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 net
*net
, struct sock
*sk
, struct sk_buff
*skb
);
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_FAKE_RTABLE 0x0040
58 #define DST_XFRM_TUNNEL 0x0080
59 #define DST_XFRM_QUEUE 0x0100
60 #define DST_METADATA 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
89 * Align __refcnt to a 64 bytes alignment
90 * (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
;
101 struct lwtunnel_state
*lwtstate
;
103 struct dst_entry
*next
;
104 struct rtable __rcu
*rt_next
;
105 struct rt6_info
*rt6_next
;
106 struct dn_route __rcu
*dn_next
;
110 u32
*dst_cow_metrics_generic(struct dst_entry
*dst
, unsigned long old
);
111 extern const u32 dst_default_metrics
[];
113 #define DST_METRICS_READ_ONLY 0x1UL
114 #define DST_METRICS_FLAGS 0x3UL
115 #define __DST_METRICS_PTR(Y) \
116 ((u32 *)((Y) & ~DST_METRICS_FLAGS))
117 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
119 static inline bool dst_metrics_read_only(const struct dst_entry
*dst
)
121 return dst
->_metrics
& DST_METRICS_READ_ONLY
;
124 void __dst_destroy_metrics_generic(struct dst_entry
*dst
, unsigned long old
);
126 static inline void dst_destroy_metrics_generic(struct dst_entry
*dst
)
128 unsigned long val
= dst
->_metrics
;
129 if (!(val
& DST_METRICS_READ_ONLY
))
130 __dst_destroy_metrics_generic(dst
, val
);
133 static inline u32
*dst_metrics_write_ptr(struct dst_entry
*dst
)
135 unsigned long p
= dst
->_metrics
;
139 if (p
& DST_METRICS_READ_ONLY
)
140 return dst
->ops
->cow_metrics(dst
, p
);
141 return __DST_METRICS_PTR(p
);
144 /* This may only be invoked before the entry has reached global
147 static inline void dst_init_metrics(struct dst_entry
*dst
,
148 const u32
*src_metrics
,
151 dst
->_metrics
= ((unsigned long) src_metrics
) |
152 (read_only
? DST_METRICS_READ_ONLY
: 0);
155 static inline void dst_copy_metrics(struct dst_entry
*dest
, const struct dst_entry
*src
)
157 u32
*dst_metrics
= dst_metrics_write_ptr(dest
);
160 u32
*src_metrics
= DST_METRICS_PTR(src
);
162 memcpy(dst_metrics
, src_metrics
, RTAX_MAX
* sizeof(u32
));
166 static inline u32
*dst_metrics_ptr(struct dst_entry
*dst
)
168 return DST_METRICS_PTR(dst
);
172 dst_metric_raw(const struct dst_entry
*dst
, const int metric
)
174 u32
*p
= DST_METRICS_PTR(dst
);
180 dst_metric(const struct dst_entry
*dst
, const int metric
)
182 WARN_ON_ONCE(metric
== RTAX_HOPLIMIT
||
183 metric
== RTAX_ADVMSS
||
185 return dst_metric_raw(dst
, metric
);
189 dst_metric_advmss(const struct dst_entry
*dst
)
191 u32 advmss
= dst_metric_raw(dst
, RTAX_ADVMSS
);
194 advmss
= dst
->ops
->default_advmss(dst
);
199 static inline void dst_metric_set(struct dst_entry
*dst
, int metric
, u32 val
)
201 u32
*p
= dst_metrics_write_ptr(dst
);
207 /* Kernel-internal feature bits that are unallocated in user space. */
208 #define DST_FEATURE_ECN_CA (1 << 31)
210 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA)
211 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
214 dst_feature(const struct dst_entry
*dst
, u32 feature
)
216 return dst_metric(dst
, RTAX_FEATURES
) & feature
;
219 static inline u32
dst_mtu(const struct dst_entry
*dst
)
221 return dst
->ops
->mtu(dst
);
224 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
225 static inline unsigned long dst_metric_rtt(const struct dst_entry
*dst
, int metric
)
227 return msecs_to_jiffies(dst_metric(dst
, metric
));
231 dst_allfrag(const struct dst_entry
*dst
)
233 int ret
= dst_feature(dst
, RTAX_FEATURE_ALLFRAG
);
238 dst_metric_locked(const struct dst_entry
*dst
, int metric
)
240 return dst_metric(dst
, RTAX_LOCK
) & (1<<metric
);
243 static inline void dst_hold(struct dst_entry
*dst
)
246 * If your kernel compilation stops here, please check
247 * __pad_to_align_refcnt declaration in struct dst_entry
249 BUILD_BUG_ON(offsetof(struct dst_entry
, __refcnt
) & 63);
250 atomic_inc(&dst
->__refcnt
);
253 static inline void dst_use(struct dst_entry
*dst
, unsigned long time
)
260 static inline void dst_use_noref(struct dst_entry
*dst
, unsigned long time
)
266 static inline struct dst_entry
*dst_clone(struct dst_entry
*dst
)
269 atomic_inc(&dst
->__refcnt
);
273 void dst_release(struct dst_entry
*dst
);
275 static inline void refdst_drop(unsigned long refdst
)
277 if (!(refdst
& SKB_DST_NOREF
))
278 dst_release((struct dst_entry
*)(refdst
& SKB_DST_PTRMASK
));
282 * skb_dst_drop - drops skb dst
285 * Drops dst reference count if a reference was taken.
287 static inline void skb_dst_drop(struct sk_buff
*skb
)
289 if (skb
->_skb_refdst
) {
290 refdst_drop(skb
->_skb_refdst
);
291 skb
->_skb_refdst
= 0UL;
295 static inline void __skb_dst_copy(struct sk_buff
*nskb
, unsigned long refdst
)
297 nskb
->_skb_refdst
= refdst
;
298 if (!(nskb
->_skb_refdst
& SKB_DST_NOREF
))
299 dst_clone(skb_dst(nskb
));
302 static inline void skb_dst_copy(struct sk_buff
*nskb
, const struct sk_buff
*oskb
)
304 __skb_dst_copy(nskb
, oskb
->_skb_refdst
);
308 * skb_dst_force - makes sure skb dst is refcounted
311 * If dst is not yet refcounted, let's do it
313 static inline void skb_dst_force(struct sk_buff
*skb
)
315 if (skb_dst_is_noref(skb
)) {
316 WARN_ON(!rcu_read_lock_held());
317 skb
->_skb_refdst
&= ~SKB_DST_NOREF
;
318 dst_clone(skb_dst(skb
));
323 * dst_hold_safe - Take a reference on a dst if possible
324 * @dst: pointer to dst entry
326 * This helper returns false if it could not safely
327 * take a reference on a dst.
329 static inline bool dst_hold_safe(struct dst_entry
*dst
)
331 if (dst
->flags
& DST_NOCACHE
)
332 return atomic_inc_not_zero(&dst
->__refcnt
);
338 * skb_dst_force_safe - makes sure skb dst is refcounted
341 * If dst is not yet refcounted and not destroyed, grab a ref on it.
343 static inline void skb_dst_force_safe(struct sk_buff
*skb
)
345 if (skb_dst_is_noref(skb
)) {
346 struct dst_entry
*dst
= skb_dst(skb
);
348 if (!dst_hold_safe(dst
))
351 skb
->_skb_refdst
= (unsigned long)dst
;
357 * __skb_tunnel_rx - prepare skb for rx reinsert
359 * @dev: tunnel device
360 * @net: netns for packet i/o
362 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
363 * so make some cleanups. (no accounting done)
365 static inline void __skb_tunnel_rx(struct sk_buff
*skb
, struct net_device
*dev
,
371 * Clear hash so that we can recalulate the hash for the
372 * encapsulated packet, unless we have already determine the hash
373 * over the L4 4-tuple.
375 skb_clear_hash_if_not_l4(skb
);
376 skb_set_queue_mapping(skb
, 0);
377 skb_scrub_packet(skb
, !net_eq(net
, dev_net(dev
)));
381 * skb_tunnel_rx - prepare skb for rx reinsert
383 * @dev: tunnel device
385 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
386 * so make some cleanups, and perform accounting.
387 * Note: this accounting is not SMP safe.
389 static inline void skb_tunnel_rx(struct sk_buff
*skb
, struct net_device
*dev
,
392 /* TODO : stats should be SMP safe */
393 dev
->stats
.rx_packets
++;
394 dev
->stats
.rx_bytes
+= skb
->len
;
395 __skb_tunnel_rx(skb
, dev
, net
);
398 static inline u32
dst_tclassid(const struct sk_buff
*skb
)
400 #ifdef CONFIG_IP_ROUTE_CLASSID
401 const struct dst_entry
*dst
;
405 return dst
->tclassid
;
410 int dst_discard_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
);
411 static inline int dst_discard(struct sk_buff
*skb
)
413 return dst_discard_out(&init_net
, skb
->sk
, skb
);
415 void *dst_alloc(struct dst_ops
*ops
, struct net_device
*dev
, int initial_ref
,
416 int initial_obsolete
, unsigned short flags
);
417 void dst_init(struct dst_entry
*dst
, struct dst_ops
*ops
,
418 struct net_device
*dev
, int initial_ref
, int initial_obsolete
,
419 unsigned short flags
);
420 void __dst_free(struct dst_entry
*dst
);
421 struct dst_entry
*dst_destroy(struct dst_entry
*dst
);
423 static inline void dst_free(struct dst_entry
*dst
)
425 if (dst
->obsolete
> 0)
427 if (!atomic_read(&dst
->__refcnt
)) {
428 dst
= dst_destroy(dst
);
435 static inline void dst_rcu_free(struct rcu_head
*head
)
437 struct dst_entry
*dst
= container_of(head
, struct dst_entry
, rcu_head
);
441 static inline void dst_confirm(struct dst_entry
*dst
)
443 dst
->pending_confirm
= 1;
446 static inline int dst_neigh_output(struct dst_entry
*dst
, struct neighbour
*n
,
449 const struct hh_cache
*hh
;
451 if (dst
->pending_confirm
) {
452 unsigned long now
= jiffies
;
454 dst
->pending_confirm
= 0;
455 /* avoid dirtying neighbour */
456 if (n
->confirmed
!= now
)
461 if ((n
->nud_state
& NUD_CONNECTED
) && hh
->hh_len
)
462 return neigh_hh_output(hh
, skb
);
464 return n
->output(n
, skb
);
467 static inline struct neighbour
*dst_neigh_lookup(const struct dst_entry
*dst
, const void *daddr
)
469 struct neighbour
*n
= dst
->ops
->neigh_lookup(dst
, NULL
, daddr
);
470 return IS_ERR(n
) ? NULL
: n
;
473 static inline struct neighbour
*dst_neigh_lookup_skb(const struct dst_entry
*dst
,
476 struct neighbour
*n
= dst
->ops
->neigh_lookup(dst
, skb
, NULL
);
477 return IS_ERR(n
) ? NULL
: n
;
480 static inline void dst_link_failure(struct sk_buff
*skb
)
482 struct dst_entry
*dst
= skb_dst(skb
);
483 if (dst
&& dst
->ops
&& dst
->ops
->link_failure
)
484 dst
->ops
->link_failure(skb
);
487 static inline void dst_set_expires(struct dst_entry
*dst
, int timeout
)
489 unsigned long expires
= jiffies
+ timeout
;
494 if (dst
->expires
== 0 || time_before(expires
, dst
->expires
))
495 dst
->expires
= expires
;
498 /* Output packet to network from transport. */
499 static inline int dst_output(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
501 return skb_dst(skb
)->output(net
, sk
, skb
);
504 /* Input packet from network to transport. */
505 static inline int dst_input(struct sk_buff
*skb
)
507 return skb_dst(skb
)->input(skb
);
510 static inline struct dst_entry
*dst_check(struct dst_entry
*dst
, u32 cookie
)
513 dst
= dst
->ops
->check(dst
, cookie
);
517 void dst_subsys_init(void);
519 /* Flags for xfrm_lookup flags argument. */
521 XFRM_LOOKUP_ICMP
= 1 << 0,
522 XFRM_LOOKUP_QUEUE
= 1 << 1,
523 XFRM_LOOKUP_KEEP_DST_REF
= 1 << 2,
528 static inline struct dst_entry
*xfrm_lookup(struct net
*net
,
529 struct dst_entry
*dst_orig
,
530 const struct flowi
*fl
,
531 const struct sock
*sk
,
537 static inline struct dst_entry
*xfrm_lookup_route(struct net
*net
,
538 struct dst_entry
*dst_orig
,
539 const struct flowi
*fl
,
540 const struct sock
*sk
,
546 static inline struct xfrm_state
*dst_xfrm(const struct dst_entry
*dst
)
552 struct dst_entry
*xfrm_lookup(struct net
*net
, struct dst_entry
*dst_orig
,
553 const struct flowi
*fl
, const struct sock
*sk
,
556 struct dst_entry
*xfrm_lookup_route(struct net
*net
, struct dst_entry
*dst_orig
,
557 const struct flowi
*fl
, const struct sock
*sk
,
560 /* skb attached with this dst needs transformation if dst->xfrm is valid */
561 static inline struct xfrm_state
*dst_xfrm(const struct dst_entry
*dst
)
567 #endif /* _NET_DST_H */
This page took 0.045493 seconds and 6 git commands to generate.