]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - include/net/dst.h
Merge branch 'tip/perf/core-2' of git://git.kernel.org/pub/scm/linux/kernel/git/roste...
[mirror_ubuntu-zesty-kernel.git] / include / net / dst.h
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 union {
40 unsigned long expires;
41 /* point to where the dst_entry copied from */
42 struct dst_entry *from;
43 };
44 struct dst_entry *path;
45 struct neighbour __rcu *_neighbour;
46 #ifdef CONFIG_XFRM
47 struct xfrm_state *xfrm;
48 #else
49 void *__pad1;
50 #endif
51 int (*input)(struct sk_buff*);
52 int (*output)(struct sk_buff*);
53
54 int flags;
55 #define DST_HOST 0x0001
56 #define DST_NOXFRM 0x0002
57 #define DST_NOPOLICY 0x0004
58 #define DST_NOHASH 0x0008
59 #define DST_NOCACHE 0x0010
60 #define DST_NOCOUNT 0x0020
61 #define DST_NOPEER 0x0040
62
63 short error;
64 short obsolete;
65 unsigned short header_len; /* more space at head required */
66 unsigned short trailer_len; /* space to reserve at tail */
67 #ifdef CONFIG_IP_ROUTE_CLASSID
68 __u32 tclassid;
69 #else
70 __u32 __pad2;
71 #endif
72
73 /*
74 * Align __refcnt to a 64 bytes alignment
75 * (L1_CACHE_SIZE would be too much)
76 */
77 #ifdef CONFIG_64BIT
78 long __pad_to_align_refcnt[2];
79 #endif
80 /*
81 * __refcnt wants to be on a different cache line from
82 * input/output/ops or performance tanks badly
83 */
84 atomic_t __refcnt; /* client references */
85 int __use;
86 unsigned long lastuse;
87 union {
88 struct dst_entry *next;
89 struct rtable __rcu *rt_next;
90 struct rt6_info *rt6_next;
91 struct dn_route __rcu *dn_next;
92 };
93 };
94
95 static inline struct neighbour *dst_get_neighbour_noref(struct dst_entry *dst)
96 {
97 return rcu_dereference(dst->_neighbour);
98 }
99
100 static inline struct neighbour *dst_get_neighbour_noref_raw(struct dst_entry *dst)
101 {
102 return rcu_dereference_raw(dst->_neighbour);
103 }
104
105 static inline void dst_set_neighbour(struct dst_entry *dst, struct neighbour *neigh)
106 {
107 rcu_assign_pointer(dst->_neighbour, neigh);
108 }
109
110 extern u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
111 extern const u32 dst_default_metrics[RTAX_MAX];
112
113 #define DST_METRICS_READ_ONLY 0x1UL
114 #define __DST_METRICS_PTR(Y) \
115 ((u32 *)((Y) & ~DST_METRICS_READ_ONLY))
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 extern 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 void set_dst_metric_rtt(struct dst_entry *dst, int metric,
224 unsigned long rtt)
225 {
226 dst_metric_set(dst, metric, jiffies_to_msecs(rtt));
227 }
228
229 static inline u32
230 dst_allfrag(const struct dst_entry *dst)
231 {
232 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG);
233 return ret;
234 }
235
236 static inline int
237 dst_metric_locked(const struct dst_entry *dst, int metric)
238 {
239 return dst_metric(dst, RTAX_LOCK) & (1<<metric);
240 }
241
242 static inline void dst_hold(struct dst_entry * dst)
243 {
244 /*
245 * If your kernel compilation stops here, please check
246 * __pad_to_align_refcnt declaration in struct dst_entry
247 */
248 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
249 atomic_inc(&dst->__refcnt);
250 }
251
252 static inline void dst_use(struct dst_entry *dst, unsigned long time)
253 {
254 dst_hold(dst);
255 dst->__use++;
256 dst->lastuse = time;
257 }
258
259 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
260 {
261 dst->__use++;
262 dst->lastuse = time;
263 }
264
265 static inline
266 struct dst_entry * dst_clone(struct dst_entry * dst)
267 {
268 if (dst)
269 atomic_inc(&dst->__refcnt);
270 return dst;
271 }
272
273 extern void dst_release(struct dst_entry *dst);
274
275 static 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 */
287 static 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
295 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
296 {
297 nskb->_skb_refdst = oskb->_skb_refdst;
298 if (!(nskb->_skb_refdst & SKB_DST_NOREF))
299 dst_clone(skb_dst(nskb));
300 }
301
302 /**
303 * skb_dst_force - makes sure skb dst is refcounted
304 * @skb: buffer
305 *
306 * If dst is not yet refcounted, let's do it
307 */
308 static inline void skb_dst_force(struct sk_buff *skb)
309 {
310 if (skb_dst_is_noref(skb)) {
311 WARN_ON(!rcu_read_lock_held());
312 skb->_skb_refdst &= ~SKB_DST_NOREF;
313 dst_clone(skb_dst(skb));
314 }
315 }
316
317
318 /**
319 * __skb_tunnel_rx - prepare skb for rx reinsert
320 * @skb: buffer
321 * @dev: tunnel device
322 *
323 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
324 * so make some cleanups. (no accounting done)
325 */
326 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
327 {
328 skb->dev = dev;
329
330 /*
331 * Clear rxhash 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 if (!skb->l4_rxhash)
336 skb->rxhash = 0;
337 skb_set_queue_mapping(skb, 0);
338 skb_dst_drop(skb);
339 nf_reset(skb);
340 }
341
342 /**
343 * skb_tunnel_rx - prepare skb for rx reinsert
344 * @skb: buffer
345 * @dev: tunnel device
346 *
347 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
348 * so make some cleanups, and perform accounting.
349 * Note: this accounting is not SMP safe.
350 */
351 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
352 {
353 /* TODO : stats should be SMP safe */
354 dev->stats.rx_packets++;
355 dev->stats.rx_bytes += skb->len;
356 __skb_tunnel_rx(skb, dev);
357 }
358
359 /* Children define the path of the packet through the
360 * Linux networking. Thus, destinations are stackable.
361 */
362
363 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
364 {
365 struct dst_entry *child = dst_clone(skb_dst(skb)->child);
366
367 skb_dst_drop(skb);
368 return child;
369 }
370
371 extern int dst_discard(struct sk_buff *skb);
372 extern void *dst_alloc(struct dst_ops * ops, struct net_device *dev,
373 int initial_ref, int initial_obsolete, int flags);
374 extern void __dst_free(struct dst_entry * dst);
375 extern struct dst_entry *dst_destroy(struct dst_entry * dst);
376
377 static inline void dst_free(struct dst_entry * dst)
378 {
379 if (dst->obsolete > 1)
380 return;
381 if (!atomic_read(&dst->__refcnt)) {
382 dst = dst_destroy(dst);
383 if (!dst)
384 return;
385 }
386 __dst_free(dst);
387 }
388
389 static inline void dst_rcu_free(struct rcu_head *head)
390 {
391 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
392 dst_free(dst);
393 }
394
395 static inline void dst_confirm(struct dst_entry *dst)
396 {
397 if (dst) {
398 struct neighbour *n;
399
400 rcu_read_lock();
401 n = dst_get_neighbour_noref(dst);
402 neigh_confirm(n);
403 rcu_read_unlock();
404 }
405 }
406
407 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
408 {
409 return dst->ops->neigh_lookup(dst, daddr);
410 }
411
412 static inline void dst_link_failure(struct sk_buff *skb)
413 {
414 struct dst_entry *dst = skb_dst(skb);
415 if (dst && dst->ops && dst->ops->link_failure)
416 dst->ops->link_failure(skb);
417 }
418
419 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
420 {
421 unsigned long expires = jiffies + timeout;
422
423 if (expires == 0)
424 expires = 1;
425
426 if (dst->expires == 0 || time_before(expires, dst->expires))
427 dst->expires = expires;
428 }
429
430 /* Output packet to network from transport. */
431 static inline int dst_output(struct sk_buff *skb)
432 {
433 return skb_dst(skb)->output(skb);
434 }
435
436 /* Input packet from network to transport. */
437 static inline int dst_input(struct sk_buff *skb)
438 {
439 return skb_dst(skb)->input(skb);
440 }
441
442 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
443 {
444 if (dst->obsolete)
445 dst = dst->ops->check(dst, cookie);
446 return dst;
447 }
448
449 extern void dst_init(void);
450
451 /* Flags for xfrm_lookup flags argument. */
452 enum {
453 XFRM_LOOKUP_ICMP = 1 << 0,
454 };
455
456 struct flowi;
457 #ifndef CONFIG_XFRM
458 static inline struct dst_entry *xfrm_lookup(struct net *net,
459 struct dst_entry *dst_orig,
460 const struct flowi *fl, struct sock *sk,
461 int flags)
462 {
463 return dst_orig;
464 }
465 #else
466 extern struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
467 const struct flowi *fl, struct sock *sk,
468 int flags);
469 #endif
470
471 #endif /* _NET_DST_H */