projects/linux·blob master · refs

436 lines · 12.6 KiB

1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef _NET_IP6_ROUTE_H
3#define _NET_IP6_ROUTE_H
4
5#include <net/addrconf.h>
6#include <net/flow.h>
7#include <net/ip6_fib.h>
8#include <net/sock.h>
9#include <net/lwtunnel.h>
10#include <linux/ip.h>
11#include <linux/ipv6.h>
12#include <linux/route.h>
13#include <net/nexthop.h>
14
15struct route_info {
16 __u8 type;
17 __u8 length;
18 __u8 prefix_len;
19#if defined(__BIG_ENDIAN_BITFIELD)
20 __u8 reserved_h:3,
21 route_pref:2,
22 reserved_l:3;
23#elif defined(__LITTLE_ENDIAN_BITFIELD)
24 __u8 reserved_l:3,
25 route_pref:2,
26 reserved_h:3;
27#endif
28 __be32 lifetime;
29 __u8 prefix[]; /* 0,8 or 16 */
30};
31
32#define RT6_LOOKUP_F_IFACE 0x00000001
33#define RT6_LOOKUP_F_REACHABLE 0x00000002
34#define RT6_LOOKUP_F_HAS_SADDR 0x00000004
35#define RT6_LOOKUP_F_SRCPREF_TMP 0x00000008
36#define RT6_LOOKUP_F_SRCPREF_PUBLIC 0x00000010
37#define RT6_LOOKUP_F_SRCPREF_COA 0x00000020
38#define RT6_LOOKUP_F_IGNORE_LINKSTATE 0x00000040
39#define RT6_LOOKUP_F_DST_NOREF 0x00000080
40
41/* We do not (yet ?) support IPv6 jumbograms (RFC 2675)
42 * Unlike IPv4, hdr->seg_len doesn't include the IPv6 header
43 */
44#define IP6_MAX_MTU (0xFFFF + sizeof(struct ipv6hdr))
45
46/*
47 * rt6_srcprefs2flags() and rt6_flags2srcprefs() translate
48 * between IPV6_ADDR_PREFERENCES socket option values
49 * IPV6_PREFER_SRC_TMP = 0x1
50 * IPV6_PREFER_SRC_PUBLIC = 0x2
51 * IPV6_PREFER_SRC_COA = 0x4
52 * and above RT6_LOOKUP_F_SRCPREF_xxx flags.
53 */
54static inline int rt6_srcprefs2flags(unsigned int srcprefs)
55{
56 return (srcprefs & IPV6_PREFER_SRC_MASK) << 3;
57}
58
59static inline unsigned int rt6_flags2srcprefs(int flags)
60{
61 return (flags >> 3) & IPV6_PREFER_SRC_MASK;
62}
63
64static inline bool rt6_need_strict(const struct in6_addr *daddr)
65{
66 return ipv6_addr_type(daddr) &
67 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
68}
69
70/* fib entries using a nexthop object can not be coalesced into
71 * a multipath route
72 */
73static inline bool rt6_qualify_for_ecmp(const struct fib6_info *f6i)
74{
75 /* the RTF_ADDRCONF flag filters out RA's */
76 return !(f6i->fib6_flags & RTF_ADDRCONF) && !f6i->nh &&
77 f6i->fib6_nh->fib_nh_gw_family;
78}
79
80#if IS_ENABLED(CONFIG_IPV6)
81void ip6_route_input(struct sk_buff *skb);
82#else
83static inline void ip6_route_input(struct sk_buff *skb)
84{
85}
86#endif
87
88struct dst_entry *ip6_route_input_lookup(struct net *net,
89 struct net_device *dev,
90 struct flowi6 *fl6,
91 const struct sk_buff *skb, int flags);
92
93struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
94 struct flowi6 *fl6, int flags);
95
96static inline struct dst_entry *ip6_route_output(struct net *net,
97 const struct sock *sk,
98 struct flowi6 *fl6)
99{
100 return ip6_route_output_flags(net, sk, fl6, 0);
101}
102
103/* Only conditionally release dst if flags indicates
104 * !RT6_LOOKUP_F_DST_NOREF or dst is uncached.
105 */
106static inline void ip6_rt_put_flags(struct rt6_info *rt, int flags)
107{
108 if (!(flags & RT6_LOOKUP_F_DST_NOREF) ||
109 rt->dst.rt_uncached_list)
110 ip6_rt_put(rt);
111}
112
113struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
114 const struct sk_buff *skb, int flags);
115struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
116 int ifindex, struct flowi6 *fl6,
117 const struct sk_buff *skb, int flags);
118
119void ip6_route_init_special_entries(void);
120int ip6_route_init(void);
121void ip6_route_cleanup(void);
122
123int ipv6_route_ioctl(struct net *net, unsigned int cmd,
124 struct in6_rtmsg *rtmsg);
125
126int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
127 struct netlink_ext_ack *extack);
128int ip6_ins_rt(struct net *net, struct fib6_info *f6i);
129int ip6_del_rt_reason(struct net *net, struct fib6_info *f6i,
130 enum rt_del_reason del_reason);
131#if IS_ENABLED(CONFIG_IPV6)
132int ip6_del_rt(struct net *net, struct fib6_info *f6i, bool skip_notify);
133#else
134static inline int ip6_del_rt(struct net *net, struct fib6_info *f6i,
135 bool skip_notify)
136{
137 return -EAFNOSUPPORT;
138}
139#endif
140
141void rt6_flush_exceptions(struct fib6_info *f6i);
142void rt6_age_exceptions(struct fib6_info *f6i, struct fib6_gc_args *gc_args,
143 unsigned long now);
144
145static inline int ip6_route_get_saddr(struct net *net, struct fib6_info *f6i,
146 const struct in6_addr *daddr,
147 unsigned int prefs, int l3mdev_index,
148 struct in6_addr *saddr)
149{
150 struct net_device *l3mdev;
151 struct net_device *dev;
152 bool same_vrf;
153 int err = 0;
154
155 rcu_read_lock();
156
157 l3mdev = dev_get_by_index_rcu(net, l3mdev_index);
158 if (!f6i || !f6i->fib6_prefsrc.plen || l3mdev)
159 dev = f6i ? fib6_info_nh_dev(f6i) : NULL;
160 same_vrf = !l3mdev || l3mdev_master_dev_rcu(dev) == l3mdev;
161 if (f6i && f6i->fib6_prefsrc.plen && same_vrf)
162 *saddr = f6i->fib6_prefsrc.addr;
163 else
164 err = ipv6_dev_get_saddr(net, same_vrf ? dev : l3mdev, daddr, prefs, saddr);
165
166 rcu_read_unlock();
167
168 return err;
169}
170
171struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
172 const struct in6_addr *saddr, int oif,
173 const struct sk_buff *skb, int flags);
174u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
175 const struct sk_buff *skb, struct flow_keys *hkeys);
176
177struct dst_entry *icmp6_dst_alloc(struct net_device *dev, struct flowi6 *fl6);
178
179void fib6_force_start_gc(struct net *net);
180
181struct fib6_info *addrconf_f6i_alloc(struct net *net, struct inet6_dev *idev,
182 const struct in6_addr *addr, bool anycast,
183 gfp_t gfp_flags, struct netlink_ext_ack *extack);
184
185struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
186 int flags);
187
188/*
189 * support functions for ND
190 *
191 */
192struct fib6_info *rt6_get_dflt_router(struct net *net,
193 const struct in6_addr *addr,
194 struct net_device *dev);
195struct fib6_info *rt6_add_dflt_router(struct net *net,
196 const struct in6_addr *gwaddr,
197 struct net_device *dev, unsigned int pref,
198 u32 defrtr_usr_metric,
199 int lifetime);
200
201void rt6_purge_dflt_routers(struct net *net);
202
203int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
204 const struct in6_addr *gwaddr);
205
206void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu, int oif,
207 u32 mark, kuid_t uid);
208void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu);
209void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
210 kuid_t uid);
211void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif);
212void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk);
213
214struct netlink_callback;
215
216struct rt6_rtnl_dump_arg {
217 struct sk_buff *skb;
218 struct netlink_callback *cb;
219 struct net *net;
220 struct fib_dump_filter filter;
221};
222
223int rt6_dump_route(struct fib6_info *f6i, void *p_arg, unsigned int skip);
224void rt6_mtu_change(struct net_device *dev, unsigned int mtu);
225void rt6_remove_prefsrc(struct inet6_ifaddr *ifp);
226void rt6_clean_tohost(struct net *net, struct in6_addr *gateway);
227void rt6_sync_up(struct net_device *dev, unsigned char nh_flags);
228void rt6_disable_ip(struct net_device *dev, unsigned long event);
229void rt6_sync_down_dev(struct net_device *dev, unsigned long event);
230void rt6_multipath_rebalance(struct fib6_info *f6i);
231
232void rt6_uncached_list_add(struct rt6_info *rt);
233void rt6_uncached_list_del(struct rt6_info *rt);
234
235static inline const struct rt6_info *skb_rt6_info(const struct sk_buff *skb)
236{
237 const struct dst_entry *dst = skb_dst(skb);
238
239 if (dst)
240 return dst_rt6_info(dst);
241
242 return NULL;
243}
244
245/*
246 * Store a destination cache entry in a socket
247 */
248static inline void ip6_dst_store(struct sock *sk, struct dst_entry *dst,
249 bool daddr_set,
250 bool saddr_set)
251{
252 struct ipv6_pinfo *np = inet6_sk(sk);
253
254 np->dst_cookie = rt6_get_cookie(dst_rt6_info(dst));
255 sk_setup_caps(sk, dst);
256 np->daddr_cache = daddr_set;
257#ifdef CONFIG_IPV6_SUBTREES
258 np->saddr_cache = saddr_set;
259#endif
260}
261
262void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
263 const struct flowi6 *fl6);
264
265static inline bool ipv6_unicast_destination(const struct sk_buff *skb)
266{
267 const struct rt6_info *rt = dst_rt6_info(skb_dst(skb));
268
269 return rt->rt6i_flags & RTF_LOCAL;
270}
271
272static inline bool __ipv6_anycast_destination(const struct rt6key *rt6i_dst,
273 u32 rt6i_flags,
274 const struct in6_addr *daddr)
275{
276 return rt6i_flags & RTF_ANYCAST ||
277 (rt6i_dst->plen < 127 &&
278 !(rt6i_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) &&
279 ipv6_addr_equal(&rt6i_dst->addr, daddr));
280}
281
282static inline bool ipv6_anycast_destination(const struct dst_entry *dst,
283 const struct in6_addr *daddr)
284{
285 const struct rt6_info *rt = dst_rt6_info(dst);
286
287 return __ipv6_anycast_destination(&rt->rt6i_dst, rt->rt6i_flags, daddr);
288}
289
290#if IS_ENABLED(CONFIG_IPV6)
291int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
292 int (*output)(struct net *, struct sock *, struct sk_buff *));
293#else
294static inline int ip6_fragment(struct net *net, struct sock *sk,
295 struct sk_buff *skb,
296 int (*output)(struct net *, struct sock *,
297 struct sk_buff *))
298{
299 kfree_skb(skb);
300 return -EAFNOSUPPORT;
301}
302#endif
303
304/* Variant of dst_mtu() for IPv6 users */
305static inline u32 dst6_mtu(const struct dst_entry *dst)
306{
307 return INDIRECT_CALL_1(dst->ops->mtu, ip6_mtu, dst);
308}
309
310static inline unsigned int ip6_skb_dst_mtu(const struct sk_buff *skb)
311{
312 const struct ipv6_pinfo *np = skb->sk && !dev_recursion_level() ?
313 inet6_sk(skb->sk) : NULL;
314 const struct dst_entry *dst = skb_dst(skb);
315 unsigned int mtu;
316
317 if (np && READ_ONCE(np->pmtudisc) >= IPV6_PMTUDISC_PROBE) {
318 mtu = READ_ONCE(dst_dev(dst)->mtu);
319 mtu -= lwtunnel_headroom(dst->lwtstate, mtu);
320 } else {
321 mtu = dst_mtu(dst);
322 }
323 return mtu;
324}
325
326static inline bool ip6_sk_accept_pmtu(const struct sock *sk)
327{
328 u8 pmtudisc = READ_ONCE(inet6_sk(sk)->pmtudisc);
329
330 return pmtudisc != IPV6_PMTUDISC_INTERFACE &&
331 pmtudisc != IPV6_PMTUDISC_OMIT;
332}
333
334static inline bool ip6_sk_ignore_df(const struct sock *sk)
335{
336 u8 pmtudisc = READ_ONCE(inet6_sk(sk)->pmtudisc);
337
338 return pmtudisc < IPV6_PMTUDISC_DO ||
339 pmtudisc == IPV6_PMTUDISC_OMIT;
340}
341
342static inline const struct in6_addr *rt6_nexthop(const struct rt6_info *rt,
343 const struct in6_addr *daddr)
344{
345 if (rt->rt6i_flags & RTF_GATEWAY)
346 return &rt->rt6i_gateway;
347 else if (unlikely(rt->rt6i_flags & RTF_CACHE))
348 return &rt->rt6i_dst.addr;
349 else
350 return daddr;
351}
352
353static inline bool rt6_duplicate_nexthop(struct fib6_info *a, struct fib6_info *b)
354{
355 struct fib6_nh *nha, *nhb;
356
357 if (a->nh || b->nh)
358 return nexthop_cmp(a->nh, b->nh);
359
360 nha = a->fib6_nh;
361 nhb = b->fib6_nh;
362 return nha->fib_nh_dev == nhb->fib_nh_dev &&
363 ipv6_addr_equal(&nha->fib_nh_gw6, &nhb->fib_nh_gw6) &&
364 !lwtunnel_cmp_encap(nha->fib_nh_lws, nhb->fib_nh_lws);
365}
366
367static inline unsigned int ip6_dst_mtu_maybe_forward(const struct dst_entry *dst,
368 bool forwarding)
369{
370 struct inet6_dev *idev;
371 unsigned int mtu;
372
373 if (!forwarding || dst_metric_locked(dst, RTAX_MTU)) {
374 mtu = dst_metric_raw(dst, RTAX_MTU);
375 if (mtu)
376 goto out;
377 }
378
379 mtu = IPV6_MIN_MTU;
380 rcu_read_lock();
381 idev = __in6_dev_get(dst_dev_rcu(dst));
382 if (idev)
383 mtu = READ_ONCE(idev->cnf.mtu6);
384 rcu_read_unlock();
385
386out:
387 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
388
389 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
390}
391
392/* Configured/administrative MTU of a route, for advertising the TCP MSS.
393 *
394 * Unlike ip6_dst_mtu_maybe_forward(), this ignores any ICMPv6-learned path
395 * MTU (which is kept on the RTF_CACHE exception route) and returns the MTU of
396 * the underlying route (fib6_pmtu) or the egress device. The advertised MSS
397 * bounds what the peer may send to us and must reflect our receive
398 * capability, not a path MTU learned on the reverse (send) direction. See
399 * RFC 2923 section 2.3 and the comment above tcp_advertise_mss().
400 */
401static inline unsigned int ip6_dst_mtu_configured(const struct dst_entry *dst)
402{
403 const struct rt6_info *rt = dst_rt6_info(dst);
404 const struct fib6_info *from;
405 struct inet6_dev *idev;
406 unsigned int mtu = 0;
407
408 rcu_read_lock();
409 /* IPv6 keeps the learned PMTU and the configured MTU in the same
410 * RTAX_MTU slot: the learned value sits on this (possibly RTF_CACHE)
411 * dst, the configured one on the underlying route. Reach the latter
412 * via ->from (fib6_pmtu), populated by ip6_route_info_create().
413 */
414 from = rcu_dereference(rt->from);
415 if (from)
416 mtu = from->fib6_pmtu;
417 if (!mtu) {
418 mtu = IPV6_MIN_MTU;
419 idev = __in6_dev_get(dst_dev_rcu(dst));
420 if (idev)
421 mtu = max_t(unsigned int, mtu, READ_ONCE(idev->cnf.mtu6));
422 }
423 rcu_read_unlock();
424
425 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
426 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
427}
428
429u32 ip6_mtu_from_fib6(const struct fib6_result *res,
430 const struct in6_addr *daddr,
431 const struct in6_addr *saddr);
432
433struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
434 struct net_device *dev, struct sk_buff *skb,
435 const void *daddr);
436#endif