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 netdevice_tracker dev_tracker; 81 }; 82 83 struct dst_metrics { 84 u32 metrics[RTAX_MAX]; 85 refcount_t refcnt; 86 } __aligned(4); /* Low pointer bits contain DST_METRICS_FLAGS */ 87 extern const struct dst_metrics dst_default_metrics; 88 89 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old); 90 91 #define DST_METRICS_READ_ONLY 0x1UL 92 #define DST_METRICS_REFCOUNTED 0x2UL 93 #define DST_METRICS_FLAGS 0x3UL 94 #define __DST_METRICS_PTR(Y) \ 95 ((u32 *)((Y) & ~DST_METRICS_FLAGS)) 96 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics) 97 98 static inline bool dst_metrics_read_only(const struct dst_entry *dst) 99 { 100 return dst->_metrics & DST_METRICS_READ_ONLY; 101 } 102 103 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old); 104 105 static inline void dst_destroy_metrics_generic(struct dst_entry *dst) 106 { 107 unsigned long val = dst->_metrics; 108 if (!(val & DST_METRICS_READ_ONLY)) 109 __dst_destroy_metrics_generic(dst, val); 110 } 111 112 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst) 113 { 114 unsigned long p = dst->_metrics; 115 116 BUG_ON(!p); 117 118 if (p & DST_METRICS_READ_ONLY) 119 return dst->ops->cow_metrics(dst, p); 120 return __DST_METRICS_PTR(p); 121 } 122 123 /* This may only be invoked before the entry has reached global 124 * visibility. 125 */ 126 static inline void dst_init_metrics(struct dst_entry *dst, 127 const u32 *src_metrics, 128 bool read_only) 129 { 130 dst->_metrics = ((unsigned long) src_metrics) | 131 (read_only ? DST_METRICS_READ_ONLY : 0); 132 } 133 134 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src) 135 { 136 u32 *dst_metrics = dst_metrics_write_ptr(dest); 137 138 if (dst_metrics) { 139 u32 *src_metrics = DST_METRICS_PTR(src); 140 141 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32)); 142 } 143 } 144 145 static inline u32 *dst_metrics_ptr(struct dst_entry *dst) 146 { 147 return DST_METRICS_PTR(dst); 148 } 149 150 static inline u32 151 dst_metric_raw(const struct dst_entry *dst, const int metric) 152 { 153 u32 *p = DST_METRICS_PTR(dst); 154 155 return p[metric-1]; 156 } 157 158 static inline u32 159 dst_metric(const struct dst_entry *dst, const int metric) 160 { 161 WARN_ON_ONCE(metric == RTAX_HOPLIMIT || 162 metric == RTAX_ADVMSS || 163 metric == RTAX_MTU); 164 return dst_metric_raw(dst, metric); 165 } 166 167 static inline u32 168 dst_metric_advmss(const struct dst_entry *dst) 169 { 170 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS); 171 172 if (!advmss) 173 advmss = dst->ops->default_advmss(dst); 174 175 return advmss; 176 } 177 178 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val) 179 { 180 u32 *p = dst_metrics_write_ptr(dst); 181 182 if (p) 183 p[metric-1] = val; 184 } 185 186 /* Kernel-internal feature bits that are unallocated in user space. */ 187 #define DST_FEATURE_ECN_CA (1U << 31) 188 189 #define DST_FEATURE_MASK (DST_FEATURE_ECN_CA) 190 #define DST_FEATURE_ECN_MASK (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN) 191 192 static inline u32 193 dst_feature(const struct dst_entry *dst, u32 feature) 194 { 195 return dst_metric(dst, RTAX_FEATURES) & feature; 196 } 197 198 INDIRECT_CALLABLE_DECLARE(unsigned int ip6_mtu(const struct dst_entry *)); 199 INDIRECT_CALLABLE_DECLARE(unsigned int ipv4_mtu(const struct dst_entry *)); 200 static inline u32 dst_mtu(const struct dst_entry *dst) 201 { 202 return INDIRECT_CALL_INET(dst->ops->mtu, ip6_mtu, ipv4_mtu, dst); 203 } 204 205 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */ 206 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric) 207 { 208 return msecs_to_jiffies(dst_metric(dst, metric)); 209 } 210 211 static inline u32 212 dst_allfrag(const struct dst_entry *dst) 213 { 214 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG); 215 return ret; 216 } 217 218 static inline int 219 dst_metric_locked(const struct dst_entry *dst, int metric) 220 { 221 return dst_metric(dst, RTAX_LOCK) & (1 << metric); 222 } 223 224 static inline void dst_hold(struct dst_entry *dst) 225 { 226 /* 227 * If your kernel compilation stops here, please check 228 * the placement of __refcnt in struct dst_entry 229 */ 230 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63); 231 WARN_ON(atomic_inc_not_zero(&dst->__refcnt) == 0); 232 } 233 234 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time) 235 { 236 if (unlikely(time != dst->lastuse)) { 237 dst->__use++; 238 dst->lastuse = time; 239 } 240 } 241 242 static inline void dst_hold_and_use(struct dst_entry *dst, unsigned long time) 243 { 244 dst_hold(dst); 245 dst_use_noref(dst, time); 246 } 247 248 static inline struct dst_entry *dst_clone(struct dst_entry *dst) 249 { 250 if (dst) 251 dst_hold(dst); 252 return dst; 253 } 254 255 void dst_release(struct dst_entry *dst); 256 257 void dst_release_immediate(struct dst_entry *dst); 258 259 static inline void refdst_drop(unsigned long refdst) 260 { 261 if (!(refdst & SKB_DST_NOREF)) 262 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK)); 263 } 264 265 /** 266 * skb_dst_drop - drops skb dst 267 * @skb: buffer 268 * 269 * Drops dst reference count if a reference was taken. 270 */ 271 static inline void skb_dst_drop(struct sk_buff *skb) 272 { 273 if (skb->_skb_refdst) { 274 refdst_drop(skb->_skb_refdst); 275 skb->_skb_refdst = 0UL; 276 } 277 } 278 279 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst) 280 { 281 nskb->slow_gro |= !!refdst; 282 nskb->_skb_refdst = refdst; 283 if (!(nskb->_skb_refdst & SKB_DST_NOREF)) 284 dst_clone(skb_dst(nskb)); 285 } 286 287 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb) 288 { 289 __skb_dst_copy(nskb, oskb->_skb_refdst); 290 } 291 292 /** 293 * dst_hold_safe - Take a reference on a dst if possible 294 * @dst: pointer to dst entry 295 * 296 * This helper returns false if it could not safely 297 * take a reference on a dst. 298 */ 299 static inline bool dst_hold_safe(struct dst_entry *dst) 300 { 301 return atomic_inc_not_zero(&dst->__refcnt); 302 } 303 304 /** 305 * skb_dst_force - makes sure skb dst is refcounted 306 * @skb: buffer 307 * 308 * If dst is not yet refcounted and not destroyed, grab a ref on it. 309 * Returns true if dst is refcounted. 310 */ 311 static inline bool skb_dst_force(struct sk_buff *skb) 312 { 313 if (skb_dst_is_noref(skb)) { 314 struct dst_entry *dst = skb_dst(skb); 315 316 WARN_ON(!rcu_read_lock_held()); 317 if (!dst_hold_safe(dst)) 318 dst = NULL; 319 320 skb->_skb_refdst = (unsigned long)dst; 321 skb->slow_gro |= !!dst; 322 } 323 324 return skb->_skb_refdst != 0UL; 325 } 326 327 328 /** 329 * __skb_tunnel_rx - prepare skb for rx reinsert 330 * @skb: buffer 331 * @dev: tunnel device 332 * @net: netns for packet i/o 333 * 334 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 335 * so make some cleanups. (no accounting done) 336 */ 337 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev, 338 struct net *net) 339 { 340 skb->dev = dev; 341 342 /* 343 * Clear hash so that we can recalulate the hash for the 344 * encapsulated packet, unless we have already determine the hash 345 * over the L4 4-tuple. 346 */ 347 skb_clear_hash_if_not_l4(skb); 348 skb_set_queue_mapping(skb, 0); 349 skb_scrub_packet(skb, !net_eq(net, dev_net(dev))); 350 } 351 352 /** 353 * skb_tunnel_rx - prepare skb for rx reinsert 354 * @skb: buffer 355 * @dev: tunnel device 356 * @net: netns for packet i/o 357 * 358 * After decapsulation, packet is going to re-enter (netif_rx()) our stack, 359 * so make some cleanups, and perform accounting. 360 * Note: this accounting is not SMP safe. 361 */ 362 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev, 363 struct net *net) 364 { 365 /* TODO : stats should be SMP safe */ 366 dev->stats.rx_packets++; 367 dev->stats.rx_bytes += skb->len; 368 __skb_tunnel_rx(skb, dev, net); 369 } 370 371 static inline u32 dst_tclassid(const struct sk_buff *skb) 372 { 373 #ifdef CONFIG_IP_ROUTE_CLASSID 374 const struct dst_entry *dst; 375 376 dst = skb_dst(skb); 377 if (dst) 378 return dst->tclassid; 379 #endif 380 return 0; 381 } 382 383 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb); 384 static inline int dst_discard(struct sk_buff *skb) 385 { 386 return dst_discard_out(&init_net, skb->sk, skb); 387 } 388 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref, 389 int initial_obsolete, unsigned short flags); 390 void dst_init(struct dst_entry *dst, struct dst_ops *ops, 391 struct net_device *dev, int initial_ref, int initial_obsolete, 392 unsigned short flags); 393 struct dst_entry *dst_destroy(struct dst_entry *dst); 394 void dst_dev_put(struct dst_entry *dst); 395 396 static inline void dst_confirm(struct dst_entry *dst) 397 { 398 } 399 400 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr) 401 { 402 struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr); 403 return IS_ERR(n) ? NULL : n; 404 } 405 406 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst, 407 struct sk_buff *skb) 408 { 409 struct neighbour *n; 410 411 if (WARN_ON_ONCE(!dst->ops->neigh_lookup)) 412 return NULL; 413 414 n = dst->ops->neigh_lookup(dst, skb, NULL); 415 416 return IS_ERR(n) ? NULL : n; 417 } 418 419 static inline void dst_confirm_neigh(const struct dst_entry *dst, 420 const void *daddr) 421 { 422 if (dst->ops->confirm_neigh) 423 dst->ops->confirm_neigh(dst, daddr); 424 } 425 426 static inline void dst_link_failure(struct sk_buff *skb) 427 { 428 struct dst_entry *dst = skb_dst(skb); 429 if (dst && dst->ops && dst->ops->link_failure) 430 dst->ops->link_failure(skb); 431 } 432 433 static inline void dst_set_expires(struct dst_entry *dst, int timeout) 434 { 435 unsigned long expires = jiffies + timeout; 436 437 if (expires == 0) 438 expires = 1; 439 440 if (dst->expires == 0 || time_before(expires, dst->expires)) 441 dst->expires = expires; 442 } 443 444 INDIRECT_CALLABLE_DECLARE(int ip6_output(struct net *, struct sock *, 445 struct sk_buff *)); 446 INDIRECT_CALLABLE_DECLARE(int ip_output(struct net *, struct sock *, 447 struct sk_buff *)); 448 /* Output packet to network from transport. */ 449 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb) 450 { 451 return INDIRECT_CALL_INET(skb_dst(skb)->output, 452 ip6_output, ip_output, 453 net, sk, skb); 454 } 455 456 INDIRECT_CALLABLE_DECLARE(int ip6_input(struct sk_buff *)); 457 INDIRECT_CALLABLE_DECLARE(int ip_local_deliver(struct sk_buff *)); 458 /* Input packet from network to transport. */ 459 static inline int dst_input(struct sk_buff *skb) 460 { 461 return INDIRECT_CALL_INET(skb_dst(skb)->input, 462 ip6_input, ip_local_deliver, skb); 463 } 464 465 INDIRECT_CALLABLE_DECLARE(struct dst_entry *ip6_dst_check(struct dst_entry *, 466 u32)); 467 INDIRECT_CALLABLE_DECLARE(struct dst_entry *ipv4_dst_check(struct dst_entry *, 468 u32)); 469 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie) 470 { 471 if (dst->obsolete) 472 dst = INDIRECT_CALL_INET(dst->ops->check, ip6_dst_check, 473 ipv4_dst_check, dst, cookie); 474 return dst; 475 } 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 * 496 xfrm_lookup_with_ifid(struct net *net, struct dst_entry *dst_orig, 497 const struct flowi *fl, const struct sock *sk, 498 int flags, u32 if_id) 499 { 500 return dst_orig; 501 } 502 503 static inline struct dst_entry *xfrm_lookup_route(struct net *net, 504 struct dst_entry *dst_orig, 505 const struct flowi *fl, 506 const struct sock *sk, 507 int flags) 508 { 509 return dst_orig; 510 } 511 512 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst) 513 { 514 return NULL; 515 } 516 517 #else 518 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig, 519 const struct flowi *fl, const struct sock *sk, 520 int flags); 521 522 struct dst_entry *xfrm_lookup_with_ifid(struct net *net, 523 struct dst_entry *dst_orig, 524 const struct flowi *fl, 525 const struct sock *sk, int flags, 526 u32 if_id); 527 528 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig, 529 const struct flowi *fl, const struct sock *sk, 530 int flags); 531 532 /* skb attached with this dst needs transformation if dst->xfrm is valid */ 533 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst) 534 { 535 return dst->xfrm; 536 } 537 #endif 538 539 static inline void skb_dst_update_pmtu(struct sk_buff *skb, u32 mtu) 540 { 541 struct dst_entry *dst = skb_dst(skb); 542 543 if (dst && dst->ops->update_pmtu) 544 dst->ops->update_pmtu(dst, NULL, skb, mtu, true); 545 } 546 547 /* update dst pmtu but not do neighbor confirm */ 548 static inline void skb_dst_update_pmtu_no_confirm(struct sk_buff *skb, u32 mtu) 549 { 550 struct dst_entry *dst = skb_dst(skb); 551 552 if (dst && dst->ops->update_pmtu) 553 dst->ops->update_pmtu(dst, NULL, skb, mtu, false); 554 } 555 556 struct dst_entry *dst_blackhole_check(struct dst_entry *dst, u32 cookie); 557 void dst_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk, 558 struct sk_buff *skb, u32 mtu, bool confirm_neigh); 559 void dst_blackhole_redirect(struct dst_entry *dst, struct sock *sk, 560 struct sk_buff *skb); 561 u32 *dst_blackhole_cow_metrics(struct dst_entry *dst, unsigned long old); 562 struct neighbour *dst_blackhole_neigh_lookup(const struct dst_entry *dst, 563 struct sk_buff *skb, 564 const void *daddr); 565 unsigned int dst_blackhole_mtu(const struct dst_entry *dst); 566 567 #endif /* _NET_DST_H */ 568