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 #include <linux/indirect_call_wrapper.h> 22 23 struct sk_buff; 24 25 struct dst_entry { 26 struct net_device *dev; 27 struct dst_ops *ops; 28 unsigned long _metrics; 29 unsigned long expires; 30 #ifdef CONFIG_XFRM 31 struct xfrm_state *xfrm; 32 #else 33 void *__pad1; 34 #endif 35 int (*input)(struct sk_buff *); 36 int (*output)(struct net *net, struct sock *sk, struct sk_buff *skb); 37 38 unsigned short flags; 39 #define DST_NOXFRM 0x0002 40 #define DST_NOPOLICY 0x0004 41 #define DST_NOCOUNT 0x0008 42 #define DST_FAKE_RTABLE 0x0010 43 #define DST_XFRM_TUNNEL 0x0020 44 #define DST_XFRM_QUEUE 0x0040 45 #define DST_METADATA 0x0080 46 47 /* A non-zero value of dst->obsolete forces by-hand validation 48 * of the route entry. Positive values are set by the generic 49 * dst layer to indicate that the entry has been forcefully 50 * destroyed. 51 * 52 * Negative values are used by the implementation layer code to 53 * force invocation of the dst_ops->check() method. 54 */ 55 short obsolete; 56 #define DST_OBSOLETE_NONE 0 57 #define DST_OBSOLETE_DEAD 2 58 #define DST_OBSOLETE_FORCE_CHK -1 59 #define DST_OBSOLETE_KILL -2 60 unsigned short header_len; /* more space at head required */ 61 unsigned short trailer_len; /* space to reserve at tail */ 62 63 /* 64 * __refcnt wants to be on a different cache line from 65 * input/output/ops or performance tanks badly 66 */ 67 #ifdef CONFIG_64BIT 68 atomic_t __refcnt; /* 64-bit offset 64 */ 69 #endif 70 int __use; 71 unsigned long lastuse; 72 struct lwtunnel_state *lwtstate; 73 struct rcu_head rcu_head; 74 short error; 75 short __pad; 76 __u32 tclassid; 77 #ifndef CONFIG_64BIT 78 atomic_t __refcnt; /* 32-bit offset 64 */ 79 #endif 80 }; 81 82 struct dst_metrics { 83 u32 metrics[RTAX_MAX]; 84 refcount_t refcnt; 85 } __aligned(4); /* Low pointer bits contain DST_METRICS_FLAGS */ 86 extern const struct dst_metrics dst_default_metrics; 87 88 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old); 89 90 #define DST_METRICS_READ_ONLY 0x1UL 91 #define DST_METRICS_REFCOUNTED 0x2UL 92 #define DST_METRICS_FLAGS 0x3UL 93 #define __DST_METRICS_PTR(Y) \ 94 ((u32 *)((Y) & ~DST_METRICS_FLAGS)) 95 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics) 96 97 static inline bool dst_metrics_read_only(const struct dst_entry *dst) 98 { 99 return dst->_metrics & DST_METRICS_READ_ONLY; 100 } 101 102 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old); 103 104 static inline void dst_destroy_metrics_generic(struct dst_entry *dst) 105 { 106 unsigned long val = dst->_metrics; 107 if (!(val & DST_METRICS_READ_ONLY)) 108 __dst_destroy_metrics_generic(dst, val); 109 } 110 111 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst) 112 { 113 unsigned long p = dst->_metrics; 114 115 BUG_ON(!p); 116 117 if (p & DST_METRICS_READ_ONLY) 118 return dst->ops->cow_metrics(dst, p); 119 return __DST_METRICS_PTR(p); 120 } 121 122 /* This may only be invoked before the entry has reached global 123 * visibility. 124 */ 125 static inline void dst_init_metrics(struct dst_entry *dst, 126 const u32 *src_metrics, 127 bool read_only) 128 { 129 dst->_metrics = ((unsigned long) src_metrics) | 130 (read_only ? DST_METRICS_READ_ONLY : 0); 131 } 132 133 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src) 134 { 135 u32 *dst_metrics = dst_metrics_write_ptr(dest); 136 137 if (dst_metrics) { 138 u32 *src_metrics = DST_METRICS_PTR(src); 139 140 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32)); 141 } 142 } 143 144 static inline u32 *dst_metrics_ptr(struct dst_entry *dst) 145 { 146 return DST_METRICS_PTR(dst); 147 } 148 149 static inline u32 150 dst_metric_raw(const struct dst_entry *dst, const int metric) 151 { 152 u32 *p = DST_METRICS_PTR(dst); 153 154 return p[metric-1]; 155 } 156 157 static inline u32 158 dst_metric(const struct dst_entry *dst, const int metric) 159 { 160 WARN_ON_ONCE(metric == RTAX_HOPLIMIT || 161 metric == RTAX_ADVMSS || 162 metric == RTAX_MTU); 163 return dst_metric_raw(dst, metric); 164 } 165 166 static inline u32 167 dst_metric_advmss(const struct dst_entry *dst) 168 { 169 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS); 170 171 if (!advmss) 172 advmss = dst->ops->default_advmss(dst); 173 174 return advmss; 175 } 176 177 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val) 178 { 179 u32 *p = dst_metrics_write_ptr(dst); 180 181 if (p) 182 p[metric-1] = val; 183 } 184 185 /* Kernel-internal feature bits that are unallocated in user space. */ 186 #define DST_FEATURE_ECN_CA (1U << 31) 187 188 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA) 189 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN) 190 191 static inline u32 192 dst_feature(const struct dst_entry *dst, u32 feature) 193 { 194 return dst_metric(dst, RTAX_FEATURES) & feature; 195 } 196 197 INDIRECT_CALLABLE_DECLARE(unsigned int ip6_mtu(const struct dst_entry *)); 198 INDIRECT_CALLABLE_DECLARE(unsigned int ipv4_mtu(const struct dst_entry *)); 199 static inline u32 dst_mtu(const struct dst_entry *dst) 200 { 201 return INDIRECT_CALL_INET(dst->ops->mtu, ip6_mtu, ipv4_mtu, dst); 202 } 203 204 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */ 205 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric) 206 { 207 return msecs_to_jiffies(dst_metric(dst, metric)); 208 } 209 210 static inline u32 211 dst_allfrag(const struct dst_entry *dst) 212 { 213 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG); 214 return ret; 215 } 216 217 static inline int 218 dst_metric_locked(const struct dst_entry *dst, int metric) 219 { 220 return dst_metric(dst, RTAX_LOCK) & (1 << metric); 221 } 222 223 static inline void dst_hold(struct dst_entry *dst) 224 { 225 /* 226 * If your kernel compilation stops here, please check 227 * the placement of __refcnt in struct dst_entry 228 */ 229 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63); 230 WARN_ON(atomic_inc_not_zero(&dst->__refcnt) == 0); 231 } 232 233 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time) 234 { 235 if (unlikely(time != dst->lastuse)) { 236 dst->__use++; 237 dst->lastuse = time; 238 } 239 } 240 241 static inline void dst_hold_and_use(struct dst_entry *dst, unsigned long time) 242 { 243 dst_hold(dst); 244 dst_use_noref(dst, time); 245 } 246 247 static inline struct dst_entry *dst_clone(struct dst_entry *dst) 248 { 249 if (dst) 250 dst_hold(dst); 251 return dst; 252 } 253 254 void dst_release(struct dst_entry *dst); 255 256 void dst_release_immediate(struct dst_entry *dst); 257 258 static inline void refdst_drop(unsigned long refdst) 259 { 260 if (!(refdst & SKB_DST_NOREF)) 261 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK)); 262 } 263 264 /** 265 * skb_dst_drop - drops skb dst 266 * @skb: buffer 267 * 268 * Drops dst reference count if a reference was taken. 269 */ 270 static inline void skb_dst_drop(struct sk_buff *skb) 271 { 272 if (skb->_skb_refdst) { 273 refdst_drop(skb->_skb_refdst); 274 skb->_skb_refdst = 0UL; 275 } 276 } 277 278 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst) 279 { 280 nskb->_skb_refdst = refdst; 281 if (!(nskb->_skb_refdst & SKB_DST_NOREF)) 282 dst_clone(skb_dst(nskb)); 283 } 284 285 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb) 286 { 287 __skb_dst_copy(nskb, oskb->_skb_refdst); 288 } 289 290 /** 291 * dst_hold_safe - Take a reference on a dst if possible 292 * @dst: pointer to dst entry 293 * 294 * This helper returns false if it could not safely 295 * take a reference on a dst. 296 */ 297 static inline bool dst_hold_safe(struct dst_entry *dst) 298 { 299 return atomic_inc_not_zero(&dst->__refcnt); 300 } 301 302 /** 303 * skb_dst_force - makes sure skb dst is refcounted 304 * @skb: buffer 305 * 306 * If dst is not yet refcounted and not destroyed, grab a ref on it. 307 * Returns true if dst is refcounted. 308 */ 309 static inline bool skb_dst_force(struct sk_buff *skb) 310 { 311 if (skb_dst_is_noref(skb)) { 312 struct dst_entry *dst = skb_dst(skb); 313 314 WARN_ON(!rcu_read_lock_held()); 315 if (!dst_hold_safe(dst)) 316 dst = NULL; 317 318 skb->_skb_refdst = (unsigned long)dst; 319 } 320 321 return skb->_skb_refdst != 0UL; 322 } 323 324 325 /** 326 * __skb_tunnel_rx - prepare skb for rx reinsert 327 * @skb: buffer 328 * @dev: tunnel device 329 * @net: netns for packet i/o 330 * 331 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 332 * so make some cleanups. (no accounting done) 333 */ 334 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev, 335 struct net *net) 336 { 337 skb->dev = dev; 338 339 /* 340 * Clear hash so that we can recalulate the hash for the 341 * encapsulated packet, unless we have already determine the hash 342 * over the L4 4-tuple. 343 */ 344 skb_clear_hash_if_not_l4(skb); 345 skb_set_queue_mapping(skb, 0); 346 skb_scrub_packet(skb, !net_eq(net, dev_net(dev))); 347 } 348 349 /** 350 * skb_tunnel_rx - prepare skb for rx reinsert 351 * @skb: buffer 352 * @dev: tunnel device 353 * @net: netns for packet i/o 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 static inline u32 dst_tclassid(const struct sk_buff *skb) 369 { 370 #ifdef CONFIG_IP_ROUTE_CLASSID 371 const struct dst_entry *dst; 372 373 dst = skb_dst(skb); 374 if (dst) 375 return dst->tclassid; 376 #endif 377 return 0; 378 } 379 380 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb); 381 static inline int dst_discard(struct sk_buff *skb) 382 { 383 return dst_discard_out(&init_net, skb->sk, skb); 384 } 385 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref, 386 int initial_obsolete, unsigned short flags); 387 void dst_init(struct dst_entry *dst, struct dst_ops *ops, 388 struct net_device *dev, int initial_ref, int initial_obsolete, 389 unsigned short flags); 390 struct dst_entry *dst_destroy(struct dst_entry *dst); 391 void dst_dev_put(struct dst_entry *dst); 392 393 static inline void dst_confirm(struct dst_entry *dst) 394 { 395 } 396 397 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr) 398 { 399 struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr); 400 return IS_ERR(n) ? NULL : n; 401 } 402 403 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst, 404 struct sk_buff *skb) 405 { 406 struct neighbour *n; 407 408 if (WARN_ON_ONCE(!dst->ops->neigh_lookup)) 409 return NULL; 410 411 n = dst->ops->neigh_lookup(dst, skb, NULL); 412 413 return IS_ERR(n) ? NULL : n; 414 } 415 416 static inline void dst_confirm_neigh(const struct dst_entry *dst, 417 const void *daddr) 418 { 419 if (dst->ops->confirm_neigh) 420 dst->ops->confirm_neigh(dst, daddr); 421 } 422 423 static inline void dst_link_failure(struct sk_buff *skb) 424 { 425 struct dst_entry *dst = skb_dst(skb); 426 if (dst && dst->ops && dst->ops->link_failure) 427 dst->ops->link_failure(skb); 428 } 429 430 static inline void dst_set_expires(struct dst_entry *dst, int timeout) 431 { 432 unsigned long expires = jiffies + timeout; 433 434 if (expires == 0) 435 expires = 1; 436 437 if (dst->expires == 0 || time_before(expires, dst->expires)) 438 dst->expires = expires; 439 } 440 441 INDIRECT_CALLABLE_DECLARE(int ip6_output(struct net *, struct sock *, 442 struct sk_buff *)); 443 INDIRECT_CALLABLE_DECLARE(int ip_output(struct net *, struct sock *, 444 struct sk_buff *)); 445 /* Output packet to network from transport. */ 446 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb) 447 { 448 return INDIRECT_CALL_INET(skb_dst(skb)->output, 449 ip6_output, ip_output, 450 net, sk, skb); 451 } 452 453 INDIRECT_CALLABLE_DECLARE(int ip6_input(struct sk_buff *)); 454 INDIRECT_CALLABLE_DECLARE(int ip_local_deliver(struct sk_buff *)); 455 /* Input packet from network to transport. */ 456 static inline int dst_input(struct sk_buff *skb) 457 { 458 return INDIRECT_CALL_INET(skb_dst(skb)->input, 459 ip6_input, ip_local_deliver, skb); 460 } 461 462 INDIRECT_CALLABLE_DECLARE(struct dst_entry *ip6_dst_check(struct dst_entry *, 463 u32)); 464 INDIRECT_CALLABLE_DECLARE(struct dst_entry *ipv4_dst_check(struct dst_entry *, 465 u32)); 466 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie) 467 { 468 if (dst->obsolete) 469 dst = INDIRECT_CALL_INET(dst->ops->check, ip6_dst_check, 470 ipv4_dst_check, dst, cookie); 471 return dst; 472 } 473 474 /* Flags for xfrm_lookup flags argument. */ 475 enum { 476 XFRM_LOOKUP_ICMP = 1 << 0, 477 XFRM_LOOKUP_QUEUE = 1 << 1, 478 XFRM_LOOKUP_KEEP_DST_REF = 1 << 2, 479 }; 480 481 struct flowi; 482 #ifndef CONFIG_XFRM 483 static inline struct dst_entry *xfrm_lookup(struct net *net, 484 struct dst_entry *dst_orig, 485 const struct flowi *fl, 486 const struct sock *sk, 487 int flags) 488 { 489 return dst_orig; 490 } 491 492 static inline struct dst_entry * 493 xfrm_lookup_with_ifid(struct net *net, struct dst_entry *dst_orig, 494 const struct flowi *fl, const struct sock *sk, 495 int flags, u32 if_id) 496 { 497 return dst_orig; 498 } 499 500 static inline struct dst_entry *xfrm_lookup_route(struct net *net, 501 struct dst_entry *dst_orig, 502 const struct flowi *fl, 503 const struct sock *sk, 504 int flags) 505 { 506 return dst_orig; 507 } 508 509 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst) 510 { 511 return NULL; 512 } 513 514 #else 515 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig, 516 const struct flowi *fl, const struct sock *sk, 517 int flags); 518 519 struct dst_entry *xfrm_lookup_with_ifid(struct net *net, 520 struct dst_entry *dst_orig, 521 const struct flowi *fl, 522 const struct sock *sk, int flags, 523 u32 if_id); 524 525 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig, 526 const struct flowi *fl, const struct sock *sk, 527 int flags); 528 529 /* skb attached with this dst needs transformation if dst->xfrm is valid */ 530 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst) 531 { 532 return dst->xfrm; 533 } 534 #endif 535 536 static inline void skb_dst_update_pmtu(struct sk_buff *skb, u32 mtu) 537 { 538 struct dst_entry *dst = skb_dst(skb); 539 540 if (dst && dst->ops->update_pmtu) 541 dst->ops->update_pmtu(dst, NULL, skb, mtu, true); 542 } 543 544 /* update dst pmtu but not do neighbor confirm */ 545 static inline void skb_dst_update_pmtu_no_confirm(struct sk_buff *skb, u32 mtu) 546 { 547 struct dst_entry *dst = skb_dst(skb); 548 549 if (dst && dst->ops->update_pmtu) 550 dst->ops->update_pmtu(dst, NULL, skb, mtu, false); 551 } 552 553 struct dst_entry *dst_blackhole_check(struct dst_entry *dst, u32 cookie); 554 void dst_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk, 555 struct sk_buff *skb, u32 mtu, bool confirm_neigh); 556 void dst_blackhole_redirect(struct dst_entry *dst, struct sock *sk, 557 struct sk_buff *skb); 558 u32 *dst_blackhole_cow_metrics(struct dst_entry *dst, unsigned long old); 559 struct neighbour *dst_blackhole_neigh_lookup(const struct dst_entry *dst, 560 struct sk_buff *skb, 561 const void *daddr); 562 unsigned int dst_blackhole_mtu(const struct dst_entry *dst); 563 564 #endif /* _NET_DST_H */ 565