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>
22 #define DST_GC_MIN (HZ/10)
23 #define DST_GC_INC (HZ/2)
24 #define DST_GC_MAX (120*HZ)
26 /* Each dst_entry has reference count and sits in some parent list(s).
27 * When it is removed from parent list, it is "freed" (dst_free).
28 * After this it enters dead state (dst->obsolete > 0) and if its refcnt
29 * is zero, it can be destroyed immediately, otherwise it is added
30 * to gc list and garbage collector periodically checks the refcnt.
36 struct net_device
*dev
;
38 unsigned long _metrics
;
39 unsigned long expires
;
41 struct xfrm_state
*xfrm
;
45 int (*input
)(struct sk_buff
*);
46 int (*output
)(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
);
49 #define DST_HOST 0x0001
50 #define DST_NOXFRM 0x0002
51 #define DST_NOPOLICY 0x0004
52 #define DST_NOCOUNT 0x0008
53 #define DST_FAKE_RTABLE 0x0010
54 #define DST_XFRM_TUNNEL 0x0020
55 #define DST_XFRM_QUEUE 0x0040
56 #define DST_METADATA 0x0080
58 /* A non-zero value of dst->obsolete forces by-hand validation
59 * of the route entry. Positive values are set by the generic
60 * dst layer to indicate that the entry has been forcefully
63 * Negative values are used by the implementation layer code to
64 * force invocation of the dst_ops->check() method.
67 #define DST_OBSOLETE_NONE 0
68 #define DST_OBSOLETE_DEAD 2
69 #define DST_OBSOLETE_FORCE_CHK -1
70 #define DST_OBSOLETE_KILL -2
71 unsigned short header_len
; /* more space at head required */
72 unsigned short trailer_len
; /* space to reserve at tail */
75 * __refcnt wants to be on a different cache line from
76 * input/output/ops or performance tanks badly
79 atomic_t __refcnt
; /* 64-bit offset 64 */
82 unsigned long lastuse
;
83 struct lwtunnel_state
*lwtstate
;
84 struct rcu_head rcu_head
;
89 atomic_t __refcnt
; /* 32-bit offset 64 */
94 u32 metrics
[RTAX_MAX
];
97 extern const struct dst_metrics dst_default_metrics
;
99 u32
*dst_cow_metrics_generic(struct dst_entry
*dst
, unsigned long old
);
101 #define DST_METRICS_READ_ONLY 0x1UL
102 #define DST_METRICS_REFCOUNTED 0x2UL
103 #define DST_METRICS_FLAGS 0x3UL
104 #define __DST_METRICS_PTR(Y) \
105 ((u32 *)((Y) & ~DST_METRICS_FLAGS))
106 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
108 static inline bool dst_metrics_read_only(const struct dst_entry
*dst
)
110 return dst
->_metrics
& DST_METRICS_READ_ONLY
;
113 void __dst_destroy_metrics_generic(struct dst_entry
*dst
, unsigned long old
);
115 static inline void dst_destroy_metrics_generic(struct dst_entry
*dst
)
117 unsigned long val
= dst
->_metrics
;
118 if (!(val
& DST_METRICS_READ_ONLY
))
119 __dst_destroy_metrics_generic(dst
, val
);
122 static inline u32
*dst_metrics_write_ptr(struct dst_entry
*dst
)
124 unsigned long p
= dst
->_metrics
;
128 if (p
& DST_METRICS_READ_ONLY
)
129 return dst
->ops
->cow_metrics(dst
, p
);
130 return __DST_METRICS_PTR(p
);
133 /* This may only be invoked before the entry has reached global
136 static inline void dst_init_metrics(struct dst_entry
*dst
,
137 const u32
*src_metrics
,
140 dst
->_metrics
= ((unsigned long) src_metrics
) |
141 (read_only
? DST_METRICS_READ_ONLY
: 0);
144 static inline void dst_copy_metrics(struct dst_entry
*dest
, const struct dst_entry
*src
)
146 u32
*dst_metrics
= dst_metrics_write_ptr(dest
);
149 u32
*src_metrics
= DST_METRICS_PTR(src
);
151 memcpy(dst_metrics
, src_metrics
, RTAX_MAX
* sizeof(u32
));
155 static inline u32
*dst_metrics_ptr(struct dst_entry
*dst
)
157 return DST_METRICS_PTR(dst
);
161 dst_metric_raw(const struct dst_entry
*dst
, const int metric
)
163 u32
*p
= DST_METRICS_PTR(dst
);
169 dst_metric(const struct dst_entry
*dst
, const int metric
)
171 WARN_ON_ONCE(metric
== RTAX_HOPLIMIT
||
172 metric
== RTAX_ADVMSS
||
174 return dst_metric_raw(dst
, metric
);
178 dst_metric_advmss(const struct dst_entry
*dst
)
180 u32 advmss
= dst_metric_raw(dst
, RTAX_ADVMSS
);
183 advmss
= dst
->ops
->default_advmss(dst
);
188 static inline void dst_metric_set(struct dst_entry
*dst
, int metric
, u32 val
)
190 u32
*p
= dst_metrics_write_ptr(dst
);
196 /* Kernel-internal feature bits that are unallocated in user space. */
197 #define DST_FEATURE_ECN_CA (1 << 31)
199 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA)
200 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
203 dst_feature(const struct dst_entry
*dst
, u32 feature
)
205 return dst_metric(dst
, RTAX_FEATURES
) & feature
;
208 static inline u32
dst_mtu(const struct dst_entry
*dst
)
210 return dst
->ops
->mtu(dst
);
213 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
214 static inline unsigned long dst_metric_rtt(const struct dst_entry
*dst
, int metric
)
216 return msecs_to_jiffies(dst_metric(dst
, metric
));
220 dst_allfrag(const struct dst_entry
*dst
)
222 int ret
= dst_feature(dst
, RTAX_FEATURE_ALLFRAG
);
227 dst_metric_locked(const struct dst_entry
*dst
, int metric
)
229 return dst_metric(dst
, RTAX_LOCK
) & (1<<metric
);
232 static inline void dst_hold(struct dst_entry
*dst
)
235 * If your kernel compilation stops here, please check
236 * the placement of __refcnt in struct dst_entry
238 BUILD_BUG_ON(offsetof(struct dst_entry
, __refcnt
) & 63);
239 WARN_ON(atomic_inc_not_zero(&dst
->__refcnt
) == 0);
242 static inline void dst_use_noref(struct dst_entry
*dst
, unsigned long time
)
244 if (unlikely(time
!= dst
->lastuse
)) {
250 static inline void dst_hold_and_use(struct dst_entry
*dst
, unsigned long time
)
253 dst_use_noref(dst
, time
);
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
->_skb_refdst
= refdst
;
290 if (!(nskb
->_skb_refdst
& SKB_DST_NOREF
))
291 dst_clone(skb_dst(nskb
));
294 static inline void skb_dst_copy(struct sk_buff
*nskb
, const struct sk_buff
*oskb
)
296 __skb_dst_copy(nskb
, oskb
->_skb_refdst
);
300 * dst_hold_safe - Take a reference on a dst if possible
301 * @dst: pointer to dst entry
303 * This helper returns false if it could not safely
304 * take a reference on a dst.
306 static inline bool dst_hold_safe(struct dst_entry
*dst
)
308 return atomic_inc_not_zero(&dst
->__refcnt
);
312 * skb_dst_force - makes sure skb dst is refcounted
315 * If dst is not yet refcounted and not destroyed, grab a ref on it.
317 static inline void skb_dst_force(struct sk_buff
*skb
)
319 if (skb_dst_is_noref(skb
)) {
320 struct dst_entry
*dst
= skb_dst(skb
);
322 WARN_ON(!rcu_read_lock_held());
323 if (!dst_hold_safe(dst
))
326 skb
->_skb_refdst
= (unsigned long)dst
;
332 * __skb_tunnel_rx - prepare skb for rx reinsert
334 * @dev: tunnel device
335 * @net: netns for packet i/o
337 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
338 * so make some cleanups. (no accounting done)
340 static inline void __skb_tunnel_rx(struct sk_buff
*skb
, struct net_device
*dev
,
346 * Clear hash so that we can recalulate the hash for the
347 * encapsulated packet, unless we have already determine the hash
348 * over the L4 4-tuple.
350 skb_clear_hash_if_not_l4(skb
);
351 skb_set_queue_mapping(skb
, 0);
352 skb_scrub_packet(skb
, !net_eq(net
, dev_net(dev
)));
356 * skb_tunnel_rx - prepare skb for rx reinsert
358 * @dev: tunnel device
360 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
361 * so make some cleanups, and perform accounting.
362 * Note: this accounting is not SMP safe.
364 static inline void skb_tunnel_rx(struct sk_buff
*skb
, struct net_device
*dev
,
367 /* TODO : stats should be SMP safe */
368 dev
->stats
.rx_packets
++;
369 dev
->stats
.rx_bytes
+= skb
->len
;
370 __skb_tunnel_rx(skb
, dev
, net
);
373 static inline u32
dst_tclassid(const struct sk_buff
*skb
)
375 #ifdef CONFIG_IP_ROUTE_CLASSID
376 const struct dst_entry
*dst
;
380 return dst
->tclassid
;
385 int dst_discard_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
);
386 static inline int dst_discard(struct sk_buff
*skb
)
388 return dst_discard_out(&init_net
, skb
->sk
, skb
);
390 void *dst_alloc(struct dst_ops
*ops
, struct net_device
*dev
, int initial_ref
,
391 int initial_obsolete
, unsigned short flags
);
392 void dst_init(struct dst_entry
*dst
, struct dst_ops
*ops
,
393 struct net_device
*dev
, int initial_ref
, int initial_obsolete
,
394 unsigned short flags
);
395 struct dst_entry
*dst_destroy(struct dst_entry
*dst
);
396 void dst_dev_put(struct dst_entry
*dst
);
398 static inline void dst_confirm(struct dst_entry
*dst
)
402 static inline struct neighbour
*dst_neigh_lookup(const struct dst_entry
*dst
, const void *daddr
)
404 struct neighbour
*n
= dst
->ops
->neigh_lookup(dst
, NULL
, daddr
);
405 return IS_ERR(n
) ? NULL
: n
;
408 static inline struct neighbour
*dst_neigh_lookup_skb(const struct dst_entry
*dst
,
411 struct neighbour
*n
= dst
->ops
->neigh_lookup(dst
, skb
, NULL
);
412 return IS_ERR(n
) ? NULL
: n
;
415 static inline void dst_confirm_neigh(const struct dst_entry
*dst
,
418 if (dst
->ops
->confirm_neigh
)
419 dst
->ops
->confirm_neigh(dst
, daddr
);
422 static inline void dst_link_failure(struct sk_buff
*skb
)
424 struct dst_entry
*dst
= skb_dst(skb
);
425 if (dst
&& dst
->ops
&& dst
->ops
->link_failure
)
426 dst
->ops
->link_failure(skb
);
429 static inline void dst_set_expires(struct dst_entry
*dst
, int timeout
)
431 unsigned long expires
= jiffies
+ timeout
;
436 if (dst
->expires
== 0 || time_before(expires
, dst
->expires
))
437 dst
->expires
= expires
;
440 /* Output packet to network from transport. */
441 static inline int dst_output(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
443 return skb_dst(skb
)->output(net
, sk
, skb
);
446 /* Input packet from network to transport. */
447 static inline int dst_input(struct sk_buff
*skb
)
449 return skb_dst(skb
)->input(skb
);
452 static inline struct dst_entry
*dst_check(struct dst_entry
*dst
, u32 cookie
)
455 dst
= dst
->ops
->check(dst
, cookie
);
459 /* Flags for xfrm_lookup flags argument. */
461 XFRM_LOOKUP_ICMP
= 1 << 0,
462 XFRM_LOOKUP_QUEUE
= 1 << 1,
463 XFRM_LOOKUP_KEEP_DST_REF
= 1 << 2,
468 static inline struct dst_entry
*xfrm_lookup(struct net
*net
,
469 struct dst_entry
*dst_orig
,
470 const struct flowi
*fl
,
471 const struct sock
*sk
,
477 static inline struct dst_entry
*xfrm_lookup_route(struct net
*net
,
478 struct dst_entry
*dst_orig
,
479 const struct flowi
*fl
,
480 const struct sock
*sk
,
486 static inline struct xfrm_state
*dst_xfrm(const struct dst_entry
*dst
)
492 struct dst_entry
*xfrm_lookup(struct net
*net
, struct dst_entry
*dst_orig
,
493 const struct flowi
*fl
, const struct sock
*sk
,
496 struct dst_entry
*xfrm_lookup_route(struct net
*net
, struct dst_entry
*dst_orig
,
497 const struct flowi
*fl
, const struct sock
*sk
,
500 /* skb attached with this dst needs transformation if dst->xfrm is valid */
501 static inline struct xfrm_state
*dst_xfrm(const struct dst_entry
*dst
)
507 static inline void skb_dst_update_pmtu(struct sk_buff
*skb
, u32 mtu
)
509 struct dst_entry
*dst
= skb_dst(skb
);
511 if (dst
&& dst
->ops
->update_pmtu
)
512 dst
->ops
->update_pmtu(dst
, NULL
, skb
, mtu
);
515 #endif /* _NET_DST_H */