1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * net/dst.h Protocol independent destination cache definitions. 4 * 5 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 6 * 7 */ 8 9 #ifndef _NET_DST_H 10 #define _NET_DST_H 11 12 #include <net/dst_ops.h> 13 #include <linux/netdevice.h> 14 #include <linux/rtnetlink.h> 15 #include <linux/rcupdate.h> 16 #include <linux/bug.h> 17 #include <linux/jiffies.h> 18 #include <linux/refcount.h> 19 #include <net/neighbour.h> 20 #include <asm/processor.h> 21 22 #define DST_GC_MIN (HZ/10) 23 #define DST_GC_INC (HZ/2) 24 #define DST_GC_MAX (120*HZ) 25 26 /* Each dst_entry has reference count and sits in some parent list(s). 27 * When it is removed from parent list, it is "freed" (dst_free). 28 * After this it enters dead state (dst->obsolete > 0) and if its refcnt 29 * is zero, it can be destroyed immediately, otherwise it is added 30 * to gc list and garbage collector periodically checks the refcnt. 31 */ 32 33 struct sk_buff; 34 35 struct dst_entry { 36 struct net_device *dev; 37 struct rcu_head rcu_head; 38 struct dst_entry *child; 39 struct dst_ops *ops; 40 unsigned long _metrics; 41 unsigned long expires; 42 struct dst_entry *path; 43 struct dst_entry *from; 44 #ifdef CONFIG_XFRM 45 struct xfrm_state *xfrm; 46 #else 47 void *__pad1; 48 #endif 49 int (*input)(struct sk_buff *); 50 int (*output)(struct net *net, struct sock *sk, struct sk_buff *skb); 51 52 unsigned short flags; 53 #define DST_HOST 0x0001 54 #define DST_NOXFRM 0x0002 55 #define DST_NOPOLICY 0x0004 56 #define DST_NOCOUNT 0x0008 57 #define DST_FAKE_RTABLE 0x0010 58 #define DST_XFRM_TUNNEL 0x0020 59 #define DST_XFRM_QUEUE 0x0040 60 #define DST_METADATA 0x0080 61 62 short error; 63 64 /* A non-zero value of dst->obsolete forces by-hand validation 65 * of the route entry. Positive values are set by the generic 66 * dst layer to indicate that the entry has been forcefully 67 * destroyed. 68 * 69 * Negative values are used by the implementation layer code to 70 * force invocation of the dst_ops->check() method. 71 */ 72 short obsolete; 73 #define DST_OBSOLETE_NONE 0 74 #define DST_OBSOLETE_DEAD 2 75 #define DST_OBSOLETE_FORCE_CHK -1 76 #define DST_OBSOLETE_KILL -2 77 unsigned short header_len; /* more space at head required */ 78 unsigned short trailer_len; /* space to reserve at tail */ 79 unsigned short __pad3; 80 81 #ifdef CONFIG_IP_ROUTE_CLASSID 82 __u32 tclassid; 83 #else 84 __u32 __pad2; 85 #endif 86 87 #ifdef CONFIG_64BIT 88 /* 89 * Align __refcnt to a 64 bytes alignment 90 * (L1_CACHE_SIZE would be too much) 91 */ 92 long __pad_to_align_refcnt[2]; 93 #endif 94 /* 95 * __refcnt wants to be on a different cache line from 96 * input/output/ops or performance tanks badly 97 */ 98 atomic_t __refcnt; /* client references */ 99 int __use; 100 unsigned long lastuse; 101 struct lwtunnel_state *lwtstate; 102 union { 103 struct dst_entry *next; 104 struct rtable __rcu *rt_next; 105 struct rt6_info __rcu *rt6_next; 106 struct dn_route __rcu *dn_next; 107 }; 108 }; 109 110 struct dst_metrics { 111 u32 metrics[RTAX_MAX]; 112 refcount_t refcnt; 113 }; 114 extern const struct dst_metrics dst_default_metrics; 115 116 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old); 117 118 #define DST_METRICS_READ_ONLY 0x1UL 119 #define DST_METRICS_REFCOUNTED 0x2UL 120 #define DST_METRICS_FLAGS 0x3UL 121 #define __DST_METRICS_PTR(Y) \ 122 ((u32 *)((Y) & ~DST_METRICS_FLAGS)) 123 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics) 124 125 static inline bool dst_metrics_read_only(const struct dst_entry *dst) 126 { 127 return dst->_metrics & DST_METRICS_READ_ONLY; 128 } 129 130 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old); 131 132 static inline void dst_destroy_metrics_generic(struct dst_entry *dst) 133 { 134 unsigned long val = dst->_metrics; 135 if (!(val & DST_METRICS_READ_ONLY)) 136 __dst_destroy_metrics_generic(dst, val); 137 } 138 139 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst) 140 { 141 unsigned long p = dst->_metrics; 142 143 BUG_ON(!p); 144 145 if (p & DST_METRICS_READ_ONLY) 146 return dst->ops->cow_metrics(dst, p); 147 return __DST_METRICS_PTR(p); 148 } 149 150 /* This may only be invoked before the entry has reached global 151 * visibility. 152 */ 153 static inline void dst_init_metrics(struct dst_entry *dst, 154 const u32 *src_metrics, 155 bool read_only) 156 { 157 dst->_metrics = ((unsigned long) src_metrics) | 158 (read_only ? DST_METRICS_READ_ONLY : 0); 159 } 160 161 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src) 162 { 163 u32 *dst_metrics = dst_metrics_write_ptr(dest); 164 165 if (dst_metrics) { 166 u32 *src_metrics = DST_METRICS_PTR(src); 167 168 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32)); 169 } 170 } 171 172 static inline u32 *dst_metrics_ptr(struct dst_entry *dst) 173 { 174 return DST_METRICS_PTR(dst); 175 } 176 177 static inline u32 178 dst_metric_raw(const struct dst_entry *dst, const int metric) 179 { 180 u32 *p = DST_METRICS_PTR(dst); 181 182 return p[metric-1]; 183 } 184 185 static inline u32 186 dst_metric(const struct dst_entry *dst, const int metric) 187 { 188 WARN_ON_ONCE(metric == RTAX_HOPLIMIT || 189 metric == RTAX_ADVMSS || 190 metric == RTAX_MTU); 191 return dst_metric_raw(dst, metric); 192 } 193 194 static inline u32 195 dst_metric_advmss(const struct dst_entry *dst) 196 { 197 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS); 198 199 if (!advmss) 200 advmss = dst->ops->default_advmss(dst); 201 202 return advmss; 203 } 204 205 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val) 206 { 207 u32 *p = dst_metrics_write_ptr(dst); 208 209 if (p) 210 p[metric-1] = val; 211 } 212 213 /* Kernel-internal feature bits that are unallocated in user space. */ 214 #define DST_FEATURE_ECN_CA (1 << 31) 215 216 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA) 217 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN) 218 219 static inline u32 220 dst_feature(const struct dst_entry *dst, u32 feature) 221 { 222 return dst_metric(dst, RTAX_FEATURES) & feature; 223 } 224 225 static inline u32 dst_mtu(const struct dst_entry *dst) 226 { 227 return dst->ops->mtu(dst); 228 } 229 230 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */ 231 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric) 232 { 233 return msecs_to_jiffies(dst_metric(dst, metric)); 234 } 235 236 static inline u32 237 dst_allfrag(const struct dst_entry *dst) 238 { 239 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG); 240 return ret; 241 } 242 243 static inline int 244 dst_metric_locked(const struct dst_entry *dst, int metric) 245 { 246 return dst_metric(dst, RTAX_LOCK) & (1<<metric); 247 } 248 249 static inline void dst_hold(struct dst_entry *dst) 250 { 251 /* 252 * If your kernel compilation stops here, please check 253 * __pad_to_align_refcnt declaration in struct dst_entry 254 */ 255 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63); 256 WARN_ON(atomic_inc_not_zero(&dst->__refcnt) == 0); 257 } 258 259 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time) 260 { 261 if (unlikely(time != dst->lastuse)) { 262 dst->__use++; 263 dst->lastuse = time; 264 } 265 } 266 267 static inline void dst_hold_and_use(struct dst_entry *dst, unsigned long time) 268 { 269 dst_hold(dst); 270 dst_use_noref(dst, time); 271 } 272 273 static inline struct dst_entry *dst_clone(struct dst_entry *dst) 274 { 275 if (dst) 276 dst_hold(dst); 277 return dst; 278 } 279 280 void dst_release(struct dst_entry *dst); 281 282 void dst_release_immediate(struct dst_entry *dst); 283 284 static inline void refdst_drop(unsigned long refdst) 285 { 286 if (!(refdst & SKB_DST_NOREF)) 287 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK)); 288 } 289 290 /** 291 * skb_dst_drop - drops skb dst 292 * @skb: buffer 293 * 294 * Drops dst reference count if a reference was taken. 295 */ 296 static inline void skb_dst_drop(struct sk_buff *skb) 297 { 298 if (skb->_skb_refdst) { 299 refdst_drop(skb->_skb_refdst); 300 skb->_skb_refdst = 0UL; 301 } 302 } 303 304 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst) 305 { 306 nskb->_skb_refdst = refdst; 307 if (!(nskb->_skb_refdst & SKB_DST_NOREF)) 308 dst_clone(skb_dst(nskb)); 309 } 310 311 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb) 312 { 313 __skb_dst_copy(nskb, oskb->_skb_refdst); 314 } 315 316 /** 317 * dst_hold_safe - Take a reference on a dst if possible 318 * @dst: pointer to dst entry 319 * 320 * This helper returns false if it could not safely 321 * take a reference on a dst. 322 */ 323 static inline bool dst_hold_safe(struct dst_entry *dst) 324 { 325 return atomic_inc_not_zero(&dst->__refcnt); 326 } 327 328 /** 329 * skb_dst_force - makes sure skb dst is refcounted 330 * @skb: buffer 331 * 332 * If dst is not yet refcounted and not destroyed, grab a ref on it. 333 */ 334 static inline void skb_dst_force(struct sk_buff *skb) 335 { 336 if (skb_dst_is_noref(skb)) { 337 struct dst_entry *dst = skb_dst(skb); 338 339 WARN_ON(!rcu_read_lock_held()); 340 if (!dst_hold_safe(dst)) 341 dst = NULL; 342 343 skb->_skb_refdst = (unsigned long)dst; 344 } 345 } 346 347 348 /** 349 * __skb_tunnel_rx - prepare skb for rx reinsert 350 * @skb: buffer 351 * @dev: tunnel device 352 * @net: netns for packet i/o 353 * 354 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 355 * so make some cleanups. (no accounting done) 356 */ 357 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev, 358 struct net *net) 359 { 360 skb->dev = dev; 361 362 /* 363 * Clear hash so that we can recalulate the hash for the 364 * encapsulated packet, unless we have already determine the hash 365 * over the L4 4-tuple. 366 */ 367 skb_clear_hash_if_not_l4(skb); 368 skb_set_queue_mapping(skb, 0); 369 skb_scrub_packet(skb, !net_eq(net, dev_net(dev))); 370 } 371 372 /** 373 * skb_tunnel_rx - prepare skb for rx reinsert 374 * @skb: buffer 375 * @dev: tunnel device 376 * 377 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 378 * so make some cleanups, and perform accounting. 379 * Note: this accounting is not SMP safe. 380 */ 381 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev, 382 struct net *net) 383 { 384 /* TODO : stats should be SMP safe */ 385 dev->stats.rx_packets++; 386 dev->stats.rx_bytes += skb->len; 387 __skb_tunnel_rx(skb, dev, net); 388 } 389 390 static inline u32 dst_tclassid(const struct sk_buff *skb) 391 { 392 #ifdef CONFIG_IP_ROUTE_CLASSID 393 const struct dst_entry *dst; 394 395 dst = skb_dst(skb); 396 if (dst) 397 return dst->tclassid; 398 #endif 399 return 0; 400 } 401 402 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb); 403 static inline int dst_discard(struct sk_buff *skb) 404 { 405 return dst_discard_out(&init_net, skb->sk, skb); 406 } 407 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref, 408 int initial_obsolete, unsigned short flags); 409 void dst_init(struct dst_entry *dst, struct dst_ops *ops, 410 struct net_device *dev, int initial_ref, int initial_obsolete, 411 unsigned short flags); 412 struct dst_entry *dst_destroy(struct dst_entry *dst); 413 void dst_dev_put(struct dst_entry *dst); 414 415 static inline void dst_confirm(struct dst_entry *dst) 416 { 417 } 418 419 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr) 420 { 421 struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr); 422 return IS_ERR(n) ? NULL : n; 423 } 424 425 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst, 426 struct sk_buff *skb) 427 { 428 struct neighbour *n = dst->ops->neigh_lookup(dst, skb, NULL); 429 return IS_ERR(n) ? NULL : n; 430 } 431 432 static inline void dst_confirm_neigh(const struct dst_entry *dst, 433 const void *daddr) 434 { 435 if (dst->ops->confirm_neigh) 436 dst->ops->confirm_neigh(dst, daddr); 437 } 438 439 static inline void dst_link_failure(struct sk_buff *skb) 440 { 441 struct dst_entry *dst = skb_dst(skb); 442 if (dst && dst->ops && dst->ops->link_failure) 443 dst->ops->link_failure(skb); 444 } 445 446 static inline void dst_set_expires(struct dst_entry *dst, int timeout) 447 { 448 unsigned long expires = jiffies + timeout; 449 450 if (expires == 0) 451 expires = 1; 452 453 if (dst->expires == 0 || time_before(expires, dst->expires)) 454 dst->expires = expires; 455 } 456 457 /* Output packet to network from transport. */ 458 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb) 459 { 460 return skb_dst(skb)->output(net, sk, skb); 461 } 462 463 /* Input packet from network to transport. */ 464 static inline int dst_input(struct sk_buff *skb) 465 { 466 return skb_dst(skb)->input(skb); 467 } 468 469 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie) 470 { 471 if (dst->obsolete) 472 dst = dst->ops->check(dst, cookie); 473 return dst; 474 } 475 476 /* Flags for xfrm_lookup flags argument. */ 477 enum { 478 XFRM_LOOKUP_ICMP = 1 << 0, 479 XFRM_LOOKUP_QUEUE = 1 << 1, 480 XFRM_LOOKUP_KEEP_DST_REF = 1 << 2, 481 }; 482 483 struct flowi; 484 #ifndef CONFIG_XFRM 485 static inline struct dst_entry *xfrm_lookup(struct net *net, 486 struct dst_entry *dst_orig, 487 const struct flowi *fl, 488 const struct sock *sk, 489 int flags) 490 { 491 return dst_orig; 492 } 493 494 static inline struct dst_entry *xfrm_lookup_route(struct net *net, 495 struct dst_entry *dst_orig, 496 const struct flowi *fl, 497 const struct sock *sk, 498 int flags) 499 { 500 return dst_orig; 501 } 502 503 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst) 504 { 505 return NULL; 506 } 507 508 #else 509 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig, 510 const struct flowi *fl, const struct sock *sk, 511 int flags); 512 513 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig, 514 const struct flowi *fl, const struct sock *sk, 515 int flags); 516 517 /* skb attached with this dst needs transformation if dst->xfrm is valid */ 518 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst) 519 { 520 return dst->xfrm; 521 } 522 #endif 523 524 #endif /* _NET_DST_H */ 525