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
88 struct lwtunnel_state *lwtstate;
90 * Align __refcnt to a 64 bytes alignment
91 * (L1_CACHE_SIZE would be too much)
93 long __pad_to_align_refcnt[1];
96 * __refcnt wants to be on a different cache line from
97 * input/output/ops or performance tanks badly
99 atomic_t __refcnt; /* client references */
101 unsigned long lastuse;
103 struct lwtunnel_state *lwtstate;
106 struct dst_entry *next;
107 struct rtable __rcu *rt_next;
108 struct rt6_info *rt6_next;
109 struct dn_route __rcu *dn_next;
113 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
114 extern const u32 dst_default_metrics[];
116 #define DST_METRICS_READ_ONLY 0x1UL
117 #define DST_METRICS_FLAGS 0x3UL
118 #define __DST_METRICS_PTR(Y) \
119 ((u32 *)((Y) & ~DST_METRICS_FLAGS))
120 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
122 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
124 return dst->_metrics & DST_METRICS_READ_ONLY;
127 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
129 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
131 unsigned long val = dst->_metrics;
132 if (!(val & DST_METRICS_READ_ONLY))
133 __dst_destroy_metrics_generic(dst, val);
136 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
138 unsigned long p = dst->_metrics;
142 if (p & DST_METRICS_READ_ONLY)
143 return dst->ops->cow_metrics(dst, p);
144 return __DST_METRICS_PTR(p);
147 /* This may only be invoked before the entry has reached global
150 static inline void dst_init_metrics(struct dst_entry *dst,
151 const u32 *src_metrics,
154 dst->_metrics = ((unsigned long) src_metrics) |
155 (read_only ? DST_METRICS_READ_ONLY : 0);
158 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
160 u32 *dst_metrics = dst_metrics_write_ptr(dest);
163 u32 *src_metrics = DST_METRICS_PTR(src);
165 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
169 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
171 return DST_METRICS_PTR(dst);
175 dst_metric_raw(const struct dst_entry *dst, const int metric)
177 u32 *p = DST_METRICS_PTR(dst);
183 dst_metric(const struct dst_entry *dst, const int metric)
185 WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
186 metric == RTAX_ADVMSS ||
188 return dst_metric_raw(dst, metric);
192 dst_metric_advmss(const struct dst_entry *dst)
194 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
197 advmss = dst->ops->default_advmss(dst);
202 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
204 u32 *p = dst_metrics_write_ptr(dst);
210 /* Kernel-internal feature bits that are unallocated in user space. */
211 #define DST_FEATURE_ECN_CA (1 << 31)
213 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA)
214 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
217 dst_feature(const struct dst_entry *dst, u32 feature)
219 return dst_metric(dst, RTAX_FEATURES) & feature;
222 static inline u32 dst_mtu(const struct dst_entry *dst)
224 return dst->ops->mtu(dst);
227 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
228 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
230 return msecs_to_jiffies(dst_metric(dst, metric));
234 dst_allfrag(const struct dst_entry *dst)
236 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG);
241 dst_metric_locked(const struct dst_entry *dst, int metric)
243 return dst_metric(dst, RTAX_LOCK) & (1<<metric);
246 static inline void dst_hold(struct dst_entry *dst)
249 * If your kernel compilation stops here, please check
250 * __pad_to_align_refcnt declaration in struct dst_entry
252 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
253 atomic_inc(&dst->__refcnt);
256 static inline void dst_use(struct dst_entry *dst, unsigned long time)
263 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
269 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
272 atomic_inc(&dst->__refcnt);
276 void dst_release(struct dst_entry *dst);
278 static inline void refdst_drop(unsigned long refdst)
280 if (!(refdst & SKB_DST_NOREF))
281 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
285 * skb_dst_drop - drops skb dst
288 * Drops dst reference count if a reference was taken.
290 static inline void skb_dst_drop(struct sk_buff *skb)
292 if (skb->_skb_refdst) {
293 refdst_drop(skb->_skb_refdst);
294 skb->_skb_refdst = 0UL;
298 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
300 nskb->_skb_refdst = refdst;
301 if (!(nskb->_skb_refdst & SKB_DST_NOREF))
302 dst_clone(skb_dst(nskb));
305 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
307 __skb_dst_copy(nskb, oskb->_skb_refdst);
311 * skb_dst_force - makes sure skb dst is refcounted
314 * If dst is not yet refcounted, let's do it
316 static inline void skb_dst_force(struct sk_buff *skb)
318 if (skb_dst_is_noref(skb)) {
319 WARN_ON(!rcu_read_lock_held());
320 skb->_skb_refdst &= ~SKB_DST_NOREF;
321 dst_clone(skb_dst(skb));
326 * dst_hold_safe - Take a reference on a dst if possible
327 * @dst: pointer to dst entry
329 * This helper returns false if it could not safely
330 * take a reference on a dst.
332 static inline bool dst_hold_safe(struct dst_entry *dst)
334 if (dst->flags & DST_NOCACHE)
335 return atomic_inc_not_zero(&dst->__refcnt);
341 * skb_dst_force_safe - makes sure skb dst is refcounted
344 * If dst is not yet refcounted and not destroyed, grab a ref on it.
346 static inline void skb_dst_force_safe(struct sk_buff *skb)
348 if (skb_dst_is_noref(skb)) {
349 struct dst_entry *dst = skb_dst(skb);
351 if (!dst_hold_safe(dst))
354 skb->_skb_refdst = (unsigned long)dst;
360 * __skb_tunnel_rx - prepare skb for rx reinsert
362 * @dev: tunnel device
363 * @net: netns for packet i/o
365 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
366 * so make some cleanups. (no accounting done)
368 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
374 * Clear hash so that we can recalulate the hash for the
375 * encapsulated packet, unless we have already determine the hash
376 * over the L4 4-tuple.
378 skb_clear_hash_if_not_l4(skb);
379 skb_set_queue_mapping(skb, 0);
380 skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
384 * skb_tunnel_rx - prepare skb for rx reinsert
386 * @dev: tunnel device
388 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
389 * so make some cleanups, and perform accounting.
390 * Note: this accounting is not SMP safe.
392 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
395 /* TODO : stats should be SMP safe */
396 dev->stats.rx_packets++;
397 dev->stats.rx_bytes += skb->len;
398 __skb_tunnel_rx(skb, dev, net);
401 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
402 static inline int dst_discard(struct sk_buff *skb)
404 return dst_discard_out(&init_net, skb->sk, skb);
406 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
407 int initial_obsolete, unsigned short flags);
408 void dst_init(struct dst_entry *dst, struct dst_ops *ops,
409 struct net_device *dev, int initial_ref, int initial_obsolete,
410 unsigned short flags);
411 void __dst_free(struct dst_entry *dst);
412 struct dst_entry *dst_destroy(struct dst_entry *dst);
414 static inline void dst_free(struct dst_entry *dst)
416 if (dst->obsolete > 0)
418 if (!atomic_read(&dst->__refcnt)) {
419 dst = dst_destroy(dst);
426 static inline void dst_rcu_free(struct rcu_head *head)
428 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
432 static inline void dst_confirm(struct dst_entry *dst)
434 dst->pending_confirm = 1;
437 static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n,
440 const struct hh_cache *hh;
442 if (dst->pending_confirm) {
443 unsigned long now = jiffies;
445 dst->pending_confirm = 0;
446 /* avoid dirtying neighbour */
447 if (n->confirmed != now)
452 if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
453 return neigh_hh_output(hh, skb);
455 return n->output(n, skb);
458 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
460 struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
461 return IS_ERR(n) ? NULL : n;
464 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
467 struct neighbour *n = dst->ops->neigh_lookup(dst, skb, NULL);
468 return IS_ERR(n) ? NULL : n;
471 static inline void dst_link_failure(struct sk_buff *skb)
473 struct dst_entry *dst = skb_dst(skb);
474 if (dst && dst->ops && dst->ops->link_failure)
475 dst->ops->link_failure(skb);
478 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
480 unsigned long expires = jiffies + timeout;
485 if (dst->expires == 0 || time_before(expires, dst->expires))
486 dst->expires = expires;
489 /* Output packet to network from transport. */
490 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
492 return skb_dst(skb)->output(net, sk, skb);
495 /* Input packet from network to transport. */
496 static inline int dst_input(struct sk_buff *skb)
498 return skb_dst(skb)->input(skb);
501 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
504 dst = dst->ops->check(dst, cookie);
508 void dst_subsys_init(void);
510 /* Flags for xfrm_lookup flags argument. */
512 XFRM_LOOKUP_ICMP = 1 << 0,
513 XFRM_LOOKUP_QUEUE = 1 << 1,
514 XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
519 static inline struct dst_entry *xfrm_lookup(struct net *net,
520 struct dst_entry *dst_orig,
521 const struct flowi *fl,
522 const struct sock *sk,
528 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
529 struct dst_entry *dst_orig,
530 const struct flowi *fl,
531 const struct sock *sk,
537 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
543 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
544 const struct flowi *fl, const struct sock *sk,
547 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
548 const struct flowi *fl, const struct sock *sk,
551 /* skb attached with this dst needs transformation if dst->xfrm is valid */
552 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
558 #endif /* _NET_DST_H */