Linux Audio

Check our new training course

Loading...
v6.13.7
  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 <linux/rcuref.h>
 20#include <net/neighbour.h>
 21#include <asm/processor.h>
 22#include <linux/indirect_call_wrapper.h>
 
 
 
 
 
 
 
 
 
 
 23
 24struct sk_buff;
 25
 26struct dst_entry {
 
 
 27	struct net_device       *dev;
 28	struct  dst_ops	        *ops;
 29	unsigned long		_metrics;
 30	unsigned long           expires;
 
 
 31#ifdef CONFIG_XFRM
 32	struct xfrm_state	*xfrm;
 33#else
 34	void			*__pad1;
 35#endif
 36	int			(*input)(struct sk_buff *);
 37	int			(*output)(struct net *net, struct sock *sk, struct sk_buff *skb);
 38
 39	unsigned short		flags;
 
 40#define DST_NOXFRM		0x0002
 41#define DST_NOPOLICY		0x0004
 42#define DST_NOCOUNT		0x0008
 43#define DST_FAKE_RTABLE		0x0010
 44#define DST_XFRM_TUNNEL		0x0020
 45#define DST_XFRM_QUEUE		0x0040
 46#define DST_METADATA		0x0080
 
 
 
 
 
 
 47
 48	/* A non-zero value of dst->obsolete forces by-hand validation
 49	 * of the route entry.  Positive values are set by the generic
 50	 * dst layer to indicate that the entry has been forcefully
 51	 * destroyed.
 52	 *
 53	 * Negative values are used by the implementation layer code to
 54	 * force invocation of the dst_ops->check() method.
 55	 */
 56	short			obsolete;
 57#define DST_OBSOLETE_NONE	0
 58#define DST_OBSOLETE_DEAD	2
 59#define DST_OBSOLETE_FORCE_CHK	-1
 60#define DST_OBSOLETE_KILL	-2
 61	unsigned short		header_len;	/* more space at head required */
 62	unsigned short		trailer_len;	/* space to reserve at tail */
 
 
 
 
 
 63
 
 64	/*
 65	 * __rcuref wants to be on a different cache line from
 66	 * input/output/ops or performance tanks badly
 67	 */
 68#ifdef CONFIG_64BIT
 69	rcuref_t		__rcuref;	/* 64-bit offset 64 */
 70#endif
 71	int			__use;
 72	unsigned long		lastuse;
 73	struct rcu_head		rcu_head;
 74	short			error;
 75	short			__pad;
 76	__u32			tclassid;
 77#ifndef CONFIG_64BIT
 78	struct lwtunnel_state   *lwtstate;
 79	rcuref_t		__rcuref;	/* 32-bit offset 64 */
 80#endif
 81	netdevice_tracker	dev_tracker;
 82
 83	/*
 84	 * Used by rtable and rt6_info. Moves lwtstate into the next cache
 85	 * line on 64bit so that lwtstate does not cause false sharing with
 86	 * __rcuref under contention of __rcuref. This also puts the
 87	 * frequently accessed members of rtable and rt6_info out of the
 88	 * __rcuref cache line.
 89	 */
 90	struct list_head	rt_uncached;
 91	struct uncached_list	*rt_uncached_list;
 92#ifdef CONFIG_64BIT
 93	struct lwtunnel_state   *lwtstate;
 94#endif
 
 
 
 
 
 95};
 96
 97struct dst_metrics {
 98	u32		metrics[RTAX_MAX];
 99	refcount_t	refcnt;
100} __aligned(4);		/* Low pointer bits contain DST_METRICS_FLAGS */
101extern const struct dst_metrics dst_default_metrics;
102
103u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
 
104
105#define DST_METRICS_READ_ONLY		0x1UL
106#define DST_METRICS_REFCOUNTED		0x2UL
107#define DST_METRICS_FLAGS		0x3UL
108#define __DST_METRICS_PTR(Y)	\
109	((u32 *)((Y) & ~DST_METRICS_FLAGS))
110#define DST_METRICS_PTR(X)	__DST_METRICS_PTR((X)->_metrics)
111
112static inline bool dst_metrics_read_only(const struct dst_entry *dst)
113{
114	return dst->_metrics & DST_METRICS_READ_ONLY;
115}
116
117void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
118
119static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
120{
121	unsigned long val = dst->_metrics;
122	if (!(val & DST_METRICS_READ_ONLY))
123		__dst_destroy_metrics_generic(dst, val);
124}
125
126static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
127{
128	unsigned long p = dst->_metrics;
129
130	BUG_ON(!p);
131
132	if (p & DST_METRICS_READ_ONLY)
133		return dst->ops->cow_metrics(dst, p);
134	return __DST_METRICS_PTR(p);
135}
136
137/* This may only be invoked before the entry has reached global
138 * visibility.
139 */
140static inline void dst_init_metrics(struct dst_entry *dst,
141				    const u32 *src_metrics,
142				    bool read_only)
143{
144	dst->_metrics = ((unsigned long) src_metrics) |
145		(read_only ? DST_METRICS_READ_ONLY : 0);
146}
147
148static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
149{
150	u32 *dst_metrics = dst_metrics_write_ptr(dest);
151
152	if (dst_metrics) {
153		u32 *src_metrics = DST_METRICS_PTR(src);
154
155		memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
156	}
157}
158
159static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
160{
161	return DST_METRICS_PTR(dst);
162}
163
164static inline u32
165dst_metric_raw(const struct dst_entry *dst, const int metric)
166{
167	u32 *p = DST_METRICS_PTR(dst);
168
169	return p[metric-1];
170}
171
172static inline u32
173dst_metric(const struct dst_entry *dst, const int metric)
174{
175	WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
176		     metric == RTAX_ADVMSS ||
177		     metric == RTAX_MTU);
178	return dst_metric_raw(dst, metric);
179}
180
181static inline u32
182dst_metric_advmss(const struct dst_entry *dst)
183{
184	u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
185
186	if (!advmss)
187		advmss = dst->ops->default_advmss(dst);
188
189	return advmss;
190}
191
192static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
193{
194	u32 *p = dst_metrics_write_ptr(dst);
195
196	if (p)
197		p[metric-1] = val;
198}
199
200/* Kernel-internal feature bits that are unallocated in user space. */
201#define DST_FEATURE_ECN_CA	(1U << 31)
202
203#define DST_FEATURE_MASK	(DST_FEATURE_ECN_CA)
204#define DST_FEATURE_ECN_MASK	(DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
205
206static inline u32
207dst_feature(const struct dst_entry *dst, u32 feature)
208{
209	return dst_metric(dst, RTAX_FEATURES) & feature;
210}
211
212INDIRECT_CALLABLE_DECLARE(unsigned int ip6_mtu(const struct dst_entry *));
213INDIRECT_CALLABLE_DECLARE(unsigned int ipv4_mtu(const struct dst_entry *));
214static inline u32 dst_mtu(const struct dst_entry *dst)
215{
216	return INDIRECT_CALL_INET(dst->ops->mtu, ip6_mtu, ipv4_mtu, dst);
217}
218
219/* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
220static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
221{
222	return msecs_to_jiffies(dst_metric(dst, metric));
223}
224
 
 
 
 
 
 
 
225static inline int
226dst_metric_locked(const struct dst_entry *dst, int metric)
227{
228	return dst_metric(dst, RTAX_LOCK) & (1 << metric);
229}
230
231static inline void dst_hold(struct dst_entry *dst)
232{
233	/*
234	 * If your kernel compilation stops here, please check
235	 * the placement of __rcuref in struct dst_entry
236	 */
237	BUILD_BUG_ON(offsetof(struct dst_entry, __rcuref) & 63);
238	WARN_ON(!rcuref_get(&dst->__rcuref));
 
 
 
 
 
 
 
239}
240
241static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
242{
243	if (unlikely(time != dst->lastuse)) {
244		dst->__use++;
245		dst->lastuse = time;
246	}
247}
248
249static inline struct dst_entry *dst_clone(struct dst_entry *dst)
250{
251	if (dst)
252		dst_hold(dst);
253	return dst;
254}
255
256void dst_release(struct dst_entry *dst);
257
258void dst_release_immediate(struct dst_entry *dst);
259
260static inline void refdst_drop(unsigned long refdst)
261{
262	if (!(refdst & SKB_DST_NOREF))
263		dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
264}
265
266/**
267 * skb_dst_drop - drops skb dst
268 * @skb: buffer
269 *
270 * Drops dst reference count if a reference was taken.
271 */
272static inline void skb_dst_drop(struct sk_buff *skb)
273{
274	if (skb->_skb_refdst) {
275		refdst_drop(skb->_skb_refdst);
276		skb->_skb_refdst = 0UL;
277	}
278}
279
280static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
281{
282	nskb->slow_gro |= !!refdst;
283	nskb->_skb_refdst = refdst;
284	if (!(nskb->_skb_refdst & SKB_DST_NOREF))
285		dst_clone(skb_dst(nskb));
286}
287
288static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
289{
290	__skb_dst_copy(nskb, oskb->_skb_refdst);
291}
292
293/**
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
294 * dst_hold_safe - Take a reference on a dst if possible
295 * @dst: pointer to dst entry
296 *
297 * This helper returns false if it could not safely
298 * take a reference on a dst.
299 */
300static inline bool dst_hold_safe(struct dst_entry *dst)
301{
302	return rcuref_get(&dst->__rcuref);
 
 
 
303}
304
305/**
306 * skb_dst_force - makes sure skb dst is refcounted
307 * @skb: buffer
308 *
309 * If dst is not yet refcounted and not destroyed, grab a ref on it.
310 * Returns true if dst is refcounted.
311 */
312static inline bool skb_dst_force(struct sk_buff *skb)
313{
314	if (skb_dst_is_noref(skb)) {
315		struct dst_entry *dst = skb_dst(skb);
316
317		WARN_ON(!rcu_read_lock_held());
318		if (!dst_hold_safe(dst))
319			dst = NULL;
320
321		skb->_skb_refdst = (unsigned long)dst;
322		skb->slow_gro |= !!dst;
323	}
324
325	return skb->_skb_refdst != 0UL;
326}
327
328
329/**
330 *	__skb_tunnel_rx - prepare skb for rx reinsert
331 *	@skb: buffer
332 *	@dev: tunnel device
333 *	@net: netns for packet i/o
334 *
335 *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
336 *	so make some cleanups. (no accounting done)
337 */
338static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
339				   struct net *net)
340{
341	skb->dev = dev;
342
343	/*
344	 * Clear hash so that we can recalculate the hash for the
345	 * encapsulated packet, unless we have already determine the hash
346	 * over the L4 4-tuple.
347	 */
348	skb_clear_hash_if_not_l4(skb);
349	skb_set_queue_mapping(skb, 0);
350	skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
351}
352
353/**
354 *	skb_tunnel_rx - prepare skb for rx reinsert
355 *	@skb: buffer
356 *	@dev: tunnel device
357 *	@net: netns for packet i/o
358 *
359 *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
360 *	so make some cleanups, and perform accounting.
361 *	Note: this accounting is not SMP safe.
362 */
363static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
364				 struct net *net)
365{
366	DEV_STATS_INC(dev, rx_packets);
367	DEV_STATS_ADD(dev, rx_bytes, skb->len);
 
368	__skb_tunnel_rx(skb, dev, net);
369}
370
371static inline u32 dst_tclassid(const struct sk_buff *skb)
372{
373#ifdef CONFIG_IP_ROUTE_CLASSID
374	const struct dst_entry *dst;
375
376	dst = skb_dst(skb);
377	if (dst)
378		return dst->tclassid;
379#endif
380	return 0;
381}
382
383int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
384static inline int dst_discard(struct sk_buff *skb)
385{
386	return dst_discard_out(&init_net, skb->sk, skb);
387}
388void *dst_alloc(struct dst_ops *ops, struct net_device *dev,
389		int initial_obsolete, unsigned short flags);
390void dst_init(struct dst_entry *dst, struct dst_ops *ops,
391	      struct net_device *dev, int initial_obsolete,
392	      unsigned short flags);
393void dst_dev_put(struct dst_entry *dst);
 
394
395static inline void dst_confirm(struct dst_entry *dst)
396{
 
 
 
 
 
 
 
 
397}
398
399static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
400{
401	struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
402	return IS_ERR(n) ? NULL : n;
403}
404
405static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
406						     struct sk_buff *skb)
 
 
 
 
 
407{
408	struct neighbour *n;
409
410	if (WARN_ON_ONCE(!dst->ops->neigh_lookup))
411		return NULL;
412
413	n = dst->ops->neigh_lookup(dst, skb, NULL);
 
 
 
 
414
 
 
 
 
 
 
 
 
 
 
415	return IS_ERR(n) ? NULL : n;
416}
417
418static inline void dst_confirm_neigh(const struct dst_entry *dst,
419				     const void *daddr)
420{
421	if (dst->ops->confirm_neigh)
422		dst->ops->confirm_neigh(dst, daddr);
423}
424
425static inline void dst_link_failure(struct sk_buff *skb)
426{
427	struct dst_entry *dst = skb_dst(skb);
428	if (dst && dst->ops && dst->ops->link_failure)
429		dst->ops->link_failure(skb);
430}
431
432static inline void dst_set_expires(struct dst_entry *dst, int timeout)
433{
434	unsigned long expires = jiffies + timeout;
435
436	if (expires == 0)
437		expires = 1;
438
439	if (dst->expires == 0 || time_before(expires, dst->expires))
440		dst->expires = expires;
441}
442
443static inline unsigned int dst_dev_overhead(struct dst_entry *dst,
444					    struct sk_buff *skb)
445{
446	if (likely(dst))
447		return LL_RESERVED_SPACE(dst->dev);
448
449	return skb->mac_len;
450}
451
452INDIRECT_CALLABLE_DECLARE(int ip6_output(struct net *, struct sock *,
453					 struct sk_buff *));
454INDIRECT_CALLABLE_DECLARE(int ip_output(struct net *, struct sock *,
455					 struct sk_buff *));
456/* Output packet to network from transport.  */
457static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
458{
459	return INDIRECT_CALL_INET(skb_dst(skb)->output,
460				  ip6_output, ip_output,
461				  net, sk, skb);
462}
463
464INDIRECT_CALLABLE_DECLARE(int ip6_input(struct sk_buff *));
465INDIRECT_CALLABLE_DECLARE(int ip_local_deliver(struct sk_buff *));
466/* Input packet from network to transport.  */
467static inline int dst_input(struct sk_buff *skb)
468{
469	return INDIRECT_CALL_INET(skb_dst(skb)->input,
470				  ip6_input, ip_local_deliver, skb);
471}
472
473INDIRECT_CALLABLE_DECLARE(struct dst_entry *ip6_dst_check(struct dst_entry *,
474							  u32));
475INDIRECT_CALLABLE_DECLARE(struct dst_entry *ipv4_dst_check(struct dst_entry *,
476							   u32));
477static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
478{
479	if (dst->obsolete)
480		dst = INDIRECT_CALL_INET(dst->ops->check, ip6_dst_check,
481					 ipv4_dst_check, dst, cookie);
482	return dst;
483}
484
 
 
485/* Flags for xfrm_lookup flags argument. */
486enum {
487	XFRM_LOOKUP_ICMP = 1 << 0,
488	XFRM_LOOKUP_QUEUE = 1 << 1,
489	XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
490};
491
492struct flowi;
493#ifndef CONFIG_XFRM
494static inline struct dst_entry *xfrm_lookup(struct net *net,
495					    struct dst_entry *dst_orig,
496					    const struct flowi *fl,
497					    const struct sock *sk,
498					    int flags)
499{
500	return dst_orig;
501}
502
503static inline struct dst_entry *
504xfrm_lookup_with_ifid(struct net *net, struct dst_entry *dst_orig,
505		      const struct flowi *fl, const struct sock *sk,
506		      int flags, u32 if_id)
507{
508	return dst_orig;
509}
510
511static inline struct dst_entry *xfrm_lookup_route(struct net *net,
512						  struct dst_entry *dst_orig,
513						  const struct flowi *fl,
514						  const struct sock *sk,
515						  int flags)
516{
517	return dst_orig;
518}
519
520static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
521{
522	return NULL;
523}
524
525#else
526struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
527			      const struct flowi *fl, const struct sock *sk,
528			      int flags);
529
530struct dst_entry *xfrm_lookup_with_ifid(struct net *net,
531					struct dst_entry *dst_orig,
532					const struct flowi *fl,
533					const struct sock *sk, int flags,
534					u32 if_id);
535
536struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
537				    const struct flowi *fl, const struct sock *sk,
538				    int flags);
539
540/* skb attached with this dst needs transformation if dst->xfrm is valid */
541static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
542{
543	return dst->xfrm;
544}
545#endif
546
547static inline void skb_dst_update_pmtu(struct sk_buff *skb, u32 mtu)
548{
549	struct dst_entry *dst = skb_dst(skb);
550
551	if (dst && dst->ops->update_pmtu)
552		dst->ops->update_pmtu(dst, NULL, skb, mtu, true);
553}
554
555/* update dst pmtu but not do neighbor confirm */
556static inline void skb_dst_update_pmtu_no_confirm(struct sk_buff *skb, u32 mtu)
557{
558	struct dst_entry *dst = skb_dst(skb);
559
560	if (dst && dst->ops->update_pmtu)
561		dst->ops->update_pmtu(dst, NULL, skb, mtu, false);
562}
563
564struct dst_entry *dst_blackhole_check(struct dst_entry *dst, u32 cookie);
565void dst_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
566			       struct sk_buff *skb, u32 mtu, bool confirm_neigh);
567void dst_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
568			    struct sk_buff *skb);
569u32 *dst_blackhole_cow_metrics(struct dst_entry *dst, unsigned long old);
570struct neighbour *dst_blackhole_neigh_lookup(const struct dst_entry *dst,
571					     struct sk_buff *skb,
572					     const void *daddr);
573unsigned int dst_blackhole_mtu(const struct dst_entry *dst);
574
575#endif /* _NET_DST_H */
v4.10.11
 
  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
 31struct sk_buff;
 32
 33struct 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 net *net, 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#ifdef CONFIG_64BIT
 88	/*
 89	 * Align __refcnt to a 64 bytes alignment
 90	 * (L1_CACHE_SIZE would be too much)
 91	 */
 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	struct lwtunnel_state   *lwtstate;
102	union {
103		struct dst_entry	*next;
104		struct rtable __rcu	*rt_next;
105		struct rt6_info		*rt6_next;
106		struct dn_route __rcu	*dn_next;
107	};
108};
109
 
 
 
 
 
 
110u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
111extern const u32 dst_default_metrics[];
112
113#define DST_METRICS_READ_ONLY		0x1UL
 
114#define DST_METRICS_FLAGS		0x3UL
115#define __DST_METRICS_PTR(Y)	\
116	((u32 *)((Y) & ~DST_METRICS_FLAGS))
117#define DST_METRICS_PTR(X)	__DST_METRICS_PTR((X)->_metrics)
118
119static inline bool dst_metrics_read_only(const struct dst_entry *dst)
120{
121	return dst->_metrics & DST_METRICS_READ_ONLY;
122}
123
124void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
125
126static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
127{
128	unsigned long val = dst->_metrics;
129	if (!(val & DST_METRICS_READ_ONLY))
130		__dst_destroy_metrics_generic(dst, val);
131}
132
133static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
134{
135	unsigned long p = dst->_metrics;
136
137	BUG_ON(!p);
138
139	if (p & DST_METRICS_READ_ONLY)
140		return dst->ops->cow_metrics(dst, p);
141	return __DST_METRICS_PTR(p);
142}
143
144/* This may only be invoked before the entry has reached global
145 * visibility.
146 */
147static inline void dst_init_metrics(struct dst_entry *dst,
148				    const u32 *src_metrics,
149				    bool read_only)
150{
151	dst->_metrics = ((unsigned long) src_metrics) |
152		(read_only ? DST_METRICS_READ_ONLY : 0);
153}
154
155static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
156{
157	u32 *dst_metrics = dst_metrics_write_ptr(dest);
158
159	if (dst_metrics) {
160		u32 *src_metrics = DST_METRICS_PTR(src);
161
162		memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
163	}
164}
165
166static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
167{
168	return DST_METRICS_PTR(dst);
169}
170
171static inline u32
172dst_metric_raw(const struct dst_entry *dst, const int metric)
173{
174	u32 *p = DST_METRICS_PTR(dst);
175
176	return p[metric-1];
177}
178
179static inline u32
180dst_metric(const struct dst_entry *dst, const int metric)
181{
182	WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
183		     metric == RTAX_ADVMSS ||
184		     metric == RTAX_MTU);
185	return dst_metric_raw(dst, metric);
186}
187
188static inline u32
189dst_metric_advmss(const struct dst_entry *dst)
190{
191	u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
192
193	if (!advmss)
194		advmss = dst->ops->default_advmss(dst);
195
196	return advmss;
197}
198
199static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
200{
201	u32 *p = dst_metrics_write_ptr(dst);
202
203	if (p)
204		p[metric-1] = val;
205}
206
207/* Kernel-internal feature bits that are unallocated in user space. */
208#define DST_FEATURE_ECN_CA	(1 << 31)
209
210#define DST_FEATURE_MASK	(DST_FEATURE_ECN_CA)
211#define DST_FEATURE_ECN_MASK	(DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
212
213static inline u32
214dst_feature(const struct dst_entry *dst, u32 feature)
215{
216	return dst_metric(dst, RTAX_FEATURES) & feature;
217}
218
 
 
219static inline u32 dst_mtu(const struct dst_entry *dst)
220{
221	return dst->ops->mtu(dst);
222}
223
224/* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
225static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
226{
227	return msecs_to_jiffies(dst_metric(dst, metric));
228}
229
230static inline u32
231dst_allfrag(const struct dst_entry *dst)
232{
233	int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
234	return ret;
235}
236
237static inline int
238dst_metric_locked(const struct dst_entry *dst, int metric)
239{
240	return dst_metric(dst, RTAX_LOCK) & (1<<metric);
241}
242
243static inline void dst_hold(struct dst_entry *dst)
244{
245	/*
246	 * If your kernel compilation stops here, please check
247	 * __pad_to_align_refcnt declaration in struct dst_entry
248	 */
249	BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
250	atomic_inc(&dst->__refcnt);
251}
252
253static inline void dst_use(struct dst_entry *dst, unsigned long time)
254{
255	dst_hold(dst);
256	dst->__use++;
257	dst->lastuse = time;
258}
259
260static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
261{
262	dst->__use++;
263	dst->lastuse = time;
 
 
264}
265
266static inline struct dst_entry *dst_clone(struct dst_entry *dst)
267{
268	if (dst)
269		atomic_inc(&dst->__refcnt);
270	return dst;
271}
272
273void dst_release(struct dst_entry *dst);
274
 
 
275static inline void refdst_drop(unsigned long refdst)
276{
277	if (!(refdst & SKB_DST_NOREF))
278		dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
279}
280
281/**
282 * skb_dst_drop - drops skb dst
283 * @skb: buffer
284 *
285 * Drops dst reference count if a reference was taken.
286 */
287static inline void skb_dst_drop(struct sk_buff *skb)
288{
289	if (skb->_skb_refdst) {
290		refdst_drop(skb->_skb_refdst);
291		skb->_skb_refdst = 0UL;
292	}
293}
294
295static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
296{
 
297	nskb->_skb_refdst = refdst;
298	if (!(nskb->_skb_refdst & SKB_DST_NOREF))
299		dst_clone(skb_dst(nskb));
300}
301
302static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
303{
304	__skb_dst_copy(nskb, oskb->_skb_refdst);
305}
306
307/**
308 * skb_dst_force - makes sure skb dst is refcounted
309 * @skb: buffer
310 *
311 * If dst is not yet refcounted, let's do it
312 */
313static inline void skb_dst_force(struct sk_buff *skb)
314{
315	if (skb_dst_is_noref(skb)) {
316		WARN_ON(!rcu_read_lock_held());
317		skb->_skb_refdst &= ~SKB_DST_NOREF;
318		dst_clone(skb_dst(skb));
319	}
320}
321
322/**
323 * dst_hold_safe - Take a reference on a dst if possible
324 * @dst: pointer to dst entry
325 *
326 * This helper returns false if it could not safely
327 * take a reference on a dst.
328 */
329static inline bool dst_hold_safe(struct dst_entry *dst)
330{
331	if (dst->flags & DST_NOCACHE)
332		return atomic_inc_not_zero(&dst->__refcnt);
333	dst_hold(dst);
334	return true;
335}
336
337/**
338 * skb_dst_force_safe - makes sure skb dst is refcounted
339 * @skb: buffer
340 *
341 * If dst is not yet refcounted and not destroyed, grab a ref on it.
 
342 */
343static inline void skb_dst_force_safe(struct sk_buff *skb)
344{
345	if (skb_dst_is_noref(skb)) {
346		struct dst_entry *dst = skb_dst(skb);
347
 
348		if (!dst_hold_safe(dst))
349			dst = NULL;
350
351		skb->_skb_refdst = (unsigned long)dst;
 
352	}
 
 
353}
354
355
356/**
357 *	__skb_tunnel_rx - prepare skb for rx reinsert
358 *	@skb: buffer
359 *	@dev: tunnel device
360 *	@net: netns for packet i/o
361 *
362 *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
363 *	so make some cleanups. (no accounting done)
364 */
365static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
366				   struct net *net)
367{
368	skb->dev = dev;
369
370	/*
371	 * Clear hash so that we can recalulate the hash for the
372	 * encapsulated packet, unless we have already determine the hash
373	 * over the L4 4-tuple.
374	 */
375	skb_clear_hash_if_not_l4(skb);
376	skb_set_queue_mapping(skb, 0);
377	skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
378}
379
380/**
381 *	skb_tunnel_rx - prepare skb for rx reinsert
382 *	@skb: buffer
383 *	@dev: tunnel device
 
384 *
385 *	After decapsulation, packet is going to re-enter (netif_rx()) our stack,
386 *	so make some cleanups, and perform accounting.
387 *	Note: this accounting is not SMP safe.
388 */
389static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
390				 struct net *net)
391{
392	/* TODO : stats should be SMP safe */
393	dev->stats.rx_packets++;
394	dev->stats.rx_bytes += skb->len;
395	__skb_tunnel_rx(skb, dev, net);
396}
397
398static inline u32 dst_tclassid(const struct sk_buff *skb)
399{
400#ifdef CONFIG_IP_ROUTE_CLASSID
401	const struct dst_entry *dst;
402
403	dst = skb_dst(skb);
404	if (dst)
405		return dst->tclassid;
406#endif
407	return 0;
408}
409
410int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
411static inline int dst_discard(struct sk_buff *skb)
412{
413	return dst_discard_out(&init_net, skb->sk, skb);
414}
415void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
416		int initial_obsolete, unsigned short flags);
417void dst_init(struct dst_entry *dst, struct dst_ops *ops,
418	      struct net_device *dev, int initial_ref, int initial_obsolete,
419	      unsigned short flags);
420void __dst_free(struct dst_entry *dst);
421struct dst_entry *dst_destroy(struct dst_entry *dst);
422
423static inline void dst_free(struct dst_entry *dst)
424{
425	if (dst->obsolete > 0)
426		return;
427	if (!atomic_read(&dst->__refcnt)) {
428		dst = dst_destroy(dst);
429		if (!dst)
430			return;
431	}
432	__dst_free(dst);
433}
434
435static inline void dst_rcu_free(struct rcu_head *head)
436{
437	struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
438	dst_free(dst);
439}
440
441static inline void dst_confirm(struct dst_entry *dst)
442{
443	dst->pending_confirm = 1;
444}
445
446static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n,
447				   struct sk_buff *skb)
448{
449	const struct hh_cache *hh;
450
451	if (dst->pending_confirm) {
452		unsigned long now = jiffies;
453
454		dst->pending_confirm = 0;
455		/* avoid dirtying neighbour */
456		if (n->confirmed != now)
457			n->confirmed = now;
458	}
459
460	hh = &n->hh;
461	if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
462		return neigh_hh_output(hh, skb);
463	else
464		return n->output(n, skb);
465}
466
467static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
468{
469	struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
470	return IS_ERR(n) ? NULL : n;
471}
472
473static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
474						     struct sk_buff *skb)
475{
476	struct neighbour *n =  dst->ops->neigh_lookup(dst, skb, NULL);
477	return IS_ERR(n) ? NULL : n;
478}
479
480static inline void dst_link_failure(struct sk_buff *skb)
481{
482	struct dst_entry *dst = skb_dst(skb);
483	if (dst && dst->ops && dst->ops->link_failure)
484		dst->ops->link_failure(skb);
485}
486
487static inline void dst_set_expires(struct dst_entry *dst, int timeout)
488{
489	unsigned long expires = jiffies + timeout;
490
491	if (expires == 0)
492		expires = 1;
493
494	if (dst->expires == 0 || time_before(expires, dst->expires))
495		dst->expires = expires;
496}
497
 
 
 
 
 
 
 
 
 
 
 
 
 
498/* Output packet to network from transport.  */
499static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
500{
501	return skb_dst(skb)->output(net, sk, skb);
 
 
502}
503
 
 
504/* Input packet from network to transport.  */
505static inline int dst_input(struct sk_buff *skb)
506{
507	return skb_dst(skb)->input(skb);
 
508}
509
 
 
 
 
510static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
511{
512	if (dst->obsolete)
513		dst = dst->ops->check(dst, cookie);
 
514	return dst;
515}
516
517void dst_subsys_init(void);
518
519/* Flags for xfrm_lookup flags argument. */
520enum {
521	XFRM_LOOKUP_ICMP = 1 << 0,
522	XFRM_LOOKUP_QUEUE = 1 << 1,
523	XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
524};
525
526struct flowi;
527#ifndef CONFIG_XFRM
528static inline struct dst_entry *xfrm_lookup(struct net *net,
529					    struct dst_entry *dst_orig,
530					    const struct flowi *fl,
531					    const struct sock *sk,
532					    int flags)
533{
534	return dst_orig;
535}
536
 
 
 
 
 
 
 
 
537static inline struct dst_entry *xfrm_lookup_route(struct net *net,
538						  struct dst_entry *dst_orig,
539						  const struct flowi *fl,
540						  const struct sock *sk,
541						  int flags)
542{
543	return dst_orig;
544}
545
546static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
547{
548	return NULL;
549}
550
551#else
552struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
553			      const struct flowi *fl, const struct sock *sk,
554			      int flags);
555
 
 
 
 
 
 
556struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
557				    const struct flowi *fl, const struct sock *sk,
558				    int flags);
559
560/* skb attached with this dst needs transformation if dst->xfrm is valid */
561static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
562{
563	return dst->xfrm;
564}
565#endif
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
566
567#endif /* _NET_DST_H */