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 #define DST_FAKE_RTABLE 0x0080 63 64 short error; 65 short obsolete; 66 unsigned short header_len; /* more space at head required */ 67 unsigned short trailer_len; /* space to reserve at tail */ 68 #ifdef CONFIG_IP_ROUTE_CLASSID 69 __u32 tclassid; 70 #else 71 __u32 __pad2; 72 #endif 73 74 /* 75 * Align __refcnt to a 64 bytes alignment 76 * (L1_CACHE_SIZE would be too much) 77 */ 78 #ifdef CONFIG_64BIT 79 long __pad_to_align_refcnt[2]; 80 #endif 81 /* 82 * __refcnt wants to be on a different cache line from 83 * input/output/ops or performance tanks badly 84 */ 85 atomic_t __refcnt; /* client references */ 86 int __use; 87 unsigned long lastuse; 88 union { 89 struct dst_entry *next; 90 struct rtable __rcu *rt_next; 91 struct rt6_info *rt6_next; 92 struct dn_route __rcu *dn_next; 93 }; 94 }; 95 96 static inline struct neighbour *dst_get_neighbour_noref(struct dst_entry *dst) 97 { 98 return rcu_dereference(dst->_neighbour); 99 } 100 101 static inline struct neighbour *dst_get_neighbour_noref_raw(struct dst_entry *dst) 102 { 103 return rcu_dereference_raw(dst->_neighbour); 104 } 105 106 static inline void dst_set_neighbour(struct dst_entry *dst, struct neighbour *neigh) 107 { 108 rcu_assign_pointer(dst->_neighbour, neigh); 109 } 110 111 extern u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old); 112 extern const u32 dst_default_metrics[RTAX_MAX]; 113 114 #define DST_METRICS_READ_ONLY 0x1UL 115 #define __DST_METRICS_PTR(Y) \ 116 ((u32 *)((Y) & ~DST_METRICS_READ_ONLY)) 117 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics) 118 119 static inline bool dst_metrics_read_only(const struct dst_entry *dst) 120 { 121 return dst->_metrics & DST_METRICS_READ_ONLY; 122 } 123 124 extern void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old); 125 126 static 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 } 132 133 static 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 */ 147 static 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 } 154 155 static 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 } 165 166 static inline u32 *dst_metrics_ptr(struct dst_entry *dst) 167 { 168 return DST_METRICS_PTR(dst); 169 } 170 171 static inline u32 172 dst_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 } 178 179 static inline u32 180 dst_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 } 187 188 static inline u32 189 dst_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 } 198 199 static 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 static inline u32 208 dst_feature(const struct dst_entry *dst, u32 feature) 209 { 210 return dst_metric(dst, RTAX_FEATURES) & feature; 211 } 212 213 static inline u32 dst_mtu(const struct dst_entry *dst) 214 { 215 return dst->ops->mtu(dst); 216 } 217 218 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */ 219 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric) 220 { 221 return msecs_to_jiffies(dst_metric(dst, metric)); 222 } 223 224 static inline void set_dst_metric_rtt(struct dst_entry *dst, int metric, 225 unsigned long rtt) 226 { 227 dst_metric_set(dst, metric, jiffies_to_msecs(rtt)); 228 } 229 230 static inline u32 231 dst_allfrag(const struct dst_entry *dst) 232 { 233 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG); 234 return ret; 235 } 236 237 static inline int 238 dst_metric_locked(const struct dst_entry *dst, int metric) 239 { 240 return dst_metric(dst, RTAX_LOCK) & (1<<metric); 241 } 242 243 static 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 } 252 253 static inline void dst_use(struct dst_entry *dst, unsigned long time) 254 { 255 dst_hold(dst); 256 dst->__use++; 257 dst->lastuse = time; 258 } 259 260 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time) 261 { 262 dst->__use++; 263 dst->lastuse = time; 264 } 265 266 static inline 267 struct dst_entry * dst_clone(struct dst_entry * dst) 268 { 269 if (dst) 270 atomic_inc(&dst->__refcnt); 271 return dst; 272 } 273 274 extern void dst_release(struct dst_entry *dst); 275 276 static inline void refdst_drop(unsigned long refdst) 277 { 278 if (!(refdst & SKB_DST_NOREF)) 279 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK)); 280 } 281 282 /** 283 * skb_dst_drop - drops skb dst 284 * @skb: buffer 285 * 286 * Drops dst reference count if a reference was taken. 287 */ 288 static inline void skb_dst_drop(struct sk_buff *skb) 289 { 290 if (skb->_skb_refdst) { 291 refdst_drop(skb->_skb_refdst); 292 skb->_skb_refdst = 0UL; 293 } 294 } 295 296 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb) 297 { 298 nskb->_skb_refdst = oskb->_skb_refdst; 299 if (!(nskb->_skb_refdst & SKB_DST_NOREF)) 300 dst_clone(skb_dst(nskb)); 301 } 302 303 /** 304 * skb_dst_force - makes sure skb dst is refcounted 305 * @skb: buffer 306 * 307 * If dst is not yet refcounted, let's do it 308 */ 309 static inline void skb_dst_force(struct sk_buff *skb) 310 { 311 if (skb_dst_is_noref(skb)) { 312 WARN_ON(!rcu_read_lock_held()); 313 skb->_skb_refdst &= ~SKB_DST_NOREF; 314 dst_clone(skb_dst(skb)); 315 } 316 } 317 318 319 /** 320 * __skb_tunnel_rx - prepare skb for rx reinsert 321 * @skb: buffer 322 * @dev: tunnel device 323 * 324 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 325 * so make some cleanups. (no accounting done) 326 */ 327 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev) 328 { 329 skb->dev = dev; 330 331 /* 332 * Clear rxhash so that we can recalulate the hash for the 333 * encapsulated packet, unless we have already determine the hash 334 * over the L4 4-tuple. 335 */ 336 if (!skb->l4_rxhash) 337 skb->rxhash = 0; 338 skb_set_queue_mapping(skb, 0); 339 skb_dst_drop(skb); 340 nf_reset(skb); 341 } 342 343 /** 344 * skb_tunnel_rx - prepare skb for rx reinsert 345 * @skb: buffer 346 * @dev: tunnel device 347 * 348 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 349 * so make some cleanups, and perform accounting. 350 * Note: this accounting is not SMP safe. 351 */ 352 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev) 353 { 354 /* TODO : stats should be SMP safe */ 355 dev->stats.rx_packets++; 356 dev->stats.rx_bytes += skb->len; 357 __skb_tunnel_rx(skb, dev); 358 } 359 360 /* Children define the path of the packet through the 361 * Linux networking. Thus, destinations are stackable. 362 */ 363 364 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb) 365 { 366 struct dst_entry *child = dst_clone(skb_dst(skb)->child); 367 368 skb_dst_drop(skb); 369 return child; 370 } 371 372 extern int dst_discard(struct sk_buff *skb); 373 extern void *dst_alloc(struct dst_ops * ops, struct net_device *dev, 374 int initial_ref, int initial_obsolete, int flags); 375 extern void __dst_free(struct dst_entry * dst); 376 extern struct dst_entry *dst_destroy(struct dst_entry * dst); 377 378 static inline void dst_free(struct dst_entry * dst) 379 { 380 if (dst->obsolete > 1) 381 return; 382 if (!atomic_read(&dst->__refcnt)) { 383 dst = dst_destroy(dst); 384 if (!dst) 385 return; 386 } 387 __dst_free(dst); 388 } 389 390 static inline void dst_rcu_free(struct rcu_head *head) 391 { 392 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head); 393 dst_free(dst); 394 } 395 396 static inline void dst_confirm(struct dst_entry *dst) 397 { 398 if (dst) { 399 struct neighbour *n; 400 401 rcu_read_lock(); 402 n = dst_get_neighbour_noref(dst); 403 neigh_confirm(n); 404 rcu_read_unlock(); 405 } 406 } 407 408 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr) 409 { 410 return dst->ops->neigh_lookup(dst, daddr); 411 } 412 413 static inline void dst_link_failure(struct sk_buff *skb) 414 { 415 struct dst_entry *dst = skb_dst(skb); 416 if (dst && dst->ops && dst->ops->link_failure) 417 dst->ops->link_failure(skb); 418 } 419 420 static inline void dst_set_expires(struct dst_entry *dst, int timeout) 421 { 422 unsigned long expires = jiffies + timeout; 423 424 if (expires == 0) 425 expires = 1; 426 427 if (dst->expires == 0 || time_before(expires, dst->expires)) 428 dst->expires = expires; 429 } 430 431 /* Output packet to network from transport. */ 432 static inline int dst_output(struct sk_buff *skb) 433 { 434 return skb_dst(skb)->output(skb); 435 } 436 437 /* Input packet from network to transport. */ 438 static inline int dst_input(struct sk_buff *skb) 439 { 440 return skb_dst(skb)->input(skb); 441 } 442 443 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie) 444 { 445 if (dst->obsolete) 446 dst = dst->ops->check(dst, cookie); 447 return dst; 448 } 449 450 extern void dst_init(void); 451 452 /* Flags for xfrm_lookup flags argument. */ 453 enum { 454 XFRM_LOOKUP_ICMP = 1 << 0, 455 }; 456 457 struct flowi; 458 #ifndef CONFIG_XFRM 459 static inline struct dst_entry *xfrm_lookup(struct net *net, 460 struct dst_entry *dst_orig, 461 const struct flowi *fl, struct sock *sk, 462 int flags) 463 { 464 return dst_orig; 465 } 466 #else 467 extern struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig, 468 const struct flowi *fl, struct sock *sk, 469 int flags); 470 #endif 471 472 #endif /* _NET_DST_H */ 473