xref: /linux/include/net/dst.h (revision 83a37b3292f4aca799b355179ad6fbdd78a08e10)
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 <linux/refcount.h>
18 #include <net/neighbour.h>
19 #include <asm/processor.h>
20 
21 #define DST_GC_MIN	(HZ/10)
22 #define DST_GC_INC	(HZ/2)
23 #define DST_GC_MAX	(120*HZ)
24 
25 /* Each dst_entry has reference count and sits in some parent list(s).
26  * When it is removed from parent list, it is "freed" (dst_free).
27  * After this it enters dead state (dst->obsolete > 0) and if its refcnt
28  * is zero, it can be destroyed immediately, otherwise it is added
29  * to gc list and garbage collector periodically checks the refcnt.
30  */
31 
32 struct sk_buff;
33 
34 struct dst_entry {
35 	struct net_device       *dev;
36 	struct rcu_head		rcu_head;
37 	struct dst_entry	*child;
38 	struct  dst_ops	        *ops;
39 	unsigned long		_metrics;
40 	unsigned long           expires;
41 	struct dst_entry	*path;
42 	struct dst_entry	*from;
43 #ifdef CONFIG_XFRM
44 	struct xfrm_state	*xfrm;
45 #else
46 	void			*__pad1;
47 #endif
48 	int			(*input)(struct sk_buff *);
49 	int			(*output)(struct net *net, struct sock *sk, struct sk_buff *skb);
50 
51 	unsigned short		flags;
52 #define DST_HOST		0x0001
53 #define DST_NOXFRM		0x0002
54 #define DST_NOPOLICY		0x0004
55 #define DST_NOCOUNT		0x0008
56 #define DST_FAKE_RTABLE		0x0010
57 #define DST_XFRM_TUNNEL		0x0020
58 #define DST_XFRM_QUEUE		0x0040
59 #define DST_METADATA		0x0080
60 
61 	short			error;
62 
63 	/* A non-zero value of dst->obsolete forces by-hand validation
64 	 * of the route entry.  Positive values are set by the generic
65 	 * dst layer to indicate that the entry has been forcefully
66 	 * destroyed.
67 	 *
68 	 * Negative values are used by the implementation layer code to
69 	 * force invocation of the dst_ops->check() method.
70 	 */
71 	short			obsolete;
72 #define DST_OBSOLETE_NONE	0
73 #define DST_OBSOLETE_DEAD	2
74 #define DST_OBSOLETE_FORCE_CHK	-1
75 #define DST_OBSOLETE_KILL	-2
76 	unsigned short		header_len;	/* more space at head required */
77 	unsigned short		trailer_len;	/* space to reserve at tail */
78 	unsigned short		__pad3;
79 
80 #ifdef CONFIG_IP_ROUTE_CLASSID
81 	__u32			tclassid;
82 #else
83 	__u32			__pad2;
84 #endif
85 
86 #ifdef CONFIG_64BIT
87 	/*
88 	 * Align __refcnt to a 64 bytes alignment
89 	 * (L1_CACHE_SIZE would be too much)
90 	 */
91 	long			__pad_to_align_refcnt[2];
92 #endif
93 	/*
94 	 * __refcnt wants to be on a different cache line from
95 	 * input/output/ops or performance tanks badly
96 	 */
97 	atomic_t		__refcnt;	/* client references	*/
98 	int			__use;
99 	unsigned long		lastuse;
100 	struct lwtunnel_state   *lwtstate;
101 	union {
102 		struct dst_entry	*next;
103 		struct rtable __rcu	*rt_next;
104 		struct rt6_info __rcu	*rt6_next;
105 		struct dn_route __rcu	*dn_next;
106 	};
107 };
108 
109 struct dst_metrics {
110 	u32		metrics[RTAX_MAX];
111 	refcount_t	refcnt;
112 };
113 extern const struct dst_metrics dst_default_metrics;
114 
115 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
116 
117 #define DST_METRICS_READ_ONLY		0x1UL
118 #define DST_METRICS_REFCOUNTED		0x2UL
119 #define DST_METRICS_FLAGS		0x3UL
120 #define __DST_METRICS_PTR(Y)	\
121 	((u32 *)((Y) & ~DST_METRICS_FLAGS))
122 #define DST_METRICS_PTR(X)	__DST_METRICS_PTR((X)->_metrics)
123 
124 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
125 {
126 	return dst->_metrics & DST_METRICS_READ_ONLY;
127 }
128 
129 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
130 
131 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
132 {
133 	unsigned long val = dst->_metrics;
134 	if (!(val & DST_METRICS_READ_ONLY))
135 		__dst_destroy_metrics_generic(dst, val);
136 }
137 
138 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
139 {
140 	unsigned long p = dst->_metrics;
141 
142 	BUG_ON(!p);
143 
144 	if (p & DST_METRICS_READ_ONLY)
145 		return dst->ops->cow_metrics(dst, p);
146 	return __DST_METRICS_PTR(p);
147 }
148 
149 /* This may only be invoked before the entry has reached global
150  * visibility.
151  */
152 static inline void dst_init_metrics(struct dst_entry *dst,
153 				    const u32 *src_metrics,
154 				    bool read_only)
155 {
156 	dst->_metrics = ((unsigned long) src_metrics) |
157 		(read_only ? DST_METRICS_READ_ONLY : 0);
158 }
159 
160 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
161 {
162 	u32 *dst_metrics = dst_metrics_write_ptr(dest);
163 
164 	if (dst_metrics) {
165 		u32 *src_metrics = DST_METRICS_PTR(src);
166 
167 		memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
168 	}
169 }
170 
171 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
172 {
173 	return DST_METRICS_PTR(dst);
174 }
175 
176 static inline u32
177 dst_metric_raw(const struct dst_entry *dst, const int metric)
178 {
179 	u32 *p = DST_METRICS_PTR(dst);
180 
181 	return p[metric-1];
182 }
183 
184 static inline u32
185 dst_metric(const struct dst_entry *dst, const int metric)
186 {
187 	WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
188 		     metric == RTAX_ADVMSS ||
189 		     metric == RTAX_MTU);
190 	return dst_metric_raw(dst, metric);
191 }
192 
193 static inline u32
194 dst_metric_advmss(const struct dst_entry *dst)
195 {
196 	u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
197 
198 	if (!advmss)
199 		advmss = dst->ops->default_advmss(dst);
200 
201 	return advmss;
202 }
203 
204 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
205 {
206 	u32 *p = dst_metrics_write_ptr(dst);
207 
208 	if (p)
209 		p[metric-1] = val;
210 }
211 
212 /* Kernel-internal feature bits that are unallocated in user space. */
213 #define DST_FEATURE_ECN_CA	(1 << 31)
214 
215 #define DST_FEATURE_MASK	(DST_FEATURE_ECN_CA)
216 #define DST_FEATURE_ECN_MASK	(DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
217 
218 static inline u32
219 dst_feature(const struct dst_entry *dst, u32 feature)
220 {
221 	return dst_metric(dst, RTAX_FEATURES) & feature;
222 }
223 
224 static inline u32 dst_mtu(const struct dst_entry *dst)
225 {
226 	return dst->ops->mtu(dst);
227 }
228 
229 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
230 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
231 {
232 	return msecs_to_jiffies(dst_metric(dst, metric));
233 }
234 
235 static inline u32
236 dst_allfrag(const struct dst_entry *dst)
237 {
238 	int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
239 	return ret;
240 }
241 
242 static inline int
243 dst_metric_locked(const struct dst_entry *dst, int metric)
244 {
245 	return dst_metric(dst, RTAX_LOCK) & (1<<metric);
246 }
247 
248 static inline void dst_hold(struct dst_entry *dst)
249 {
250 	/*
251 	 * If your kernel compilation stops here, please check
252 	 * __pad_to_align_refcnt declaration in struct dst_entry
253 	 */
254 	BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
255 	WARN_ON(atomic_inc_not_zero(&dst->__refcnt) == 0);
256 }
257 
258 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
259 {
260 	if (time != dst->lastuse) {
261 		dst->__use++;
262 		dst->lastuse = time;
263 	}
264 }
265 
266 static inline void dst_hold_and_use(struct dst_entry *dst, unsigned long time)
267 {
268 	dst_hold(dst);
269 	dst_use_noref(dst, time);
270 }
271 
272 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
273 {
274 	if (dst)
275 		dst_hold(dst);
276 	return dst;
277 }
278 
279 void dst_release(struct dst_entry *dst);
280 
281 void dst_release_immediate(struct dst_entry *dst);
282 
283 static inline void refdst_drop(unsigned long refdst)
284 {
285 	if (!(refdst & SKB_DST_NOREF))
286 		dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
287 }
288 
289 /**
290  * skb_dst_drop - drops skb dst
291  * @skb: buffer
292  *
293  * Drops dst reference count if a reference was taken.
294  */
295 static inline void skb_dst_drop(struct sk_buff *skb)
296 {
297 	if (skb->_skb_refdst) {
298 		refdst_drop(skb->_skb_refdst);
299 		skb->_skb_refdst = 0UL;
300 	}
301 }
302 
303 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
304 {
305 	nskb->_skb_refdst = refdst;
306 	if (!(nskb->_skb_refdst & SKB_DST_NOREF))
307 		dst_clone(skb_dst(nskb));
308 }
309 
310 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
311 {
312 	__skb_dst_copy(nskb, oskb->_skb_refdst);
313 }
314 
315 /**
316  * dst_hold_safe - Take a reference on a dst if possible
317  * @dst: pointer to dst entry
318  *
319  * This helper returns false if it could not safely
320  * take a reference on a dst.
321  */
322 static inline bool dst_hold_safe(struct dst_entry *dst)
323 {
324 	return atomic_inc_not_zero(&dst->__refcnt);
325 }
326 
327 /**
328  * skb_dst_force - makes sure skb dst is refcounted
329  * @skb: buffer
330  *
331  * If dst is not yet refcounted and not destroyed, grab a ref on it.
332  */
333 static inline void skb_dst_force(struct sk_buff *skb)
334 {
335 	if (skb_dst_is_noref(skb)) {
336 		struct dst_entry *dst = skb_dst(skb);
337 
338 		WARN_ON(!rcu_read_lock_held());
339 		if (!dst_hold_safe(dst))
340 			dst = NULL;
341 
342 		skb->_skb_refdst = (unsigned long)dst;
343 	}
344 }
345 
346 
347 /**
348  *	__skb_tunnel_rx - prepare skb for rx reinsert
349  *	@skb: buffer
350  *	@dev: tunnel device
351  *	@net: netns for packet i/o
352  *
353  *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
354  *	so make some cleanups. (no accounting done)
355  */
356 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
357 				   struct net *net)
358 {
359 	skb->dev = dev;
360 
361 	/*
362 	 * Clear hash so that we can recalulate the hash for the
363 	 * encapsulated packet, unless we have already determine the hash
364 	 * over the L4 4-tuple.
365 	 */
366 	skb_clear_hash_if_not_l4(skb);
367 	skb_set_queue_mapping(skb, 0);
368 	skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
369 }
370 
371 /**
372  *	skb_tunnel_rx - prepare skb for rx reinsert
373  *	@skb: buffer
374  *	@dev: tunnel device
375  *
376  *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
377  *	so make some cleanups, and perform accounting.
378  *	Note: this accounting is not SMP safe.
379  */
380 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
381 				 struct net *net)
382 {
383 	/* TODO : stats should be SMP safe */
384 	dev->stats.rx_packets++;
385 	dev->stats.rx_bytes += skb->len;
386 	__skb_tunnel_rx(skb, dev, net);
387 }
388 
389 static inline u32 dst_tclassid(const struct sk_buff *skb)
390 {
391 #ifdef CONFIG_IP_ROUTE_CLASSID
392 	const struct dst_entry *dst;
393 
394 	dst = skb_dst(skb);
395 	if (dst)
396 		return dst->tclassid;
397 #endif
398 	return 0;
399 }
400 
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)
403 {
404 	return dst_discard_out(&init_net, skb->sk, skb);
405 }
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 struct dst_entry *dst_destroy(struct dst_entry *dst);
412 void dst_dev_put(struct dst_entry *dst);
413 
414 static inline void dst_confirm(struct dst_entry *dst)
415 {
416 }
417 
418 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
419 {
420 	struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
421 	return IS_ERR(n) ? NULL : n;
422 }
423 
424 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
425 						     struct sk_buff *skb)
426 {
427 	struct neighbour *n =  dst->ops->neigh_lookup(dst, skb, NULL);
428 	return IS_ERR(n) ? NULL : n;
429 }
430 
431 static inline void dst_confirm_neigh(const struct dst_entry *dst,
432 				     const void *daddr)
433 {
434 	if (dst->ops->confirm_neigh)
435 		dst->ops->confirm_neigh(dst, daddr);
436 }
437 
438 static inline void dst_link_failure(struct sk_buff *skb)
439 {
440 	struct dst_entry *dst = skb_dst(skb);
441 	if (dst && dst->ops && dst->ops->link_failure)
442 		dst->ops->link_failure(skb);
443 }
444 
445 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
446 {
447 	unsigned long expires = jiffies + timeout;
448 
449 	if (expires == 0)
450 		expires = 1;
451 
452 	if (dst->expires == 0 || time_before(expires, dst->expires))
453 		dst->expires = expires;
454 }
455 
456 /* Output packet to network from transport.  */
457 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
458 {
459 	return skb_dst(skb)->output(net, sk, skb);
460 }
461 
462 /* Input packet from network to transport.  */
463 static inline int dst_input(struct sk_buff *skb)
464 {
465 	return skb_dst(skb)->input(skb);
466 }
467 
468 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
469 {
470 	if (dst->obsolete)
471 		dst = dst->ops->check(dst, cookie);
472 	return dst;
473 }
474 
475 /* Flags for xfrm_lookup flags argument. */
476 enum {
477 	XFRM_LOOKUP_ICMP = 1 << 0,
478 	XFRM_LOOKUP_QUEUE = 1 << 1,
479 	XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
480 };
481 
482 struct flowi;
483 #ifndef CONFIG_XFRM
484 static inline struct dst_entry *xfrm_lookup(struct net *net,
485 					    struct dst_entry *dst_orig,
486 					    const struct flowi *fl,
487 					    const struct sock *sk,
488 					    int flags)
489 {
490 	return dst_orig;
491 }
492 
493 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
494 						  struct dst_entry *dst_orig,
495 						  const struct flowi *fl,
496 						  const struct sock *sk,
497 						  int flags)
498 {
499 	return dst_orig;
500 }
501 
502 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
503 {
504 	return NULL;
505 }
506 
507 #else
508 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
509 			      const struct flowi *fl, const struct sock *sk,
510 			      int flags);
511 
512 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
513 				    const struct flowi *fl, const struct sock *sk,
514 				    int flags);
515 
516 /* skb attached with this dst needs transformation if dst->xfrm is valid */
517 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
518 {
519 	return dst->xfrm;
520 }
521 #endif
522 
523 #endif /* _NET_DST_H */
524