xref: /openbmc/linux/include/net/dst.h (revision bbde9fc1824aab58bc78c084163007dd6c03fe5b)
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 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 	unsigned short		pending_confirm;
63 
64 	short			error;
65 
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
69 	 * destroyed.
70 	 *
71 	 * Negative values are used by the implementation layer code to
72 	 * force invocation of the dst_ops->check() method.
73 	 */
74 	short			obsolete;
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
82 	__u32			tclassid;
83 #else
84 	__u32			__pad2;
85 #endif
86 
87 	/*
88 	 * Align __refcnt to a 64 bytes alignment
89 	 * (L1_CACHE_SIZE would be too much)
90 	 */
91 #ifdef CONFIG_64BIT
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 	union {
102 		struct dst_entry	*next;
103 		struct rtable __rcu	*rt_next;
104 		struct rt6_info		*rt6_next;
105 		struct dn_route __rcu	*dn_next;
106 	};
107 };
108 
109 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
110 extern const u32 dst_default_metrics[];
111 
112 #define DST_METRICS_READ_ONLY		0x1UL
113 #define DST_METRICS_FLAGS		0x3UL
114 #define __DST_METRICS_PTR(Y)	\
115 	((u32 *)((Y) & ~DST_METRICS_FLAGS))
116 #define DST_METRICS_PTR(X)	__DST_METRICS_PTR((X)->_metrics)
117 
118 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
119 {
120 	return dst->_metrics & DST_METRICS_READ_ONLY;
121 }
122 
123 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
124 
125 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
126 {
127 	unsigned long val = dst->_metrics;
128 	if (!(val & DST_METRICS_READ_ONLY))
129 		__dst_destroy_metrics_generic(dst, val);
130 }
131 
132 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
133 {
134 	unsigned long p = dst->_metrics;
135 
136 	BUG_ON(!p);
137 
138 	if (p & DST_METRICS_READ_ONLY)
139 		return dst->ops->cow_metrics(dst, p);
140 	return __DST_METRICS_PTR(p);
141 }
142 
143 /* This may only be invoked before the entry has reached global
144  * visibility.
145  */
146 static inline void dst_init_metrics(struct dst_entry *dst,
147 				    const u32 *src_metrics,
148 				    bool read_only)
149 {
150 	dst->_metrics = ((unsigned long) src_metrics) |
151 		(read_only ? DST_METRICS_READ_ONLY : 0);
152 }
153 
154 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
155 {
156 	u32 *dst_metrics = dst_metrics_write_ptr(dest);
157 
158 	if (dst_metrics) {
159 		u32 *src_metrics = DST_METRICS_PTR(src);
160 
161 		memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
162 	}
163 }
164 
165 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
166 {
167 	return DST_METRICS_PTR(dst);
168 }
169 
170 static inline u32
171 dst_metric_raw(const struct dst_entry *dst, const int metric)
172 {
173 	u32 *p = DST_METRICS_PTR(dst);
174 
175 	return p[metric-1];
176 }
177 
178 static inline u32
179 dst_metric(const struct dst_entry *dst, const int metric)
180 {
181 	WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
182 		     metric == RTAX_ADVMSS ||
183 		     metric == RTAX_MTU);
184 	return dst_metric_raw(dst, metric);
185 }
186 
187 static inline u32
188 dst_metric_advmss(const struct dst_entry *dst)
189 {
190 	u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
191 
192 	if (!advmss)
193 		advmss = dst->ops->default_advmss(dst);
194 
195 	return advmss;
196 }
197 
198 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
199 {
200 	u32 *p = dst_metrics_write_ptr(dst);
201 
202 	if (p)
203 		p[metric-1] = val;
204 }
205 
206 static inline u32
207 dst_feature(const struct dst_entry *dst, u32 feature)
208 {
209 	return dst_metric(dst, RTAX_FEATURES) & feature;
210 }
211 
212 static inline u32 dst_mtu(const struct dst_entry *dst)
213 {
214 	return dst->ops->mtu(dst);
215 }
216 
217 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
218 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
219 {
220 	return msecs_to_jiffies(dst_metric(dst, metric));
221 }
222 
223 static inline u32
224 dst_allfrag(const struct dst_entry *dst)
225 {
226 	int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
227 	return ret;
228 }
229 
230 static inline int
231 dst_metric_locked(const struct dst_entry *dst, int metric)
232 {
233 	return dst_metric(dst, RTAX_LOCK) & (1<<metric);
234 }
235 
236 static inline void dst_hold(struct dst_entry *dst)
237 {
238 	/*
239 	 * If your kernel compilation stops here, please check
240 	 * __pad_to_align_refcnt declaration in struct dst_entry
241 	 */
242 	BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
243 	atomic_inc(&dst->__refcnt);
244 }
245 
246 static inline void dst_use(struct dst_entry *dst, unsigned long time)
247 {
248 	dst_hold(dst);
249 	dst->__use++;
250 	dst->lastuse = time;
251 }
252 
253 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
254 {
255 	dst->__use++;
256 	dst->lastuse = time;
257 }
258 
259 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
260 {
261 	if (dst)
262 		atomic_inc(&dst->__refcnt);
263 	return dst;
264 }
265 
266 void dst_release(struct dst_entry *dst);
267 
268 static inline void refdst_drop(unsigned long refdst)
269 {
270 	if (!(refdst & SKB_DST_NOREF))
271 		dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
272 }
273 
274 /**
275  * skb_dst_drop - drops skb dst
276  * @skb: buffer
277  *
278  * Drops dst reference count if a reference was taken.
279  */
280 static inline void skb_dst_drop(struct sk_buff *skb)
281 {
282 	if (skb->_skb_refdst) {
283 		refdst_drop(skb->_skb_refdst);
284 		skb->_skb_refdst = 0UL;
285 	}
286 }
287 
288 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
289 {
290 	nskb->_skb_refdst = oskb->_skb_refdst;
291 	if (!(nskb->_skb_refdst & SKB_DST_NOREF))
292 		dst_clone(skb_dst(nskb));
293 }
294 
295 /**
296  * skb_dst_force - makes sure skb dst is refcounted
297  * @skb: buffer
298  *
299  * If dst is not yet refcounted, let's do it
300  */
301 static inline void skb_dst_force(struct sk_buff *skb)
302 {
303 	if (skb_dst_is_noref(skb)) {
304 		WARN_ON(!rcu_read_lock_held());
305 		skb->_skb_refdst &= ~SKB_DST_NOREF;
306 		dst_clone(skb_dst(skb));
307 	}
308 }
309 
310 
311 /**
312  *	__skb_tunnel_rx - prepare skb for rx reinsert
313  *	@skb: buffer
314  *	@dev: tunnel device
315  *	@net: netns for packet i/o
316  *
317  *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
318  *	so make some cleanups. (no accounting done)
319  */
320 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
321 				   struct net *net)
322 {
323 	skb->dev = dev;
324 
325 	/*
326 	 * Clear hash so that we can recalulate the hash for the
327 	 * encapsulated packet, unless we have already determine the hash
328 	 * over the L4 4-tuple.
329 	 */
330 	skb_clear_hash_if_not_l4(skb);
331 	skb_set_queue_mapping(skb, 0);
332 	skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
333 }
334 
335 /**
336  *	skb_tunnel_rx - prepare skb for rx reinsert
337  *	@skb: buffer
338  *	@dev: tunnel device
339  *
340  *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
341  *	so make some cleanups, and perform accounting.
342  *	Note: this accounting is not SMP safe.
343  */
344 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
345 				 struct net *net)
346 {
347 	/* TODO : stats should be SMP safe */
348 	dev->stats.rx_packets++;
349 	dev->stats.rx_bytes += skb->len;
350 	__skb_tunnel_rx(skb, dev, net);
351 }
352 
353 int dst_discard_sk(struct sock *sk, struct sk_buff *skb);
354 static inline int dst_discard(struct sk_buff *skb)
355 {
356 	return dst_discard_sk(skb->sk, skb);
357 }
358 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
359 		int initial_obsolete, unsigned short flags);
360 void dst_init(struct dst_entry *dst, struct dst_ops *ops,
361 	      struct net_device *dev, int initial_ref, int initial_obsolete,
362 	      unsigned short flags);
363 void __dst_free(struct dst_entry *dst);
364 struct dst_entry *dst_destroy(struct dst_entry *dst);
365 
366 static inline void dst_free(struct dst_entry *dst)
367 {
368 	if (dst->obsolete > 0)
369 		return;
370 	if (!atomic_read(&dst->__refcnt)) {
371 		dst = dst_destroy(dst);
372 		if (!dst)
373 			return;
374 	}
375 	__dst_free(dst);
376 }
377 
378 static inline void dst_rcu_free(struct rcu_head *head)
379 {
380 	struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
381 	dst_free(dst);
382 }
383 
384 static inline void dst_confirm(struct dst_entry *dst)
385 {
386 	dst->pending_confirm = 1;
387 }
388 
389 static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n,
390 				   struct sk_buff *skb)
391 {
392 	const struct hh_cache *hh;
393 
394 	if (dst->pending_confirm) {
395 		unsigned long now = jiffies;
396 
397 		dst->pending_confirm = 0;
398 		/* avoid dirtying neighbour */
399 		if (n->confirmed != now)
400 			n->confirmed = now;
401 	}
402 
403 	hh = &n->hh;
404 	if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
405 		return neigh_hh_output(hh, skb);
406 	else
407 		return n->output(n, skb);
408 }
409 
410 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
411 {
412 	struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
413 	return IS_ERR(n) ? NULL : n;
414 }
415 
416 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
417 						     struct sk_buff *skb)
418 {
419 	struct neighbour *n =  dst->ops->neigh_lookup(dst, skb, NULL);
420 	return IS_ERR(n) ? NULL : n;
421 }
422 
423 static inline void dst_link_failure(struct sk_buff *skb)
424 {
425 	struct dst_entry *dst = skb_dst(skb);
426 	if (dst && dst->ops && dst->ops->link_failure)
427 		dst->ops->link_failure(skb);
428 }
429 
430 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
431 {
432 	unsigned long expires = jiffies + timeout;
433 
434 	if (expires == 0)
435 		expires = 1;
436 
437 	if (dst->expires == 0 || time_before(expires, dst->expires))
438 		dst->expires = expires;
439 }
440 
441 /* Output packet to network from transport.  */
442 static inline int dst_output_sk(struct sock *sk, struct sk_buff *skb)
443 {
444 	return skb_dst(skb)->output(sk, skb);
445 }
446 static inline int dst_output(struct sk_buff *skb)
447 {
448 	return dst_output_sk(skb->sk, skb);
449 }
450 
451 /* Input packet from network to transport.  */
452 static inline int dst_input(struct sk_buff *skb)
453 {
454 	return skb_dst(skb)->input(skb);
455 }
456 
457 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
458 {
459 	if (dst->obsolete)
460 		dst = dst->ops->check(dst, cookie);
461 	return dst;
462 }
463 
464 void dst_subsys_init(void);
465 
466 /* Flags for xfrm_lookup flags argument. */
467 enum {
468 	XFRM_LOOKUP_ICMP = 1 << 0,
469 	XFRM_LOOKUP_QUEUE = 1 << 1,
470 	XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
471 };
472 
473 struct flowi;
474 #ifndef CONFIG_XFRM
475 static inline struct dst_entry *xfrm_lookup(struct net *net,
476 					    struct dst_entry *dst_orig,
477 					    const struct flowi *fl, struct sock *sk,
478 					    int flags)
479 {
480 	return dst_orig;
481 }
482 
483 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
484 						  struct dst_entry *dst_orig,
485 						  const struct flowi *fl,
486 						  struct sock *sk,
487 						  int flags)
488 {
489 	return dst_orig;
490 }
491 
492 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
493 {
494 	return NULL;
495 }
496 
497 #else
498 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
499 			      const struct flowi *fl, struct sock *sk,
500 			      int flags);
501 
502 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
503 				    const struct flowi *fl, struct sock *sk,
504 				    int flags);
505 
506 /* skb attached with this dst needs transformation if dst->xfrm is valid */
507 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
508 {
509 	return dst->xfrm;
510 }
511 #endif
512 
513 #endif /* _NET_DST_H */
514