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/jiffies.h> 16 #include <net/neighbour.h> 17 #include <asm/processor.h> 18 19 #define DST_GC_MIN (HZ/10) 20 #define DST_GC_INC (HZ/2) 21 #define DST_GC_MAX (120*HZ) 22 23 /* Each dst_entry has reference count and sits in some parent list(s). 24 * When it is removed from parent list, it is "freed" (dst_free). 25 * After this it enters dead state (dst->obsolete > 0) and if its refcnt 26 * is zero, it can be destroyed immediately, otherwise it is added 27 * to gc list and garbage collector periodically checks the refcnt. 28 */ 29 30 struct sk_buff; 31 32 struct dst_entry { 33 struct rcu_head rcu_head; 34 struct dst_entry *child; 35 struct net_device *dev; 36 struct dst_ops *ops; 37 unsigned long _metrics; 38 unsigned long expires; 39 struct dst_entry *path; 40 struct neighbour __rcu *_neighbour; 41 #ifdef CONFIG_XFRM 42 struct xfrm_state *xfrm; 43 #else 44 void *__pad1; 45 #endif 46 int (*input)(struct sk_buff*); 47 int (*output)(struct sk_buff*); 48 49 int flags; 50 #define DST_HOST 0x0001 51 #define DST_NOXFRM 0x0002 52 #define DST_NOPOLICY 0x0004 53 #define DST_NOHASH 0x0008 54 #define DST_NOCACHE 0x0010 55 #define DST_NOCOUNT 0x0020 56 #define DST_NOPEER 0x0040 57 58 short error; 59 short obsolete; 60 unsigned short header_len; /* more space at head required */ 61 unsigned short trailer_len; /* space to reserve at tail */ 62 #ifdef CONFIG_IP_ROUTE_CLASSID 63 __u32 tclassid; 64 #else 65 __u32 __pad2; 66 #endif 67 68 /* 69 * Align __refcnt to a 64 bytes alignment 70 * (L1_CACHE_SIZE would be too much) 71 */ 72 #ifdef CONFIG_64BIT 73 long __pad_to_align_refcnt[2]; 74 #endif 75 /* 76 * __refcnt wants to be on a different cache line from 77 * input/output/ops or performance tanks badly 78 */ 79 atomic_t __refcnt; /* client references */ 80 int __use; 81 unsigned long lastuse; 82 union { 83 struct dst_entry *next; 84 struct rtable __rcu *rt_next; 85 struct rt6_info *rt6_next; 86 struct dn_route __rcu *dn_next; 87 }; 88 }; 89 90 static inline struct neighbour *dst_get_neighbour_noref(struct dst_entry *dst) 91 { 92 return rcu_dereference(dst->_neighbour); 93 } 94 95 static inline struct neighbour *dst_get_neighbour_noref_raw(struct dst_entry *dst) 96 { 97 return rcu_dereference_raw(dst->_neighbour); 98 } 99 100 static inline void dst_set_neighbour(struct dst_entry *dst, struct neighbour *neigh) 101 { 102 rcu_assign_pointer(dst->_neighbour, neigh); 103 } 104 105 extern u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old); 106 extern const u32 dst_default_metrics[RTAX_MAX]; 107 108 #define DST_METRICS_READ_ONLY 0x1UL 109 #define __DST_METRICS_PTR(Y) \ 110 ((u32 *)((Y) & ~DST_METRICS_READ_ONLY)) 111 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics) 112 113 static inline bool dst_metrics_read_only(const struct dst_entry *dst) 114 { 115 return dst->_metrics & DST_METRICS_READ_ONLY; 116 } 117 118 extern void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old); 119 120 static inline void dst_destroy_metrics_generic(struct dst_entry *dst) 121 { 122 unsigned long val = dst->_metrics; 123 if (!(val & DST_METRICS_READ_ONLY)) 124 __dst_destroy_metrics_generic(dst, val); 125 } 126 127 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst) 128 { 129 unsigned long p = dst->_metrics; 130 131 BUG_ON(!p); 132 133 if (p & DST_METRICS_READ_ONLY) 134 return dst->ops->cow_metrics(dst, p); 135 return __DST_METRICS_PTR(p); 136 } 137 138 /* This may only be invoked before the entry has reached global 139 * visibility. 140 */ 141 static inline void dst_init_metrics(struct dst_entry *dst, 142 const u32 *src_metrics, 143 bool read_only) 144 { 145 dst->_metrics = ((unsigned long) src_metrics) | 146 (read_only ? DST_METRICS_READ_ONLY : 0); 147 } 148 149 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src) 150 { 151 u32 *dst_metrics = dst_metrics_write_ptr(dest); 152 153 if (dst_metrics) { 154 u32 *src_metrics = DST_METRICS_PTR(src); 155 156 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32)); 157 } 158 } 159 160 static inline u32 *dst_metrics_ptr(struct dst_entry *dst) 161 { 162 return DST_METRICS_PTR(dst); 163 } 164 165 static inline u32 166 dst_metric_raw(const struct dst_entry *dst, const int metric) 167 { 168 u32 *p = DST_METRICS_PTR(dst); 169 170 return p[metric-1]; 171 } 172 173 static inline u32 174 dst_metric(const struct dst_entry *dst, const int metric) 175 { 176 WARN_ON_ONCE(metric == RTAX_HOPLIMIT || 177 metric == RTAX_ADVMSS || 178 metric == RTAX_MTU); 179 return dst_metric_raw(dst, metric); 180 } 181 182 static inline u32 183 dst_metric_advmss(const struct dst_entry *dst) 184 { 185 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS); 186 187 if (!advmss) 188 advmss = dst->ops->default_advmss(dst); 189 190 return advmss; 191 } 192 193 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val) 194 { 195 u32 *p = dst_metrics_write_ptr(dst); 196 197 if (p) 198 p[metric-1] = val; 199 } 200 201 static inline u32 202 dst_feature(const struct dst_entry *dst, u32 feature) 203 { 204 return dst_metric(dst, RTAX_FEATURES) & feature; 205 } 206 207 static inline u32 dst_mtu(const struct dst_entry *dst) 208 { 209 return dst->ops->mtu(dst); 210 } 211 212 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */ 213 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric) 214 { 215 return msecs_to_jiffies(dst_metric(dst, metric)); 216 } 217 218 static inline void set_dst_metric_rtt(struct dst_entry *dst, int metric, 219 unsigned long rtt) 220 { 221 dst_metric_set(dst, metric, jiffies_to_msecs(rtt)); 222 } 223 224 static inline u32 225 dst_allfrag(const struct dst_entry *dst) 226 { 227 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG); 228 return ret; 229 } 230 231 static inline int 232 dst_metric_locked(const struct dst_entry *dst, int metric) 233 { 234 return dst_metric(dst, RTAX_LOCK) & (1<<metric); 235 } 236 237 static inline void dst_hold(struct dst_entry * dst) 238 { 239 /* 240 * If your kernel compilation stops here, please check 241 * __pad_to_align_refcnt declaration in struct dst_entry 242 */ 243 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63); 244 atomic_inc(&dst->__refcnt); 245 } 246 247 static inline void dst_use(struct dst_entry *dst, unsigned long time) 248 { 249 dst_hold(dst); 250 dst->__use++; 251 dst->lastuse = time; 252 } 253 254 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time) 255 { 256 dst->__use++; 257 dst->lastuse = time; 258 } 259 260 static inline 261 struct dst_entry * dst_clone(struct dst_entry * dst) 262 { 263 if (dst) 264 atomic_inc(&dst->__refcnt); 265 return dst; 266 } 267 268 extern void dst_release(struct dst_entry *dst); 269 270 static inline void refdst_drop(unsigned long refdst) 271 { 272 if (!(refdst & SKB_DST_NOREF)) 273 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK)); 274 } 275 276 /** 277 * skb_dst_drop - drops skb dst 278 * @skb: buffer 279 * 280 * Drops dst reference count if a reference was taken. 281 */ 282 static inline void skb_dst_drop(struct sk_buff *skb) 283 { 284 if (skb->_skb_refdst) { 285 refdst_drop(skb->_skb_refdst); 286 skb->_skb_refdst = 0UL; 287 } 288 } 289 290 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb) 291 { 292 nskb->_skb_refdst = oskb->_skb_refdst; 293 if (!(nskb->_skb_refdst & SKB_DST_NOREF)) 294 dst_clone(skb_dst(nskb)); 295 } 296 297 /** 298 * skb_dst_force - makes sure skb dst is refcounted 299 * @skb: buffer 300 * 301 * If dst is not yet refcounted, let's do it 302 */ 303 static inline void skb_dst_force(struct sk_buff *skb) 304 { 305 if (skb_dst_is_noref(skb)) { 306 WARN_ON(!rcu_read_lock_held()); 307 skb->_skb_refdst &= ~SKB_DST_NOREF; 308 dst_clone(skb_dst(skb)); 309 } 310 } 311 312 313 /** 314 * __skb_tunnel_rx - prepare skb for rx reinsert 315 * @skb: buffer 316 * @dev: tunnel device 317 * 318 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 319 * so make some cleanups. (no accounting done) 320 */ 321 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev) 322 { 323 skb->dev = dev; 324 325 /* 326 * Clear rxhash so that we can recalulate the hash for the 327 * encapsulated packet, unless we have already determine the hash 328 * over the L4 4-tuple. 329 */ 330 if (!skb->l4_rxhash) 331 skb->rxhash = 0; 332 skb_set_queue_mapping(skb, 0); 333 skb_dst_drop(skb); 334 nf_reset(skb); 335 } 336 337 /** 338 * skb_tunnel_rx - prepare skb for rx reinsert 339 * @skb: buffer 340 * @dev: tunnel device 341 * 342 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 343 * so make some cleanups, and perform accounting. 344 * Note: this accounting is not SMP safe. 345 */ 346 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev) 347 { 348 /* TODO : stats should be SMP safe */ 349 dev->stats.rx_packets++; 350 dev->stats.rx_bytes += skb->len; 351 __skb_tunnel_rx(skb, dev); 352 } 353 354 /* Children define the path of the packet through the 355 * Linux networking. Thus, destinations are stackable. 356 */ 357 358 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb) 359 { 360 struct dst_entry *child = dst_clone(skb_dst(skb)->child); 361 362 skb_dst_drop(skb); 363 return child; 364 } 365 366 extern int dst_discard(struct sk_buff *skb); 367 extern void *dst_alloc(struct dst_ops * ops, struct net_device *dev, 368 int initial_ref, int initial_obsolete, int flags); 369 extern void __dst_free(struct dst_entry * dst); 370 extern struct dst_entry *dst_destroy(struct dst_entry * dst); 371 372 static inline void dst_free(struct dst_entry * dst) 373 { 374 if (dst->obsolete > 1) 375 return; 376 if (!atomic_read(&dst->__refcnt)) { 377 dst = dst_destroy(dst); 378 if (!dst) 379 return; 380 } 381 __dst_free(dst); 382 } 383 384 static inline void dst_rcu_free(struct rcu_head *head) 385 { 386 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head); 387 dst_free(dst); 388 } 389 390 static inline void dst_confirm(struct dst_entry *dst) 391 { 392 if (dst) { 393 struct neighbour *n; 394 395 rcu_read_lock(); 396 n = dst_get_neighbour_noref(dst); 397 neigh_confirm(n); 398 rcu_read_unlock(); 399 } 400 } 401 402 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr) 403 { 404 return dst->ops->neigh_lookup(dst, daddr); 405 } 406 407 static inline void dst_link_failure(struct sk_buff *skb) 408 { 409 struct dst_entry *dst = skb_dst(skb); 410 if (dst && dst->ops && dst->ops->link_failure) 411 dst->ops->link_failure(skb); 412 } 413 414 static inline void dst_set_expires(struct dst_entry *dst, int timeout) 415 { 416 unsigned long expires = jiffies + timeout; 417 418 if (expires == 0) 419 expires = 1; 420 421 if (dst->expires == 0 || time_before(expires, dst->expires)) 422 dst->expires = expires; 423 } 424 425 /* Output packet to network from transport. */ 426 static inline int dst_output(struct sk_buff *skb) 427 { 428 return skb_dst(skb)->output(skb); 429 } 430 431 /* Input packet from network to transport. */ 432 static inline int dst_input(struct sk_buff *skb) 433 { 434 return skb_dst(skb)->input(skb); 435 } 436 437 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie) 438 { 439 if (dst->obsolete) 440 dst = dst->ops->check(dst, cookie); 441 return dst; 442 } 443 444 extern void dst_init(void); 445 446 /* Flags for xfrm_lookup flags argument. */ 447 enum { 448 XFRM_LOOKUP_ICMP = 1 << 0, 449 }; 450 451 struct flowi; 452 #ifndef CONFIG_XFRM 453 static inline struct dst_entry *xfrm_lookup(struct net *net, 454 struct dst_entry *dst_orig, 455 const struct flowi *fl, struct sock *sk, 456 int flags) 457 { 458 return dst_orig; 459 } 460 #else 461 extern struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig, 462 const struct flowi *fl, struct sock *sk, 463 int flags); 464 #endif 465 466 #endif /* _NET_DST_H */ 467