinfosize != IPT_ALIGN(sizeof(struct ipt_udp))) { |
| 1634 | duprintf("ipt_udp: matchsize %u != %u\n", |
| 1635 | matchinfosize, IPT_ALIGN(sizeof(struct ipt_udp))); |
| 1636 | return 0; |
| 1637 | } |
| 1638 | if (udpinfo->invflags & ~IPT_UDP_INV_MASK) { |
| 1639 | duprintf("ipt_udp: unknown flags %X\n", |
| 1640 | udpinfo->invflags); |
| 1641 | return 0; |
| 1642 | } |
| 1643 | |
| 1644 | return 1; |
| 1645 | } |
| 1646 | |
| 1647 | /* Returns 1 if the type and code is matched by the range, 0 otherwise */ |
| 1648 | static inline int |
| 1649 | icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code, |
| 1650 | u_int8_t type, u_int8_t code, |
| 1651 | int invert) |
| 1652 | { |
| 1653 | return (type == test_type && code >= min_code && code <= max_code) |
| 1654 | ^ invert; |
| 1655 | } |
| 1656 | |
| 1657 | static int |
| 1658 | icmp_match(const struct sk_buff *skb, |
| 1659 | const struct net_device *in, |
| 1660 | const struct net_device *out, |
| 1661 | const void *matchinfo, |
| 1662 | int offset, |
| 1663 | const void *hdr, |
| 1664 | u_int16_t datalen, |
| 1665 | int *hotdrop) |
| 1666 | { |
| 1667 | const struct icmphdr *icmp = hdr; |
| 1668 | const struct ipt_icmp *icmpinfo = matchinfo; |
| 1669 | |
| 1670 | if (offset == 0 && datalen < 2) { |
| 1671 | /* We've been asked to examine this packet, and we |
| 1672 | can't. Hence, no choice but to drop. */ |
| 1673 | duprintf("Dropping evil ICMP tinygram.\n"); |
| 1674 | *hotdrop = 1; |
| 1675 | return 0; |
| 1676 | } |
| 1677 | |
| 1678 | /* Must not be a fragment. */ |
| 1679 | return !offset |
| 1680 | && icmp_type_code_match(icmpinfo->type, |
| 1681 | icmpinfo->code[0], |
| 1682 | icmpinfo->code[1], |
| 1683 | icmp->type, icmp->code, |
| 1684 | !!(icmpinfo->invflags&IPT_ICMP_INV)); |
| 1685 | } |
| 1686 | |
| 1687 | /* Called when user tries to insert an entry of this type. */ |
| 1688 | static int |
| 1689 | icmp_checkentry(const char *tablename, |
| 1690 | const struct ipt_ip *ip, |
| 1691 | void *matchinfo, |
| 1692 | unsigned int matchsize, |
| 1693 | unsigned int hook_mask) |
| 1694 | { |
| 1695 | const struct ipt_icmp *icmpinfo = matchinfo; |
| 1696 | |
| 1697 | /* Must specify proto == ICMP, and no unknown invflags */ |
| 1698 | return ip->proto == IPPROTO_ICMP |
| 1699 | && !(ip->invflags & IPT_INV_PROTO) |
| 1700 | && matchsize == IPT_ALIGN(sizeof(struct ipt_icmp)) |
| 1701 | && !(icmpinfo->invflags & ~IPT_ICMP_INV); |
| 1702 | } |
| 1703 | |
| 1704 | /* The built-in targets: standard (NULL) and error. */ |
| 1705 | static struct ipt_target ipt_standard_target |
| 1706 | = { { NULL, NULL }, IPT_STANDARD_TARGET, NULL, NULL, NULL }; |
| 1707 | static struct ipt_target ipt_error_target |
| 1708 | = { { NULL, NULL }, IPT_ERROR_TARGET, ipt_error, NULL, NULL }; |
| 1709 | |
| 1710 | static struct nf_sockopt_ops ipt_sockopts |
| 1711 | = { { NULL, NULL }, PF_INET, IPT_BASE_CTL, IPT_SO_SET_MAX+1, do_ipt_set_ctl, |
| 1712 | IPT_BASE_CTL, IPT_SO_GET_MAX+1, do_ipt_get_ctl, 0, NULL }; |
| 1713 | |
| 1714 | static struct ipt_match tcp_matchstruct |
| 1715 | = { { NULL, NULL }, "tcp", &tcp_match, &tcp_checkentry, NULL }; |
| 1716 | static struct ipt_match udp_matchstruct |
| 1717 | = { { NULL, NULL }, "udp", &udp_match, &udp_checkentry, NULL }; |
| 1718 | static struct ipt_match icmp_matchstruct |
| 1719 | = { { NULL, NULL }, "icmp", &icmp_match, &icmp_checkentry, NULL }; |
| 1720 | |
| 1721 | #ifdef CONFIG_PROC_FS |
| 1722 | static inline int print_name(const struct ipt_table *t, |
| 1723 | off_t start_offset, char *buffer, int length, |
| 1724 | off_t *pos, unsigned int *count) |
| 1725 | { |
| 1726 | if ((*count)++ >= start_offset) { |
| 1727 | unsigned int namelen; |
| 1728 | |
| 1729 | namelen = sprintf(buffer + *pos, "%s\n", t->name); |
| 1730 | if (*pos + namelen > length) { |
| 1731 | /* Stop iterating */ |
| 1732 | return 1; |
| 1733 | } |
| 1734 | *pos += namelen; |
| 1735 | } |
| 1736 | return 0; |
| 1737 | } |
| 1738 | |
| 1739 | static int ipt_get_tables(char *buffer, char **start, off_t offset, int length) |
| 1740 | { |
| 1741 | off_t pos = 0; |
| 1742 | unsigned int count = 0; |
| 1743 | |
| 1744 | if (down_interruptible(&ipt_mutex) != 0) |
| 1745 | return 0; |
| 1746 | |
| 1747 | LIST_FIND(&ipt_tables, print_name, struct ipt_table *, |
| 1748 | offset, buffer, length, &pos, &count); |
| 1749 | |
| 1750 | up(&ipt_mutex); |
| 1751 | |
| 1752 | /* `start' hack - see fs/proc/generic.c line ~105 */ |
| 1753 | *start=(char *)((unsigned long)count-offset); |
| 1754 | return pos; |
| 1755 | } |
| 1756 | #endif /*CONFIG_PROC_FS*/ |
| 1757 | |
| 1758 | static int __init init(void) |
| 1759 | { |
| 1760 | int ret; |
| 1761 | |
| 1762 | /* Noone else will be downing sem now, so we won't sleep */ |
| 1763 | down(&ipt_mutex); |
| 1764 | list_append(&ipt_target, &ipt_standard_target); |
| 1765 | list_append(&ipt_target, &ipt_error_target); |
| 1766 | list_append(&ipt_match, &tcp_matchstruct); |
| 1767 | list_append(&ipt_match, &udp_matchstruct); |
| 1768 | list_append(&ipt_match, &icmp_matchstruct); |
| 1769 | up(&ipt_mutex); |
| 1770 | |
| 1771 | /* Register setsockopt */ |
| 1772 | ret = nf_register_sockopt(&ipt_sockopts); |
| 1773 | if (ret < 0) { |
| 1774 | duprintf("Unable to register sockopts.\n"); |
| 1775 | return ret; |
| 1776 | } |
| 1777 | |
| 1778 | #ifdef CONFIG_PROC_FS |
| 1779 | { |
| 1780 | struct proc_dir_entry *proc; |
| 1781 | |
| 1782 | proc = proc_net_create("ip_tables_names", 0, ipt_get_tables); |
| 1783 | if (!proc) { |
| 1784 | nf_unregister_sockopt(&ipt_sockopts); |
| 1785 | return -ENOMEM; |
| 1786 | } |
| 1787 | proc->owner = THIS_MODULE; |
| 1788 | } |
| 1789 | #endif |
| 1790 | |
| 1791 | printk("ip_tables: (C) 2000-2002 Netfilter core team\n"); |
| 1792 | return 0; |
| 1793 | } |
| 1794 | |
| 1795 | static void __exit fini(void) |
| 1796 | { |
| 1797 | nf_unregister_sockopt(&ipt_sockopts); |
| 1798 | #ifdef CONFIG_PROC_FS |
| 1799 | proc_net_remove("ip_tables_names"); |
| 1800 | #endif |
| 1801 | } |
| 1802 | |
| 1803 | EXPORT_SYMBOL(ipt_register_table); |
| 1804 | EXPORT_SYMBOL(ipt_unregister_table); |
| 1805 | EXPORT_SYMBOL(ipt_register_match); |
| 1806 | EXPORT_SYMBOL(ipt_unregister_match); |
| 1807 | EXPORT_SYMBOL(ipt_do_table); |
| 1808 | EXPORT_SYMBOL(ipt_register_target); |
| 1809 | EXPORT_SYMBOL(ipt_unregister_target); |
| 1810 | |
| 1811 | module_init(init); |
| 1812 | module_exit(fini); |
| 1813 | MODULE_LICENSE("GPL"); |