]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/net/dst.h
Merge tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
[mirror_ubuntu-bionic-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 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 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
61 unsigned short pending_confirm;
62
63 short error;
64
65 /* A non-zero value of dst->obsolete forces by-hand validation
66 * of the route entry. Positive values are set by the generic
67 * dst layer to indicate that the entry has been forcefully
68 * destroyed.
69 *
70 * Negative values are used by the implementation layer code to
71 * force invocation of the dst_ops->check() method.
72 */
73 short obsolete;
74 #define DST_OBSOLETE_NONE 0
75 #define DST_OBSOLETE_DEAD 2
76 #define DST_OBSOLETE_FORCE_CHK -1
77 #define DST_OBSOLETE_KILL -2
78 unsigned short header_len; /* more space at head required */
79 unsigned short trailer_len; /* space to reserve at tail */
80 #ifdef CONFIG_IP_ROUTE_CLASSID
81 __u32 tclassid;
82 #else
83 __u32 __pad2;
84 #endif
85
86 /*
87 * Align __refcnt to a 64 bytes alignment
88 * (L1_CACHE_SIZE would be too much)
89 */
90 #ifdef CONFIG_64BIT
91 long __pad_to_align_refcnt[2];
92 #endif
93 /*
94 * __refcnt wants to be on a different cache line from
95 * input/output/ops or performance tanks badly
96 */
97 atomic_t __refcnt; /* client references */
98 int __use;
99 unsigned long lastuse;
100 union {
101 struct dst_entry *next;
102 struct rtable __rcu *rt_next;
103 struct rt6_info *rt6_next;
104 struct dn_route __rcu *dn_next;
105 };
106 };
107
108 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
109 extern const u32 dst_default_metrics[];
110
111 #define DST_METRICS_READ_ONLY 0x1UL
112 #define DST_METRICS_FORCE_OVERWRITE 0x2UL
113 #define DST_METRICS_FLAGS 0x3UL
114 #define __DST_METRICS_PTR(Y) \
115 ((u32 *)((Y) & ~DST_METRICS_FLAGS))
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 static inline void dst_metrics_set_force_overwrite(struct dst_entry *dst)
124 {
125 dst->_metrics |= DST_METRICS_FORCE_OVERWRITE;
126 }
127
128 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
129
130 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
131 {
132 unsigned long val = dst->_metrics;
133 if (!(val & DST_METRICS_READ_ONLY))
134 __dst_destroy_metrics_generic(dst, val);
135 }
136
137 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
138 {
139 unsigned long p = dst->_metrics;
140
141 BUG_ON(!p);
142
143 if (p & DST_METRICS_READ_ONLY)
144 return dst->ops->cow_metrics(dst, p);
145 return __DST_METRICS_PTR(p);
146 }
147
148 /* This may only be invoked before the entry has reached global
149 * visibility.
150 */
151 static inline void dst_init_metrics(struct dst_entry *dst,
152 const u32 *src_metrics,
153 bool read_only)
154 {
155 dst->_metrics = ((unsigned long) src_metrics) |
156 (read_only ? DST_METRICS_READ_ONLY : 0);
157 }
158
159 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
160 {
161 u32 *dst_metrics = dst_metrics_write_ptr(dest);
162
163 if (dst_metrics) {
164 u32 *src_metrics = DST_METRICS_PTR(src);
165
166 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
167 }
168 }
169
170 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
171 {
172 return DST_METRICS_PTR(dst);
173 }
174
175 static inline u32
176 dst_metric_raw(const struct dst_entry *dst, const int metric)
177 {
178 u32 *p = DST_METRICS_PTR(dst);
179
180 return p[metric-1];
181 }
182
183 static inline u32
184 dst_metric(const struct dst_entry *dst, const int metric)
185 {
186 WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
187 metric == RTAX_ADVMSS ||
188 metric == RTAX_MTU);
189 return dst_metric_raw(dst, metric);
190 }
191
192 static inline u32
193 dst_metric_advmss(const struct dst_entry *dst)
194 {
195 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
196
197 if (!advmss)
198 advmss = dst->ops->default_advmss(dst);
199
200 return advmss;
201 }
202
203 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
204 {
205 u32 *p = dst_metrics_write_ptr(dst);
206
207 if (p)
208 p[metric-1] = val;
209 }
210
211 static inline u32
212 dst_feature(const struct dst_entry *dst, u32 feature)
213 {
214 return dst_metric(dst, RTAX_FEATURES) & feature;
215 }
216
217 static inline u32 dst_mtu(const struct dst_entry *dst)
218 {
219 return dst->ops->mtu(dst);
220 }
221
222 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
223 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
224 {
225 return msecs_to_jiffies(dst_metric(dst, metric));
226 }
227
228 static inline u32
229 dst_allfrag(const struct dst_entry *dst)
230 {
231 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG);
232 return ret;
233 }
234
235 static inline int
236 dst_metric_locked(const struct dst_entry *dst, int metric)
237 {
238 return dst_metric(dst, RTAX_LOCK) & (1<<metric);
239 }
240
241 static inline void dst_hold(struct dst_entry *dst)
242 {
243 /*
244 * If your kernel compilation stops here, please check
245 * __pad_to_align_refcnt declaration in struct dst_entry
246 */
247 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
248 atomic_inc(&dst->__refcnt);
249 }
250
251 static inline void dst_use(struct dst_entry *dst, unsigned long time)
252 {
253 dst_hold(dst);
254 dst->__use++;
255 dst->lastuse = time;
256 }
257
258 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
259 {
260 dst->__use++;
261 dst->lastuse = time;
262 }
263
264 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
265 {
266 if (dst)
267 atomic_inc(&dst->__refcnt);
268 return dst;
269 }
270
271 void dst_release(struct dst_entry *dst);
272
273 static inline void refdst_drop(unsigned long refdst)
274 {
275 if (!(refdst & SKB_DST_NOREF))
276 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
277 }
278
279 /**
280 * skb_dst_drop - drops skb dst
281 * @skb: buffer
282 *
283 * Drops dst reference count if a reference was taken.
284 */
285 static inline void skb_dst_drop(struct sk_buff *skb)
286 {
287 if (skb->_skb_refdst) {
288 refdst_drop(skb->_skb_refdst);
289 skb->_skb_refdst = 0UL;
290 }
291 }
292
293 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
294 {
295 nskb->_skb_refdst = oskb->_skb_refdst;
296 if (!(nskb->_skb_refdst & SKB_DST_NOREF))
297 dst_clone(skb_dst(nskb));
298 }
299
300 /**
301 * skb_dst_force - makes sure skb dst is refcounted
302 * @skb: buffer
303 *
304 * If dst is not yet refcounted, let's do it
305 */
306 static inline void skb_dst_force(struct sk_buff *skb)
307 {
308 if (skb_dst_is_noref(skb)) {
309 WARN_ON(!rcu_read_lock_held());
310 skb->_skb_refdst &= ~SKB_DST_NOREF;
311 dst_clone(skb_dst(skb));
312 }
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 */
325 static 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 *
345 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
346 * so make some cleanups, and perform accounting.
347 * Note: this accounting is not SMP safe.
348 */
349 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
350 struct net *net)
351 {
352 /* TODO : stats should be SMP safe */
353 dev->stats.rx_packets++;
354 dev->stats.rx_bytes += skb->len;
355 __skb_tunnel_rx(skb, dev, net);
356 }
357
358 /* Children define the path of the packet through the
359 * Linux networking. Thus, destinations are stackable.
360 */
361
362 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
363 {
364 struct dst_entry *child = dst_clone(skb_dst(skb)->child);
365
366 skb_dst_drop(skb);
367 return child;
368 }
369
370 int dst_discard_sk(struct sock *sk, struct sk_buff *skb);
371 static inline int dst_discard(struct sk_buff *skb)
372 {
373 return dst_discard_sk(skb->sk, skb);
374 }
375 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
376 int initial_obsolete, unsigned short flags);
377 void __dst_free(struct dst_entry *dst);
378 struct dst_entry *dst_destroy(struct dst_entry *dst);
379
380 static inline void dst_free(struct dst_entry *dst)
381 {
382 if (dst->obsolete > 0)
383 return;
384 if (!atomic_read(&dst->__refcnt)) {
385 dst = dst_destroy(dst);
386 if (!dst)
387 return;
388 }
389 __dst_free(dst);
390 }
391
392 static inline void dst_rcu_free(struct rcu_head *head)
393 {
394 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
395 dst_free(dst);
396 }
397
398 static inline void dst_confirm(struct dst_entry *dst)
399 {
400 dst->pending_confirm = 1;
401 }
402
403 static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n,
404 struct sk_buff *skb)
405 {
406 const struct hh_cache *hh;
407
408 if (dst->pending_confirm) {
409 unsigned long now = jiffies;
410
411 dst->pending_confirm = 0;
412 /* avoid dirtying neighbour */
413 if (n->confirmed != now)
414 n->confirmed = now;
415 }
416
417 hh = &n->hh;
418 if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
419 return neigh_hh_output(hh, skb);
420 else
421 return n->output(n, skb);
422 }
423
424 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
425 {
426 struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
427 return IS_ERR(n) ? NULL : n;
428 }
429
430 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
431 struct sk_buff *skb)
432 {
433 struct neighbour *n = dst->ops->neigh_lookup(dst, skb, NULL);
434 return IS_ERR(n) ? NULL : n;
435 }
436
437 static inline void dst_link_failure(struct sk_buff *skb)
438 {
439 struct dst_entry *dst = skb_dst(skb);
440 if (dst && dst->ops && dst->ops->link_failure)
441 dst->ops->link_failure(skb);
442 }
443
444 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
445 {
446 unsigned long expires = jiffies + timeout;
447
448 if (expires == 0)
449 expires = 1;
450
451 if (dst->expires == 0 || time_before(expires, dst->expires))
452 dst->expires = expires;
453 }
454
455 /* Output packet to network from transport. */
456 static inline int dst_output_sk(struct sock *sk, struct sk_buff *skb)
457 {
458 return skb_dst(skb)->output(sk, skb);
459 }
460 static inline int dst_output(struct sk_buff *skb)
461 {
462 return dst_output_sk(skb->sk, skb);
463 }
464
465 /* Input packet from network to transport. */
466 static inline int dst_input(struct sk_buff *skb)
467 {
468 return skb_dst(skb)->input(skb);
469 }
470
471 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
472 {
473 if (dst->obsolete)
474 dst = dst->ops->check(dst, cookie);
475 return dst;
476 }
477
478 void dst_init(void);
479
480 /* Flags for xfrm_lookup flags argument. */
481 enum {
482 XFRM_LOOKUP_ICMP = 1 << 0,
483 XFRM_LOOKUP_QUEUE = 1 << 1,
484 XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
485 };
486
487 struct flowi;
488 #ifndef CONFIG_XFRM
489 static inline struct dst_entry *xfrm_lookup(struct net *net,
490 struct dst_entry *dst_orig,
491 const struct flowi *fl, struct sock *sk,
492 int flags)
493 {
494 return dst_orig;
495 }
496
497 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
498 struct dst_entry *dst_orig,
499 const struct flowi *fl,
500 struct sock *sk,
501 int flags)
502 {
503 return dst_orig;
504 }
505
506 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
507 {
508 return NULL;
509 }
510
511 #else
512 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
513 const struct flowi *fl, struct sock *sk,
514 int flags);
515
516 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
517 const struct flowi *fl, struct sock *sk,
518 int flags);
519
520 /* skb attached with this dst needs transformation if dst->xfrm is valid */
521 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
522 {
523 return dst->xfrm;
524 }
525 #endif
526
527 #endif /* _NET_DST_H */