• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  * net/dst.h	Protocol independent destination cache definitions.
4  *
5  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
6  *
7  */
8 
9 #ifndef _NET_DST_H
10 #define _NET_DST_H
11 
12 #include <net/dst_ops.h>
13 #include <linux/netdevice.h>
14 #include <linux/rtnetlink.h>
15 #include <linux/rcupdate.h>
16 #include <linux/bug.h>
17 #include <linux/jiffies.h>
18 #include <linux/refcount.h>
19 #include <net/neighbour.h>
20 #include <asm/processor.h>
21 
22 struct sk_buff;
23 
24 struct dst_entry {
25 	struct net_device       *dev;
26 	struct  dst_ops	        *ops;
27 	unsigned long		_metrics;
28 	unsigned long           expires;
29 #ifdef CONFIG_XFRM
30 	struct xfrm_state	*xfrm;
31 #else
32 	void			*__pad1;
33 #endif
34 	int			(*input)(struct sk_buff *);
35 	int			(*output)(struct net *net, struct sock *sk, struct sk_buff *skb);
36 
37 	unsigned short		flags;
38 #define DST_NOXFRM		0x0002
39 #define DST_NOPOLICY		0x0004
40 #define DST_NOCOUNT		0x0008
41 #define DST_FAKE_RTABLE		0x0010
42 #define DST_XFRM_TUNNEL		0x0020
43 #define DST_XFRM_QUEUE		0x0040
44 #define DST_METADATA		0x0080
45 
46 	/* A non-zero value of dst->obsolete forces by-hand validation
47 	 * of the route entry.  Positive values are set by the generic
48 	 * dst layer to indicate that the entry has been forcefully
49 	 * destroyed.
50 	 *
51 	 * Negative values are used by the implementation layer code to
52 	 * force invocation of the dst_ops->check() method.
53 	 */
54 	short			obsolete;
55 #define DST_OBSOLETE_NONE	0
56 #define DST_OBSOLETE_DEAD	2
57 #define DST_OBSOLETE_FORCE_CHK	-1
58 #define DST_OBSOLETE_KILL	-2
59 	unsigned short		header_len;	/* more space at head required */
60 	unsigned short		trailer_len;	/* space to reserve at tail */
61 
62 	/*
63 	 * __refcnt wants to be on a different cache line from
64 	 * input/output/ops or performance tanks badly
65 	 */
66 #ifdef CONFIG_64BIT
67 	atomic_t		__refcnt;	/* 64-bit offset 64 */
68 #endif
69 	int			__use;
70 	unsigned long		lastuse;
71 	struct lwtunnel_state   *lwtstate;
72 	struct rcu_head		rcu_head;
73 	short			error;
74 	short			__pad;
75 	__u32			tclassid;
76 #ifndef CONFIG_64BIT
77 	atomic_t		__refcnt;	/* 32-bit offset 64 */
78 #endif
79 };
80 
81 struct dst_metrics {
82 	u32		metrics[RTAX_MAX];
83 	refcount_t	refcnt;
84 } __aligned(4);		/* Low pointer bits contain DST_METRICS_FLAGS */
85 extern const struct dst_metrics dst_default_metrics;
86 
87 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
88 
89 #define DST_METRICS_READ_ONLY		0x1UL
90 #define DST_METRICS_REFCOUNTED		0x2UL
91 #define DST_METRICS_FLAGS		0x3UL
92 #define __DST_METRICS_PTR(Y)	\
93 	((u32 *)((Y) & ~DST_METRICS_FLAGS))
94 #define DST_METRICS_PTR(X)	__DST_METRICS_PTR((X)->_metrics)
95 
dst_metrics_read_only(const struct dst_entry * dst)96 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
97 {
98 	return dst->_metrics & DST_METRICS_READ_ONLY;
99 }
100 
101 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
102 
dst_destroy_metrics_generic(struct dst_entry * dst)103 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
104 {
105 	unsigned long val = dst->_metrics;
106 	if (!(val & DST_METRICS_READ_ONLY))
107 		__dst_destroy_metrics_generic(dst, val);
108 }
109 
dst_metrics_write_ptr(struct dst_entry * dst)110 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
111 {
112 	unsigned long p = dst->_metrics;
113 
114 	BUG_ON(!p);
115 
116 	if (p & DST_METRICS_READ_ONLY)
117 		return dst->ops->cow_metrics(dst, p);
118 	return __DST_METRICS_PTR(p);
119 }
120 
121 /* This may only be invoked before the entry has reached global
122  * visibility.
123  */
dst_init_metrics(struct dst_entry * dst,const u32 * src_metrics,bool read_only)124 static inline void dst_init_metrics(struct dst_entry *dst,
125 				    const u32 *src_metrics,
126 				    bool read_only)
127 {
128 	dst->_metrics = ((unsigned long) src_metrics) |
129 		(read_only ? DST_METRICS_READ_ONLY : 0);
130 }
131 
dst_copy_metrics(struct dst_entry * dest,const struct dst_entry * src)132 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
133 {
134 	u32 *dst_metrics = dst_metrics_write_ptr(dest);
135 
136 	if (dst_metrics) {
137 		u32 *src_metrics = DST_METRICS_PTR(src);
138 
139 		memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
140 	}
141 }
142 
dst_metrics_ptr(struct dst_entry * dst)143 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
144 {
145 	return DST_METRICS_PTR(dst);
146 }
147 
148 static inline u32
dst_metric_raw(const struct dst_entry * dst,const int metric)149 dst_metric_raw(const struct dst_entry *dst, const int metric)
150 {
151 	u32 *p = DST_METRICS_PTR(dst);
152 
153 	return p[metric-1];
154 }
155 
156 static inline u32
dst_metric(const struct dst_entry * dst,const int metric)157 dst_metric(const struct dst_entry *dst, const int metric)
158 {
159 	WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
160 		     metric == RTAX_ADVMSS ||
161 		     metric == RTAX_MTU);
162 	return dst_metric_raw(dst, metric);
163 }
164 
165 static inline u32
dst_metric_advmss(const struct dst_entry * dst)166 dst_metric_advmss(const struct dst_entry *dst)
167 {
168 	u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
169 
170 	if (!advmss)
171 		advmss = dst->ops->default_advmss(dst);
172 
173 	return advmss;
174 }
175 
dst_metric_set(struct dst_entry * dst,int metric,u32 val)176 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
177 {
178 	u32 *p = dst_metrics_write_ptr(dst);
179 
180 	if (p)
181 		p[metric-1] = val;
182 }
183 
184 /* Kernel-internal feature bits that are unallocated in user space. */
185 #define DST_FEATURE_ECN_CA	(1U << 31)
186 
187 #define DST_FEATURE_MASK	(DST_FEATURE_ECN_CA)
188 #define DST_FEATURE_ECN_MASK	(DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
189 
190 static inline u32
dst_feature(const struct dst_entry * dst,u32 feature)191 dst_feature(const struct dst_entry *dst, u32 feature)
192 {
193 	return dst_metric(dst, RTAX_FEATURES) & feature;
194 }
195 
dst_mtu(const struct dst_entry * dst)196 static inline u32 dst_mtu(const struct dst_entry *dst)
197 {
198 	return dst->ops->mtu(dst);
199 }
200 
201 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
dst_metric_rtt(const struct dst_entry * dst,int metric)202 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
203 {
204 	return msecs_to_jiffies(dst_metric(dst, metric));
205 }
206 
207 static inline u32
dst_allfrag(const struct dst_entry * dst)208 dst_allfrag(const struct dst_entry *dst)
209 {
210 	int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
211 	return ret;
212 }
213 
214 static inline int
dst_metric_locked(const struct dst_entry * dst,int metric)215 dst_metric_locked(const struct dst_entry *dst, int metric)
216 {
217 	return dst_metric(dst, RTAX_LOCK) & (1 << metric);
218 }
219 
dst_hold(struct dst_entry * dst)220 static inline void dst_hold(struct dst_entry *dst)
221 {
222 	/*
223 	 * If your kernel compilation stops here, please check
224 	 * the placement of __refcnt in struct dst_entry
225 	 */
226 	BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
227 	WARN_ON(atomic_inc_not_zero(&dst->__refcnt) == 0);
228 }
229 
dst_use_noref(struct dst_entry * dst,unsigned long time)230 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
231 {
232 	if (unlikely(time != dst->lastuse)) {
233 		dst->__use++;
234 		dst->lastuse = time;
235 	}
236 }
237 
dst_hold_and_use(struct dst_entry * dst,unsigned long time)238 static inline void dst_hold_and_use(struct dst_entry *dst, unsigned long time)
239 {
240 	dst_hold(dst);
241 	dst_use_noref(dst, time);
242 }
243 
dst_clone(struct dst_entry * dst)244 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
245 {
246 	if (dst)
247 		dst_hold(dst);
248 	return dst;
249 }
250 
251 void dst_release(struct dst_entry *dst);
252 
253 void dst_release_immediate(struct dst_entry *dst);
254 
refdst_drop(unsigned long refdst)255 static inline void refdst_drop(unsigned long refdst)
256 {
257 	if (!(refdst & SKB_DST_NOREF))
258 		dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
259 }
260 
261 /**
262  * skb_dst_drop - drops skb dst
263  * @skb: buffer
264  *
265  * Drops dst reference count if a reference was taken.
266  */
skb_dst_drop(struct sk_buff * skb)267 static inline void skb_dst_drop(struct sk_buff *skb)
268 {
269 	if (skb->_skb_refdst) {
270 		refdst_drop(skb->_skb_refdst);
271 		skb->_skb_refdst = 0UL;
272 	}
273 }
274 
__skb_dst_copy(struct sk_buff * nskb,unsigned long refdst)275 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
276 {
277 	nskb->_skb_refdst = refdst;
278 	if (!(nskb->_skb_refdst & SKB_DST_NOREF))
279 		dst_clone(skb_dst(nskb));
280 }
281 
skb_dst_copy(struct sk_buff * nskb,const struct sk_buff * oskb)282 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
283 {
284 	__skb_dst_copy(nskb, oskb->_skb_refdst);
285 }
286 
287 /**
288  * dst_hold_safe - Take a reference on a dst if possible
289  * @dst: pointer to dst entry
290  *
291  * This helper returns false if it could not safely
292  * take a reference on a dst.
293  */
dst_hold_safe(struct dst_entry * dst)294 static inline bool dst_hold_safe(struct dst_entry *dst)
295 {
296 	return atomic_inc_not_zero(&dst->__refcnt);
297 }
298 
299 /**
300  * skb_dst_force - makes sure skb dst is refcounted
301  * @skb: buffer
302  *
303  * If dst is not yet refcounted and not destroyed, grab a ref on it.
304  * Returns true if dst is refcounted.
305  */
skb_dst_force(struct sk_buff * skb)306 static inline bool skb_dst_force(struct sk_buff *skb)
307 {
308 	if (skb_dst_is_noref(skb)) {
309 		struct dst_entry *dst = skb_dst(skb);
310 
311 		WARN_ON(!rcu_read_lock_held());
312 		if (!dst_hold_safe(dst))
313 			dst = NULL;
314 
315 		skb->_skb_refdst = (unsigned long)dst;
316 	}
317 
318 	return skb->_skb_refdst != 0UL;
319 }
320 
321 
322 /**
323  *	__skb_tunnel_rx - prepare skb for rx reinsert
324  *	@skb: buffer
325  *	@dev: tunnel device
326  *	@net: netns for packet i/o
327  *
328  *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
329  *	so make some cleanups. (no accounting done)
330  */
__skb_tunnel_rx(struct sk_buff * skb,struct net_device * dev,struct net * net)331 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
332 				   struct net *net)
333 {
334 	skb->dev = dev;
335 
336 	/*
337 	 * Clear hash so that we can recalulate the hash for the
338 	 * encapsulated packet, unless we have already determine the hash
339 	 * over the L4 4-tuple.
340 	 */
341 	skb_clear_hash_if_not_l4(skb);
342 	skb_set_queue_mapping(skb, 0);
343 	skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
344 }
345 
346 /**
347  *	skb_tunnel_rx - prepare skb for rx reinsert
348  *	@skb: buffer
349  *	@dev: tunnel device
350  *	@net: netns for packet i/o
351  *
352  *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
353  *	so make some cleanups, and perform accounting.
354  *	Note: this accounting is not SMP safe.
355  */
skb_tunnel_rx(struct sk_buff * skb,struct net_device * dev,struct net * net)356 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
357 				 struct net *net)
358 {
359 	/* TODO : stats should be SMP safe */
360 	dev->stats.rx_packets++;
361 	dev->stats.rx_bytes += skb->len;
362 	__skb_tunnel_rx(skb, dev, net);
363 }
364 
dst_tclassid(const struct sk_buff * skb)365 static inline u32 dst_tclassid(const struct sk_buff *skb)
366 {
367 #ifdef CONFIG_IP_ROUTE_CLASSID
368 	const struct dst_entry *dst;
369 
370 	dst = skb_dst(skb);
371 	if (dst)
372 		return dst->tclassid;
373 #endif
374 	return 0;
375 }
376 
377 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
dst_discard(struct sk_buff * skb)378 static inline int dst_discard(struct sk_buff *skb)
379 {
380 	return dst_discard_out(&init_net, skb->sk, skb);
381 }
382 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
383 		int initial_obsolete, unsigned short flags);
384 void dst_init(struct dst_entry *dst, struct dst_ops *ops,
385 	      struct net_device *dev, int initial_ref, int initial_obsolete,
386 	      unsigned short flags);
387 struct dst_entry *dst_destroy(struct dst_entry *dst);
388 void dst_dev_put(struct dst_entry *dst);
389 
dst_confirm(struct dst_entry * dst)390 static inline void dst_confirm(struct dst_entry *dst)
391 {
392 }
393 
dst_neigh_lookup(const struct dst_entry * dst,const void * daddr)394 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
395 {
396 	struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
397 	return IS_ERR(n) ? NULL : n;
398 }
399 
dst_neigh_lookup_skb(const struct dst_entry * dst,struct sk_buff * skb)400 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
401 						     struct sk_buff *skb)
402 {
403 	struct neighbour *n = NULL;
404 
405 	/* The packets from tunnel devices (eg bareudp) may have only
406 	 * metadata in the dst pointer of skb. Hence a pointer check of
407 	 * neigh_lookup is needed.
408 	 */
409 	if (dst->ops->neigh_lookup)
410 		n = dst->ops->neigh_lookup(dst, skb, NULL);
411 
412 	return IS_ERR(n) ? NULL : n;
413 }
414 
dst_confirm_neigh(const struct dst_entry * dst,const void * daddr)415 static inline void dst_confirm_neigh(const struct dst_entry *dst,
416 				     const void *daddr)
417 {
418 	if (dst->ops->confirm_neigh)
419 		dst->ops->confirm_neigh(dst, daddr);
420 }
421 
dst_link_failure(struct sk_buff * skb)422 static inline void dst_link_failure(struct sk_buff *skb)
423 {
424 	struct dst_entry *dst = skb_dst(skb);
425 	if (dst && dst->ops && dst->ops->link_failure)
426 		dst->ops->link_failure(skb);
427 }
428 
dst_set_expires(struct dst_entry * dst,int timeout)429 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
430 {
431 	unsigned long expires = jiffies + timeout;
432 
433 	if (expires == 0)
434 		expires = 1;
435 
436 	if (dst->expires == 0 || time_before(expires, dst->expires))
437 		dst->expires = expires;
438 }
439 
440 /* Output packet to network from transport.  */
dst_output(struct net * net,struct sock * sk,struct sk_buff * skb)441 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
442 {
443 	return skb_dst(skb)->output(net, sk, skb);
444 }
445 
446 /* Input packet from network to transport.  */
dst_input(struct sk_buff * skb)447 static inline int dst_input(struct sk_buff *skb)
448 {
449 	return skb_dst(skb)->input(skb);
450 }
451 
dst_check(struct dst_entry * dst,u32 cookie)452 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
453 {
454 	if (dst->obsolete)
455 		dst = dst->ops->check(dst, cookie);
456 	return dst;
457 }
458 
459 /* Flags for xfrm_lookup flags argument. */
460 enum {
461 	XFRM_LOOKUP_ICMP = 1 << 0,
462 	XFRM_LOOKUP_QUEUE = 1 << 1,
463 	XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
464 };
465 
466 struct flowi;
467 #ifndef CONFIG_XFRM
xfrm_lookup(struct net * net,struct dst_entry * dst_orig,const struct flowi * fl,const struct sock * sk,int flags)468 static inline struct dst_entry *xfrm_lookup(struct net *net,
469 					    struct dst_entry *dst_orig,
470 					    const struct flowi *fl,
471 					    const struct sock *sk,
472 					    int flags)
473 {
474 	return dst_orig;
475 }
476 
477 static inline struct dst_entry *
xfrm_lookup_with_ifid(struct net * net,struct dst_entry * dst_orig,const struct flowi * fl,const struct sock * sk,int flags,u32 if_id)478 xfrm_lookup_with_ifid(struct net *net, struct dst_entry *dst_orig,
479 		      const struct flowi *fl, const struct sock *sk,
480 		      int flags, u32 if_id)
481 {
482 	return dst_orig;
483 }
484 
xfrm_lookup_route(struct net * net,struct dst_entry * dst_orig,const struct flowi * fl,const struct sock * sk,int flags)485 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
486 						  struct dst_entry *dst_orig,
487 						  const struct flowi *fl,
488 						  const struct sock *sk,
489 						  int flags)
490 {
491 	return dst_orig;
492 }
493 
dst_xfrm(const struct dst_entry * dst)494 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
495 {
496 	return NULL;
497 }
498 
499 #else
500 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
501 			      const struct flowi *fl, const struct sock *sk,
502 			      int flags);
503 
504 struct dst_entry *xfrm_lookup_with_ifid(struct net *net,
505 					struct dst_entry *dst_orig,
506 					const struct flowi *fl,
507 					const struct sock *sk, int flags,
508 					u32 if_id);
509 
510 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
511 				    const struct flowi *fl, const struct sock *sk,
512 				    int flags);
513 
514 /* skb attached with this dst needs transformation if dst->xfrm is valid */
dst_xfrm(const struct dst_entry * dst)515 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
516 {
517 	return dst->xfrm;
518 }
519 #endif
520 
skb_dst_update_pmtu(struct sk_buff * skb,u32 mtu)521 static inline void skb_dst_update_pmtu(struct sk_buff *skb, u32 mtu)
522 {
523 	struct dst_entry *dst = skb_dst(skb);
524 
525 	if (dst && dst->ops->update_pmtu)
526 		dst->ops->update_pmtu(dst, NULL, skb, mtu, true);
527 }
528 
529 /* update dst pmtu but not do neighbor confirm */
skb_dst_update_pmtu_no_confirm(struct sk_buff * skb,u32 mtu)530 static inline void skb_dst_update_pmtu_no_confirm(struct sk_buff *skb, u32 mtu)
531 {
532 	struct dst_entry *dst = skb_dst(skb);
533 
534 	if (dst && dst->ops->update_pmtu)
535 		dst->ops->update_pmtu(dst, NULL, skb, mtu, false);
536 }
537 
538 struct dst_entry *dst_blackhole_check(struct dst_entry *dst, u32 cookie);
539 void dst_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
540 			       struct sk_buff *skb, u32 mtu, bool confirm_neigh);
541 void dst_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
542 			    struct sk_buff *skb);
543 u32 *dst_blackhole_cow_metrics(struct dst_entry *dst, unsigned long old);
544 struct neighbour *dst_blackhole_neigh_lookup(const struct dst_entry *dst,
545 					     struct sk_buff *skb,
546 					     const void *daddr);
547 unsigned int dst_blackhole_mtu(const struct dst_entry *dst);
548 
549 #endif /* _NET_DST_H */
550