xref: /linux/include/net/dst.h (revision e58e871becec2d3b04ed91c0c16fe8deac9c9dfa)
1 /*
2  * net/dst.h	Protocol independent destination cache definitions.
3  *
4  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
5  *
6  */
7 
8 #ifndef _NET_DST_H
9 #define _NET_DST_H
10 
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>
19 
20 #define DST_GC_MIN	(HZ/10)
21 #define DST_GC_INC	(HZ/2)
22 #define DST_GC_MAX	(120*HZ)
23 
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.
29  */
30 
31 struct sk_buff;
32 
33 struct dst_entry {
34 	struct rcu_head		rcu_head;
35 	struct dst_entry	*child;
36 	struct net_device       *dev;
37 	struct  dst_ops	        *ops;
38 	unsigned long		_metrics;
39 	unsigned long           expires;
40 	struct dst_entry	*path;
41 	struct dst_entry	*from;
42 #ifdef CONFIG_XFRM
43 	struct xfrm_state	*xfrm;
44 #else
45 	void			*__pad1;
46 #endif
47 	int			(*input)(struct sk_buff *);
48 	int			(*output)(struct net *net, struct sock *sk, struct sk_buff *skb);
49 
50 	unsigned short		flags;
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
61 
62 	short			error;
63 
64 	/* A non-zero value of dst->obsolete forces by-hand validation
65 	 * of the route entry.  Positive values are set by the generic
66 	 * dst layer to indicate that the entry has been forcefully
67 	 * destroyed.
68 	 *
69 	 * Negative values are used by the implementation layer code to
70 	 * force invocation of the dst_ops->check() method.
71 	 */
72 	short			obsolete;
73 #define DST_OBSOLETE_NONE	0
74 #define DST_OBSOLETE_DEAD	2
75 #define DST_OBSOLETE_FORCE_CHK	-1
76 #define DST_OBSOLETE_KILL	-2
77 	unsigned short		header_len;	/* more space at head required */
78 	unsigned short		trailer_len;	/* space to reserve at tail */
79 	unsigned short		__pad3;
80 
81 #ifdef CONFIG_IP_ROUTE_CLASSID
82 	__u32			tclassid;
83 #else
84 	__u32			__pad2;
85 #endif
86 
87 #ifdef CONFIG_64BIT
88 	/*
89 	 * Align __refcnt to a 64 bytes alignment
90 	 * (L1_CACHE_SIZE would be too much)
91 	 */
92 	long			__pad_to_align_refcnt[2];
93 #endif
94 	/*
95 	 * __refcnt wants to be on a different cache line from
96 	 * input/output/ops or performance tanks badly
97 	 */
98 	atomic_t		__refcnt;	/* client references	*/
99 	int			__use;
100 	unsigned long		lastuse;
101 	struct lwtunnel_state   *lwtstate;
102 	union {
103 		struct dst_entry	*next;
104 		struct rtable __rcu	*rt_next;
105 		struct rt6_info		*rt6_next;
106 		struct dn_route __rcu	*dn_next;
107 	};
108 };
109 
110 struct dst_metrics {
111 	u32		metrics[RTAX_MAX];
112 	atomic_t	refcnt;
113 };
114 extern const struct dst_metrics dst_default_metrics;
115 
116 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
117 
118 #define DST_METRICS_READ_ONLY		0x1UL
119 #define DST_METRICS_REFCOUNTED		0x2UL
120 #define DST_METRICS_FLAGS		0x3UL
121 #define __DST_METRICS_PTR(Y)	\
122 	((u32 *)((Y) & ~DST_METRICS_FLAGS))
123 #define DST_METRICS_PTR(X)	__DST_METRICS_PTR((X)->_metrics)
124 
125 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
126 {
127 	return dst->_metrics & DST_METRICS_READ_ONLY;
128 }
129 
130 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
131 
132 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
133 {
134 	unsigned long val = dst->_metrics;
135 	if (!(val & DST_METRICS_READ_ONLY))
136 		__dst_destroy_metrics_generic(dst, val);
137 }
138 
139 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
140 {
141 	unsigned long p = dst->_metrics;
142 
143 	BUG_ON(!p);
144 
145 	if (p & DST_METRICS_READ_ONLY)
146 		return dst->ops->cow_metrics(dst, p);
147 	return __DST_METRICS_PTR(p);
148 }
149 
150 /* This may only be invoked before the entry has reached global
151  * visibility.
152  */
153 static inline void dst_init_metrics(struct dst_entry *dst,
154 				    const u32 *src_metrics,
155 				    bool read_only)
156 {
157 	dst->_metrics = ((unsigned long) src_metrics) |
158 		(read_only ? DST_METRICS_READ_ONLY : 0);
159 }
160 
161 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
162 {
163 	u32 *dst_metrics = dst_metrics_write_ptr(dest);
164 
165 	if (dst_metrics) {
166 		u32 *src_metrics = DST_METRICS_PTR(src);
167 
168 		memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
169 	}
170 }
171 
172 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
173 {
174 	return DST_METRICS_PTR(dst);
175 }
176 
177 static inline u32
178 dst_metric_raw(const struct dst_entry *dst, const int metric)
179 {
180 	u32 *p = DST_METRICS_PTR(dst);
181 
182 	return p[metric-1];
183 }
184 
185 static inline u32
186 dst_metric(const struct dst_entry *dst, const int metric)
187 {
188 	WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
189 		     metric == RTAX_ADVMSS ||
190 		     metric == RTAX_MTU);
191 	return dst_metric_raw(dst, metric);
192 }
193 
194 static inline u32
195 dst_metric_advmss(const struct dst_entry *dst)
196 {
197 	u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
198 
199 	if (!advmss)
200 		advmss = dst->ops->default_advmss(dst);
201 
202 	return advmss;
203 }
204 
205 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
206 {
207 	u32 *p = dst_metrics_write_ptr(dst);
208 
209 	if (p)
210 		p[metric-1] = val;
211 }
212 
213 /* Kernel-internal feature bits that are unallocated in user space. */
214 #define DST_FEATURE_ECN_CA	(1 << 31)
215 
216 #define DST_FEATURE_MASK	(DST_FEATURE_ECN_CA)
217 #define DST_FEATURE_ECN_MASK	(DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
218 
219 static inline u32
220 dst_feature(const struct dst_entry *dst, u32 feature)
221 {
222 	return dst_metric(dst, RTAX_FEATURES) & feature;
223 }
224 
225 static inline u32 dst_mtu(const struct dst_entry *dst)
226 {
227 	return dst->ops->mtu(dst);
228 }
229 
230 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
231 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
232 {
233 	return msecs_to_jiffies(dst_metric(dst, metric));
234 }
235 
236 static inline u32
237 dst_allfrag(const struct dst_entry *dst)
238 {
239 	int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
240 	return ret;
241 }
242 
243 static inline int
244 dst_metric_locked(const struct dst_entry *dst, int metric)
245 {
246 	return dst_metric(dst, RTAX_LOCK) & (1<<metric);
247 }
248 
249 static inline void dst_hold(struct dst_entry *dst)
250 {
251 	/*
252 	 * If your kernel compilation stops here, please check
253 	 * __pad_to_align_refcnt declaration in struct dst_entry
254 	 */
255 	BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
256 	atomic_inc(&dst->__refcnt);
257 }
258 
259 static inline void dst_use(struct dst_entry *dst, unsigned long time)
260 {
261 	dst_hold(dst);
262 	dst->__use++;
263 	dst->lastuse = time;
264 }
265 
266 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
267 {
268 	dst->__use++;
269 	dst->lastuse = time;
270 }
271 
272 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
273 {
274 	if (dst)
275 		atomic_inc(&dst->__refcnt);
276 	return dst;
277 }
278 
279 void dst_release(struct dst_entry *dst);
280 
281 static inline void refdst_drop(unsigned long refdst)
282 {
283 	if (!(refdst & SKB_DST_NOREF))
284 		dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
285 }
286 
287 /**
288  * skb_dst_drop - drops skb dst
289  * @skb: buffer
290  *
291  * Drops dst reference count if a reference was taken.
292  */
293 static inline void skb_dst_drop(struct sk_buff *skb)
294 {
295 	if (skb->_skb_refdst) {
296 		refdst_drop(skb->_skb_refdst);
297 		skb->_skb_refdst = 0UL;
298 	}
299 }
300 
301 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
302 {
303 	nskb->_skb_refdst = refdst;
304 	if (!(nskb->_skb_refdst & SKB_DST_NOREF))
305 		dst_clone(skb_dst(nskb));
306 }
307 
308 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
309 {
310 	__skb_dst_copy(nskb, oskb->_skb_refdst);
311 }
312 
313 /**
314  * skb_dst_force - makes sure skb dst is refcounted
315  * @skb: buffer
316  *
317  * If dst is not yet refcounted, let's do it
318  */
319 static inline void skb_dst_force(struct sk_buff *skb)
320 {
321 	if (skb_dst_is_noref(skb)) {
322 		WARN_ON(!rcu_read_lock_held());
323 		skb->_skb_refdst &= ~SKB_DST_NOREF;
324 		dst_clone(skb_dst(skb));
325 	}
326 }
327 
328 /**
329  * dst_hold_safe - Take a reference on a dst if possible
330  * @dst: pointer to dst entry
331  *
332  * This helper returns false if it could not safely
333  * take a reference on a dst.
334  */
335 static inline bool dst_hold_safe(struct dst_entry *dst)
336 {
337 	if (dst->flags & DST_NOCACHE)
338 		return atomic_inc_not_zero(&dst->__refcnt);
339 	dst_hold(dst);
340 	return true;
341 }
342 
343 /**
344  * skb_dst_force_safe - makes sure skb dst is refcounted
345  * @skb: buffer
346  *
347  * If dst is not yet refcounted and not destroyed, grab a ref on it.
348  */
349 static inline void skb_dst_force_safe(struct sk_buff *skb)
350 {
351 	if (skb_dst_is_noref(skb)) {
352 		struct dst_entry *dst = skb_dst(skb);
353 
354 		if (!dst_hold_safe(dst))
355 			dst = NULL;
356 
357 		skb->_skb_refdst = (unsigned long)dst;
358 	}
359 }
360 
361 
362 /**
363  *	__skb_tunnel_rx - prepare skb for rx reinsert
364  *	@skb: buffer
365  *	@dev: tunnel device
366  *	@net: netns for packet i/o
367  *
368  *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
369  *	so make some cleanups. (no accounting done)
370  */
371 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
372 				   struct net *net)
373 {
374 	skb->dev = dev;
375 
376 	/*
377 	 * Clear hash so that we can recalulate the hash for the
378 	 * encapsulated packet, unless we have already determine the hash
379 	 * over the L4 4-tuple.
380 	 */
381 	skb_clear_hash_if_not_l4(skb);
382 	skb_set_queue_mapping(skb, 0);
383 	skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
384 }
385 
386 /**
387  *	skb_tunnel_rx - prepare skb for rx reinsert
388  *	@skb: buffer
389  *	@dev: tunnel device
390  *
391  *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
392  *	so make some cleanups, and perform accounting.
393  *	Note: this accounting is not SMP safe.
394  */
395 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
396 				 struct net *net)
397 {
398 	/* TODO : stats should be SMP safe */
399 	dev->stats.rx_packets++;
400 	dev->stats.rx_bytes += skb->len;
401 	__skb_tunnel_rx(skb, dev, net);
402 }
403 
404 static inline u32 dst_tclassid(const struct sk_buff *skb)
405 {
406 #ifdef CONFIG_IP_ROUTE_CLASSID
407 	const struct dst_entry *dst;
408 
409 	dst = skb_dst(skb);
410 	if (dst)
411 		return dst->tclassid;
412 #endif
413 	return 0;
414 }
415 
416 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
417 static inline int dst_discard(struct sk_buff *skb)
418 {
419 	return dst_discard_out(&init_net, skb->sk, skb);
420 }
421 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
422 		int initial_obsolete, unsigned short flags);
423 void dst_init(struct dst_entry *dst, struct dst_ops *ops,
424 	      struct net_device *dev, int initial_ref, int initial_obsolete,
425 	      unsigned short flags);
426 void __dst_free(struct dst_entry *dst);
427 struct dst_entry *dst_destroy(struct dst_entry *dst);
428 
429 static inline void dst_free(struct dst_entry *dst)
430 {
431 	if (dst->obsolete > 0)
432 		return;
433 	if (!atomic_read(&dst->__refcnt)) {
434 		dst = dst_destroy(dst);
435 		if (!dst)
436 			return;
437 	}
438 	__dst_free(dst);
439 }
440 
441 static inline void dst_rcu_free(struct rcu_head *head)
442 {
443 	struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
444 	dst_free(dst);
445 }
446 
447 static inline void dst_confirm(struct dst_entry *dst)
448 {
449 }
450 
451 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
452 {
453 	struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
454 	return IS_ERR(n) ? NULL : n;
455 }
456 
457 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
458 						     struct sk_buff *skb)
459 {
460 	struct neighbour *n =  dst->ops->neigh_lookup(dst, skb, NULL);
461 	return IS_ERR(n) ? NULL : n;
462 }
463 
464 static inline void dst_confirm_neigh(const struct dst_entry *dst,
465 				     const void *daddr)
466 {
467 	if (dst->ops->confirm_neigh)
468 		dst->ops->confirm_neigh(dst, daddr);
469 }
470 
471 static inline void dst_link_failure(struct sk_buff *skb)
472 {
473 	struct dst_entry *dst = skb_dst(skb);
474 	if (dst && dst->ops && dst->ops->link_failure)
475 		dst->ops->link_failure(skb);
476 }
477 
478 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
479 {
480 	unsigned long expires = jiffies + timeout;
481 
482 	if (expires == 0)
483 		expires = 1;
484 
485 	if (dst->expires == 0 || time_before(expires, dst->expires))
486 		dst->expires = expires;
487 }
488 
489 /* Output packet to network from transport.  */
490 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
491 {
492 	return skb_dst(skb)->output(net, sk, skb);
493 }
494 
495 /* Input packet from network to transport.  */
496 static inline int dst_input(struct sk_buff *skb)
497 {
498 	return skb_dst(skb)->input(skb);
499 }
500 
501 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
502 {
503 	if (dst->obsolete)
504 		dst = dst->ops->check(dst, cookie);
505 	return dst;
506 }
507 
508 void dst_subsys_init(void);
509 
510 /* Flags for xfrm_lookup flags argument. */
511 enum {
512 	XFRM_LOOKUP_ICMP = 1 << 0,
513 	XFRM_LOOKUP_QUEUE = 1 << 1,
514 	XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
515 };
516 
517 struct flowi;
518 #ifndef CONFIG_XFRM
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,
523 					    int flags)
524 {
525 	return dst_orig;
526 }
527 
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,
532 						  int flags)
533 {
534 	return dst_orig;
535 }
536 
537 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
538 {
539 	return NULL;
540 }
541 
542 #else
543 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
544 			      const struct flowi *fl, const struct sock *sk,
545 			      int flags);
546 
547 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
548 				    const struct flowi *fl, const struct sock *sk,
549 				    int flags);
550 
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)
553 {
554 	return dst->xfrm;
555 }
556 #endif
557 
558 #endif /* _NET_DST_H */
559