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;
114 u32 metrics[RTAX_MAX];
116 } __aligned(4); /* Low pointer bits contain DST_METRICS_FLAGS */
117 extern const struct dst_metrics dst_default_metrics;
119 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
121 #define DST_METRICS_READ_ONLY 0x1UL
122 #define DST_METRICS_REFCOUNTED 0x2UL
123 #define DST_METRICS_FLAGS 0x3UL
124 #define __DST_METRICS_PTR(Y) \
125 ((u32 *)((Y) & ~DST_METRICS_FLAGS))
126 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
128 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
130 return dst->_metrics & DST_METRICS_READ_ONLY;
133 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
135 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
137 unsigned long val = dst->_metrics;
138 if (!(val & DST_METRICS_READ_ONLY))
139 __dst_destroy_metrics_generic(dst, val);
142 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
144 unsigned long p = dst->_metrics;
148 if (p & DST_METRICS_READ_ONLY)
149 return dst->ops->cow_metrics(dst, p);
150 return __DST_METRICS_PTR(p);
153 /* This may only be invoked before the entry has reached global
156 static inline void dst_init_metrics(struct dst_entry *dst,
157 const u32 *src_metrics,
160 dst->_metrics = ((unsigned long) src_metrics) |
161 (read_only ? DST_METRICS_READ_ONLY : 0);
164 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
166 u32 *dst_metrics = dst_metrics_write_ptr(dest);
169 u32 *src_metrics = DST_METRICS_PTR(src);
171 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
175 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
177 return DST_METRICS_PTR(dst);
181 dst_metric_raw(const struct dst_entry *dst, const int metric)
183 u32 *p = DST_METRICS_PTR(dst);
189 dst_metric(const struct dst_entry *dst, const int metric)
191 WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
192 metric == RTAX_ADVMSS ||
194 return dst_metric_raw(dst, metric);
198 dst_metric_advmss(const struct dst_entry *dst)
200 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
203 advmss = dst->ops->default_advmss(dst);
208 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
210 u32 *p = dst_metrics_write_ptr(dst);
216 /* Kernel-internal feature bits that are unallocated in user space. */
217 #define DST_FEATURE_ECN_CA (1 << 31)
219 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA)
220 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
223 dst_feature(const struct dst_entry *dst, u32 feature)
225 return dst_metric(dst, RTAX_FEATURES) & feature;
228 static inline u32 dst_mtu(const struct dst_entry *dst)
230 return dst->ops->mtu(dst);
233 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
234 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
236 return msecs_to_jiffies(dst_metric(dst, metric));
240 dst_allfrag(const struct dst_entry *dst)
242 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG);
247 dst_metric_locked(const struct dst_entry *dst, int metric)
249 return dst_metric(dst, RTAX_LOCK) & (1<<metric);
252 static inline void dst_hold(struct dst_entry *dst)
255 * If your kernel compilation stops here, please check
256 * __pad_to_align_refcnt declaration in struct dst_entry
258 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
259 atomic_inc(&dst->__refcnt);
262 static inline void dst_use(struct dst_entry *dst, unsigned long time)
269 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
275 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
278 atomic_inc(&dst->__refcnt);
282 void dst_release(struct dst_entry *dst);
284 static inline void refdst_drop(unsigned long refdst)
286 if (!(refdst & SKB_DST_NOREF))
287 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
291 * skb_dst_drop - drops skb dst
294 * Drops dst reference count if a reference was taken.
296 static inline void skb_dst_drop(struct sk_buff *skb)
298 if (skb->_skb_refdst) {
299 refdst_drop(skb->_skb_refdst);
300 skb->_skb_refdst = 0UL;
304 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
306 nskb->_skb_refdst = refdst;
307 if (!(nskb->_skb_refdst & SKB_DST_NOREF))
308 dst_clone(skb_dst(nskb));
311 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
313 __skb_dst_copy(nskb, oskb->_skb_refdst);
317 * skb_dst_force - makes sure skb dst is refcounted
320 * If dst is not yet refcounted, let's do it
322 static inline void skb_dst_force(struct sk_buff *skb)
324 if (skb_dst_is_noref(skb)) {
325 WARN_ON(!rcu_read_lock_held());
326 skb->_skb_refdst &= ~SKB_DST_NOREF;
327 dst_clone(skb_dst(skb));
332 * dst_hold_safe - Take a reference on a dst if possible
333 * @dst: pointer to dst entry
335 * This helper returns false if it could not safely
336 * take a reference on a dst.
338 static inline bool dst_hold_safe(struct dst_entry *dst)
340 if (dst->flags & DST_NOCACHE)
341 return atomic_inc_not_zero(&dst->__refcnt);
347 * skb_dst_force_safe - makes sure skb dst is refcounted
350 * If dst is not yet refcounted and not destroyed, grab a ref on it.
352 static inline void skb_dst_force_safe(struct sk_buff *skb)
354 if (skb_dst_is_noref(skb)) {
355 struct dst_entry *dst = skb_dst(skb);
357 if (!dst_hold_safe(dst))
360 skb->_skb_refdst = (unsigned long)dst;
366 * __skb_tunnel_rx - prepare skb for rx reinsert
368 * @dev: tunnel device
369 * @net: netns for packet i/o
371 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
372 * so make some cleanups. (no accounting done)
374 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
380 * Clear hash so that we can recalulate the hash for the
381 * encapsulated packet, unless we have already determine the hash
382 * over the L4 4-tuple.
384 skb_clear_hash_if_not_l4(skb);
385 skb_set_queue_mapping(skb, 0);
386 skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
390 * skb_tunnel_rx - prepare skb for rx reinsert
392 * @dev: tunnel device
394 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
395 * so make some cleanups, and perform accounting.
396 * Note: this accounting is not SMP safe.
398 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
401 /* TODO : stats should be SMP safe */
402 dev->stats.rx_packets++;
403 dev->stats.rx_bytes += skb->len;
404 __skb_tunnel_rx(skb, dev, net);
407 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
408 static inline int dst_discard(struct sk_buff *skb)
410 return dst_discard_out(&init_net, skb->sk, skb);
412 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
413 int initial_obsolete, unsigned short flags);
414 void dst_init(struct dst_entry *dst, struct dst_ops *ops,
415 struct net_device *dev, int initial_ref, int initial_obsolete,
416 unsigned short flags);
417 void __dst_free(struct dst_entry *dst);
418 struct dst_entry *dst_destroy(struct dst_entry *dst);
420 static inline void dst_free(struct dst_entry *dst)
422 if (dst->obsolete > 0)
424 if (!atomic_read(&dst->__refcnt)) {
425 dst = dst_destroy(dst);
432 static inline void dst_rcu_free(struct rcu_head *head)
434 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
438 static inline void dst_confirm(struct dst_entry *dst)
440 dst->pending_confirm = 1;
443 static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n,
446 const struct hh_cache *hh;
448 if (dst->pending_confirm) {
449 unsigned long now = jiffies;
451 dst->pending_confirm = 0;
452 /* avoid dirtying neighbour */
453 if (n->confirmed != now)
458 if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
459 return neigh_hh_output(hh, skb);
461 return n->output(n, skb);
464 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
466 struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
467 return IS_ERR(n) ? NULL : n;
470 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
473 struct neighbour *n = NULL;
475 /* The packets from tunnel devices (eg bareudp) may have only
476 * metadata in the dst pointer of skb. Hence a pointer check of
477 * neigh_lookup is needed.
479 if (dst->ops->neigh_lookup)
480 n = dst->ops->neigh_lookup(dst, skb, NULL);
482 return IS_ERR(n) ? NULL : n;
485 static inline void dst_link_failure(struct sk_buff *skb)
487 struct dst_entry *dst = skb_dst(skb);
488 if (dst && dst->ops && dst->ops->link_failure)
489 dst->ops->link_failure(skb);
492 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
494 unsigned long expires = jiffies + timeout;
499 if (dst->expires == 0 || time_before(expires, dst->expires))
500 dst->expires = expires;
503 /* Output packet to network from transport. */
504 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
506 return skb_dst(skb)->output(net, sk, skb);
509 /* Input packet from network to transport. */
510 static inline int dst_input(struct sk_buff *skb)
512 return skb_dst(skb)->input(skb);
515 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
518 dst = dst->ops->check(dst, cookie);
522 void dst_subsys_init(void);
524 /* Flags for xfrm_lookup flags argument. */
526 XFRM_LOOKUP_ICMP = 1 << 0,
527 XFRM_LOOKUP_QUEUE = 1 << 1,
528 XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
533 static inline struct dst_entry *xfrm_lookup(struct net *net,
534 struct dst_entry *dst_orig,
535 const struct flowi *fl,
536 const struct sock *sk,
542 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
543 struct dst_entry *dst_orig,
544 const struct flowi *fl,
545 const struct sock *sk,
551 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
557 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
558 const struct flowi *fl, const struct sock *sk,
561 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
562 const struct flowi *fl, const struct sock *sk,
565 /* skb attached with this dst needs transformation if dst->xfrm is valid */
566 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
572 #endif /* _NET_DST_H */