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/jiffies.h>
16 #include <net/neighbour.h>
17 #include <asm/processor.h>
20 * 0 - no debugging messages
21 * 1 - rare events and bugs (default)
24 #define RT_CACHE_DEBUG 0
26 #define DST_GC_MIN (HZ/10)
27 #define DST_GC_INC (HZ/2)
28 #define DST_GC_MAX (120*HZ)
30 /* Each dst_entry has reference count and sits in some parent list(s).
31 * When it is removed from parent list, it is "freed" (dst_free).
32 * After this it enters dead state (dst->obsolete > 0) and if its refcnt
33 * is zero, it can be destroyed immediately, otherwise it is added
34 * to gc list and garbage collector periodically checks the refcnt.
40 struct rcu_head rcu_head;
41 struct dst_entry *child;
42 struct net_device *dev;
46 #define DST_HOST 0x0001
47 #define DST_NOXFRM 0x0002
48 #define DST_NOPOLICY 0x0004
49 #define DST_NOHASH 0x0008
50 #define DST_NOCACHE 0x0010
51 unsigned long expires;
53 unsigned short header_len; /* more space at head required */
54 unsigned short trailer_len; /* space to reserve at tail */
56 unsigned int rate_tokens;
57 unsigned long rate_last; /* rate limiting for ICMP */
59 struct dst_entry *path;
61 struct neighbour *neighbour;
64 struct xfrm_state *xfrm;
68 int (*input)(struct sk_buff*);
69 int (*output)(struct sk_buff*);
73 u32 _metrics[RTAX_MAX];
75 #ifdef CONFIG_NET_CLS_ROUTE
83 * Align __refcnt to a 64 bytes alignment
84 * (L1_CACHE_SIZE would be too much)
87 long __pad_to_align_refcnt[1];
90 * __refcnt wants to be on a different cache line from
91 * input/output/ops or performance tanks badly
93 atomic_t __refcnt; /* client references */
95 unsigned long lastuse;
97 struct dst_entry *next;
98 struct rtable __rcu *rt_next;
99 struct rt6_info *rt6_next;
100 struct dn_route __rcu *dn_next;
107 dst_metric(const struct dst_entry *dst, int metric)
109 return dst->_metrics[metric-1];
112 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
114 dst->_metrics[metric-1] = val;
117 static inline void dst_import_metrics(struct dst_entry *dst, const u32 *src_metrics)
119 memcpy(dst->_metrics, src_metrics, RTAX_MAX * sizeof(u32));
122 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
124 dst_import_metrics(dest, src->_metrics);
127 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
129 return dst->_metrics;
133 dst_feature(const struct dst_entry *dst, u32 feature)
135 return dst_metric(dst, RTAX_FEATURES) & feature;
138 static inline u32 dst_mtu(const struct dst_entry *dst)
140 u32 mtu = dst_metric(dst, RTAX_MTU);
142 * Alexey put it here, so ask him about it :)
148 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
149 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
151 return msecs_to_jiffies(dst_metric(dst, metric));
154 static inline void set_dst_metric_rtt(struct dst_entry *dst, int metric,
157 dst_metric_set(dst, metric, jiffies_to_msecs(rtt));
161 dst_allfrag(const struct dst_entry *dst)
163 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG);
164 /* Yes, _exactly_. This is paranoia. */
170 dst_metric_locked(struct dst_entry *dst, int metric)
172 return dst_metric(dst, RTAX_LOCK) & (1<<metric);
175 static inline void dst_hold(struct dst_entry * dst)
178 * If your kernel compilation stops here, please check
179 * __pad_to_align_refcnt declaration in struct dst_entry
181 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
182 atomic_inc(&dst->__refcnt);
185 static inline void dst_use(struct dst_entry *dst, unsigned long time)
192 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
199 struct dst_entry * dst_clone(struct dst_entry * dst)
202 atomic_inc(&dst->__refcnt);
206 extern void dst_release(struct dst_entry *dst);
208 static inline void refdst_drop(unsigned long refdst)
210 if (!(refdst & SKB_DST_NOREF))
211 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
215 * skb_dst_drop - drops skb dst
218 * Drops dst reference count if a reference was taken.
220 static inline void skb_dst_drop(struct sk_buff *skb)
222 if (skb->_skb_refdst) {
223 refdst_drop(skb->_skb_refdst);
224 skb->_skb_refdst = 0UL;
228 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
230 nskb->_skb_refdst = oskb->_skb_refdst;
231 if (!(nskb->_skb_refdst & SKB_DST_NOREF))
232 dst_clone(skb_dst(nskb));
236 * skb_dst_force - makes sure skb dst is refcounted
239 * If dst is not yet refcounted, let's do it
241 static inline void skb_dst_force(struct sk_buff *skb)
243 if (skb_dst_is_noref(skb)) {
244 WARN_ON(!rcu_read_lock_held());
245 skb->_skb_refdst &= ~SKB_DST_NOREF;
246 dst_clone(skb_dst(skb));
252 * __skb_tunnel_rx - prepare skb for rx reinsert
254 * @dev: tunnel device
256 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
257 * so make some cleanups. (no accounting done)
259 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
263 skb_set_queue_mapping(skb, 0);
269 * skb_tunnel_rx - prepare skb for rx reinsert
271 * @dev: tunnel device
273 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
274 * so make some cleanups, and perform accounting.
275 * Note: this accounting is not SMP safe.
277 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
279 /* TODO : stats should be SMP safe */
280 dev->stats.rx_packets++;
281 dev->stats.rx_bytes += skb->len;
282 __skb_tunnel_rx(skb, dev);
285 /* Children define the path of the packet through the
286 * Linux networking. Thus, destinations are stackable.
289 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
291 struct dst_entry *child = skb_dst(skb)->child;
297 extern int dst_discard(struct sk_buff *skb);
298 extern void * dst_alloc(struct dst_ops * ops);
299 extern void __dst_free(struct dst_entry * dst);
300 extern struct dst_entry *dst_destroy(struct dst_entry * dst);
302 static inline void dst_free(struct dst_entry * dst)
304 if (dst->obsolete > 1)
306 if (!atomic_read(&dst->__refcnt)) {
307 dst = dst_destroy(dst);
314 static inline void dst_rcu_free(struct rcu_head *head)
316 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
320 static inline void dst_confirm(struct dst_entry *dst)
323 neigh_confirm(dst->neighbour);
326 static inline void dst_link_failure(struct sk_buff *skb)
328 struct dst_entry *dst = skb_dst(skb);
329 if (dst && dst->ops && dst->ops->link_failure)
330 dst->ops->link_failure(skb);
333 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
335 unsigned long expires = jiffies + timeout;
340 if (dst->expires == 0 || time_before(expires, dst->expires))
341 dst->expires = expires;
344 /* Output packet to network from transport. */
345 static inline int dst_output(struct sk_buff *skb)
347 return skb_dst(skb)->output(skb);
350 /* Input packet from network to transport. */
351 static inline int dst_input(struct sk_buff *skb)
353 return skb_dst(skb)->input(skb);
356 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
359 dst = dst->ops->check(dst, cookie);
363 extern void dst_init(void);
365 /* Flags for xfrm_lookup flags argument. */
367 XFRM_LOOKUP_WAIT = 1 << 0,
368 XFRM_LOOKUP_ICMP = 1 << 1,
373 static inline int xfrm_lookup(struct net *net, struct dst_entry **dst_p,
374 struct flowi *fl, struct sock *sk, int flags)
378 static inline int __xfrm_lookup(struct net *net, struct dst_entry **dst_p,
379 struct flowi *fl, struct sock *sk, int flags)
384 extern int xfrm_lookup(struct net *net, struct dst_entry **dst_p,
385 struct flowi *fl, struct sock *sk, int flags);
386 extern int __xfrm_lookup(struct net *net, struct dst_entry **dst_p,
387 struct flowi *fl, struct sock *sk, int flags);
391 #endif /* _NET_DST_H */