dst.h 11 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478
  1. /*
  2. * net/dst.h Protocol independent destination cache definitions.
  3. *
  4. * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  5. *
  6. */
  7. #ifndef _NET_DST_H
  8. #define _NET_DST_H
  9. #include <net/dst_ops.h>
  10. #include <linux/netdevice.h>
  11. #include <linux/rtnetlink.h>
  12. #include <linux/rcupdate.h>
  13. #include <linux/bug.h>
  14. #include <linux/jiffies.h>
  15. #include <net/neighbour.h>
  16. #include <asm/processor.h>
  17. #define DST_GC_MIN (HZ/10)
  18. #define DST_GC_INC (HZ/2)
  19. #define DST_GC_MAX (120*HZ)
  20. /* Each dst_entry has reference count and sits in some parent list(s).
  21. * When it is removed from parent list, it is "freed" (dst_free).
  22. * After this it enters dead state (dst->obsolete > 0) and if its refcnt
  23. * is zero, it can be destroyed immediately, otherwise it is added
  24. * to gc list and garbage collector periodically checks the refcnt.
  25. */
  26. struct sk_buff;
  27. struct dst_entry {
  28. struct rcu_head rcu_head;
  29. struct dst_entry *child;
  30. struct net_device *dev;
  31. struct dst_ops *ops;
  32. unsigned long _metrics;
  33. union {
  34. unsigned long expires;
  35. /* point to where the dst_entry copied from */
  36. struct dst_entry *from;
  37. };
  38. struct dst_entry *path;
  39. void *__pad0;
  40. #ifdef CONFIG_XFRM
  41. struct xfrm_state *xfrm;
  42. #else
  43. void *__pad1;
  44. #endif
  45. int (*input)(struct sk_buff *);
  46. int (*output)(struct sk_buff *);
  47. unsigned short flags;
  48. #define DST_HOST 0x0001
  49. #define DST_NOXFRM 0x0002
  50. #define DST_NOPOLICY 0x0004
  51. #define DST_NOHASH 0x0008
  52. #define DST_NOCACHE 0x0010
  53. #define DST_NOCOUNT 0x0020
  54. #define DST_NOPEER 0x0040
  55. #define DST_FAKE_RTABLE 0x0080
  56. #define DST_XFRM_TUNNEL 0x0100
  57. #define DST_RCU_FREE 0x0200
  58. unsigned short pending_confirm;
  59. short error;
  60. /* A non-zero value of dst->obsolete forces by-hand validation
  61. * of the route entry. Positive values are set by the generic
  62. * dst layer to indicate that the entry has been forcefully
  63. * destroyed.
  64. *
  65. * Negative values are used by the implementation layer code to
  66. * force invocation of the dst_ops->check() method.
  67. */
  68. short obsolete;
  69. #define DST_OBSOLETE_NONE 0
  70. #define DST_OBSOLETE_DEAD 2
  71. #define DST_OBSOLETE_FORCE_CHK -1
  72. #define DST_OBSOLETE_KILL -2
  73. unsigned short header_len; /* more space at head required */
  74. unsigned short trailer_len; /* space to reserve at tail */
  75. #ifdef CONFIG_IP_ROUTE_CLASSID
  76. __u32 tclassid;
  77. #else
  78. __u32 __pad2;
  79. #endif
  80. /*
  81. * Align __refcnt to a 64 bytes alignment
  82. * (L1_CACHE_SIZE would be too much)
  83. */
  84. #ifdef CONFIG_64BIT
  85. long __pad_to_align_refcnt[2];
  86. #endif
  87. /*
  88. * __refcnt wants to be on a different cache line from
  89. * input/output/ops or performance tanks badly
  90. */
  91. atomic_t __refcnt; /* client references */
  92. int __use;
  93. unsigned long lastuse;
  94. union {
  95. struct dst_entry *next;
  96. struct rtable __rcu *rt_next;
  97. struct rt6_info *rt6_next;
  98. struct dn_route __rcu *dn_next;
  99. };
  100. };
  101. extern u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
  102. extern const u32 dst_default_metrics[RTAX_MAX];
  103. #define DST_METRICS_READ_ONLY 0x1UL
  104. #define __DST_METRICS_PTR(Y) \
  105. ((u32 *)((Y) & ~DST_METRICS_READ_ONLY))
  106. #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
  107. static inline bool dst_metrics_read_only(const struct dst_entry *dst)
  108. {
  109. return dst->_metrics & DST_METRICS_READ_ONLY;
  110. }
  111. extern void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
  112. static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
  113. {
  114. unsigned long val = dst->_metrics;
  115. if (!(val & DST_METRICS_READ_ONLY))
  116. __dst_destroy_metrics_generic(dst, val);
  117. }
  118. static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
  119. {
  120. unsigned long p = dst->_metrics;
  121. BUG_ON(!p);
  122. if (p & DST_METRICS_READ_ONLY)
  123. return dst->ops->cow_metrics(dst, p);
  124. return __DST_METRICS_PTR(p);
  125. }
  126. /* This may only be invoked before the entry has reached global
  127. * visibility.
  128. */
  129. static inline void dst_init_metrics(struct dst_entry *dst,
  130. const u32 *src_metrics,
  131. bool read_only)
  132. {
  133. dst->_metrics = ((unsigned long) src_metrics) |
  134. (read_only ? DST_METRICS_READ_ONLY : 0);
  135. }
  136. static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
  137. {
  138. u32 *dst_metrics = dst_metrics_write_ptr(dest);
  139. if (dst_metrics) {
  140. u32 *src_metrics = DST_METRICS_PTR(src);
  141. memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
  142. }
  143. }
  144. static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
  145. {
  146. return DST_METRICS_PTR(dst);
  147. }
  148. static inline u32
  149. dst_metric_raw(const struct dst_entry *dst, const int metric)
  150. {
  151. u32 *p = DST_METRICS_PTR(dst);
  152. return p[metric-1];
  153. }
  154. static inline u32
  155. dst_metric(const struct dst_entry *dst, const int metric)
  156. {
  157. WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
  158. metric == RTAX_ADVMSS ||
  159. metric == RTAX_MTU);
  160. return dst_metric_raw(dst, metric);
  161. }
  162. static inline u32
  163. dst_metric_advmss(const struct dst_entry *dst)
  164. {
  165. u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
  166. if (!advmss)
  167. advmss = dst->ops->default_advmss(dst);
  168. return advmss;
  169. }
  170. static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
  171. {
  172. u32 *p = dst_metrics_write_ptr(dst);
  173. if (p)
  174. p[metric-1] = val;
  175. }
  176. static inline u32
  177. dst_feature(const struct dst_entry *dst, u32 feature)
  178. {
  179. return dst_metric(dst, RTAX_FEATURES) & feature;
  180. }
  181. static inline u32 dst_mtu(const struct dst_entry *dst)
  182. {
  183. return dst->ops->mtu(dst);
  184. }
  185. /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
  186. static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
  187. {
  188. return msecs_to_jiffies(dst_metric(dst, metric));
  189. }
  190. static inline u32
  191. dst_allfrag(const struct dst_entry *dst)
  192. {
  193. int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG);
  194. return ret;
  195. }
  196. static inline int
  197. dst_metric_locked(const struct dst_entry *dst, int metric)
  198. {
  199. return dst_metric(dst, RTAX_LOCK) & (1<<metric);
  200. }
  201. static inline void dst_hold(struct dst_entry *dst)
  202. {
  203. /*
  204. * If your kernel compilation stops here, please check
  205. * __pad_to_align_refcnt declaration in struct dst_entry
  206. */
  207. BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
  208. atomic_inc(&dst->__refcnt);
  209. }
  210. static inline void dst_use(struct dst_entry *dst, unsigned long time)
  211. {
  212. dst_hold(dst);
  213. dst->__use++;
  214. dst->lastuse = time;
  215. }
  216. static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
  217. {
  218. dst->__use++;
  219. dst->lastuse = time;
  220. }
  221. static inline struct dst_entry *dst_clone(struct dst_entry *dst)
  222. {
  223. if (dst)
  224. atomic_inc(&dst->__refcnt);
  225. return dst;
  226. }
  227. extern void dst_release(struct dst_entry *dst);
  228. static inline void refdst_drop(unsigned long refdst)
  229. {
  230. if (!(refdst & SKB_DST_NOREF))
  231. dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
  232. }
  233. /**
  234. * skb_dst_drop - drops skb dst
  235. * @skb: buffer
  236. *
  237. * Drops dst reference count if a reference was taken.
  238. */
  239. static inline void skb_dst_drop(struct sk_buff *skb)
  240. {
  241. if (skb->_skb_refdst) {
  242. refdst_drop(skb->_skb_refdst);
  243. skb->_skb_refdst = 0UL;
  244. }
  245. }
  246. static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
  247. {
  248. nskb->_skb_refdst = oskb->_skb_refdst;
  249. if (!(nskb->_skb_refdst & SKB_DST_NOREF))
  250. dst_clone(skb_dst(nskb));
  251. }
  252. /**
  253. * skb_dst_force - makes sure skb dst is refcounted
  254. * @skb: buffer
  255. *
  256. * If dst is not yet refcounted, let's do it
  257. */
  258. static inline void skb_dst_force(struct sk_buff *skb)
  259. {
  260. if (skb_dst_is_noref(skb)) {
  261. WARN_ON(!rcu_read_lock_held());
  262. skb->_skb_refdst &= ~SKB_DST_NOREF;
  263. dst_clone(skb_dst(skb));
  264. }
  265. }
  266. /**
  267. * __skb_tunnel_rx - prepare skb for rx reinsert
  268. * @skb: buffer
  269. * @dev: tunnel device
  270. *
  271. * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
  272. * so make some cleanups. (no accounting done)
  273. */
  274. static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
  275. {
  276. skb->dev = dev;
  277. /*
  278. * Clear rxhash so that we can recalulate the hash for the
  279. * encapsulated packet, unless we have already determine the hash
  280. * over the L4 4-tuple.
  281. */
  282. if (!skb->l4_rxhash)
  283. skb->rxhash = 0;
  284. skb_set_queue_mapping(skb, 0);
  285. skb_dst_drop(skb);
  286. nf_reset(skb);
  287. }
  288. /**
  289. * skb_tunnel_rx - prepare skb for rx reinsert
  290. * @skb: buffer
  291. * @dev: tunnel device
  292. *
  293. * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
  294. * so make some cleanups, and perform accounting.
  295. * Note: this accounting is not SMP safe.
  296. */
  297. static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
  298. {
  299. /* TODO : stats should be SMP safe */
  300. dev->stats.rx_packets++;
  301. dev->stats.rx_bytes += skb->len;
  302. __skb_tunnel_rx(skb, dev);
  303. }
  304. /* Children define the path of the packet through the
  305. * Linux networking. Thus, destinations are stackable.
  306. */
  307. static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
  308. {
  309. struct dst_entry *child = dst_clone(skb_dst(skb)->child);
  310. skb_dst_drop(skb);
  311. return child;
  312. }
  313. extern int dst_discard(struct sk_buff *skb);
  314. extern void *dst_alloc(struct dst_ops *ops, struct net_device *dev,
  315. int initial_ref, int initial_obsolete,
  316. unsigned short flags);
  317. extern void __dst_free(struct dst_entry *dst);
  318. extern struct dst_entry *dst_destroy(struct dst_entry *dst);
  319. static inline void dst_free(struct dst_entry *dst)
  320. {
  321. if (dst->obsolete > 0)
  322. return;
  323. if (!atomic_read(&dst->__refcnt)) {
  324. dst = dst_destroy(dst);
  325. if (!dst)
  326. return;
  327. }
  328. __dst_free(dst);
  329. }
  330. static inline void dst_confirm(struct dst_entry *dst)
  331. {
  332. dst->pending_confirm = 1;
  333. }
  334. static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n,
  335. struct sk_buff *skb)
  336. {
  337. struct hh_cache *hh;
  338. if (unlikely(dst->pending_confirm)) {
  339. n->confirmed = jiffies;
  340. dst->pending_confirm = 0;
  341. }
  342. hh = &n->hh;
  343. if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
  344. return neigh_hh_output(hh, skb);
  345. else
  346. return n->output(n, skb);
  347. }
  348. static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
  349. {
  350. return dst->ops->neigh_lookup(dst, NULL, daddr);
  351. }
  352. static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
  353. struct sk_buff *skb)
  354. {
  355. return dst->ops->neigh_lookup(dst, skb, NULL);
  356. }
  357. static inline void dst_link_failure(struct sk_buff *skb)
  358. {
  359. struct dst_entry *dst = skb_dst(skb);
  360. if (dst && dst->ops && dst->ops->link_failure)
  361. dst->ops->link_failure(skb);
  362. }
  363. static inline void dst_set_expires(struct dst_entry *dst, int timeout)
  364. {
  365. unsigned long expires = jiffies + timeout;
  366. if (expires == 0)
  367. expires = 1;
  368. if (dst->expires == 0 || time_before(expires, dst->expires))
  369. dst->expires = expires;
  370. }
  371. /* Output packet to network from transport. */
  372. static inline int dst_output(struct sk_buff *skb)
  373. {
  374. return skb_dst(skb)->output(skb);
  375. }
  376. /* Input packet from network to transport. */
  377. static inline int dst_input(struct sk_buff *skb)
  378. {
  379. return skb_dst(skb)->input(skb);
  380. }
  381. static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
  382. {
  383. if (dst->obsolete)
  384. dst = dst->ops->check(dst, cookie);
  385. return dst;
  386. }
  387. extern void dst_init(void);
  388. /* Flags for xfrm_lookup flags argument. */
  389. enum {
  390. XFRM_LOOKUP_ICMP = 1 << 0,
  391. };
  392. struct flowi;
  393. #ifndef CONFIG_XFRM
  394. static inline struct dst_entry *xfrm_lookup(struct net *net,
  395. struct dst_entry *dst_orig,
  396. const struct flowi *fl, struct sock *sk,
  397. int flags)
  398. {
  399. return dst_orig;
  400. }
  401. #else
  402. extern struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
  403. const struct flowi *fl, struct sock *sk,
  404. int flags);
  405. #endif
  406. #endif /* _NET_DST_H */