2 * net/dst.h Protocol independent destination cache definitions.
4 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
11 #include <net/dst_ops.h>
12 #include <linux/netdevice.h>
13 #include <linux/rtnetlink.h>
14 #include <linux/rcupdate.h>
15 #include <linux/bug.h>
16 #include <linux/jiffies.h>
17 #include <net/neighbour.h>
18 #include <asm/processor.h>
20 #define DST_GC_MIN (HZ/10)
21 #define DST_GC_INC (HZ/2)
22 #define DST_GC_MAX (120*HZ)
24 /* Each dst_entry has reference count and sits in some parent list(s).
25 * When it is removed from parent list, it is "freed" (dst_free).
26 * After this it enters dead state (dst->obsolete > 0) and if its refcnt
27 * is zero, it can be destroyed immediately, otherwise it is added
28 * to gc list and garbage collector periodically checks the refcnt.
34 struct rcu_head rcu_head
;
35 struct dst_entry
*child
;
36 struct net_device
*dev
;
38 unsigned long _metrics
;
39 unsigned long expires
;
40 struct dst_entry
*path
;
41 struct dst_entry
*from
;
43 struct xfrm_state
*xfrm
;
47 int (*input
)(struct sk_buff
*);
48 int (*output
)(struct sk_buff
*);
51 #define DST_HOST 0x0001
52 #define DST_NOXFRM 0x0002
53 #define DST_NOPOLICY 0x0004
54 #define DST_NOHASH 0x0008
55 #define DST_NOCACHE 0x0010
56 #define DST_NOCOUNT 0x0020
57 #define DST_NOPEER 0x0040
58 #define DST_FAKE_RTABLE 0x0080
59 #define DST_XFRM_TUNNEL 0x0100
60 #define DST_XFRM_QUEUE 0x0200
62 unsigned short pending_confirm
;
66 /* A non-zero value of dst->obsolete forces by-hand validation
67 * of the route entry. Positive values are set by the generic
68 * dst layer to indicate that the entry has been forcefully
71 * Negative values are used by the implementation layer code to
72 * force invocation of the dst_ops->check() method.
75 #define DST_OBSOLETE_NONE 0
76 #define DST_OBSOLETE_DEAD 2
77 #define DST_OBSOLETE_FORCE_CHK -1
78 #define DST_OBSOLETE_KILL -2
79 unsigned short header_len
; /* more space at head required */
80 unsigned short trailer_len
; /* space to reserve at tail */
81 #ifdef CONFIG_IP_ROUTE_CLASSID
88 * Align __refcnt to a 64 bytes alignment
89 * (L1_CACHE_SIZE would be too much)
92 long __pad_to_align_refcnt
[2];
95 * __refcnt wants to be on a different cache line from
96 * input/output/ops or performance tanks badly
98 atomic_t __refcnt
; /* client references */
100 unsigned long lastuse
;
102 struct dst_entry
*next
;
103 struct rtable __rcu
*rt_next
;
104 struct rt6_info
*rt6_next
;
105 struct dn_route __rcu
*dn_next
;
109 extern u32
*dst_cow_metrics_generic(struct dst_entry
*dst
, unsigned long old
);
110 extern const u32 dst_default_metrics
[];
112 #define DST_METRICS_READ_ONLY 0x1UL
113 #define __DST_METRICS_PTR(Y) \
114 ((u32 *)((Y) & ~DST_METRICS_READ_ONLY))
115 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
117 static inline bool dst_metrics_read_only(const struct dst_entry
*dst
)
119 return dst
->_metrics
& DST_METRICS_READ_ONLY
;
122 extern void __dst_destroy_metrics_generic(struct dst_entry
*dst
, unsigned long old
);
124 static inline void dst_destroy_metrics_generic(struct dst_entry
*dst
)
126 unsigned long val
= dst
->_metrics
;
127 if (!(val
& DST_METRICS_READ_ONLY
))
128 __dst_destroy_metrics_generic(dst
, val
);
131 static inline u32
*dst_metrics_write_ptr(struct dst_entry
*dst
)
133 unsigned long p
= dst
->_metrics
;
137 if (p
& DST_METRICS_READ_ONLY
)
138 return dst
->ops
->cow_metrics(dst
, p
);
139 return __DST_METRICS_PTR(p
);
142 /* This may only be invoked before the entry has reached global
145 static inline void dst_init_metrics(struct dst_entry
*dst
,
146 const u32
*src_metrics
,
149 dst
->_metrics
= ((unsigned long) src_metrics
) |
150 (read_only
? DST_METRICS_READ_ONLY
: 0);
153 static inline void dst_copy_metrics(struct dst_entry
*dest
, const struct dst_entry
*src
)
155 u32
*dst_metrics
= dst_metrics_write_ptr(dest
);
158 u32
*src_metrics
= DST_METRICS_PTR(src
);
160 memcpy(dst_metrics
, src_metrics
, RTAX_MAX
* sizeof(u32
));
164 static inline u32
*dst_metrics_ptr(struct dst_entry
*dst
)
166 return DST_METRICS_PTR(dst
);
170 dst_metric_raw(const struct dst_entry
*dst
, const int metric
)
172 u32
*p
= DST_METRICS_PTR(dst
);
178 dst_metric(const struct dst_entry
*dst
, const int metric
)
180 WARN_ON_ONCE(metric
== RTAX_HOPLIMIT
||
181 metric
== RTAX_ADVMSS
||
183 return dst_metric_raw(dst
, metric
);
187 dst_metric_advmss(const struct dst_entry
*dst
)
189 u32 advmss
= dst_metric_raw(dst
, RTAX_ADVMSS
);
192 advmss
= dst
->ops
->default_advmss(dst
);
197 static inline void dst_metric_set(struct dst_entry
*dst
, int metric
, u32 val
)
199 u32
*p
= dst_metrics_write_ptr(dst
);
206 dst_feature(const struct dst_entry
*dst
, u32 feature
)
208 return dst_metric(dst
, RTAX_FEATURES
) & feature
;
211 static inline u32
dst_mtu(const struct dst_entry
*dst
)
213 return dst
->ops
->mtu(dst
);
216 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
217 static inline unsigned long dst_metric_rtt(const struct dst_entry
*dst
, int metric
)
219 return msecs_to_jiffies(dst_metric(dst
, metric
));
223 dst_allfrag(const struct dst_entry
*dst
)
225 int ret
= dst_feature(dst
, RTAX_FEATURE_ALLFRAG
);
230 dst_metric_locked(const struct dst_entry
*dst
, int metric
)
232 return dst_metric(dst
, RTAX_LOCK
) & (1<<metric
);
235 static inline void dst_hold(struct dst_entry
*dst
)
238 * If your kernel compilation stops here, please check
239 * __pad_to_align_refcnt declaration in struct dst_entry
241 BUILD_BUG_ON(offsetof(struct dst_entry
, __refcnt
) & 63);
242 atomic_inc(&dst
->__refcnt
);
245 static inline void dst_use(struct dst_entry
*dst
, unsigned long time
)
252 static inline void dst_use_noref(struct dst_entry
*dst
, unsigned long time
)
258 static inline struct dst_entry
*dst_clone(struct dst_entry
*dst
)
261 atomic_inc(&dst
->__refcnt
);
265 extern void dst_release(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
, const struct sk_buff
*oskb
)
289 nskb
->_skb_refdst
= oskb
->_skb_refdst
;
290 if (!(nskb
->_skb_refdst
& SKB_DST_NOREF
))
291 dst_clone(skb_dst(nskb
));
295 * skb_dst_force - makes sure skb dst is refcounted
298 * If dst is not yet refcounted, let's do it
300 static inline void skb_dst_force(struct sk_buff
*skb
)
302 if (skb_dst_is_noref(skb
)) {
303 WARN_ON(!rcu_read_lock_held());
304 skb
->_skb_refdst
&= ~SKB_DST_NOREF
;
305 dst_clone(skb_dst(skb
));
310 * dst_hold_safe - Take a reference on a dst if possible
311 * @dst: pointer to dst entry
313 * This helper returns false if it could not safely
314 * take a reference on a dst.
316 static inline bool dst_hold_safe(struct dst_entry
*dst
)
318 if (dst
->flags
& DST_NOCACHE
)
319 return atomic_inc_not_zero(&dst
->__refcnt
);
325 * skb_dst_force_safe - makes sure skb dst is refcounted
328 * If dst is not yet refcounted and not destroyed, grab a ref on it.
330 static inline void skb_dst_force_safe(struct sk_buff
*skb
)
332 if (skb_dst_is_noref(skb
)) {
333 struct dst_entry
*dst
= skb_dst(skb
);
335 if (!dst_hold_safe(dst
))
338 skb
->_skb_refdst
= (unsigned long)dst
;
344 * __skb_tunnel_rx - prepare skb for rx reinsert
346 * @dev: tunnel device
347 * @net: netns for packet i/o
349 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
350 * so make some cleanups. (no accounting done)
352 static inline void __skb_tunnel_rx(struct sk_buff
*skb
, struct net_device
*dev
,
358 * Clear rxhash so that we can recalulate the hash for the
359 * encapsulated packet, unless we have already determine the hash
360 * over the L4 4-tuple.
364 skb_set_queue_mapping(skb
, 0);
365 skb_scrub_packet(skb
, !net_eq(net
, dev_net(dev
)));
369 * skb_tunnel_rx - prepare skb for rx reinsert
371 * @dev: tunnel device
373 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
374 * so make some cleanups, and perform accounting.
375 * Note: this accounting is not SMP safe.
377 static inline void skb_tunnel_rx(struct sk_buff
*skb
, struct net_device
*dev
,
380 /* TODO : stats should be SMP safe */
381 dev
->stats
.rx_packets
++;
382 dev
->stats
.rx_bytes
+= skb
->len
;
383 __skb_tunnel_rx(skb
, dev
, net
);
386 /* Children define the path of the packet through the
387 * Linux networking. Thus, destinations are stackable.
390 static inline struct dst_entry
*skb_dst_pop(struct sk_buff
*skb
)
392 struct dst_entry
*child
= dst_clone(skb_dst(skb
)->child
);
398 extern int dst_discard(struct sk_buff
*skb
);
399 extern void *dst_alloc(struct dst_ops
*ops
, struct net_device
*dev
,
400 int initial_ref
, int initial_obsolete
,
401 unsigned short flags
);
402 extern void __dst_free(struct dst_entry
*dst
);
403 extern struct dst_entry
*dst_destroy(struct dst_entry
*dst
);
405 static inline void dst_free(struct dst_entry
*dst
)
407 if (dst
->obsolete
> 0)
409 if (!atomic_read(&dst
->__refcnt
)) {
410 dst
= dst_destroy(dst
);
417 static inline void dst_rcu_free(struct rcu_head
*head
)
419 struct dst_entry
*dst
= container_of(head
, struct dst_entry
, rcu_head
);
423 static inline void dst_confirm(struct dst_entry
*dst
)
425 dst
->pending_confirm
= 1;
428 static inline int dst_neigh_output(struct dst_entry
*dst
, struct neighbour
*n
,
431 const struct hh_cache
*hh
;
433 if (dst
->pending_confirm
) {
434 unsigned long now
= jiffies
;
436 dst
->pending_confirm
= 0;
437 /* avoid dirtying neighbour */
438 if (n
->confirmed
!= now
)
443 if ((n
->nud_state
& NUD_CONNECTED
) && hh
->hh_len
)
444 return neigh_hh_output(hh
, skb
);
446 return n
->output(n
, skb
);
449 static inline struct neighbour
*dst_neigh_lookup(const struct dst_entry
*dst
, const void *daddr
)
451 struct neighbour
*n
= dst
->ops
->neigh_lookup(dst
, NULL
, daddr
);
452 return IS_ERR(n
) ? NULL
: n
;
455 static inline struct neighbour
*dst_neigh_lookup_skb(const struct dst_entry
*dst
,
458 struct neighbour
*n
= dst
->ops
->neigh_lookup(dst
, skb
, NULL
);
459 return IS_ERR(n
) ? NULL
: n
;
462 static inline void dst_link_failure(struct sk_buff
*skb
)
464 struct dst_entry
*dst
= skb_dst(skb
);
465 if (dst
&& dst
->ops
&& dst
->ops
->link_failure
)
466 dst
->ops
->link_failure(skb
);
469 static inline void dst_set_expires(struct dst_entry
*dst
, int timeout
)
471 unsigned long expires
= jiffies
+ timeout
;
476 if (dst
->expires
== 0 || time_before(expires
, dst
->expires
))
477 dst
->expires
= expires
;
480 /* Output packet to network from transport. */
481 static inline int dst_output(struct sk_buff
*skb
)
483 return skb_dst(skb
)->output(skb
);
486 /* Input packet from network to transport. */
487 static inline int dst_input(struct sk_buff
*skb
)
489 return skb_dst(skb
)->input(skb
);
492 static inline struct dst_entry
*dst_check(struct dst_entry
*dst
, u32 cookie
)
495 dst
= dst
->ops
->check(dst
, cookie
);
499 extern void dst_init(void);
501 /* Flags for xfrm_lookup flags argument. */
503 XFRM_LOOKUP_ICMP
= 1 << 0,
504 XFRM_LOOKUP_QUEUE
= 1 << 1,
505 XFRM_LOOKUP_KEEP_DST_REF
= 1 << 2,
510 static inline struct dst_entry
*xfrm_lookup(struct net
*net
,
511 struct dst_entry
*dst_orig
,
512 const struct flowi
*fl
, struct sock
*sk
,
518 static inline struct dst_entry
*xfrm_lookup_route(struct net
*net
,
519 struct dst_entry
*dst_orig
,
520 const struct flowi
*fl
,
527 static inline struct xfrm_state
*dst_xfrm(const struct dst_entry
*dst
)
533 extern struct dst_entry
*xfrm_lookup(struct net
*net
, struct dst_entry
*dst_orig
,
534 const struct flowi
*fl
, struct sock
*sk
,
537 struct dst_entry
*xfrm_lookup_route(struct net
*net
, struct dst_entry
*dst_orig
,
538 const struct flowi
*fl
, struct sock
*sk
,
541 /* skb attached with this dst needs transformation if dst->xfrm is valid */
542 static inline struct xfrm_state
*dst_xfrm(const struct dst_entry
*dst
)
548 #endif /* _NET_DST_H */