1 /* SPDX-License-Identifier: GPL-2.0 */
3 * net/dst.h Protocol independent destination cache definitions.
5 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
12 #include <net/dst_ops.h>
13 #include <linux/netdevice.h>
14 #include <linux/rtnetlink.h>
15 #include <linux/rcupdate.h>
16 #include <linux/bug.h>
17 #include <linux/jiffies.h>
18 #include <linux/refcount.h>
19 #include <linux/rcuref.h>
20 #include <net/neighbour.h>
21 #include <asm/processor.h>
22 #include <linux/indirect_call_wrapper.h>
27 struct net_device *dev;
29 unsigned long _metrics;
30 unsigned long expires;
32 struct xfrm_state *xfrm;
36 int (*input)(struct sk_buff *);
37 int (*output)(struct net *net, struct sock *sk, struct sk_buff *skb);
40 #define DST_NOXFRM 0x0002
41 #define DST_NOPOLICY 0x0004
42 #define DST_NOCOUNT 0x0008
43 #define DST_FAKE_RTABLE 0x0010
44 #define DST_XFRM_TUNNEL 0x0020
45 #define DST_XFRM_QUEUE 0x0040
46 #define DST_METADATA 0x0080
48 /* A non-zero value of dst->obsolete forces by-hand validation
49 * of the route entry. Positive values are set by the generic
50 * dst layer to indicate that the entry has been forcefully
53 * Negative values are used by the implementation layer code to
54 * force invocation of the dst_ops->check() method.
57 #define DST_OBSOLETE_NONE 0
58 #define DST_OBSOLETE_DEAD 2
59 #define DST_OBSOLETE_FORCE_CHK -1
60 #define DST_OBSOLETE_KILL -2
61 unsigned short header_len; /* more space at head required */
62 unsigned short trailer_len; /* space to reserve at tail */
65 * __rcuref wants to be on a different cache line from
66 * input/output/ops or performance tanks badly
69 rcuref_t __rcuref; /* 64-bit offset 64 */
72 unsigned long lastuse;
73 struct rcu_head rcu_head;
78 struct lwtunnel_state *lwtstate;
79 rcuref_t __rcuref; /* 32-bit offset 64 */
81 netdevice_tracker dev_tracker;
84 * Used by rtable and rt6_info. Moves lwtstate into the next cache
85 * line on 64bit so that lwtstate does not cause false sharing with
86 * __rcuref under contention of __rcuref. This also puts the
87 * frequently accessed members of rtable and rt6_info out of the
88 * __rcuref cache line.
90 struct list_head rt_uncached;
91 struct uncached_list *rt_uncached_list;
93 struct lwtunnel_state *lwtstate;
98 u32 metrics[RTAX_MAX];
100 } __aligned(4); /* Low pointer bits contain DST_METRICS_FLAGS */
101 extern const struct dst_metrics dst_default_metrics;
103 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
105 #define DST_METRICS_READ_ONLY 0x1UL
106 #define DST_METRICS_REFCOUNTED 0x2UL
107 #define DST_METRICS_FLAGS 0x3UL
108 #define __DST_METRICS_PTR(Y) \
109 ((u32 *)((Y) & ~DST_METRICS_FLAGS))
110 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
112 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
114 return dst->_metrics & DST_METRICS_READ_ONLY;
117 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
119 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
121 unsigned long val = dst->_metrics;
122 if (!(val & DST_METRICS_READ_ONLY))
123 __dst_destroy_metrics_generic(dst, val);
126 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
128 unsigned long p = dst->_metrics;
132 if (p & DST_METRICS_READ_ONLY)
133 return dst->ops->cow_metrics(dst, p);
134 return __DST_METRICS_PTR(p);
137 /* This may only be invoked before the entry has reached global
140 static inline void dst_init_metrics(struct dst_entry *dst,
141 const u32 *src_metrics,
144 dst->_metrics = ((unsigned long) src_metrics) |
145 (read_only ? DST_METRICS_READ_ONLY : 0);
148 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
150 u32 *dst_metrics = dst_metrics_write_ptr(dest);
153 u32 *src_metrics = DST_METRICS_PTR(src);
155 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
159 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
161 return DST_METRICS_PTR(dst);
165 dst_metric_raw(const struct dst_entry *dst, const int metric)
167 u32 *p = DST_METRICS_PTR(dst);
173 dst_metric(const struct dst_entry *dst, const int metric)
175 WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
176 metric == RTAX_ADVMSS ||
178 return dst_metric_raw(dst, metric);
182 dst_metric_advmss(const struct dst_entry *dst)
184 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
187 advmss = dst->ops->default_advmss(dst);
192 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
194 u32 *p = dst_metrics_write_ptr(dst);
200 /* Kernel-internal feature bits that are unallocated in user space. */
201 #define DST_FEATURE_ECN_CA (1U << 31)
203 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA)
204 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
207 dst_feature(const struct dst_entry *dst, u32 feature)
209 return dst_metric(dst, RTAX_FEATURES) & feature;
212 INDIRECT_CALLABLE_DECLARE(unsigned int ip6_mtu(const struct dst_entry *));
213 INDIRECT_CALLABLE_DECLARE(unsigned int ipv4_mtu(const struct dst_entry *));
214 static inline u32 dst_mtu(const struct dst_entry *dst)
216 return INDIRECT_CALL_INET(dst->ops->mtu, ip6_mtu, ipv4_mtu, dst);
219 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
220 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
222 return msecs_to_jiffies(dst_metric(dst, metric));
226 dst_allfrag(const struct dst_entry *dst)
228 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG);
233 dst_metric_locked(const struct dst_entry *dst, int metric)
235 return dst_metric(dst, RTAX_LOCK) & (1 << metric);
238 static inline void dst_hold(struct dst_entry *dst)
241 * If your kernel compilation stops here, please check
242 * the placement of __rcuref in struct dst_entry
244 BUILD_BUG_ON(offsetof(struct dst_entry, __rcuref) & 63);
245 WARN_ON(!rcuref_get(&dst->__rcuref));
248 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
250 if (unlikely(time != dst->lastuse)) {
256 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
263 void dst_release(struct dst_entry *dst);
265 void dst_release_immediate(struct dst_entry *dst);
267 static inline void refdst_drop(unsigned long refdst)
269 if (!(refdst & SKB_DST_NOREF))
270 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
274 * skb_dst_drop - drops skb dst
277 * Drops dst reference count if a reference was taken.
279 static inline void skb_dst_drop(struct sk_buff *skb)
281 if (skb->_skb_refdst) {
282 refdst_drop(skb->_skb_refdst);
283 skb->_skb_refdst = 0UL;
287 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
289 nskb->slow_gro |= !!refdst;
290 nskb->_skb_refdst = refdst;
291 if (!(nskb->_skb_refdst & SKB_DST_NOREF))
292 dst_clone(skb_dst(nskb));
295 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
297 __skb_dst_copy(nskb, oskb->_skb_refdst);
301 * dst_hold_safe - Take a reference on a dst if possible
302 * @dst: pointer to dst entry
304 * This helper returns false if it could not safely
305 * take a reference on a dst.
307 static inline bool dst_hold_safe(struct dst_entry *dst)
309 return rcuref_get(&dst->__rcuref);
313 * skb_dst_force - makes sure skb dst is refcounted
316 * If dst is not yet refcounted and not destroyed, grab a ref on it.
317 * Returns true if dst is refcounted.
319 static inline bool skb_dst_force(struct sk_buff *skb)
321 if (skb_dst_is_noref(skb)) {
322 struct dst_entry *dst = skb_dst(skb);
324 WARN_ON(!rcu_read_lock_held());
325 if (!dst_hold_safe(dst))
328 skb->_skb_refdst = (unsigned long)dst;
329 skb->slow_gro |= !!dst;
332 return skb->_skb_refdst != 0UL;
337 * __skb_tunnel_rx - prepare skb for rx reinsert
339 * @dev: tunnel device
340 * @net: netns for packet i/o
342 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
343 * so make some cleanups. (no accounting done)
345 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
351 * Clear hash so that we can recalulate the hash for the
352 * encapsulated packet, unless we have already determine the hash
353 * over the L4 4-tuple.
355 skb_clear_hash_if_not_l4(skb);
356 skb_set_queue_mapping(skb, 0);
357 skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
361 * skb_tunnel_rx - prepare skb for rx reinsert
363 * @dev: tunnel device
364 * @net: netns for packet i/o
366 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
367 * so make some cleanups, and perform accounting.
368 * Note: this accounting is not SMP safe.
370 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
373 DEV_STATS_INC(dev, rx_packets);
374 DEV_STATS_ADD(dev, rx_bytes, skb->len);
375 __skb_tunnel_rx(skb, dev, net);
378 static inline u32 dst_tclassid(const struct sk_buff *skb)
380 #ifdef CONFIG_IP_ROUTE_CLASSID
381 const struct dst_entry *dst;
385 return dst->tclassid;
390 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
391 static inline int dst_discard(struct sk_buff *skb)
393 return dst_discard_out(&init_net, skb->sk, skb);
395 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
396 int initial_obsolete, unsigned short flags);
397 void dst_init(struct dst_entry *dst, struct dst_ops *ops,
398 struct net_device *dev, int initial_ref, int initial_obsolete,
399 unsigned short flags);
400 struct dst_entry *dst_destroy(struct dst_entry *dst);
401 void dst_dev_put(struct dst_entry *dst);
403 static inline void dst_confirm(struct dst_entry *dst)
407 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
409 struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
410 return IS_ERR(n) ? NULL : n;
413 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
418 if (WARN_ON_ONCE(!dst->ops->neigh_lookup))
421 n = dst->ops->neigh_lookup(dst, skb, NULL);
423 return IS_ERR(n) ? NULL : n;
426 static inline void dst_confirm_neigh(const struct dst_entry *dst,
429 if (dst->ops->confirm_neigh)
430 dst->ops->confirm_neigh(dst, daddr);
433 static inline void dst_link_failure(struct sk_buff *skb)
435 struct dst_entry *dst = skb_dst(skb);
436 if (dst && dst->ops && dst->ops->link_failure)
437 dst->ops->link_failure(skb);
440 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
442 unsigned long expires = jiffies + timeout;
447 if (dst->expires == 0 || time_before(expires, dst->expires))
448 dst->expires = expires;
451 INDIRECT_CALLABLE_DECLARE(int ip6_output(struct net *, struct sock *,
453 INDIRECT_CALLABLE_DECLARE(int ip_output(struct net *, struct sock *,
455 /* Output packet to network from transport. */
456 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
458 return INDIRECT_CALL_INET(skb_dst(skb)->output,
459 ip6_output, ip_output,
463 INDIRECT_CALLABLE_DECLARE(int ip6_input(struct sk_buff *));
464 INDIRECT_CALLABLE_DECLARE(int ip_local_deliver(struct sk_buff *));
465 /* Input packet from network to transport. */
466 static inline int dst_input(struct sk_buff *skb)
468 return INDIRECT_CALL_INET(skb_dst(skb)->input,
469 ip6_input, ip_local_deliver, skb);
472 INDIRECT_CALLABLE_DECLARE(struct dst_entry *ip6_dst_check(struct dst_entry *,
474 INDIRECT_CALLABLE_DECLARE(struct dst_entry *ipv4_dst_check(struct dst_entry *,
476 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
479 dst = INDIRECT_CALL_INET(dst->ops->check, ip6_dst_check,
480 ipv4_dst_check, dst, cookie);
484 /* Flags for xfrm_lookup flags argument. */
486 XFRM_LOOKUP_ICMP = 1 << 0,
487 XFRM_LOOKUP_QUEUE = 1 << 1,
488 XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
493 static inline struct dst_entry *xfrm_lookup(struct net *net,
494 struct dst_entry *dst_orig,
495 const struct flowi *fl,
496 const struct sock *sk,
502 static inline struct dst_entry *
503 xfrm_lookup_with_ifid(struct net *net, struct dst_entry *dst_orig,
504 const struct flowi *fl, const struct sock *sk,
505 int flags, u32 if_id)
510 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
511 struct dst_entry *dst_orig,
512 const struct flowi *fl,
513 const struct sock *sk,
519 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
525 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
526 const struct flowi *fl, const struct sock *sk,
529 struct dst_entry *xfrm_lookup_with_ifid(struct net *net,
530 struct dst_entry *dst_orig,
531 const struct flowi *fl,
532 const struct sock *sk, int flags,
535 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
536 const struct flowi *fl, const struct sock *sk,
539 /* skb attached with this dst needs transformation if dst->xfrm is valid */
540 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
546 static inline void skb_dst_update_pmtu(struct sk_buff *skb, u32 mtu)
548 struct dst_entry *dst = skb_dst(skb);
550 if (dst && dst->ops->update_pmtu)
551 dst->ops->update_pmtu(dst, NULL, skb, mtu, true);
554 /* update dst pmtu but not do neighbor confirm */
555 static inline void skb_dst_update_pmtu_no_confirm(struct sk_buff *skb, u32 mtu)
557 struct dst_entry *dst = skb_dst(skb);
559 if (dst && dst->ops->update_pmtu)
560 dst->ops->update_pmtu(dst, NULL, skb, mtu, false);
563 struct dst_entry *dst_blackhole_check(struct dst_entry *dst, u32 cookie);
564 void dst_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
565 struct sk_buff *skb, u32 mtu, bool confirm_neigh);
566 void dst_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
567 struct sk_buff *skb);
568 u32 *dst_blackhole_cow_metrics(struct dst_entry *dst, unsigned long old);
569 struct neighbour *dst_blackhole_neigh_lookup(const struct dst_entry *dst,
572 unsigned int dst_blackhole_mtu(const struct dst_entry *dst);
574 #endif /* _NET_DST_H */