net/ipv4/fib_rules.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              IPv4 Forwarding Information Base: policy rules.
   7  *
   8  * Version:     $Id: fib_rules.c,v 1.17 2001/10/31 21:55:54 davem Exp $
   9  *
  10  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  11  *
  12  *              This program is free software; you can redistribute it and/or
  13  *              modify it under the terms of the GNU General Public License
  14  *              as published by the Free Software Foundation; either version
  15  *              2 of the License, or (at your option) any later version.
  16  *
  17  * Fixes:
  18  *              Rani Assaf      :       local_rule cannot be deleted
  19  *              Marc Boucher    :       routing by fwmark
  20  */
  21
  22 #include <linux/config.h>
  23 #include <asm/uaccess.h>
  24 #include <asm/system.h>
  25 #include <linux/bitops.h>
  26 #include <linux/types.h>
  27 #include <linux/kernel.h>
  28 #include <linux/sched.h>
  29 #include <linux/mm.h>
  30 #include <linux/string.h>
  31 #include <linux/socket.h>
  32 #include <linux/sockios.h>
  33 #include <linux/errno.h>
  34 #include <linux/in.h>
  35 #include <linux/inet.h>
  36 #include <linux/inetdevice.h>
  37 #include <linux/netdevice.h>
  38 #include <linux/if_arp.h>
  39 #include <linux/proc_fs.h>
  40 #include <linux/skbuff.h>
  41 #include <linux/netlink.h>
  42 #include <linux/init.h>
  43 #include <linux/list.h>
  44 #include <linux/rcupdate.h>
  45
  46 #include <net/ip.h>
  47 #include <net/protocol.h>
  48 #include <net/route.h>
  49 #include <net/tcp.h>
  50 #include <net/sock.h>
  51 #include <net/ip_fib.h>
  52
  53 #define FRprintk(a...)
  54
  55 struct fib_rule
  56 {
  57         struct hlist_node hlist;
  58         atomic_t        r_clntref;
  59         u32             r_preference;
  60         unsigned char   r_table;
  61         unsigned char   r_action;
  62         unsigned char   r_dst_len;
  63         unsigned char   r_src_len;
  64         u32             r_src;
  65         u32             r_srcmask;
  66         u32             r_dst;
  67         u32             r_dstmask;
  68         u32             r_srcmap;
  69         u8              r_flags;
  70         u8              r_tos;
  71 #ifdef CONFIG_IP_ROUTE_FWMARK
  72         u32             r_fwmark;
  73 #endif
  74         int             r_ifindex;
  75 #ifdef CONFIG_NET_CLS_ROUTE
  76         __u32           r_tclassid;
  77 #endif
  78         char            r_ifname[IFNAMSIZ];
  79         int             r_dead;
  80         struct          rcu_head rcu;
  81 };
  82
  83 static struct fib_rule default_rule = {
  84         .r_clntref =    ATOMIC_INIT(2),
  85         .r_preference = 0x7FFF,
  86         .r_table =      RT_TABLE_DEFAULT,
  87         .r_action =     RTN_UNICAST,
  88 };
  89
  90 static struct fib_rule main_rule = {
  91         .r_clntref =    ATOMIC_INIT(2),
  92         .r_preference = 0x7FFE,
  93         .r_table =      RT_TABLE_MAIN,
  94         .r_action =     RTN_UNICAST,
  95 };
  96
  97 static struct fib_rule local_rule = {
  98         .r_clntref =    ATOMIC_INIT(2),
  99         .r_table =      RT_TABLE_LOCAL,
 100         .r_action =     RTN_UNICAST,
 101 };
 102
 103 static struct hlist_head fib_rules;
 104
 105 /* writer func called from netlink -- rtnl_sem hold*/
 106
 107 int inet_rtm_delrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
 108 {
 109         struct rtattr **rta = arg;
 110         struct rtmsg *rtm = NLMSG_DATA(nlh);
 111         struct fib_rule *r;
 112         struct hlist_node *node;
 113         int err = -ESRCH;
 114
 115         hlist_for_each_entry(r, node, &fib_rules, hlist) {
 116                 if ((!rta[RTA_SRC-1] || memcmp(RTA_DATA(rta[RTA_SRC-1]), &r->r_src, 4) == 0) &&
 117                     rtm->rtm_src_len == r->r_src_len &&
 118                     rtm->rtm_dst_len == r->r_dst_len &&
 119                     (!rta[RTA_DST-1] || memcmp(RTA_DATA(rta[RTA_DST-1]), &r->r_dst, 4) == 0) &&
 120                     rtm->rtm_tos == r->r_tos &&
 121 #ifdef CONFIG_IP_ROUTE_FWMARK
 122                     (!rta[RTA_PROTOINFO-1] || memcmp(RTA_DATA(rta[RTA_PROTOINFO-1]), &r->r_fwmark, 4) == 0) &&
 123 #endif
 124                     (!rtm->rtm_type || rtm->rtm_type == r->r_action) &&
 125                     (!rta[RTA_PRIORITY-1] || memcmp(RTA_DATA(rta[RTA_PRIORITY-1]), &r->r_preference, 4) == 0) &&
 126                     (!rta[RTA_IIF-1] || rtattr_strcmp(rta[RTA_IIF-1], r->r_ifname) == 0) &&
 127                     (!rtm->rtm_table || (r && rtm->rtm_table == r->r_table))) {
 128                         err = -EPERM;
 129                         if (r == &local_rule)
 130                                 break;
 131
 132                         hlist_del_rcu(&r->hlist);
 133                         r->r_dead = 1;
 134                         fib_rule_put(r);
 135                         err = 0;
 136                         break;
 137                 }
 138         }
 139         return err;
 140 }
 141
 142 /* Allocate new unique table id */
 143
 144 static struct fib_table *fib_empty_table(void)
 145 {
 146         int id;
 147
 148         for (id = 1; id <= RT_TABLE_MAX; id++)
 149                 if (fib_tables[id] == NULL)
 150                         return __fib_new_table(id);
 151         return NULL;
 152 }
 153
 154 static inline void fib_rule_put_rcu(struct rcu_head *head)
 155 {
 156         struct fib_rule *r = container_of(head, struct fib_rule, rcu);
 157         kfree(r);
 158 }
 159
 160 void fib_rule_put(struct fib_rule *r)
 161 {
 162         if (atomic_dec_and_test(&r->r_clntref)) {
 163                 if (r->r_dead)
 164                         call_rcu(&r->rcu, fib_rule_put_rcu);
 165                 else
 166                         printk("Freeing alive rule %p\n", r);
 167         }
 168 }
 169
 170 /* writer func called from netlink -- rtnl_sem hold*/
 171
 172 int inet_rtm_newrule(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
 173 {
 174         struct rtattr **rta = arg;
 175         struct rtmsg *rtm = NLMSG_DATA(nlh);
 176         struct fib_rule *r, *new_r, *last = NULL;
 177         struct hlist_node *node = NULL;
 178         unsigned char table_id;
 179
 180         if (rtm->rtm_src_len > 32 || rtm->rtm_dst_len > 32 ||
 181             (rtm->rtm_tos & ~IPTOS_TOS_MASK))
 182                 return -EINVAL;
 183
 184         if (rta[RTA_IIF-1] && RTA_PAYLOAD(rta[RTA_IIF-1]) > IFNAMSIZ)
 185                 return -EINVAL;
 186
 187         table_id = rtm->rtm_table;
 188         if (table_id == RT_TABLE_UNSPEC) {
 189                 struct fib_table *table;
 190                 if (rtm->rtm_type == RTN_UNICAST) {
 191                         if ((table = fib_empty_table()) == NULL)
 192                                 return -ENOBUFS;
 193                         table_id = table->tb_id;
 194                 }
 195         }
 196
 197         new_r = kmalloc(sizeof(*new_r), GFP_KERNEL);
 198         if (!new_r)
 199                 return -ENOMEM;
 200         memset(new_r, 0, sizeof(*new_r));
 201
 202         if (rta[RTA_SRC-1])
 203                 memcpy(&new_r->r_src, RTA_DATA(rta[RTA_SRC-1]), 4);
 204         if (rta[RTA_DST-1])
 205                 memcpy(&new_r->r_dst, RTA_DATA(rta[RTA_DST-1]), 4);
 206         if (rta[RTA_GATEWAY-1])
 207                 memcpy(&new_r->r_srcmap, RTA_DATA(rta[RTA_GATEWAY-1]), 4);
 208         new_r->r_src_len = rtm->rtm_src_len;
 209         new_r->r_dst_len = rtm->rtm_dst_len;
 210         new_r->r_srcmask = inet_make_mask(rtm->rtm_src_len);
 211         new_r->r_dstmask = inet_make_mask(rtm->rtm_dst_len);
 212         new_r->r_tos = rtm->rtm_tos;
 213 #ifdef CONFIG_IP_ROUTE_FWMARK
 214         if (rta[RTA_PROTOINFO-1])
 215                 memcpy(&new_r->r_fwmark, RTA_DATA(rta[RTA_PROTOINFO-1]), 4);
 216 #endif
 217         new_r->r_action = rtm->rtm_type;
 218         new_r->r_flags = rtm->rtm_flags;
 219         if (rta[RTA_PRIORITY-1])
 220                 memcpy(&new_r->r_preference, RTA_DATA(rta[RTA_PRIORITY-1]), 4);
 221         new_r->r_table = table_id;
 222         if (rta[RTA_IIF-1]) {
 223                 struct net_device *dev;
 224                 rtattr_strlcpy(new_r->r_ifname, rta[RTA_IIF-1], IFNAMSIZ);
 225                 new_r->r_ifindex = -1;
 226                 dev = __dev_get_by_name(new_r->r_ifname);
 227                 if (dev)
 228                         new_r->r_ifindex = dev->ifindex;
 229         }
 230 #ifdef CONFIG_NET_CLS_ROUTE
 231         if (rta[RTA_FLOW-1])
 232                 memcpy(&new_r->r_tclassid, RTA_DATA(rta[RTA_FLOW-1]), 4);
 233 #endif
 234         r = container_of(fib_rules.first, struct fib_rule, hlist);
 235
 236         if (!new_r->r_preference) {
 237                 if (r && r->hlist.next != NULL) {
 238                         r = container_of(r->hlist.next, struct fib_rule, hlist);
 239                         if (r->r_preference)
 240                                 new_r->r_preference = r->r_preference - 1;
 241                 }
 242         }
 243
 244         hlist_for_each_entry(r, node, &fib_rules, hlist) {
 245                 if (r->r_preference > new_r->r_preference)
 246                         break;
 247                 last = r;
 248         }
 249         atomic_inc(&new_r->r_clntref);
 250
 251         if (last)
 252                 hlist_add_after_rcu(&last->hlist, &new_r->hlist);
 253         else
 254                 hlist_add_before_rcu(&new_r->hlist, &r->hlist);
 255
 256         return 0;
 257 }
 258
 259 #ifdef CONFIG_NET_CLS_ROUTE
 260 u32 fib_rules_tclass(struct fib_result *res)
 261 {
 262         if (res->r)
 263                 return res->r->r_tclassid;
 264         return 0;
 265 }
 266 #endif
 267
 268 /* callers should hold rtnl semaphore */
 269
 270 static void fib_rules_detach(struct net_device *dev)
 271 {
 272         struct hlist_node *node;
 273         struct fib_rule *r;
 274
 275         hlist_for_each_entry(r, node, &fib_rules, hlist) {
 276                 if (r->r_ifindex == dev->ifindex)
 277                         r->r_ifindex = -1;
 278
 279         }
 280 }
 281
 282 /* callers should hold rtnl semaphore */
 283
 284 static void fib_rules_attach(struct net_device *dev)
 285 {
 286         struct hlist_node *node;
 287         struct fib_rule *r;
 288
 289         hlist_for_each_entry(r, node, &fib_rules, hlist) {
 290                 if (r->r_ifindex == -1 && strcmp(dev->name, r->r_ifname) == 0)
 291                         r->r_ifindex = dev->ifindex;
 292         }
 293 }
 294
 295 int fib_lookup(const struct flowi *flp, struct fib_result *res)
 296 {
 297         int err;
 298         struct fib_rule *r, *policy;
 299         struct fib_table *tb;
 300         struct hlist_node *node;
 301
 302         u32 daddr = flp->fl4_dst;
 303         u32 saddr = flp->fl4_src;
 304
 305 FRprintk("Lookup: %u.%u.%u.%u <- %u.%u.%u.%u ",
 306         NIPQUAD(flp->fl4_dst), NIPQUAD(flp->fl4_src));
 307
 308         rcu_read_lock();
 309
 310         hlist_for_each_entry_rcu(r, node, &fib_rules, hlist) {
 311                 if (((saddr^r->r_src) & r->r_srcmask) ||
 312                     ((daddr^r->r_dst) & r->r_dstmask) ||
 313                     (r->r_tos && r->r_tos != flp->fl4_tos) ||
 314 #ifdef CONFIG_IP_ROUTE_FWMARK
 315                     (r->r_fwmark && r->r_fwmark != flp->fl4_fwmark) ||
 316 #endif
 317                     (r->r_ifindex && r->r_ifindex != flp->iif))
 318                         continue;
 319
 320 FRprintk("tb %d r %d ", r->r_table, r->r_action);
 321                 switch (r->r_action) {
 322                 case RTN_UNICAST:
 323                         policy = r;
 324                         break;
 325                 case RTN_UNREACHABLE:
 326                         rcu_read_unlock();
 327                         return -ENETUNREACH;
 328                 default:
 329                 case RTN_BLACKHOLE:
 330                         rcu_read_unlock();
 331                         return -EINVAL;
 332                 case RTN_PROHIBIT:
 333                         rcu_read_unlock();
 334                         return -EACCES;
 335                 }
 336
 337                 if ((tb = fib_get_table(r->r_table)) == NULL)
 338                         continue;
 339                 err = tb->tb_lookup(tb, flp, res);
 340                 if (err == 0) {
 341                         res->r = policy;
 342                         if (policy)
 343                                 atomic_inc(&policy->r_clntref);
 344                         rcu_read_unlock();
 345                         return 0;
 346                 }
 347                 if (err < 0 && err != -EAGAIN) {
 348                         rcu_read_unlock();
 349                         return err;
 350                 }
 351         }
 352 FRprintk("FAILURE\n");
 353         rcu_read_unlock();
 354         return -ENETUNREACH;
 355 }
 356
 357 void fib_select_default(const struct flowi *flp, struct fib_result *res)
 358 {
 359         if (res->r && res->r->r_action == RTN_UNICAST &&
 360             FIB_RES_GW(*res) && FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) {
 361                 struct fib_table *tb;
 362                 if ((tb = fib_get_table(res->r->r_table)) != NULL)
 363                         tb->tb_select_default(tb, flp, res);
 364         }
 365 }
 366
 367 static int fib_rules_event(struct notifier_block *this, unsigned long event, void *ptr)
 368 {
 369         struct net_device *dev = ptr;
 370
 371         if (event == NETDEV_UNREGISTER)
 372                 fib_rules_detach(dev);
 373         else if (event == NETDEV_REGISTER)
 374                 fib_rules_attach(dev);
 375         return NOTIFY_DONE;
 376 }
 377
 378
 379 static struct notifier_block fib_rules_notifier = {
 380         .notifier_call =fib_rules_event,
 381 };
 382
 383 static __inline__ int inet_fill_rule(struct sk_buff *skb,
 384                                      struct fib_rule *r,
 385                                      struct netlink_callback *cb,
 386                                      unsigned int flags)
 387 {
 388         struct rtmsg *rtm;
 389         struct nlmsghdr  *nlh;
 390         unsigned char    *b = skb->tail;
 391
 392         nlh = NLMSG_NEW_ANSWER(skb, cb, RTM_NEWRULE, sizeof(*rtm), flags);
 393         rtm = NLMSG_DATA(nlh);
 394         rtm->rtm_family = AF_INET;
 395         rtm->rtm_dst_len = r->r_dst_len;
 396         rtm->rtm_src_len = r->r_src_len;
 397         rtm->rtm_tos = r->r_tos;
 398 #ifdef CONFIG_IP_ROUTE_FWMARK
 399         if (r->r_fwmark)
 400                 RTA_PUT(skb, RTA_PROTOINFO, 4, &r->r_fwmark);
 401 #endif
 402         rtm->rtm_table = r->r_table;
 403         rtm->rtm_protocol = 0;
 404         rtm->rtm_scope = 0;
 405         rtm->rtm_type = r->r_action;
 406         rtm->rtm_flags = r->r_flags;
 407
 408         if (r->r_dst_len)
 409                 RTA_PUT(skb, RTA_DST, 4, &r->r_dst);
 410         if (r->r_src_len)
 411                 RTA_PUT(skb, RTA_SRC, 4, &r->r_src);
 412         if (r->r_ifname[0])
 413                 RTA_PUT(skb, RTA_IIF, IFNAMSIZ, &r->r_ifname);
 414         if (r->r_preference)
 415                 RTA_PUT(skb, RTA_PRIORITY, 4, &r->r_preference);
 416         if (r->r_srcmap)
 417                 RTA_PUT(skb, RTA_GATEWAY, 4, &r->r_srcmap);
 418 #ifdef CONFIG_NET_CLS_ROUTE
 419         if (r->r_tclassid)
 420                 RTA_PUT(skb, RTA_FLOW, 4, &r->r_tclassid);
 421 #endif
 422         nlh->nlmsg_len = skb->tail - b;
 423         return skb->len;
 424
 425 nlmsg_failure:
 426 rtattr_failure:
 427         skb_trim(skb, b - skb->data);
 428         return -1;
 429 }
 430
 431 /* callers should hold rtnl semaphore */
 432
 433 int inet_dump_rules(struct sk_buff *skb, struct netlink_callback *cb)
 434 {
 435         int idx = 0;
 436         int s_idx = cb->args[0];
 437         struct fib_rule *r;
 438         struct hlist_node *node;
 439
 440         rcu_read_lock();
 441         hlist_for_each_entry(r, node, &fib_rules, hlist) {
 442
 443                 if (idx < s_idx)
 444                         continue;
 445                 if (inet_fill_rule(skb, r, cb, NLM_F_MULTI) < 0)
 446                         break;
 447                 idx++;
 448         }
 449         rcu_read_unlock();
 450         cb->args[0] = idx;
 451
 452         return skb->len;
 453 }
 454
 455 void __init fib_rules_init(void)
 456 {
 457         INIT_HLIST_HEAD(&fib_rules);
 458         hlist_add_head(&local_rule.hlist, &fib_rules);
 459         hlist_add_after(&local_rule.hlist, &main_rule.hlist);
 460         hlist_add_after(&main_rule.hlist, &default_rule.hlist);
 461         register_netdevice_notifier(&fib_rules_notifier);
 462 }