xref: /kernel/linux/linux-5.10/include/net/dst.h (revision 8c2ecf20)
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
22struct sk_buff;
23
24struct 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
81struct dst_metrics {
82	u32		metrics[RTAX_MAX];
83	refcount_t	refcnt;
84} __aligned(4);		/* Low pointer bits contain DST_METRICS_FLAGS */
85extern const struct dst_metrics dst_default_metrics;
86
87u32 *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
96static inline bool dst_metrics_read_only(const struct dst_entry *dst)
97{
98	return dst->_metrics & DST_METRICS_READ_ONLY;
99}
100
101void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
102
103static 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
110static 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 */
124static 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
132static 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
143static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
144{
145	return DST_METRICS_PTR(dst);
146}
147
148static inline u32
149dst_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
156static inline u32
157dst_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
165static inline u32
166dst_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
176static 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
190static inline u32
191dst_feature(const struct dst_entry *dst, u32 feature)
192{
193	return dst_metric(dst, RTAX_FEATURES) & feature;
194}
195
196static 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 */
202static 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
207static inline u32
208dst_allfrag(const struct dst_entry *dst)
209{
210	int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
211	return ret;
212}
213
214static inline int
215dst_metric_locked(const struct dst_entry *dst, int metric)
216{
217	return dst_metric(dst, RTAX_LOCK) & (1 << metric);
218}
219
220static 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
230static 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
238static inline struct dst_entry *dst_clone(struct dst_entry *dst)
239{
240	if (dst)
241		dst_hold(dst);
242	return dst;
243}
244
245void dst_release(struct dst_entry *dst);
246
247void dst_release_immediate(struct dst_entry *dst);
248
249static inline void refdst_drop(unsigned long refdst)
250{
251	if (!(refdst & SKB_DST_NOREF))
252		dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
253}
254
255/**
256 * skb_dst_drop - drops skb dst
257 * @skb: buffer
258 *
259 * Drops dst reference count if a reference was taken.
260 */
261static inline void skb_dst_drop(struct sk_buff *skb)
262{
263	if (skb->_skb_refdst) {
264		refdst_drop(skb->_skb_refdst);
265		skb->_skb_refdst = 0UL;
266	}
267}
268
269static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
270{
271	nskb->_skb_refdst = refdst;
272	if (!(nskb->_skb_refdst & SKB_DST_NOREF))
273		dst_clone(skb_dst(nskb));
274}
275
276static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
277{
278	__skb_dst_copy(nskb, oskb->_skb_refdst);
279}
280
281/**
282 * dst_hold_safe - Take a reference on a dst if possible
283 * @dst: pointer to dst entry
284 *
285 * This helper returns false if it could not safely
286 * take a reference on a dst.
287 */
288static inline bool dst_hold_safe(struct dst_entry *dst)
289{
290	return atomic_inc_not_zero(&dst->__refcnt);
291}
292
293/**
294 * skb_dst_force - makes sure skb dst is refcounted
295 * @skb: buffer
296 *
297 * If dst is not yet refcounted and not destroyed, grab a ref on it.
298 * Returns true if dst is refcounted.
299 */
300static inline bool skb_dst_force(struct sk_buff *skb)
301{
302	if (skb_dst_is_noref(skb)) {
303		struct dst_entry *dst = skb_dst(skb);
304
305		WARN_ON(!rcu_read_lock_held());
306		if (!dst_hold_safe(dst))
307			dst = NULL;
308
309		skb->_skb_refdst = (unsigned long)dst;
310	}
311
312	return skb->_skb_refdst != 0UL;
313}
314
315
316/**
317 *	__skb_tunnel_rx - prepare skb for rx reinsert
318 *	@skb: buffer
319 *	@dev: tunnel device
320 *	@net: netns for packet i/o
321 *
322 *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
323 *	so make some cleanups. (no accounting done)
324 */
325static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
326				   struct net *net)
327{
328	skb->dev = dev;
329
330	/*
331	 * Clear hash so that we can recalulate the hash for the
332	 * encapsulated packet, unless we have already determine the hash
333	 * over the L4 4-tuple.
334	 */
335	skb_clear_hash_if_not_l4(skb);
336	skb_set_queue_mapping(skb, 0);
337	skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
338}
339
340/**
341 *	skb_tunnel_rx - prepare skb for rx reinsert
342 *	@skb: buffer
343 *	@dev: tunnel device
344 *	@net: netns for packet i/o
345 *
346 *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
347 *	so make some cleanups, and perform accounting.
348 *	Note: this accounting is not SMP safe.
349 */
350static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
351				 struct net *net)
352{
353	DEV_STATS_INC(dev, rx_packets);
354	DEV_STATS_ADD(dev, rx_bytes, skb->len);
355	__skb_tunnel_rx(skb, dev, net);
356}
357
358static inline u32 dst_tclassid(const struct sk_buff *skb)
359{
360#ifdef CONFIG_IP_ROUTE_CLASSID
361	const struct dst_entry *dst;
362
363	dst = skb_dst(skb);
364	if (dst)
365		return dst->tclassid;
366#endif
367	return 0;
368}
369
370int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
371static inline int dst_discard(struct sk_buff *skb)
372{
373	return dst_discard_out(&init_net, skb->sk, skb);
374}
375void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
376		int initial_obsolete, unsigned short flags);
377void dst_init(struct dst_entry *dst, struct dst_ops *ops,
378	      struct net_device *dev, int initial_ref, int initial_obsolete,
379	      unsigned short flags);
380struct dst_entry *dst_destroy(struct dst_entry *dst);
381void dst_dev_put(struct dst_entry *dst);
382
383static inline void dst_confirm(struct dst_entry *dst)
384{
385}
386
387static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
388{
389	struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
390	return IS_ERR(n) ? NULL : n;
391}
392
393static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
394						     struct sk_buff *skb)
395{
396	struct neighbour *n = NULL;
397
398	/* The packets from tunnel devices (eg bareudp) may have only
399	 * metadata in the dst pointer of skb. Hence a pointer check of
400	 * neigh_lookup is needed.
401	 */
402	if (dst->ops->neigh_lookup)
403		n = dst->ops->neigh_lookup(dst, skb, NULL);
404
405	return IS_ERR(n) ? NULL : n;
406}
407
408static inline void dst_confirm_neigh(const struct dst_entry *dst,
409				     const void *daddr)
410{
411	if (dst->ops->confirm_neigh)
412		dst->ops->confirm_neigh(dst, daddr);
413}
414
415static inline void dst_link_failure(struct sk_buff *skb)
416{
417	struct dst_entry *dst = skb_dst(skb);
418	if (dst && dst->ops && dst->ops->link_failure)
419		dst->ops->link_failure(skb);
420}
421
422static inline void dst_set_expires(struct dst_entry *dst, int timeout)
423{
424	unsigned long expires = jiffies + timeout;
425
426	if (expires == 0)
427		expires = 1;
428
429	if (dst->expires == 0 || time_before(expires, dst->expires))
430		dst->expires = expires;
431}
432
433/* Output packet to network from transport.  */
434static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
435{
436	return skb_dst(skb)->output(net, sk, skb);
437}
438
439/* Input packet from network to transport.  */
440static inline int dst_input(struct sk_buff *skb)
441{
442	return skb_dst(skb)->input(skb);
443}
444
445static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
446{
447	if (dst->obsolete)
448		dst = dst->ops->check(dst, cookie);
449	return dst;
450}
451
452/* Flags for xfrm_lookup flags argument. */
453enum {
454	XFRM_LOOKUP_ICMP = 1 << 0,
455	XFRM_LOOKUP_QUEUE = 1 << 1,
456	XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
457};
458
459struct flowi;
460#ifndef CONFIG_XFRM
461static inline struct dst_entry *xfrm_lookup(struct net *net,
462					    struct dst_entry *dst_orig,
463					    const struct flowi *fl,
464					    const struct sock *sk,
465					    int flags)
466{
467	return dst_orig;
468}
469
470static inline struct dst_entry *
471xfrm_lookup_with_ifid(struct net *net, struct dst_entry *dst_orig,
472		      const struct flowi *fl, const struct sock *sk,
473		      int flags, u32 if_id)
474{
475	return dst_orig;
476}
477
478static inline struct dst_entry *xfrm_lookup_route(struct net *net,
479						  struct dst_entry *dst_orig,
480						  const struct flowi *fl,
481						  const struct sock *sk,
482						  int flags)
483{
484	return dst_orig;
485}
486
487static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
488{
489	return NULL;
490}
491
492#else
493struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
494			      const struct flowi *fl, const struct sock *sk,
495			      int flags);
496
497struct dst_entry *xfrm_lookup_with_ifid(struct net *net,
498					struct dst_entry *dst_orig,
499					const struct flowi *fl,
500					const struct sock *sk, int flags,
501					u32 if_id);
502
503struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
504				    const struct flowi *fl, const struct sock *sk,
505				    int flags);
506
507/* skb attached with this dst needs transformation if dst->xfrm is valid */
508static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
509{
510	return dst->xfrm;
511}
512#endif
513
514static inline void skb_dst_update_pmtu(struct sk_buff *skb, u32 mtu)
515{
516	struct dst_entry *dst = skb_dst(skb);
517
518	if (dst && dst->ops->update_pmtu)
519		dst->ops->update_pmtu(dst, NULL, skb, mtu, true);
520}
521
522/* update dst pmtu but not do neighbor confirm */
523static inline void skb_dst_update_pmtu_no_confirm(struct sk_buff *skb, u32 mtu)
524{
525	struct dst_entry *dst = skb_dst(skb);
526
527	if (dst && dst->ops->update_pmtu)
528		dst->ops->update_pmtu(dst, NULL, skb, mtu, false);
529}
530
531struct dst_entry *dst_blackhole_check(struct dst_entry *dst, u32 cookie);
532void dst_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
533			       struct sk_buff *skb, u32 mtu, bool confirm_neigh);
534void dst_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
535			    struct sk_buff *skb);
536u32 *dst_blackhole_cow_metrics(struct dst_entry *dst, unsigned long old);
537struct neighbour *dst_blackhole_neigh_lookup(const struct dst_entry *dst,
538					     struct sk_buff *skb,
539					     const void *daddr);
540unsigned int dst_blackhole_mtu(const struct dst_entry *dst);
541
542#endif /* _NET_DST_H */
543