2 * net/dst.h Protocol independent destination cache definitions.
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 <linux/refcount.h>
18 #include <net/neighbour.h>
19 #include <asm/processor.h>
21 #define DST_GC_MIN (HZ/10)
22 #define DST_GC_INC (HZ/2)
23 #define DST_GC_MAX (120*HZ)
25 /* Each dst_entry has reference count and sits in some parent list(s).
26 * When it is removed from parent list, it is "freed" (dst_free).
27 * After this it enters dead state (dst->obsolete > 0) and if its refcnt
28 * is zero, it can be destroyed immediately, otherwise it is added
29 * to gc list and garbage collector periodically checks the refcnt.
35 struct net_device *dev;
36 struct rcu_head rcu_head;
37 struct dst_entry *child;
39 unsigned long _metrics;
40 unsigned long expires;
41 struct dst_entry *path;
42 struct dst_entry *from;
44 struct xfrm_state *xfrm;
48 int (*input)(struct sk_buff *);
49 int (*output)(struct net *net, struct sock *sk, struct sk_buff *skb);
52 #define DST_HOST 0x0001
53 #define DST_NOXFRM 0x0002
54 #define DST_NOPOLICY 0x0004
55 #define DST_NOCOUNT 0x0008
56 #define DST_FAKE_RTABLE 0x0010
57 #define DST_XFRM_TUNNEL 0x0020
58 #define DST_XFRM_QUEUE 0x0040
59 #define DST_METADATA 0x0080
63 /* A non-zero value of dst->obsolete forces by-hand validation
64 * of the route entry. Positive values are set by the generic
65 * dst layer to indicate that the entry has been forcefully
68 * Negative values are used by the implementation layer code to
69 * force invocation of the dst_ops->check() method.
72 #define DST_OBSOLETE_NONE 0
73 #define DST_OBSOLETE_DEAD 2
74 #define DST_OBSOLETE_FORCE_CHK -1
75 #define DST_OBSOLETE_KILL -2
76 unsigned short header_len; /* more space at head required */
77 unsigned short trailer_len; /* space to reserve at tail */
78 unsigned short __pad3;
80 #ifdef CONFIG_IP_ROUTE_CLASSID
88 * Align __refcnt to a 64 bytes alignment
89 * (L1_CACHE_SIZE would be too much)
91 long __pad_to_align_refcnt[2];
94 * __refcnt wants to be on a different cache line from
95 * input/output/ops or performance tanks badly
97 atomic_t __refcnt; /* client references */
99 unsigned long lastuse;
100 struct lwtunnel_state *lwtstate;
102 struct dst_entry *next;
103 struct rtable __rcu *rt_next;
104 struct rt6_info *rt6_next;
105 struct dn_route __rcu *dn_next;
110 u32 metrics[RTAX_MAX];
113 extern const struct dst_metrics dst_default_metrics;
115 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
117 #define DST_METRICS_READ_ONLY 0x1UL
118 #define DST_METRICS_REFCOUNTED 0x2UL
119 #define DST_METRICS_FLAGS 0x3UL
120 #define __DST_METRICS_PTR(Y) \
121 ((u32 *)((Y) & ~DST_METRICS_FLAGS))
122 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
124 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
126 return dst->_metrics & DST_METRICS_READ_ONLY;
129 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
131 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
133 unsigned long val = dst->_metrics;
134 if (!(val & DST_METRICS_READ_ONLY))
135 __dst_destroy_metrics_generic(dst, val);
138 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
140 unsigned long p = dst->_metrics;
144 if (p & DST_METRICS_READ_ONLY)
145 return dst->ops->cow_metrics(dst, p);
146 return __DST_METRICS_PTR(p);
149 /* This may only be invoked before the entry has reached global
152 static inline void dst_init_metrics(struct dst_entry *dst,
153 const u32 *src_metrics,
156 dst->_metrics = ((unsigned long) src_metrics) |
157 (read_only ? DST_METRICS_READ_ONLY : 0);
160 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
162 u32 *dst_metrics = dst_metrics_write_ptr(dest);
165 u32 *src_metrics = DST_METRICS_PTR(src);
167 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
171 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
173 return DST_METRICS_PTR(dst);
177 dst_metric_raw(const struct dst_entry *dst, const int metric)
179 u32 *p = DST_METRICS_PTR(dst);
185 dst_metric(const struct dst_entry *dst, const int metric)
187 WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
188 metric == RTAX_ADVMSS ||
190 return dst_metric_raw(dst, metric);
194 dst_metric_advmss(const struct dst_entry *dst)
196 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
199 advmss = dst->ops->default_advmss(dst);
204 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
206 u32 *p = dst_metrics_write_ptr(dst);
212 /* Kernel-internal feature bits that are unallocated in user space. */
213 #define DST_FEATURE_ECN_CA (1 << 31)
215 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA)
216 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
219 dst_feature(const struct dst_entry *dst, u32 feature)
221 return dst_metric(dst, RTAX_FEATURES) & feature;
224 static inline u32 dst_mtu(const struct dst_entry *dst)
226 return dst->ops->mtu(dst);
229 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
230 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
232 return msecs_to_jiffies(dst_metric(dst, metric));
236 dst_allfrag(const struct dst_entry *dst)
238 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG);
243 dst_metric_locked(const struct dst_entry *dst, int metric)
245 return dst_metric(dst, RTAX_LOCK) & (1<<metric);
248 static inline void dst_hold(struct dst_entry *dst)
251 * If your kernel compilation stops here, please check
252 * __pad_to_align_refcnt declaration in struct dst_entry
254 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
255 WARN_ON(atomic_inc_not_zero(&dst->__refcnt) == 0);
258 static inline void dst_use(struct dst_entry *dst, unsigned long time)
265 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
271 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
278 void dst_release(struct dst_entry *dst);
280 void dst_release_immediate(struct dst_entry *dst);
282 static inline void refdst_drop(unsigned long refdst)
284 if (!(refdst & SKB_DST_NOREF))
285 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
289 * skb_dst_drop - drops skb dst
292 * Drops dst reference count if a reference was taken.
294 static inline void skb_dst_drop(struct sk_buff *skb)
296 if (skb->_skb_refdst) {
297 refdst_drop(skb->_skb_refdst);
298 skb->_skb_refdst = 0UL;
302 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
304 nskb->_skb_refdst = refdst;
305 if (!(nskb->_skb_refdst & SKB_DST_NOREF))
306 dst_clone(skb_dst(nskb));
309 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
311 __skb_dst_copy(nskb, oskb->_skb_refdst);
315 * dst_hold_safe - Take a reference on a dst if possible
316 * @dst: pointer to dst entry
318 * This helper returns false if it could not safely
319 * take a reference on a dst.
321 static inline bool dst_hold_safe(struct dst_entry *dst)
323 return atomic_inc_not_zero(&dst->__refcnt);
327 * skb_dst_force - makes sure skb dst is refcounted
330 * If dst is not yet refcounted and not destroyed, grab a ref on it.
332 static inline void skb_dst_force(struct sk_buff *skb)
334 if (skb_dst_is_noref(skb)) {
335 struct dst_entry *dst = skb_dst(skb);
337 WARN_ON(!rcu_read_lock_held());
338 if (!dst_hold_safe(dst))
341 skb->_skb_refdst = (unsigned long)dst;
347 * __skb_tunnel_rx - prepare skb for rx reinsert
349 * @dev: tunnel device
350 * @net: netns for packet i/o
352 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
353 * so make some cleanups. (no accounting done)
355 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
361 * Clear hash so that we can recalulate the hash for the
362 * encapsulated packet, unless we have already determine the hash
363 * over the L4 4-tuple.
365 skb_clear_hash_if_not_l4(skb);
366 skb_set_queue_mapping(skb, 0);
367 skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
371 * skb_tunnel_rx - prepare skb for rx reinsert
373 * @dev: tunnel device
375 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
376 * so make some cleanups, and perform accounting.
377 * Note: this accounting is not SMP safe.
379 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
382 /* TODO : stats should be SMP safe */
383 dev->stats.rx_packets++;
384 dev->stats.rx_bytes += skb->len;
385 __skb_tunnel_rx(skb, dev, net);
388 static inline u32 dst_tclassid(const struct sk_buff *skb)
390 #ifdef CONFIG_IP_ROUTE_CLASSID
391 const struct dst_entry *dst;
395 return dst->tclassid;
400 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
401 static inline int dst_discard(struct sk_buff *skb)
403 return dst_discard_out(&init_net, skb->sk, skb);
405 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
406 int initial_obsolete, unsigned short flags);
407 void dst_init(struct dst_entry *dst, struct dst_ops *ops,
408 struct net_device *dev, int initial_ref, int initial_obsolete,
409 unsigned short flags);
410 struct dst_entry *dst_destroy(struct dst_entry *dst);
411 void dst_dev_put(struct dst_entry *dst);
413 static inline void dst_confirm(struct dst_entry *dst)
417 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
419 struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
420 return IS_ERR(n) ? NULL : n;
423 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
426 struct neighbour *n = dst->ops->neigh_lookup(dst, skb, NULL);
427 return IS_ERR(n) ? NULL : n;
430 static inline void dst_confirm_neigh(const struct dst_entry *dst,
433 if (dst->ops->confirm_neigh)
434 dst->ops->confirm_neigh(dst, daddr);
437 static inline void dst_link_failure(struct sk_buff *skb)
439 struct dst_entry *dst = skb_dst(skb);
440 if (dst && dst->ops && dst->ops->link_failure)
441 dst->ops->link_failure(skb);
444 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
446 unsigned long expires = jiffies + timeout;
451 if (dst->expires == 0 || time_before(expires, dst->expires))
452 dst->expires = expires;
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 skb_dst(skb)->output(net, sk, skb);
461 /* Input packet from network to transport. */
462 static inline int dst_input(struct sk_buff *skb)
464 return skb_dst(skb)->input(skb);
467 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
470 dst = dst->ops->check(dst, cookie);
474 /* Flags for xfrm_lookup flags argument. */
476 XFRM_LOOKUP_ICMP = 1 << 0,
477 XFRM_LOOKUP_QUEUE = 1 << 1,
478 XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
483 static inline struct dst_entry *xfrm_lookup(struct net *net,
484 struct dst_entry *dst_orig,
485 const struct flowi *fl,
486 const struct sock *sk,
492 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
493 struct dst_entry *dst_orig,
494 const struct flowi *fl,
495 const struct sock *sk,
501 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
507 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
508 const struct flowi *fl, const struct sock *sk,
511 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
512 const struct flowi *fl, const struct sock *sk,
515 /* skb attached with this dst needs transformation if dst->xfrm is valid */
516 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
522 #endif /* _NET_DST_H */