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 <net/neighbour.h>
20 #include <asm/processor.h>
21 #include <linux/indirect_call_wrapper.h>
26 struct net_device
*dev
;
28 unsigned long _metrics
;
29 unsigned long expires
;
31 struct xfrm_state
*xfrm
;
35 int (*input
)(struct sk_buff
*);
36 int (*output
)(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
);
39 #define DST_NOXFRM 0x0002
40 #define DST_NOPOLICY 0x0004
41 #define DST_NOCOUNT 0x0008
42 #define DST_FAKE_RTABLE 0x0010
43 #define DST_XFRM_TUNNEL 0x0020
44 #define DST_XFRM_QUEUE 0x0040
45 #define DST_METADATA 0x0080
47 /* A non-zero value of dst->obsolete forces by-hand validation
48 * of the route entry. Positive values are set by the generic
49 * dst layer to indicate that the entry has been forcefully
52 * Negative values are used by the implementation layer code to
53 * force invocation of the dst_ops->check() method.
56 #define DST_OBSOLETE_NONE 0
57 #define DST_OBSOLETE_DEAD 2
58 #define DST_OBSOLETE_FORCE_CHK -1
59 #define DST_OBSOLETE_KILL -2
60 unsigned short header_len
; /* more space at head required */
61 unsigned short trailer_len
; /* space to reserve at tail */
64 * __refcnt wants to be on a different cache line from
65 * input/output/ops or performance tanks badly
68 atomic_t __refcnt
; /* 64-bit offset 64 */
71 unsigned long lastuse
;
72 struct lwtunnel_state
*lwtstate
;
73 struct rcu_head rcu_head
;
78 atomic_t __refcnt
; /* 32-bit offset 64 */
83 u32 metrics
[RTAX_MAX
];
85 } __aligned(4); /* Low pointer bits contain DST_METRICS_FLAGS */
86 extern const struct dst_metrics dst_default_metrics
;
88 u32
*dst_cow_metrics_generic(struct dst_entry
*dst
, unsigned long old
);
90 #define DST_METRICS_READ_ONLY 0x1UL
91 #define DST_METRICS_REFCOUNTED 0x2UL
92 #define DST_METRICS_FLAGS 0x3UL
93 #define __DST_METRICS_PTR(Y) \
94 ((u32 *)((Y) & ~DST_METRICS_FLAGS))
95 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
97 static inline bool dst_metrics_read_only(const struct dst_entry
*dst
)
99 return dst
->_metrics
& DST_METRICS_READ_ONLY
;
102 void __dst_destroy_metrics_generic(struct dst_entry
*dst
, unsigned long old
);
104 static inline void dst_destroy_metrics_generic(struct dst_entry
*dst
)
106 unsigned long val
= dst
->_metrics
;
107 if (!(val
& DST_METRICS_READ_ONLY
))
108 __dst_destroy_metrics_generic(dst
, val
);
111 static inline u32
*dst_metrics_write_ptr(struct dst_entry
*dst
)
113 unsigned long p
= dst
->_metrics
;
117 if (p
& DST_METRICS_READ_ONLY
)
118 return dst
->ops
->cow_metrics(dst
, p
);
119 return __DST_METRICS_PTR(p
);
122 /* This may only be invoked before the entry has reached global
125 static inline void dst_init_metrics(struct dst_entry
*dst
,
126 const u32
*src_metrics
,
129 dst
->_metrics
= ((unsigned long) src_metrics
) |
130 (read_only
? DST_METRICS_READ_ONLY
: 0);
133 static inline void dst_copy_metrics(struct dst_entry
*dest
, const struct dst_entry
*src
)
135 u32
*dst_metrics
= dst_metrics_write_ptr(dest
);
138 u32
*src_metrics
= DST_METRICS_PTR(src
);
140 memcpy(dst_metrics
, src_metrics
, RTAX_MAX
* sizeof(u32
));
144 static inline u32
*dst_metrics_ptr(struct dst_entry
*dst
)
146 return DST_METRICS_PTR(dst
);
150 dst_metric_raw(const struct dst_entry
*dst
, const int metric
)
152 u32
*p
= DST_METRICS_PTR(dst
);
158 dst_metric(const struct dst_entry
*dst
, const int metric
)
160 WARN_ON_ONCE(metric
== RTAX_HOPLIMIT
||
161 metric
== RTAX_ADVMSS
||
163 return dst_metric_raw(dst
, metric
);
167 dst_metric_advmss(const struct dst_entry
*dst
)
169 u32 advmss
= dst_metric_raw(dst
, RTAX_ADVMSS
);
172 advmss
= dst
->ops
->default_advmss(dst
);
177 static inline void dst_metric_set(struct dst_entry
*dst
, int metric
, u32 val
)
179 u32
*p
= dst_metrics_write_ptr(dst
);
185 /* Kernel-internal feature bits that are unallocated in user space. */
186 #define DST_FEATURE_ECN_CA (1U << 31)
188 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA)
189 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
192 dst_feature(const struct dst_entry
*dst
, u32 feature
)
194 return dst_metric(dst
, RTAX_FEATURES
) & feature
;
197 INDIRECT_CALLABLE_DECLARE(unsigned int ip6_mtu(const struct dst_entry
*));
198 INDIRECT_CALLABLE_DECLARE(unsigned int ipv4_mtu(const struct dst_entry
*));
199 static inline u32
dst_mtu(const struct dst_entry
*dst
)
201 return INDIRECT_CALL_INET(dst
->ops
->mtu
, ip6_mtu
, ipv4_mtu
, dst
);
204 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
205 static inline unsigned long dst_metric_rtt(const struct dst_entry
*dst
, int metric
)
207 return msecs_to_jiffies(dst_metric(dst
, metric
));
211 dst_allfrag(const struct dst_entry
*dst
)
213 int ret
= dst_feature(dst
, RTAX_FEATURE_ALLFRAG
);
218 dst_metric_locked(const struct dst_entry
*dst
, int metric
)
220 return dst_metric(dst
, RTAX_LOCK
) & (1 << metric
);
223 static inline void dst_hold(struct dst_entry
*dst
)
226 * If your kernel compilation stops here, please check
227 * the placement of __refcnt in struct dst_entry
229 BUILD_BUG_ON(offsetof(struct dst_entry
, __refcnt
) & 63);
230 WARN_ON(atomic_inc_not_zero(&dst
->__refcnt
) == 0);
233 static inline void dst_use_noref(struct dst_entry
*dst
, unsigned long time
)
235 if (unlikely(time
!= dst
->lastuse
)) {
241 static inline void dst_hold_and_use(struct dst_entry
*dst
, unsigned long time
)
244 dst_use_noref(dst
, time
);
247 static inline struct dst_entry
*dst_clone(struct dst_entry
*dst
)
254 void dst_release(struct dst_entry
*dst
);
256 void dst_release_immediate(struct dst_entry
*dst
);
258 static inline void refdst_drop(unsigned long refdst
)
260 if (!(refdst
& SKB_DST_NOREF
))
261 dst_release((struct dst_entry
*)(refdst
& SKB_DST_PTRMASK
));
265 * skb_dst_drop - drops skb dst
268 * Drops dst reference count if a reference was taken.
270 static inline void skb_dst_drop(struct sk_buff
*skb
)
272 if (skb
->_skb_refdst
) {
273 refdst_drop(skb
->_skb_refdst
);
274 skb
->_skb_refdst
= 0UL;
278 static inline void __skb_dst_copy(struct sk_buff
*nskb
, unsigned long refdst
)
280 nskb
->slow_gro
|= !!refdst
;
281 nskb
->_skb_refdst
= refdst
;
282 if (!(nskb
->_skb_refdst
& SKB_DST_NOREF
))
283 dst_clone(skb_dst(nskb
));
286 static inline void skb_dst_copy(struct sk_buff
*nskb
, const struct sk_buff
*oskb
)
288 __skb_dst_copy(nskb
, oskb
->_skb_refdst
);
292 * dst_hold_safe - Take a reference on a dst if possible
293 * @dst: pointer to dst entry
295 * This helper returns false if it could not safely
296 * take a reference on a dst.
298 static inline bool dst_hold_safe(struct dst_entry
*dst
)
300 return atomic_inc_not_zero(&dst
->__refcnt
);
304 * skb_dst_force - makes sure skb dst is refcounted
307 * If dst is not yet refcounted and not destroyed, grab a ref on it.
308 * Returns true if dst is refcounted.
310 static inline bool skb_dst_force(struct sk_buff
*skb
)
312 if (skb_dst_is_noref(skb
)) {
313 struct dst_entry
*dst
= skb_dst(skb
);
315 WARN_ON(!rcu_read_lock_held());
316 if (!dst_hold_safe(dst
))
319 skb
->_skb_refdst
= (unsigned long)dst
;
320 skb
->slow_gro
|= !!dst
;
323 return skb
->_skb_refdst
!= 0UL;
328 * __skb_tunnel_rx - prepare skb for rx reinsert
330 * @dev: tunnel device
331 * @net: netns for packet i/o
333 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
334 * so make some cleanups. (no accounting done)
336 static inline void __skb_tunnel_rx(struct sk_buff
*skb
, struct net_device
*dev
,
342 * Clear hash so that we can recalulate the hash for the
343 * encapsulated packet, unless we have already determine the hash
344 * over the L4 4-tuple.
346 skb_clear_hash_if_not_l4(skb
);
347 skb_set_queue_mapping(skb
, 0);
348 skb_scrub_packet(skb
, !net_eq(net
, dev_net(dev
)));
352 * skb_tunnel_rx - prepare skb for rx reinsert
354 * @dev: tunnel device
355 * @net: netns for packet i/o
357 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
358 * so make some cleanups, and perform accounting.
359 * Note: this accounting is not SMP safe.
361 static inline void skb_tunnel_rx(struct sk_buff
*skb
, struct net_device
*dev
,
364 /* TODO : stats should be SMP safe */
365 dev
->stats
.rx_packets
++;
366 dev
->stats
.rx_bytes
+= skb
->len
;
367 __skb_tunnel_rx(skb
, dev
, net
);
370 static inline u32
dst_tclassid(const struct sk_buff
*skb
)
372 #ifdef CONFIG_IP_ROUTE_CLASSID
373 const struct dst_entry
*dst
;
377 return dst
->tclassid
;
382 int dst_discard_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
);
383 static inline int dst_discard(struct sk_buff
*skb
)
385 return dst_discard_out(&init_net
, skb
->sk
, skb
);
387 void *dst_alloc(struct dst_ops
*ops
, struct net_device
*dev
, int initial_ref
,
388 int initial_obsolete
, unsigned short flags
);
389 void dst_init(struct dst_entry
*dst
, struct dst_ops
*ops
,
390 struct net_device
*dev
, int initial_ref
, int initial_obsolete
,
391 unsigned short flags
);
392 struct dst_entry
*dst_destroy(struct dst_entry
*dst
);
393 void dst_dev_put(struct dst_entry
*dst
);
395 static inline void dst_confirm(struct dst_entry
*dst
)
399 static inline struct neighbour
*dst_neigh_lookup(const struct dst_entry
*dst
, const void *daddr
)
401 struct neighbour
*n
= dst
->ops
->neigh_lookup(dst
, NULL
, daddr
);
402 return IS_ERR(n
) ? NULL
: n
;
405 static inline struct neighbour
*dst_neigh_lookup_skb(const struct dst_entry
*dst
,
410 if (WARN_ON_ONCE(!dst
->ops
->neigh_lookup
))
413 n
= dst
->ops
->neigh_lookup(dst
, skb
, NULL
);
415 return IS_ERR(n
) ? NULL
: n
;
418 static inline void dst_confirm_neigh(const struct dst_entry
*dst
,
421 if (dst
->ops
->confirm_neigh
)
422 dst
->ops
->confirm_neigh(dst
, daddr
);
425 static inline void dst_link_failure(struct sk_buff
*skb
)
427 struct dst_entry
*dst
= skb_dst(skb
);
428 if (dst
&& dst
->ops
&& dst
->ops
->link_failure
)
429 dst
->ops
->link_failure(skb
);
432 static inline void dst_set_expires(struct dst_entry
*dst
, int timeout
)
434 unsigned long expires
= jiffies
+ timeout
;
439 if (dst
->expires
== 0 || time_before(expires
, dst
->expires
))
440 dst
->expires
= expires
;
443 INDIRECT_CALLABLE_DECLARE(int ip6_output(struct net
*, struct sock
*,
445 INDIRECT_CALLABLE_DECLARE(int ip_output(struct net
*, struct sock
*,
447 /* Output packet to network from transport. */
448 static inline int dst_output(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
450 return INDIRECT_CALL_INET(skb_dst(skb
)->output
,
451 ip6_output
, ip_output
,
455 INDIRECT_CALLABLE_DECLARE(int ip6_input(struct sk_buff
*));
456 INDIRECT_CALLABLE_DECLARE(int ip_local_deliver(struct sk_buff
*));
457 /* Input packet from network to transport. */
458 static inline int dst_input(struct sk_buff
*skb
)
460 return INDIRECT_CALL_INET(skb_dst(skb
)->input
,
461 ip6_input
, ip_local_deliver
, skb
);
464 INDIRECT_CALLABLE_DECLARE(struct dst_entry
*ip6_dst_check(struct dst_entry
*,
466 INDIRECT_CALLABLE_DECLARE(struct dst_entry
*ipv4_dst_check(struct dst_entry
*,
468 static inline struct dst_entry
*dst_check(struct dst_entry
*dst
, u32 cookie
)
471 dst
= INDIRECT_CALL_INET(dst
->ops
->check
, ip6_dst_check
,
472 ipv4_dst_check
, dst
, cookie
);
476 /* Flags for xfrm_lookup flags argument. */
478 XFRM_LOOKUP_ICMP
= 1 << 0,
479 XFRM_LOOKUP_QUEUE
= 1 << 1,
480 XFRM_LOOKUP_KEEP_DST_REF
= 1 << 2,
485 static inline struct dst_entry
*xfrm_lookup(struct net
*net
,
486 struct dst_entry
*dst_orig
,
487 const struct flowi
*fl
,
488 const struct sock
*sk
,
494 static inline struct dst_entry
*
495 xfrm_lookup_with_ifid(struct net
*net
, struct dst_entry
*dst_orig
,
496 const struct flowi
*fl
, const struct sock
*sk
,
497 int flags
, u32 if_id
)
502 static inline struct dst_entry
*xfrm_lookup_route(struct net
*net
,
503 struct dst_entry
*dst_orig
,
504 const struct flowi
*fl
,
505 const struct sock
*sk
,
511 static inline struct xfrm_state
*dst_xfrm(const struct dst_entry
*dst
)
517 struct dst_entry
*xfrm_lookup(struct net
*net
, struct dst_entry
*dst_orig
,
518 const struct flowi
*fl
, const struct sock
*sk
,
521 struct dst_entry
*xfrm_lookup_with_ifid(struct net
*net
,
522 struct dst_entry
*dst_orig
,
523 const struct flowi
*fl
,
524 const struct sock
*sk
, int flags
,
527 struct dst_entry
*xfrm_lookup_route(struct net
*net
, struct dst_entry
*dst_orig
,
528 const struct flowi
*fl
, const struct sock
*sk
,
531 /* skb attached with this dst needs transformation if dst->xfrm is valid */
532 static inline struct xfrm_state
*dst_xfrm(const struct dst_entry
*dst
)
538 static inline void skb_dst_update_pmtu(struct sk_buff
*skb
, u32 mtu
)
540 struct dst_entry
*dst
= skb_dst(skb
);
542 if (dst
&& dst
->ops
->update_pmtu
)
543 dst
->ops
->update_pmtu(dst
, NULL
, skb
, mtu
, true);
546 /* update dst pmtu but not do neighbor confirm */
547 static inline void skb_dst_update_pmtu_no_confirm(struct sk_buff
*skb
, u32 mtu
)
549 struct dst_entry
*dst
= skb_dst(skb
);
551 if (dst
&& dst
->ops
->update_pmtu
)
552 dst
->ops
->update_pmtu(dst
, NULL
, skb
, mtu
, false);
555 struct dst_entry
*dst_blackhole_check(struct dst_entry
*dst
, u32 cookie
);
556 void dst_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
557 struct sk_buff
*skb
, u32 mtu
, bool confirm_neigh
);
558 void dst_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
559 struct sk_buff
*skb
);
560 u32
*dst_blackhole_cow_metrics(struct dst_entry
*dst
, unsigned long old
);
561 struct neighbour
*dst_blackhole_neigh_lookup(const struct dst_entry
*dst
,
564 unsigned int dst_blackhole_mtu(const struct dst_entry
*dst
);
566 #endif /* _NET_DST_H */