1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * IPv4 Forwarding Information Base: policy rules.
13 * Rani Assaf : local_rule cannot be deleted
14 * Marc Boucher : routing by fwmark
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/netdevice.h>
20 #include <linux/netlink.h>
21 #include <linux/inetdevice.h>
22 #include <linux/init.h>
23 #include <linux/list.h>
24 #include <linux/rcupdate.h>
25 #include <linux/export.h>
26 #include <net/inet_dscp.h>
28 #include <net/route.h>
30 #include <net/ip_fib.h>
31 #include <net/nexthop.h>
32 #include <net/fib_rules.h>
33 #include <linux/indirect_call_wrapper.h>
36 struct fib_rule common;
40 u8 dscp_full:1; /* DSCP or TOS selector */
45 #ifdef CONFIG_IP_ROUTE_CLASSID
50 static bool fib4_rule_matchall(const struct fib_rule *rule)
52 struct fib4_rule *r = container_of(rule, struct fib4_rule, common);
54 if (r->dst_len || r->src_len || r->dscp)
56 return fib_rule_matchall(rule);
59 bool fib4_rule_default(const struct fib_rule *rule)
61 if (!fib4_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
64 if (rule->table != RT_TABLE_LOCAL && rule->table != RT_TABLE_MAIN &&
65 rule->table != RT_TABLE_DEFAULT)
69 EXPORT_SYMBOL_GPL(fib4_rule_default);
71 int fib4_rules_dump(struct net *net, struct notifier_block *nb,
72 struct netlink_ext_ack *extack)
74 return fib_rules_dump(net, nb, AF_INET, extack);
77 unsigned int fib4_rules_seq_read(const struct net *net)
79 return fib_rules_seq_read(net, AF_INET);
82 int __fib_lookup(struct net *net, struct flowi4 *flp,
83 struct fib_result *res, unsigned int flags)
85 struct fib_lookup_arg arg = {
91 /* update flow if oif or iif point to device enslaved to l3mdev */
92 l3mdev_update_flow(net, flowi4_to_flowi(flp));
94 err = fib_rules_lookup(net->ipv4.rules_ops, flowi4_to_flowi(flp), 0, &arg);
95 #ifdef CONFIG_IP_ROUTE_CLASSID
97 res->tclassid = ((struct fib4_rule *)arg.rule)->tclassid;
107 EXPORT_SYMBOL_GPL(__fib_lookup);
109 INDIRECT_CALLABLE_SCOPE int fib4_rule_action(struct fib_rule *rule,
110 struct flowi *flp, int flags,
111 struct fib_lookup_arg *arg)
114 struct fib_table *tbl;
117 switch (rule->action) {
121 case FR_ACT_UNREACHABLE:
124 case FR_ACT_PROHIBIT:
127 case FR_ACT_BLACKHOLE:
134 tb_id = fib_rule_get_table(rule, arg);
135 tbl = fib_get_table(rule->fr_net, tb_id);
137 err = fib_table_lookup(tbl, &flp->u.ip4,
138 (struct fib_result *)arg->result,
145 INDIRECT_CALLABLE_SCOPE bool fib4_rule_suppress(struct fib_rule *rule,
147 struct fib_lookup_arg *arg)
149 struct fib_result *result = arg->result;
150 struct net_device *dev = NULL;
153 struct fib_nh_common *nhc = fib_info_nhc(result->fi, 0);
158 /* do not accept result if the route does
159 * not meet the required prefix length
161 if (result->prefixlen <= rule->suppress_prefixlen)
164 /* do not accept result if the route uses a device
165 * belonging to a forbidden interface group
167 if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
173 if (!(arg->flags & FIB_LOOKUP_NOREF))
174 fib_info_put(result->fi);
178 INDIRECT_CALLABLE_SCOPE int fib4_rule_match(struct fib_rule *rule,
179 struct flowi *fl, int flags)
181 struct fib4_rule *r = (struct fib4_rule *) rule;
182 struct flowi4 *fl4 = &fl->u.ip4;
183 __be32 daddr = fl4->daddr;
184 __be32 saddr = fl4->saddr;
186 if (((saddr ^ r->src) & r->srcmask) ||
187 ((daddr ^ r->dst) & r->dstmask))
190 /* When DSCP selector is used we need to match on the entire DSCP field
191 * in the flow information structure. When TOS selector is used we need
192 * to mask the upper three DSCP bits prior to matching to maintain
195 if (r->dscp_full && r->dscp != inet_dsfield_to_dscp(fl4->flowi4_tos))
197 else if (!r->dscp_full && r->dscp &&
198 !fib_dscp_masked_match(r->dscp, fl4))
201 if (rule->ip_proto && (rule->ip_proto != fl4->flowi4_proto))
204 if (fib_rule_port_range_set(&rule->sport_range) &&
205 !fib_rule_port_inrange(&rule->sport_range, fl4->fl4_sport))
208 if (fib_rule_port_range_set(&rule->dport_range) &&
209 !fib_rule_port_inrange(&rule->dport_range, fl4->fl4_dport))
215 static struct fib_table *fib_empty_table(struct net *net)
220 if (!fib_get_table(net, id))
221 return fib_new_table(net, id);
223 if (id++ == RT_TABLE_MAX)
229 static int fib4_nl2rule_dscp(const struct nlattr *nla, struct fib4_rule *rule4,
230 struct netlink_ext_ack *extack)
233 NL_SET_ERR_MSG(extack, "Cannot specify both TOS and DSCP");
237 rule4->dscp = inet_dsfield_to_dscp(nla_get_u8(nla) << 2);
238 rule4->dscp_full = true;
243 static int fib4_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
244 struct fib_rule_hdr *frh,
246 struct netlink_ext_ack *extack)
248 struct net *net = sock_net(skb->sk);
250 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
252 if (!inet_validate_dscp(frh->tos)) {
253 NL_SET_ERR_MSG(extack,
254 "Invalid dsfield (tos): ECN bits must be 0");
257 /* IPv4 currently doesn't handle high order DSCP bits correctly */
258 if (frh->tos & ~IPTOS_TOS_MASK) {
259 NL_SET_ERR_MSG(extack, "Invalid tos");
262 rule4->dscp = inet_dsfield_to_dscp(frh->tos);
265 fib4_nl2rule_dscp(tb[FRA_DSCP], rule4, extack) < 0)
268 /* split local/main if they are not already split */
269 err = fib_unmerge(net);
273 if (rule->table == RT_TABLE_UNSPEC && !rule->l3mdev) {
274 if (rule->action == FR_ACT_TO_TBL) {
275 struct fib_table *table;
277 table = fib_empty_table(net);
283 rule->table = table->tb_id;
288 rule4->src = nla_get_in_addr(tb[FRA_SRC]);
291 rule4->dst = nla_get_in_addr(tb[FRA_DST]);
293 #ifdef CONFIG_IP_ROUTE_CLASSID
295 rule4->tclassid = nla_get_u32(tb[FRA_FLOW]);
297 atomic_inc(&net->ipv4.fib_num_tclassid_users);
301 if (fib_rule_requires_fldissect(rule))
302 net->ipv4.fib_rules_require_fldissect++;
304 rule4->src_len = frh->src_len;
305 rule4->srcmask = inet_make_mask(rule4->src_len);
306 rule4->dst_len = frh->dst_len;
307 rule4->dstmask = inet_make_mask(rule4->dst_len);
309 net->ipv4.fib_has_custom_rules = true;
316 static int fib4_rule_delete(struct fib_rule *rule)
318 struct net *net = rule->fr_net;
321 /* split local/main if they are not already split */
322 err = fib_unmerge(net);
326 #ifdef CONFIG_IP_ROUTE_CLASSID
327 if (((struct fib4_rule *)rule)->tclassid)
328 atomic_dec(&net->ipv4.fib_num_tclassid_users);
330 net->ipv4.fib_has_custom_rules = true;
332 if (net->ipv4.fib_rules_require_fldissect &&
333 fib_rule_requires_fldissect(rule))
334 net->ipv4.fib_rules_require_fldissect--;
339 static int fib4_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
342 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
344 if (frh->src_len && (rule4->src_len != frh->src_len))
347 if (frh->dst_len && (rule4->dst_len != frh->dst_len))
352 inet_dscp_to_dsfield(rule4->dscp) != frh->tos))
358 dscp = inet_dsfield_to_dscp(nla_get_u8(tb[FRA_DSCP]) << 2);
359 if (!rule4->dscp_full || rule4->dscp != dscp)
363 #ifdef CONFIG_IP_ROUTE_CLASSID
364 if (tb[FRA_FLOW] && (rule4->tclassid != nla_get_u32(tb[FRA_FLOW])))
368 if (frh->src_len && (rule4->src != nla_get_in_addr(tb[FRA_SRC])))
371 if (frh->dst_len && (rule4->dst != nla_get_in_addr(tb[FRA_DST])))
377 static int fib4_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
378 struct fib_rule_hdr *frh)
380 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
382 frh->dst_len = rule4->dst_len;
383 frh->src_len = rule4->src_len;
385 if (rule4->dscp_full) {
387 if (nla_put_u8(skb, FRA_DSCP,
388 inet_dscp_to_dsfield(rule4->dscp) >> 2))
389 goto nla_put_failure;
391 frh->tos = inet_dscp_to_dsfield(rule4->dscp);
394 if ((rule4->dst_len &&
395 nla_put_in_addr(skb, FRA_DST, rule4->dst)) ||
397 nla_put_in_addr(skb, FRA_SRC, rule4->src)))
398 goto nla_put_failure;
399 #ifdef CONFIG_IP_ROUTE_CLASSID
400 if (rule4->tclassid &&
401 nla_put_u32(skb, FRA_FLOW, rule4->tclassid))
402 goto nla_put_failure;
410 static size_t fib4_rule_nlmsg_payload(struct fib_rule *rule)
412 return nla_total_size(4) /* dst */
413 + nla_total_size(4) /* src */
414 + nla_total_size(4) /* flow */
415 + nla_total_size(1); /* dscp */
418 static void fib4_rule_flush_cache(struct fib_rules_ops *ops)
420 rt_cache_flush(ops->fro_net);
423 static const struct fib_rules_ops __net_initconst fib4_rules_ops_template = {
425 .rule_size = sizeof(struct fib4_rule),
426 .addr_size = sizeof(u32),
427 .action = fib4_rule_action,
428 .suppress = fib4_rule_suppress,
429 .match = fib4_rule_match,
430 .configure = fib4_rule_configure,
431 .delete = fib4_rule_delete,
432 .compare = fib4_rule_compare,
433 .fill = fib4_rule_fill,
434 .nlmsg_payload = fib4_rule_nlmsg_payload,
435 .flush_cache = fib4_rule_flush_cache,
436 .nlgroup = RTNLGRP_IPV4_RULE,
437 .owner = THIS_MODULE,
440 static int fib_default_rules_init(struct fib_rules_ops *ops)
444 err = fib_default_rule_add(ops, 0, RT_TABLE_LOCAL);
447 err = fib_default_rule_add(ops, 0x7FFE, RT_TABLE_MAIN);
450 err = fib_default_rule_add(ops, 0x7FFF, RT_TABLE_DEFAULT);
456 int __net_init fib4_rules_init(struct net *net)
459 struct fib_rules_ops *ops;
461 ops = fib_rules_register(&fib4_rules_ops_template, net);
465 err = fib_default_rules_init(ops);
468 net->ipv4.rules_ops = ops;
469 net->ipv4.fib_has_custom_rules = false;
470 net->ipv4.fib_rules_require_fldissect = 0;
474 /* also cleans all rules already added */
475 fib_rules_unregister(ops);
479 void __net_exit fib4_rules_exit(struct net *net)
481 fib_rules_unregister(net->ipv4.rules_ops);