1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
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.
6 *
7 * IPv4 Forwarding Information Base: policy rules.
8 *
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * Thomas Graf <tgraf@suug.ch>
11 *
12 * Fixes:
13 * Rani Assaf : local_rule cannot be deleted
14 * Marc Boucher : routing by fwmark
15 */
16
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/flow.h>
27#include <net/inet_dscp.h>
28#include <net/ip.h>
29#include <net/route.h>
30#include <net/tcp.h>
31#include <net/ip_fib.h>
32#include <net/nexthop.h>
33#include <net/fib_rules.h>
34#include <linux/indirect_call_wrapper.h>
35
36struct fib4_rule {
37 struct fib_rule common;
38 u8 dst_len;
39 u8 src_len;
40 dscp_t dscp;
41 dscp_t dscp_mask;
42 u8 dscp_full:1; /* DSCP or TOS selector */
43 __be32 src;
44 __be32 srcmask;
45 __be32 dst;
46 __be32 dstmask;
47#ifdef CONFIG_IP_ROUTE_CLASSID
48 u32 tclassid;
49#endif
50};
51
52static bool fib4_rule_matchall(const struct fib_rule *rule)
53{
54 struct fib4_rule *r = container_of(rule, struct fib4_rule, common);
55
56 if (r->dst_len || r->src_len || r->dscp)
57 return false;
58 return fib_rule_matchall(rule);
59}
60
61bool fib4_rule_default(const struct fib_rule *rule)
62{
63 if (!fib4_rule_matchall(rule) || rule->action != FR_ACT_TO_TBL ||
64 rule->l3mdev)
65 return false;
66 if (rule->table != RT_TABLE_LOCAL && rule->table != RT_TABLE_MAIN &&
67 rule->table != RT_TABLE_DEFAULT)
68 return false;
69 return true;
70}
71EXPORT_SYMBOL_GPL(fib4_rule_default);
72
73int fib4_rules_dump(struct net *net, struct notifier_block *nb,
74 struct netlink_ext_ack *extack)
75{
76 return fib_rules_dump(net, nb, AF_INET, extack);
77}
78
79unsigned int fib4_rules_seq_read(const struct net *net)
80{
81 return fib_rules_seq_read(net, AF_INET);
82}
83
84int __fib_lookup(struct net *net, struct flowi4 *flp,
85 struct fib_result *res, unsigned int flags)
86{
87 struct fib_lookup_arg arg = {
88 .result = res,
89 .flags = flags,
90 };
91 int err;
92
93 /* update flow if oif or iif point to device enslaved to l3mdev */
94 l3mdev_update_flow(net, fl: flowi4_to_flowi(fl4: flp));
95
96 err = fib_rules_lookup(net->ipv4.rules_ops, flowi4_to_flowi(fl4: flp), flags: 0, &arg);
97#ifdef CONFIG_IP_ROUTE_CLASSID
98 if (arg.rule)
99 res->tclassid = ((struct fib4_rule *)arg.rule)->tclassid;
100 else
101 res->tclassid = 0;
102#endif
103
104 if (err == -ESRCH)
105 err = -ENETUNREACH;
106
107 return err;
108}
109EXPORT_SYMBOL_GPL(__fib_lookup);
110
111INDIRECT_CALLABLE_SCOPE int fib4_rule_action(struct fib_rule *rule,
112 struct flowi *flp, int flags,
113 struct fib_lookup_arg *arg)
114{
115 int err = -EAGAIN;
116 struct fib_table *tbl;
117 u32 tb_id;
118
119 switch (rule->action) {
120 case FR_ACT_TO_TBL:
121 break;
122
123 case FR_ACT_UNREACHABLE:
124 return -ENETUNREACH;
125
126 case FR_ACT_PROHIBIT:
127 return -EACCES;
128
129 case FR_ACT_BLACKHOLE:
130 default:
131 return -EINVAL;
132 }
133
134 rcu_read_lock();
135
136 tb_id = fib_rule_get_table(rule, arg);
137 tbl = fib_get_table(net: rule->fr_net, id: tb_id);
138 if (tbl)
139 err = fib_table_lookup(tb: tbl, flp: &flp->u.ip4,
140 res: (struct fib_result *)arg->result,
141 fib_flags: arg->flags);
142
143 rcu_read_unlock();
144 return err;
145}
146
147INDIRECT_CALLABLE_SCOPE bool fib4_rule_suppress(struct fib_rule *rule,
148 int flags,
149 struct fib_lookup_arg *arg)
150{
151 struct fib_result *result = arg->result;
152 struct net_device *dev = NULL;
153
154 if (result->fi) {
155 struct fib_nh_common *nhc = fib_info_nhc(fi: result->fi, nhsel: 0);
156
157 dev = nhc->nhc_dev;
158 }
159
160 /* do not accept result if the route does
161 * not meet the required prefix length
162 */
163 if (result->prefixlen <= rule->suppress_prefixlen)
164 goto suppress_route;
165
166 /* do not accept result if the route uses a device
167 * belonging to a forbidden interface group
168 */
169 if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup)
170 goto suppress_route;
171
172 return false;
173
174suppress_route:
175 if (!(arg->flags & FIB_LOOKUP_NOREF))
176 fib_info_put(fi: result->fi);
177 return true;
178}
179
180INDIRECT_CALLABLE_SCOPE int fib4_rule_match(struct fib_rule *rule,
181 struct flowi *fl, int flags)
182{
183 struct fib4_rule *r = (struct fib4_rule *) rule;
184 struct flowi4 *fl4 = &fl->u.ip4;
185 __be32 daddr = fl4->daddr;
186 __be32 saddr = fl4->saddr;
187
188 if (((saddr ^ r->src) & r->srcmask) ||
189 ((daddr ^ r->dst) & r->dstmask))
190 return 0;
191
192 /* When DSCP selector is used we need to match on the entire DSCP field
193 * in the flow information structure. When TOS selector is used we need
194 * to mask the upper three DSCP bits prior to matching to maintain
195 * legacy behavior.
196 */
197 if (r->dscp_full && (r->dscp ^ fl4->flowi4_dscp) & r->dscp_mask)
198 return 0;
199 else if (!r->dscp_full && r->dscp &&
200 !fib_dscp_masked_match(dscp: r->dscp, fl4))
201 return 0;
202
203 if (rule->ip_proto && (rule->ip_proto != fl4->flowi4_proto))
204 return 0;
205
206 if (!fib_rule_port_match(range: &rule->sport_range, port_mask: rule->sport_mask,
207 port: fl4->fl4_sport))
208 return 0;
209
210 if (!fib_rule_port_match(range: &rule->dport_range, port_mask: rule->dport_mask,
211 port: fl4->fl4_dport))
212 return 0;
213
214 return 1;
215}
216
217static struct fib_table *fib_empty_table(struct net *net)
218{
219 u32 id = 1;
220
221 while (1) {
222 if (!fib_get_table(net, id))
223 return fib_new_table(net, id);
224
225 if (id++ == RT_TABLE_MAX)
226 break;
227 }
228 return NULL;
229}
230
231static int fib4_nl2rule_dscp(const struct nlattr *nla, struct fib4_rule *rule4,
232 struct netlink_ext_ack *extack)
233{
234 if (rule4->dscp) {
235 NL_SET_ERR_MSG(extack, "Cannot specify both TOS and DSCP");
236 return -EINVAL;
237 }
238
239 rule4->dscp = inet_dsfield_to_dscp(dsfield: nla_get_u8(nla) << 2);
240 rule4->dscp_mask = inet_dsfield_to_dscp(INET_DSCP_MASK);
241 rule4->dscp_full = true;
242
243 return 0;
244}
245
246static int fib4_nl2rule_dscp_mask(const struct nlattr *nla,
247 struct fib4_rule *rule4,
248 struct netlink_ext_ack *extack)
249{
250 dscp_t dscp_mask;
251
252 if (!rule4->dscp_full) {
253 NL_SET_ERR_MSG_ATTR(extack, nla,
254 "Cannot specify DSCP mask without DSCP value");
255 return -EINVAL;
256 }
257
258 dscp_mask = inet_dsfield_to_dscp(dsfield: nla_get_u8(nla) << 2);
259 if (rule4->dscp & ~dscp_mask) {
260 NL_SET_ERR_MSG_ATTR(extack, nla, "Invalid DSCP mask");
261 return -EINVAL;
262 }
263
264 rule4->dscp_mask = dscp_mask;
265
266 return 0;
267}
268
269static int fib4_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
270 struct fib_rule_hdr *frh,
271 struct nlattr **tb,
272 struct netlink_ext_ack *extack)
273{
274 struct fib4_rule *rule4 = (struct fib4_rule *)rule;
275 struct net *net = rule->fr_net;
276 int err = -EINVAL;
277
278 if (tb[FRA_FLOWLABEL] || tb[FRA_FLOWLABEL_MASK]) {
279 NL_SET_ERR_MSG(extack,
280 "Flow label cannot be specified for IPv4 FIB rules");
281 goto errout;
282 }
283
284 if (!inet_validate_dscp(val: frh->tos)) {
285 NL_SET_ERR_MSG(extack,
286 "Invalid dsfield (tos): ECN bits must be 0");
287 goto errout;
288 }
289 /* IPv4 currently doesn't handle high order DSCP bits correctly */
290 if (frh->tos & ~IPTOS_TOS_MASK) {
291 NL_SET_ERR_MSG(extack, "Invalid tos");
292 goto errout;
293 }
294 rule4->dscp = inet_dsfield_to_dscp(dsfield: frh->tos);
295
296 if (tb[FRA_DSCP] &&
297 fib4_nl2rule_dscp(nla: tb[FRA_DSCP], rule4, extack) < 0)
298 goto errout;
299
300 if (tb[FRA_DSCP_MASK] &&
301 fib4_nl2rule_dscp_mask(nla: tb[FRA_DSCP_MASK], rule4, extack) < 0)
302 goto errout;
303
304 /* split local/main if they are not already split */
305 err = fib_unmerge(net);
306 if (err)
307 goto errout;
308
309 if (rule->table == RT_TABLE_UNSPEC && !rule->l3mdev) {
310 if (rule->action == FR_ACT_TO_TBL) {
311 struct fib_table *table;
312
313 table = fib_empty_table(net);
314 if (!table) {
315 err = -ENOBUFS;
316 goto errout;
317 }
318
319 rule->table = table->tb_id;
320 }
321 }
322
323 if (frh->src_len)
324 rule4->src = nla_get_in_addr(nla: tb[FRA_SRC]);
325
326 if (frh->dst_len)
327 rule4->dst = nla_get_in_addr(nla: tb[FRA_DST]);
328
329#ifdef CONFIG_IP_ROUTE_CLASSID
330 if (tb[FRA_FLOW]) {
331 rule4->tclassid = nla_get_u32(tb[FRA_FLOW]);
332 if (rule4->tclassid)
333 atomic_inc(&net->ipv4.fib_num_tclassid_users);
334 }
335#endif
336
337 if (fib_rule_requires_fldissect(rule))
338 net->ipv4.fib_rules_require_fldissect++;
339
340 rule4->src_len = frh->src_len;
341 rule4->srcmask = inet_make_mask(logmask: rule4->src_len);
342 rule4->dst_len = frh->dst_len;
343 rule4->dstmask = inet_make_mask(logmask: rule4->dst_len);
344
345 net->ipv4.fib_has_custom_rules = true;
346
347 err = 0;
348errout:
349 return err;
350}
351
352static int fib4_rule_delete(struct fib_rule *rule)
353{
354 struct net *net = rule->fr_net;
355 int err;
356
357 /* split local/main if they are not already split */
358 err = fib_unmerge(net);
359 if (err)
360 goto errout;
361
362#ifdef CONFIG_IP_ROUTE_CLASSID
363 if (((struct fib4_rule *)rule)->tclassid)
364 atomic_dec(&net->ipv4.fib_num_tclassid_users);
365#endif
366 net->ipv4.fib_has_custom_rules = true;
367
368 if (net->ipv4.fib_rules_require_fldissect &&
369 fib_rule_requires_fldissect(rule))
370 net->ipv4.fib_rules_require_fldissect--;
371errout:
372 return err;
373}
374
375static int fib4_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
376 struct nlattr **tb)
377{
378 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
379
380 if (frh->src_len && (rule4->src_len != frh->src_len))
381 return 0;
382
383 if (frh->dst_len && (rule4->dst_len != frh->dst_len))
384 return 0;
385
386 if (frh->tos &&
387 (rule4->dscp_full ||
388 inet_dscp_to_dsfield(dscp: rule4->dscp) != frh->tos))
389 return 0;
390
391 if (tb[FRA_DSCP]) {
392 dscp_t dscp;
393
394 dscp = inet_dsfield_to_dscp(dsfield: nla_get_u8(nla: tb[FRA_DSCP]) << 2);
395 if (!rule4->dscp_full || rule4->dscp != dscp)
396 return 0;
397 }
398
399 if (tb[FRA_DSCP_MASK]) {
400 dscp_t dscp_mask;
401
402 dscp_mask = inet_dsfield_to_dscp(dsfield: nla_get_u8(nla: tb[FRA_DSCP_MASK]) << 2);
403 if (!rule4->dscp_full || rule4->dscp_mask != dscp_mask)
404 return 0;
405 }
406
407#ifdef CONFIG_IP_ROUTE_CLASSID
408 if (tb[FRA_FLOW] && (rule4->tclassid != nla_get_u32(tb[FRA_FLOW])))
409 return 0;
410#endif
411
412 if (frh->src_len && (rule4->src != nla_get_in_addr(nla: tb[FRA_SRC])))
413 return 0;
414
415 if (frh->dst_len && (rule4->dst != nla_get_in_addr(nla: tb[FRA_DST])))
416 return 0;
417
418 return 1;
419}
420
421static int fib4_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
422 struct fib_rule_hdr *frh)
423{
424 struct fib4_rule *rule4 = (struct fib4_rule *) rule;
425
426 frh->dst_len = rule4->dst_len;
427 frh->src_len = rule4->src_len;
428
429 if (rule4->dscp_full) {
430 frh->tos = 0;
431 if (nla_put_u8(skb, attrtype: FRA_DSCP,
432 value: inet_dscp_to_dsfield(dscp: rule4->dscp) >> 2) ||
433 nla_put_u8(skb, attrtype: FRA_DSCP_MASK,
434 value: inet_dscp_to_dsfield(dscp: rule4->dscp_mask) >> 2))
435 goto nla_put_failure;
436 } else {
437 frh->tos = inet_dscp_to_dsfield(dscp: rule4->dscp);
438 }
439
440 if ((rule4->dst_len &&
441 nla_put_in_addr(skb, attrtype: FRA_DST, addr: rule4->dst)) ||
442 (rule4->src_len &&
443 nla_put_in_addr(skb, attrtype: FRA_SRC, addr: rule4->src)))
444 goto nla_put_failure;
445#ifdef CONFIG_IP_ROUTE_CLASSID
446 if (rule4->tclassid &&
447 nla_put_u32(skb, FRA_FLOW, rule4->tclassid))
448 goto nla_put_failure;
449#endif
450 return 0;
451
452nla_put_failure:
453 return -ENOBUFS;
454}
455
456static size_t fib4_rule_nlmsg_payload(struct fib_rule *rule)
457{
458 return nla_total_size(payload: 4) /* dst */
459 + nla_total_size(payload: 4) /* src */
460 + nla_total_size(payload: 4) /* flow */
461 + nla_total_size(payload: 1) /* dscp */
462 + nla_total_size(payload: 1); /* dscp mask */
463}
464
465static void fib4_rule_flush_cache(struct fib_rules_ops *ops)
466{
467 rt_cache_flush(net: ops->fro_net);
468}
469
470static const struct fib_rules_ops __net_initconst fib4_rules_ops_template = {
471 .family = AF_INET,
472 .rule_size = sizeof(struct fib4_rule),
473 .addr_size = sizeof(u32),
474 .action = fib4_rule_action,
475 .suppress = fib4_rule_suppress,
476 .match = fib4_rule_match,
477 .configure = fib4_rule_configure,
478 .delete = fib4_rule_delete,
479 .compare = fib4_rule_compare,
480 .fill = fib4_rule_fill,
481 .nlmsg_payload = fib4_rule_nlmsg_payload,
482 .flush_cache = fib4_rule_flush_cache,
483 .nlgroup = RTNLGRP_IPV4_RULE,
484 .owner = THIS_MODULE,
485};
486
487static int fib_default_rules_init(struct fib_rules_ops *ops)
488{
489 int err;
490
491 err = fib_default_rule_add(ops, pref: 0, table: RT_TABLE_LOCAL);
492 if (err < 0)
493 return err;
494 err = fib_default_rule_add(ops, pref: 0x7FFE, table: RT_TABLE_MAIN);
495 if (err < 0)
496 return err;
497 err = fib_default_rule_add(ops, pref: 0x7FFF, table: RT_TABLE_DEFAULT);
498 if (err < 0)
499 return err;
500 return 0;
501}
502
503int __net_init fib4_rules_init(struct net *net)
504{
505 int err;
506 struct fib_rules_ops *ops;
507
508 ops = fib_rules_register(&fib4_rules_ops_template, net);
509 if (IS_ERR(ptr: ops))
510 return PTR_ERR(ptr: ops);
511
512 err = fib_default_rules_init(ops);
513 if (err < 0)
514 goto fail;
515 net->ipv4.rules_ops = ops;
516 net->ipv4.fib_has_custom_rules = false;
517 net->ipv4.fib_rules_require_fldissect = 0;
518 return 0;
519
520fail:
521 /* also cleans all rules already added */
522 fib_rules_unregister(ops);
523 return err;
524}
525
526void __net_exit fib4_rules_exit(struct net *net)
527{
528 fib_rules_unregister(net->ipv4.rules_ops);
529}
530