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
61 unsigned short pending_confirm;
65 /* A non-zero value of dst->obsolete forces by-hand validation
66 * of the route entry. Positive values are set by the generic
67 * dst layer to indicate that the entry has been forcefully
70 * Negative values are used by the implementation layer code to
71 * force invocation of the dst_ops->check() method.
74 #define DST_OBSOLETE_NONE 0
75 #define DST_OBSOLETE_DEAD 2
76 #define DST_OBSOLETE_FORCE_CHK -1
77 #define DST_OBSOLETE_KILL -2
78 unsigned short header_len; /* more space at head required */
79 unsigned short trailer_len; /* space to reserve at tail */
80 #ifdef CONFIG_IP_ROUTE_CLASSID
87 * Align __refcnt to a 64 bytes alignment
88 * (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;
101 struct dst_entry *next;
102 struct rtable __rcu *rt_next;
103 struct rt6_info *rt6_next;
104 struct dn_route __rcu *dn_next;
108 extern u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
109 extern const u32 dst_default_metrics[];
111 #define DST_METRICS_READ_ONLY 0x1UL
112 #define __DST_METRICS_PTR(Y) \
113 ((u32 *)((Y) & ~DST_METRICS_READ_ONLY))
114 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
116 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
118 return dst->_metrics & DST_METRICS_READ_ONLY;
121 extern void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
123 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
125 unsigned long val = dst->_metrics;
126 if (!(val & DST_METRICS_READ_ONLY))
127 __dst_destroy_metrics_generic(dst, val);
130 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
132 unsigned long p = dst->_metrics;
136 if (p & DST_METRICS_READ_ONLY)
137 return dst->ops->cow_metrics(dst, p);
138 return __DST_METRICS_PTR(p);
141 /* This may only be invoked before the entry has reached global
144 static inline void dst_init_metrics(struct dst_entry *dst,
145 const u32 *src_metrics,
148 dst->_metrics = ((unsigned long) src_metrics) |
149 (read_only ? DST_METRICS_READ_ONLY : 0);
152 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
154 u32 *dst_metrics = dst_metrics_write_ptr(dest);
157 u32 *src_metrics = DST_METRICS_PTR(src);
159 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
163 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
165 return DST_METRICS_PTR(dst);
169 dst_metric_raw(const struct dst_entry *dst, const int metric)
171 u32 *p = DST_METRICS_PTR(dst);
177 dst_metric(const struct dst_entry *dst, const int metric)
179 WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
180 metric == RTAX_ADVMSS ||
182 return dst_metric_raw(dst, metric);
186 dst_metric_advmss(const struct dst_entry *dst)
188 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
191 advmss = dst->ops->default_advmss(dst);
196 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
198 u32 *p = dst_metrics_write_ptr(dst);
205 dst_feature(const struct dst_entry *dst, u32 feature)
207 return dst_metric(dst, RTAX_FEATURES) & feature;
210 static inline u32 dst_mtu(const struct dst_entry *dst)
212 return dst->ops->mtu(dst);
215 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
216 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
218 return msecs_to_jiffies(dst_metric(dst, metric));
222 dst_allfrag(const struct dst_entry *dst)
224 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG);
229 dst_metric_locked(const struct dst_entry *dst, int metric)
231 return dst_metric(dst, RTAX_LOCK) & (1<<metric);
234 static inline void dst_hold(struct dst_entry *dst)
237 * If your kernel compilation stops here, please check
238 * __pad_to_align_refcnt declaration in struct dst_entry
240 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
241 atomic_inc(&dst->__refcnt);
244 static inline void dst_use(struct dst_entry *dst, unsigned long time)
251 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
257 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
260 atomic_inc(&dst->__refcnt);
264 extern void dst_release(struct dst_entry *dst);
266 static inline void refdst_drop(unsigned long refdst)
268 if (!(refdst & SKB_DST_NOREF))
269 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
273 * skb_dst_drop - drops skb dst
276 * Drops dst reference count if a reference was taken.
278 static inline void skb_dst_drop(struct sk_buff *skb)
280 if (skb->_skb_refdst) {
281 refdst_drop(skb->_skb_refdst);
282 skb->_skb_refdst = 0UL;
286 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
288 nskb->_skb_refdst = oskb->_skb_refdst;
289 if (!(nskb->_skb_refdst & SKB_DST_NOREF))
290 dst_clone(skb_dst(nskb));
294 * skb_dst_force - makes sure skb dst is refcounted
297 * If dst is not yet refcounted, let's do it
299 static inline void skb_dst_force(struct sk_buff *skb)
301 if (skb_dst_is_noref(skb)) {
302 WARN_ON(!rcu_read_lock_held());
303 skb->_skb_refdst &= ~SKB_DST_NOREF;
304 dst_clone(skb_dst(skb));
310 * __skb_tunnel_rx - prepare skb for rx reinsert
312 * @dev: tunnel device
314 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
315 * so make some cleanups. (no accounting done)
317 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
322 * Clear rxhash so that we can recalulate the hash for the
323 * encapsulated packet, unless we have already determine the hash
324 * over the L4 4-tuple.
328 skb_set_queue_mapping(skb, 0);
334 * skb_tunnel_rx - prepare skb for rx reinsert
336 * @dev: tunnel device
338 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
339 * so make some cleanups, and perform accounting.
340 * Note: this accounting is not SMP safe.
342 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
344 /* TODO : stats should be SMP safe */
345 dev->stats.rx_packets++;
346 dev->stats.rx_bytes += skb->len;
347 __skb_tunnel_rx(skb, dev);
350 /* Children define the path of the packet through the
351 * Linux networking. Thus, destinations are stackable.
354 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
356 struct dst_entry *child = dst_clone(skb_dst(skb)->child);
362 extern int dst_discard(struct sk_buff *skb);
363 extern void *dst_alloc(struct dst_ops *ops, struct net_device *dev,
364 int initial_ref, int initial_obsolete,
365 unsigned short flags);
366 extern void __dst_free(struct dst_entry *dst);
367 extern struct dst_entry *dst_destroy(struct dst_entry *dst);
369 static inline void dst_free(struct dst_entry *dst)
371 if (dst->obsolete > 0)
373 if (!atomic_read(&dst->__refcnt)) {
374 dst = dst_destroy(dst);
381 static inline void dst_rcu_free(struct rcu_head *head)
383 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
387 static inline void dst_confirm(struct dst_entry *dst)
389 dst->pending_confirm = 1;
392 static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n,
395 const struct hh_cache *hh;
397 if (dst->pending_confirm) {
398 unsigned long now = jiffies;
400 dst->pending_confirm = 0;
401 /* avoid dirtying neighbour */
402 if (n->confirmed != now)
407 if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
408 return neigh_hh_output(hh, skb);
410 return n->output(n, skb);
413 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
415 return dst->ops->neigh_lookup(dst, NULL, daddr);
418 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
421 return dst->ops->neigh_lookup(dst, skb, NULL);
424 static inline void dst_link_failure(struct sk_buff *skb)
426 struct dst_entry *dst = skb_dst(skb);
427 if (dst && dst->ops && dst->ops->link_failure)
428 dst->ops->link_failure(skb);
431 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
433 unsigned long expires = jiffies + timeout;
438 if (dst->expires == 0 || time_before(expires, dst->expires))
439 dst->expires = expires;
442 /* Output packet to network from transport. */
443 static inline int dst_output(struct sk_buff *skb)
445 return skb_dst(skb)->output(skb);
448 /* Input packet from network to transport. */
449 static inline int dst_input(struct sk_buff *skb)
451 return skb_dst(skb)->input(skb);
454 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
457 dst = dst->ops->check(dst, cookie);
461 extern void dst_init(void);
463 /* Flags for xfrm_lookup flags argument. */
465 XFRM_LOOKUP_ICMP = 1 << 0,
470 static inline struct dst_entry *xfrm_lookup(struct net *net,
471 struct dst_entry *dst_orig,
472 const struct flowi *fl, struct sock *sk,
478 extern struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
479 const struct flowi *fl, struct sock *sk,
483 #endif /* _NET_DST_H */