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 #include <linux/indirect_call_wrapper.h>
22 
23 struct sk_buff;
24 
25 struct dst_entry {
26 	struct net_device       *dev;
27 	struct  dst_ops	        *ops;
28 	unsigned long		_metrics;
29 	unsigned long           expires;
30 #ifdef CONFIG_XFRM
31 	struct xfrm_state	*xfrm;
32 #else
33 	void			*__pad1;
34 #endif
35 	int			(*input)(struct sk_buff *);
36 	int			(*output)(struct net *net, struct sock *sk, struct sk_buff *skb);
37 
38 	unsigned short		flags;
39 #define DST_NOXFRM		0x0002
40 #define DST_NOPOLICY		0x0004
41 #define DST_NOCOUNT		0x0008
42 #define DST_FAKE_RTABLE		0x0010
43 #define DST_XFRM_TUNNEL		0x0020
44 #define DST_XFRM_QUEUE		0x0040
45 #define DST_METADATA		0x0080
46 
47 	/* A non-zero value of dst->obsolete forces by-hand validation
48 	 * of the route entry.  Positive values are set by the generic
49 	 * dst layer to indicate that the entry has been forcefully
50 	 * destroyed.
51 	 *
52 	 * Negative values are used by the implementation layer code to
53 	 * force invocation of the dst_ops->check() method.
54 	 */
55 	short			obsolete;
56 #define DST_OBSOLETE_NONE	0
57 #define DST_OBSOLETE_DEAD	2
58 #define DST_OBSOLETE_FORCE_CHK	-1
59 #define DST_OBSOLETE_KILL	-2
60 	unsigned short		header_len;	/* more space at head required */
61 	unsigned short		trailer_len;	/* space to reserve at tail */
62 
63 	/*
64 	 * __refcnt wants to be on a different cache line from
65 	 * input/output/ops or performance tanks badly
66 	 */
67 #ifdef CONFIG_64BIT
68 	atomic_t		__refcnt;	/* 64-bit offset 64 */
69 #endif
70 	int			__use;
71 	unsigned long		lastuse;
72 	struct lwtunnel_state   *lwtstate;
73 	struct rcu_head		rcu_head;
74 	short			error;
75 	short			__pad;
76 	__u32			tclassid;
77 #ifndef CONFIG_64BIT
78 	atomic_t		__refcnt;	/* 32-bit offset 64 */
79 #endif
80 	netdevice_tracker	dev_tracker;
81 };
82 
83 struct dst_metrics {
84 	u32		metrics[RTAX_MAX];
85 	refcount_t	refcnt;
86 } __aligned(4);		/* Low pointer bits contain DST_METRICS_FLAGS */
87 extern const struct dst_metrics dst_default_metrics;
88 
89 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
90 
91 #define DST_METRICS_READ_ONLY		0x1UL
92 #define DST_METRICS_REFCOUNTED		0x2UL
93 #define DST_METRICS_FLAGS		0x3UL
94 #define __DST_METRICS_PTR(Y)	\
95 	((u32 *)((Y) & ~DST_METRICS_FLAGS))
96 #define DST_METRICS_PTR(X)	__DST_METRICS_PTR((X)->_metrics)
97 
dst_metrics_read_only(const struct dst_entry * dst)98 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
99 {
100 	return dst->_metrics & DST_METRICS_READ_ONLY;
101 }
102 
103 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
104 
dst_destroy_metrics_generic(struct dst_entry * dst)105 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
106 {
107 	unsigned long val = dst->_metrics;
108 	if (!(val & DST_METRICS_READ_ONLY))
109 		__dst_destroy_metrics_generic(dst, val);
110 }
111 
dst_metrics_write_ptr(struct dst_entry * dst)112 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
113 {
114 	unsigned long p = dst->_metrics;
115 
116 	BUG_ON(!p);
117 
118 	if (p & DST_METRICS_READ_ONLY)
119 		return dst->ops->cow_metrics(dst, p);
120 	return __DST_METRICS_PTR(p);
121 }
122 
123 /* This may only be invoked before the entry has reached global
124  * visibility.
125  */
dst_init_metrics(struct dst_entry * dst,const u32 * src_metrics,bool read_only)126 static inline void dst_init_metrics(struct dst_entry *dst,
127 				    const u32 *src_metrics,
128 				    bool read_only)
129 {
130 	dst->_metrics = ((unsigned long) src_metrics) |
131 		(read_only ? DST_METRICS_READ_ONLY : 0);
132 }
133 
dst_copy_metrics(struct dst_entry * dest,const struct dst_entry * src)134 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
135 {
136 	u32 *dst_metrics = dst_metrics_write_ptr(dest);
137 
138 	if (dst_metrics) {
139 		u32 *src_metrics = DST_METRICS_PTR(src);
140 
141 		memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
142 	}
143 }
144 
dst_metrics_ptr(struct dst_entry * dst)145 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
146 {
147 	return DST_METRICS_PTR(dst);
148 }
149 
150 static inline u32
dst_metric_raw(const struct dst_entry * dst,const int metric)151 dst_metric_raw(const struct dst_entry *dst, const int metric)
152 {
153 	u32 *p = DST_METRICS_PTR(dst);
154 
155 	return p[metric-1];
156 }
157 
158 static inline u32
dst_metric(const struct dst_entry * dst,const int metric)159 dst_metric(const struct dst_entry *dst, const int metric)
160 {
161 	WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
162 		     metric == RTAX_ADVMSS ||
163 		     metric == RTAX_MTU);
164 	return dst_metric_raw(dst, metric);
165 }
166 
167 static inline u32
dst_metric_advmss(const struct dst_entry * dst)168 dst_metric_advmss(const struct dst_entry *dst)
169 {
170 	u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
171 
172 	if (!advmss)
173 		advmss = dst->ops->default_advmss(dst);
174 
175 	return advmss;
176 }
177 
dst_metric_set(struct dst_entry * dst,int metric,u32 val)178 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
179 {
180 	u32 *p = dst_metrics_write_ptr(dst);
181 
182 	if (p)
183 		p[metric-1] = val;
184 }
185 
186 /* Kernel-internal feature bits that are unallocated in user space. */
187 #define DST_FEATURE_ECN_CA	(1U << 31)
188 
189 #define DST_FEATURE_MASK	(DST_FEATURE_ECN_CA)
190 #define DST_FEATURE_ECN_MASK	(DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
191 
192 static inline u32
dst_feature(const struct dst_entry * dst,u32 feature)193 dst_feature(const struct dst_entry *dst, u32 feature)
194 {
195 	return dst_metric(dst, RTAX_FEATURES) & feature;
196 }
197 
198 INDIRECT_CALLABLE_DECLARE(unsigned int ip6_mtu(const struct dst_entry *));
199 INDIRECT_CALLABLE_DECLARE(unsigned int ipv4_mtu(const struct dst_entry *));
dst_mtu(const struct dst_entry * dst)200 static inline u32 dst_mtu(const struct dst_entry *dst)
201 {
202 	return INDIRECT_CALL_INET(dst->ops->mtu, ip6_mtu, ipv4_mtu, dst);
203 }
204 
205 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
dst_metric_rtt(const struct dst_entry * dst,int metric)206 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
207 {
208 	return msecs_to_jiffies(dst_metric(dst, metric));
209 }
210 
211 static inline u32
dst_allfrag(const struct dst_entry * dst)212 dst_allfrag(const struct dst_entry *dst)
213 {
214 	int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
215 	return ret;
216 }
217 
218 static inline int
dst_metric_locked(const struct dst_entry * dst,int metric)219 dst_metric_locked(const struct dst_entry *dst, int metric)
220 {
221 	return dst_metric(dst, RTAX_LOCK) & (1 << metric);
222 }
223 
dst_hold(struct dst_entry * dst)224 static inline void dst_hold(struct dst_entry *dst)
225 {
226 	/*
227 	 * If your kernel compilation stops here, please check
228 	 * the placement of __refcnt in struct dst_entry
229 	 */
230 	BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
231 	WARN_ON(atomic_inc_not_zero(&dst->__refcnt) == 0);
232 }
233 
dst_use_noref(struct dst_entry * dst,unsigned long time)234 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
235 {
236 	if (unlikely(time != dst->lastuse)) {
237 		dst->__use++;
238 		dst->lastuse = time;
239 	}
240 }
241 
dst_clone(struct dst_entry * dst)242 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
243 {
244 	if (dst)
245 		dst_hold(dst);
246 	return dst;
247 }
248 
249 void dst_release(struct dst_entry *dst);
250 
251 void dst_release_immediate(struct dst_entry *dst);
252 
refdst_drop(unsigned long refdst)253 static inline void refdst_drop(unsigned long refdst)
254 {
255 	if (!(refdst & SKB_DST_NOREF))
256 		dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
257 }
258 
259 /**
260  * skb_dst_drop - drops skb dst
261  * @skb: buffer
262  *
263  * Drops dst reference count if a reference was taken.
264  */
skb_dst_drop(struct sk_buff * skb)265 static inline void skb_dst_drop(struct sk_buff *skb)
266 {
267 	if (skb->_skb_refdst) {
268 		refdst_drop(skb->_skb_refdst);
269 		skb->_skb_refdst = 0UL;
270 	}
271 }
272 
__skb_dst_copy(struct sk_buff * nskb,unsigned long refdst)273 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
274 {
275 	nskb->slow_gro |= !!refdst;
276 	nskb->_skb_refdst = refdst;
277 	if (!(nskb->_skb_refdst & SKB_DST_NOREF))
278 		dst_clone(skb_dst(nskb));
279 }
280 
skb_dst_copy(struct sk_buff * nskb,const struct sk_buff * oskb)281 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
282 {
283 	__skb_dst_copy(nskb, oskb->_skb_refdst);
284 }
285 
286 /**
287  * dst_hold_safe - Take a reference on a dst if possible
288  * @dst: pointer to dst entry
289  *
290  * This helper returns false if it could not safely
291  * take a reference on a dst.
292  */
dst_hold_safe(struct dst_entry * dst)293 static inline bool dst_hold_safe(struct dst_entry *dst)
294 {
295 	return atomic_inc_not_zero(&dst->__refcnt);
296 }
297 
298 /**
299  * skb_dst_force - makes sure skb dst is refcounted
300  * @skb: buffer
301  *
302  * If dst is not yet refcounted and not destroyed, grab a ref on it.
303  * Returns true if dst is refcounted.
304  */
skb_dst_force(struct sk_buff * skb)305 static inline bool skb_dst_force(struct sk_buff *skb)
306 {
307 	if (skb_dst_is_noref(skb)) {
308 		struct dst_entry *dst = skb_dst(skb);
309 
310 		WARN_ON(!rcu_read_lock_held());
311 		if (!dst_hold_safe(dst))
312 			dst = NULL;
313 
314 		skb->_skb_refdst = (unsigned long)dst;
315 		skb->slow_gro |= !!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 	DEV_STATS_INC(dev, rx_packets);
360 	DEV_STATS_ADD(dev, rx_bytes, skb->len);
361 	__skb_tunnel_rx(skb, dev, net);
362 }
363 
dst_tclassid(const struct sk_buff * skb)364 static inline u32 dst_tclassid(const struct sk_buff *skb)
365 {
366 #ifdef CONFIG_IP_ROUTE_CLASSID
367 	const struct dst_entry *dst;
368 
369 	dst = skb_dst(skb);
370 	if (dst)
371 		return dst->tclassid;
372 #endif
373 	return 0;
374 }
375 
376 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
dst_discard(struct sk_buff * skb)377 static inline int dst_discard(struct sk_buff *skb)
378 {
379 	return dst_discard_out(&init_net, skb->sk, skb);
380 }
381 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
382 		int initial_obsolete, unsigned short flags);
383 void dst_init(struct dst_entry *dst, struct dst_ops *ops,
384 	      struct net_device *dev, int initial_ref, int initial_obsolete,
385 	      unsigned short flags);
386 struct dst_entry *dst_destroy(struct dst_entry *dst);
387 void dst_dev_put(struct dst_entry *dst);
388 
dst_confirm(struct dst_entry * dst)389 static inline void dst_confirm(struct dst_entry *dst)
390 {
391 }
392 
dst_neigh_lookup(const struct dst_entry * dst,const void * daddr)393 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
394 {
395 	struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
396 	return IS_ERR(n) ? NULL : n;
397 }
398 
dst_neigh_lookup_skb(const struct dst_entry * dst,struct sk_buff * skb)399 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
400 						     struct sk_buff *skb)
401 {
402 	struct neighbour *n;
403 
404 	if (WARN_ON_ONCE(!dst->ops->neigh_lookup))
405 		return NULL;
406 
407 	n = dst->ops->neigh_lookup(dst, skb, NULL);
408 
409 	return IS_ERR(n) ? NULL : n;
410 }
411 
dst_confirm_neigh(const struct dst_entry * dst,const void * daddr)412 static inline void dst_confirm_neigh(const struct dst_entry *dst,
413 				     const void *daddr)
414 {
415 	if (dst->ops->confirm_neigh)
416 		dst->ops->confirm_neigh(dst, daddr);
417 }
418 
dst_link_failure(struct sk_buff * skb)419 static inline void dst_link_failure(struct sk_buff *skb)
420 {
421 	struct dst_entry *dst = skb_dst(skb);
422 	if (dst && dst->ops && dst->ops->link_failure)
423 		dst->ops->link_failure(skb);
424 }
425 
dst_set_expires(struct dst_entry * dst,int timeout)426 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
427 {
428 	unsigned long expires = jiffies + timeout;
429 
430 	if (expires == 0)
431 		expires = 1;
432 
433 	if (dst->expires == 0 || time_before(expires, dst->expires))
434 		dst->expires = expires;
435 }
436 
437 INDIRECT_CALLABLE_DECLARE(int ip6_output(struct net *, struct sock *,
438 					 struct sk_buff *));
439 INDIRECT_CALLABLE_DECLARE(int ip_output(struct net *, struct sock *,
440 					 struct sk_buff *));
441 /* Output packet to network from transport.  */
dst_output(struct net * net,struct sock * sk,struct sk_buff * skb)442 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
443 {
444 	return INDIRECT_CALL_INET(skb_dst(skb)->output,
445 				  ip6_output, ip_output,
446 				  net, sk, skb);
447 }
448 
449 INDIRECT_CALLABLE_DECLARE(int ip6_input(struct sk_buff *));
450 INDIRECT_CALLABLE_DECLARE(int ip_local_deliver(struct sk_buff *));
451 /* Input packet from network to transport.  */
dst_input(struct sk_buff * skb)452 static inline int dst_input(struct sk_buff *skb)
453 {
454 	return INDIRECT_CALL_INET(skb_dst(skb)->input,
455 				  ip6_input, ip_local_deliver, skb);
456 }
457 
458 INDIRECT_CALLABLE_DECLARE(struct dst_entry *ip6_dst_check(struct dst_entry *,
459 							  u32));
460 INDIRECT_CALLABLE_DECLARE(struct dst_entry *ipv4_dst_check(struct dst_entry *,
461 							   u32));
dst_check(struct dst_entry * dst,u32 cookie)462 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
463 {
464 	if (dst->obsolete)
465 		dst = INDIRECT_CALL_INET(dst->ops->check, ip6_dst_check,
466 					 ipv4_dst_check, dst, cookie);
467 	return dst;
468 }
469 
470 /* Flags for xfrm_lookup flags argument. */
471 enum {
472 	XFRM_LOOKUP_ICMP = 1 << 0,
473 	XFRM_LOOKUP_QUEUE = 1 << 1,
474 	XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
475 };
476 
477 struct flowi;
478 #ifndef CONFIG_XFRM
xfrm_lookup(struct net * net,struct dst_entry * dst_orig,const struct flowi * fl,const struct sock * sk,int flags)479 static inline struct dst_entry *xfrm_lookup(struct net *net,
480 					    struct dst_entry *dst_orig,
481 					    const struct flowi *fl,
482 					    const struct sock *sk,
483 					    int flags)
484 {
485 	return dst_orig;
486 }
487 
488 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)489 xfrm_lookup_with_ifid(struct net *net, struct dst_entry *dst_orig,
490 		      const struct flowi *fl, const struct sock *sk,
491 		      int flags, u32 if_id)
492 {
493 	return dst_orig;
494 }
495 
xfrm_lookup_route(struct net * net,struct dst_entry * dst_orig,const struct flowi * fl,const struct sock * sk,int flags)496 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
497 						  struct dst_entry *dst_orig,
498 						  const struct flowi *fl,
499 						  const struct sock *sk,
500 						  int flags)
501 {
502 	return dst_orig;
503 }
504 
dst_xfrm(const struct dst_entry * dst)505 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
506 {
507 	return NULL;
508 }
509 
510 #else
511 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
512 			      const struct flowi *fl, const struct sock *sk,
513 			      int flags);
514 
515 struct dst_entry *xfrm_lookup_with_ifid(struct net *net,
516 					struct dst_entry *dst_orig,
517 					const struct flowi *fl,
518 					const struct sock *sk, int flags,
519 					u32 if_id);
520 
521 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
522 				    const struct flowi *fl, const struct sock *sk,
523 				    int flags);
524 
525 /* skb attached with this dst needs transformation if dst->xfrm is valid */
dst_xfrm(const struct dst_entry * dst)526 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
527 {
528 	return dst->xfrm;
529 }
530 #endif
531 
skb_dst_update_pmtu(struct sk_buff * skb,u32 mtu)532 static inline void skb_dst_update_pmtu(struct sk_buff *skb, u32 mtu)
533 {
534 	struct dst_entry *dst = skb_dst(skb);
535 
536 	if (dst && dst->ops->update_pmtu)
537 		dst->ops->update_pmtu(dst, NULL, skb, mtu, true);
538 }
539 
540 /* update dst pmtu but not do neighbor confirm */
skb_dst_update_pmtu_no_confirm(struct sk_buff * skb,u32 mtu)541 static inline void skb_dst_update_pmtu_no_confirm(struct sk_buff *skb, u32 mtu)
542 {
543 	struct dst_entry *dst = skb_dst(skb);
544 
545 	if (dst && dst->ops->update_pmtu)
546 		dst->ops->update_pmtu(dst, NULL, skb, mtu, false);
547 }
548 
549 struct dst_entry *dst_blackhole_check(struct dst_entry *dst, u32 cookie);
550 void dst_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
551 			       struct sk_buff *skb, u32 mtu, bool confirm_neigh);
552 void dst_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
553 			    struct sk_buff *skb);
554 u32 *dst_blackhole_cow_metrics(struct dst_entry *dst, unsigned long old);
555 struct neighbour *dst_blackhole_neigh_lookup(const struct dst_entry *dst,
556 					     struct sk_buff *skb,
557 					     const void *daddr);
558 unsigned int dst_blackhole_mtu(const struct dst_entry *dst);
559 
560 #endif /* _NET_DST_H */
561