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 net *net, 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 #define DST_METADATA 0x0200 61 62 unsigned short pending_confirm; 63 64 short error; 65 66 /* A non-zero value of dst->obsolete forces by-hand validation 67 * of the route entry. Positive values are set by the generic 68 * dst layer to indicate that the entry has been forcefully 69 * destroyed. 70 * 71 * Negative values are used by the implementation layer code to 72 * force invocation of the dst_ops->check() method. 73 */ 74 short obsolete; 75 #define DST_OBSOLETE_NONE 0 76 #define DST_OBSOLETE_DEAD 2 77 #define DST_OBSOLETE_FORCE_CHK -1 78 #define DST_OBSOLETE_KILL -2 79 unsigned short header_len; /* more space at head required */ 80 unsigned short trailer_len; /* space to reserve at tail */ 81 #ifdef CONFIG_IP_ROUTE_CLASSID 82 __u32 tclassid; 83 #else 84 __u32 __pad2; 85 #endif 86 87 #ifdef CONFIG_64BIT 88 struct lwtunnel_state *lwtstate; 89 /* 90 * Align __refcnt to a 64 bytes alignment 91 * (L1_CACHE_SIZE would be too much) 92 */ 93 long __pad_to_align_refcnt[1]; 94 #endif 95 /* 96 * __refcnt wants to be on a different cache line from 97 * input/output/ops or performance tanks badly 98 */ 99 atomic_t __refcnt; /* client references */ 100 int __use; 101 unsigned long lastuse; 102 #ifndef CONFIG_64BIT 103 struct lwtunnel_state *lwtstate; 104 #endif 105 union { 106 struct dst_entry *next; 107 struct rtable __rcu *rt_next; 108 struct rt6_info *rt6_next; 109 struct dn_route __rcu *dn_next; 110 }; 111 }; 112 113 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old); 114 extern const u32 dst_default_metrics[]; 115 116 #define DST_METRICS_READ_ONLY 0x1UL 117 #define DST_METRICS_FLAGS 0x3UL 118 #define __DST_METRICS_PTR(Y) \ 119 ((u32 *)((Y) & ~DST_METRICS_FLAGS)) 120 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics) 121 122 static inline bool dst_metrics_read_only(const struct dst_entry *dst) 123 { 124 return dst->_metrics & DST_METRICS_READ_ONLY; 125 } 126 127 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old); 128 129 static inline void dst_destroy_metrics_generic(struct dst_entry *dst) 130 { 131 unsigned long val = dst->_metrics; 132 if (!(val & DST_METRICS_READ_ONLY)) 133 __dst_destroy_metrics_generic(dst, val); 134 } 135 136 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst) 137 { 138 unsigned long p = dst->_metrics; 139 140 BUG_ON(!p); 141 142 if (p & DST_METRICS_READ_ONLY) 143 return dst->ops->cow_metrics(dst, p); 144 return __DST_METRICS_PTR(p); 145 } 146 147 /* This may only be invoked before the entry has reached global 148 * visibility. 149 */ 150 static inline void dst_init_metrics(struct dst_entry *dst, 151 const u32 *src_metrics, 152 bool read_only) 153 { 154 dst->_metrics = ((unsigned long) src_metrics) | 155 (read_only ? DST_METRICS_READ_ONLY : 0); 156 } 157 158 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src) 159 { 160 u32 *dst_metrics = dst_metrics_write_ptr(dest); 161 162 if (dst_metrics) { 163 u32 *src_metrics = DST_METRICS_PTR(src); 164 165 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32)); 166 } 167 } 168 169 static inline u32 *dst_metrics_ptr(struct dst_entry *dst) 170 { 171 return DST_METRICS_PTR(dst); 172 } 173 174 static inline u32 175 dst_metric_raw(const struct dst_entry *dst, const int metric) 176 { 177 u32 *p = DST_METRICS_PTR(dst); 178 179 return p[metric-1]; 180 } 181 182 static inline u32 183 dst_metric(const struct dst_entry *dst, const int metric) 184 { 185 WARN_ON_ONCE(metric == RTAX_HOPLIMIT || 186 metric == RTAX_ADVMSS || 187 metric == RTAX_MTU); 188 return dst_metric_raw(dst, metric); 189 } 190 191 static inline u32 192 dst_metric_advmss(const struct dst_entry *dst) 193 { 194 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS); 195 196 if (!advmss) 197 advmss = dst->ops->default_advmss(dst); 198 199 return advmss; 200 } 201 202 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val) 203 { 204 u32 *p = dst_metrics_write_ptr(dst); 205 206 if (p) 207 p[metric-1] = val; 208 } 209 210 /* Kernel-internal feature bits that are unallocated in user space. */ 211 #define DST_FEATURE_ECN_CA (1 << 31) 212 213 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA) 214 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN) 215 216 static inline u32 217 dst_feature(const struct dst_entry *dst, u32 feature) 218 { 219 return dst_metric(dst, RTAX_FEATURES) & feature; 220 } 221 222 static inline u32 dst_mtu(const struct dst_entry *dst) 223 { 224 return dst->ops->mtu(dst); 225 } 226 227 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */ 228 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric) 229 { 230 return msecs_to_jiffies(dst_metric(dst, metric)); 231 } 232 233 static inline u32 234 dst_allfrag(const struct dst_entry *dst) 235 { 236 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG); 237 return ret; 238 } 239 240 static inline int 241 dst_metric_locked(const struct dst_entry *dst, int metric) 242 { 243 return dst_metric(dst, RTAX_LOCK) & (1<<metric); 244 } 245 246 static inline void dst_hold(struct dst_entry *dst) 247 { 248 /* 249 * If your kernel compilation stops here, please check 250 * __pad_to_align_refcnt declaration in struct dst_entry 251 */ 252 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63); 253 atomic_inc(&dst->__refcnt); 254 } 255 256 static inline void dst_use(struct dst_entry *dst, unsigned long time) 257 { 258 dst_hold(dst); 259 dst->__use++; 260 dst->lastuse = time; 261 } 262 263 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time) 264 { 265 dst->__use++; 266 dst->lastuse = time; 267 } 268 269 static inline struct dst_entry *dst_clone(struct dst_entry *dst) 270 { 271 if (dst) 272 atomic_inc(&dst->__refcnt); 273 return dst; 274 } 275 276 void dst_release(struct dst_entry *dst); 277 278 static inline void refdst_drop(unsigned long refdst) 279 { 280 if (!(refdst & SKB_DST_NOREF)) 281 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK)); 282 } 283 284 /** 285 * skb_dst_drop - drops skb dst 286 * @skb: buffer 287 * 288 * Drops dst reference count if a reference was taken. 289 */ 290 static inline void skb_dst_drop(struct sk_buff *skb) 291 { 292 if (skb->_skb_refdst) { 293 refdst_drop(skb->_skb_refdst); 294 skb->_skb_refdst = 0UL; 295 } 296 } 297 298 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst) 299 { 300 nskb->_skb_refdst = refdst; 301 if (!(nskb->_skb_refdst & SKB_DST_NOREF)) 302 dst_clone(skb_dst(nskb)); 303 } 304 305 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb) 306 { 307 __skb_dst_copy(nskb, oskb->_skb_refdst); 308 } 309 310 /** 311 * skb_dst_force - makes sure skb dst is refcounted 312 * @skb: buffer 313 * 314 * If dst is not yet refcounted, let's do it 315 */ 316 static inline void skb_dst_force(struct sk_buff *skb) 317 { 318 if (skb_dst_is_noref(skb)) { 319 WARN_ON(!rcu_read_lock_held()); 320 skb->_skb_refdst &= ~SKB_DST_NOREF; 321 dst_clone(skb_dst(skb)); 322 } 323 } 324 325 326 /** 327 * __skb_tunnel_rx - prepare skb for rx reinsert 328 * @skb: buffer 329 * @dev: tunnel device 330 * @net: netns for packet i/o 331 * 332 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 333 * so make some cleanups. (no accounting done) 334 */ 335 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev, 336 struct net *net) 337 { 338 skb->dev = dev; 339 340 /* 341 * Clear hash so that we can recalulate the hash for the 342 * encapsulated packet, unless we have already determine the hash 343 * over the L4 4-tuple. 344 */ 345 skb_clear_hash_if_not_l4(skb); 346 skb_set_queue_mapping(skb, 0); 347 skb_scrub_packet(skb, !net_eq(net, dev_net(dev))); 348 } 349 350 /** 351 * skb_tunnel_rx - prepare skb for rx reinsert 352 * @skb: buffer 353 * @dev: tunnel device 354 * 355 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 356 * so make some cleanups, and perform accounting. 357 * Note: this accounting is not SMP safe. 358 */ 359 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev, 360 struct net *net) 361 { 362 /* TODO : stats should be SMP safe */ 363 dev->stats.rx_packets++; 364 dev->stats.rx_bytes += skb->len; 365 __skb_tunnel_rx(skb, dev, net); 366 } 367 368 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb); 369 static inline int dst_discard(struct sk_buff *skb) 370 { 371 return dst_discard_out(&init_net, skb->sk, skb); 372 } 373 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref, 374 int initial_obsolete, unsigned short flags); 375 void dst_init(struct dst_entry *dst, struct dst_ops *ops, 376 struct net_device *dev, int initial_ref, int initial_obsolete, 377 unsigned short flags); 378 void __dst_free(struct dst_entry *dst); 379 struct dst_entry *dst_destroy(struct dst_entry *dst); 380 381 static inline void dst_free(struct dst_entry *dst) 382 { 383 if (dst->obsolete > 0) 384 return; 385 if (!atomic_read(&dst->__refcnt)) { 386 dst = dst_destroy(dst); 387 if (!dst) 388 return; 389 } 390 __dst_free(dst); 391 } 392 393 static inline void dst_rcu_free(struct rcu_head *head) 394 { 395 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head); 396 dst_free(dst); 397 } 398 399 static inline void dst_confirm(struct dst_entry *dst) 400 { 401 dst->pending_confirm = 1; 402 } 403 404 static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n, 405 struct sk_buff *skb) 406 { 407 const struct hh_cache *hh; 408 409 if (dst->pending_confirm) { 410 unsigned long now = jiffies; 411 412 dst->pending_confirm = 0; 413 /* avoid dirtying neighbour */ 414 if (n->confirmed != now) 415 n->confirmed = now; 416 } 417 418 hh = &n->hh; 419 if ((n->nud_state & NUD_CONNECTED) && hh->hh_len) 420 return neigh_hh_output(hh, skb); 421 else 422 return n->output(n, skb); 423 } 424 425 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr) 426 { 427 struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr); 428 return IS_ERR(n) ? NULL : n; 429 } 430 431 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst, 432 struct sk_buff *skb) 433 { 434 struct neighbour *n = dst->ops->neigh_lookup(dst, skb, NULL); 435 return IS_ERR(n) ? NULL : n; 436 } 437 438 static inline void dst_link_failure(struct sk_buff *skb) 439 { 440 struct dst_entry *dst = skb_dst(skb); 441 if (dst && dst->ops && dst->ops->link_failure) 442 dst->ops->link_failure(skb); 443 } 444 445 static inline void dst_set_expires(struct dst_entry *dst, int timeout) 446 { 447 unsigned long expires = jiffies + timeout; 448 449 if (expires == 0) 450 expires = 1; 451 452 if (dst->expires == 0 || time_before(expires, dst->expires)) 453 dst->expires = expires; 454 } 455 456 /* Output packet to network from transport. */ 457 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb) 458 { 459 return skb_dst(skb)->output(net, sk, skb); 460 } 461 462 /* Input packet from network to transport. */ 463 static inline int dst_input(struct sk_buff *skb) 464 { 465 return skb_dst(skb)->input(skb); 466 } 467 468 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie) 469 { 470 if (dst->obsolete) 471 dst = dst->ops->check(dst, cookie); 472 return dst; 473 } 474 475 void dst_subsys_init(void); 476 477 /* Flags for xfrm_lookup flags argument. */ 478 enum { 479 XFRM_LOOKUP_ICMP = 1 << 0, 480 XFRM_LOOKUP_QUEUE = 1 << 1, 481 XFRM_LOOKUP_KEEP_DST_REF = 1 << 2, 482 }; 483 484 struct flowi; 485 #ifndef CONFIG_XFRM 486 static inline struct dst_entry *xfrm_lookup(struct net *net, 487 struct dst_entry *dst_orig, 488 const struct flowi *fl, 489 const struct sock *sk, 490 int flags) 491 { 492 return dst_orig; 493 } 494 495 static inline struct dst_entry *xfrm_lookup_route(struct net *net, 496 struct dst_entry *dst_orig, 497 const struct flowi *fl, 498 const struct sock *sk, 499 int flags) 500 { 501 return dst_orig; 502 } 503 504 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst) 505 { 506 return NULL; 507 } 508 509 #else 510 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig, 511 const struct flowi *fl, const struct sock *sk, 512 int flags); 513 514 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig, 515 const struct flowi *fl, const struct sock *sk, 516 int flags); 517 518 /* skb attached with this dst needs transformation if dst->xfrm is valid */ 519 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst) 520 { 521 return dst->xfrm; 522 } 523 #endif 524 525 #endif /* _NET_DST_H */ 526