include/net/dst.h

   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 struct sk_buff;
  23
  24 struct dst_entry {
  25         struct net_device       *dev;
  26         struct  dst_ops         *ops;
  27         unsigned long           _metrics;
  28         unsigned long           expires;
  29 #ifdef CONFIG_XFRM
  30         struct xfrm_state       *xfrm;
  31 #else
  32         void                    *__pad1;
  33 #endif
  34         int                     (*input)(struct sk_buff *);
  35         int                     (*output)(struct net *net, struct sock *sk, struct sk_buff *skb);
  36
  37         unsigned short          flags;
  38 #define DST_HOST                0x0001
  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 };
  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      (1 << 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 static inline u32 dst_mtu(const struct dst_entry *dst)
 198 {
 199         return dst->ops->mtu(dst);
 200 }
 201
 202 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
 203 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
 204 {
 205         return msecs_to_jiffies(dst_metric(dst, metric));
 206 }
 207
 208 static inline u32
 209 dst_allfrag(const struct dst_entry *dst)
 210 {
 211         int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
 212         return ret;
 213 }
 214
 215 static inline int
 216 dst_metric_locked(const struct dst_entry *dst, int metric)
 217 {
 218         return dst_metric(dst, RTAX_LOCK) & (1<<metric);
 219 }
 220
 221 static inline void dst_hold(struct dst_entry *dst)
 222 {
 223         /*
 224          * If your kernel compilation stops here, please check
 225          * the placement of __refcnt in struct dst_entry
 226          */
 227         BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
 228         WARN_ON(atomic_inc_not_zero(&dst->__refcnt) == 0);
 229 }
 230
 231 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
 232 {
 233         if (unlikely(time != dst->lastuse)) {
 234                 dst->__use++;
 235                 dst->lastuse = time;
 236         }
 237 }
 238
 239 static inline void dst_hold_and_use(struct dst_entry *dst, unsigned long time)
 240 {
 241         dst_hold(dst);
 242         dst_use_noref(dst, time);
 243 }
 244
 245 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
 246 {
 247         if (dst)
 248                 dst_hold(dst);
 249         return dst;
 250 }
 251
 252 void dst_release(struct dst_entry *dst);
 253
 254 void dst_release_immediate(struct dst_entry *dst);
 255
 256 static inline void refdst_drop(unsigned long refdst)
 257 {
 258         if (!(refdst & SKB_DST_NOREF))
 259                 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
 260 }
 261
 262 /**
 263  * skb_dst_drop - drops skb dst
 264  * @skb: buffer
 265  *
 266  * Drops dst reference count if a reference was taken.
 267  */
 268 static inline void skb_dst_drop(struct sk_buff *skb)
 269 {
 270         if (skb->_skb_refdst) {
 271                 refdst_drop(skb->_skb_refdst);
 272                 skb->_skb_refdst = 0UL;
 273         }
 274 }
 275
 276 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
 277 {
 278         nskb->_skb_refdst = refdst;
 279         if (!(nskb->_skb_refdst & SKB_DST_NOREF))
 280                 dst_clone(skb_dst(nskb));
 281 }
 282
 283 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
 284 {
 285         __skb_dst_copy(nskb, oskb->_skb_refdst);
 286 }
 287
 288 /**
 289  * dst_hold_safe - Take a reference on a dst if possible
 290  * @dst: pointer to dst entry
 291  *
 292  * This helper returns false if it could not safely
 293  * take a reference on a dst.
 294  */
 295 static inline bool dst_hold_safe(struct dst_entry *dst)
 296 {
 297         return atomic_inc_not_zero(&dst->__refcnt);
 298 }
 299
 300 /**
 301  * skb_dst_force - makes sure skb dst is refcounted
 302  * @skb: buffer
 303  *
 304  * If dst is not yet refcounted and not destroyed, grab a ref on it.
 305  */
 306 static inline void skb_dst_force(struct sk_buff *skb)
 307 {
 308         if (skb_dst_is_noref(skb)) {
 309                 struct dst_entry *dst = skb_dst(skb);
 310
 311                 WARN_ON(!rcu_read_lock_held());
 312                 if (!dst_hold_safe(dst))
 313                         dst = NULL;
 314
 315                 skb->_skb_refdst = (unsigned long)dst;
 316         }
 317 }
 318
 319
 320 /**
 321  *      __skb_tunnel_rx - prepare skb for rx reinsert
 322  *      @skb: buffer
 323  *      @dev: tunnel device
 324  *      @net: netns for packet i/o
 325  *
 326  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 327  *      so make some cleanups. (no accounting done)
 328  */
 329 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
 330                                    struct net *net)
 331 {
 332         skb->dev = dev;
 333
 334         /*
 335          * Clear hash so that we can recalulate the hash for the
 336          * encapsulated packet, unless we have already determine the hash
 337          * over the L4 4-tuple.
 338          */
 339         skb_clear_hash_if_not_l4(skb);
 340         skb_set_queue_mapping(skb, 0);
 341         skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
 342 }
 343
 344 /**
 345  *      skb_tunnel_rx - prepare skb for rx reinsert
 346  *      @skb: buffer
 347  *      @dev: tunnel device
 348  *      @net: netns for packet i/o
 349  *
 350  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 351  *      so make some cleanups, and perform accounting.
 352  *      Note: this accounting is not SMP safe.
 353  */
 354 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
 355                                  struct net *net)
 356 {
 357         /* TODO : stats should be SMP safe */
 358         dev->stats.rx_packets++;
 359         dev->stats.rx_bytes += skb->len;
 360         __skb_tunnel_rx(skb, dev, net);
 361 }
 362
 363 static inline u32 dst_tclassid(const struct sk_buff *skb)
 364 {
 365 #ifdef CONFIG_IP_ROUTE_CLASSID
 366         const struct dst_entry *dst;
 367
 368         dst = skb_dst(skb);
 369         if (dst)
 370                 return dst->tclassid;
 371 #endif
 372         return 0;
 373 }
 374
 375 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
 376 static inline int dst_discard(struct sk_buff *skb)
 377 {
 378         return dst_discard_out(&init_net, skb->sk, skb);
 379 }
 380 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
 381                 int initial_obsolete, unsigned short flags);
 382 void dst_init(struct dst_entry *dst, struct dst_ops *ops,
 383               struct net_device *dev, int initial_ref, int initial_obsolete,
 384               unsigned short flags);
 385 struct dst_entry *dst_destroy(struct dst_entry *dst);
 386 void dst_dev_put(struct dst_entry *dst);
 387
 388 static inline void dst_confirm(struct dst_entry *dst)
 389 {
 390 }
 391
 392 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
 393 {
 394         struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
 395         return IS_ERR(n) ? NULL : n;
 396 }
 397
 398 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
 399                                                      struct sk_buff *skb)
 400 {
 401         struct neighbour *n =  dst->ops->neigh_lookup(dst, skb, NULL);
 402         return IS_ERR(n) ? NULL : n;
 403 }
 404
 405 static inline void dst_confirm_neigh(const struct dst_entry *dst,
 406                                      const void *daddr)
 407 {
 408         if (dst->ops->confirm_neigh)
 409                 dst->ops->confirm_neigh(dst, daddr);
 410 }
 411
 412 static inline void dst_link_failure(struct sk_buff *skb)
 413 {
 414         struct dst_entry *dst = skb_dst(skb);
 415         if (dst && dst->ops && dst->ops->link_failure)
 416                 dst->ops->link_failure(skb);
 417 }
 418
 419 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
 420 {
 421         unsigned long expires = jiffies + timeout;
 422
 423         if (expires == 0)
 424                 expires = 1;
 425
 426         if (dst->expires == 0 || time_before(expires, dst->expires))
 427                 dst->expires = expires;
 428 }
 429
 430 /* Output packet to network from transport.  */
 431 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
 432 {
 433         return skb_dst(skb)->output(net, sk, skb);
 434 }
 435
 436 /* Input packet from network to transport.  */
 437 static inline int dst_input(struct sk_buff *skb)
 438 {
 439         return skb_dst(skb)->input(skb);
 440 }
 441
 442 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
 443 {
 444         if (dst->obsolete)
 445                 dst = dst->ops->check(dst, cookie);
 446         return dst;
 447 }
 448
 449 /* Flags for xfrm_lookup flags argument. */
 450 enum {
 451         XFRM_LOOKUP_ICMP = 1 << 0,
 452         XFRM_LOOKUP_QUEUE = 1 << 1,
 453         XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
 454 };
 455
 456 struct flowi;
 457 #ifndef CONFIG_XFRM
 458 static inline struct dst_entry *xfrm_lookup(struct net *net,
 459                                             struct dst_entry *dst_orig,
 460                                             const struct flowi *fl,
 461                                             const struct sock *sk,
 462                                             int flags)
 463 {
 464         return dst_orig;
 465 }
 466
 467 static inline struct dst_entry *
 468 xfrm_lookup_with_ifid(struct net *net, struct dst_entry *dst_orig,
 469                       const struct flowi *fl, const struct sock *sk,
 470                       int flags, u32 if_id)
 471 {
 472         return dst_orig;
 473 }
 474
 475 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
 476                                                   struct dst_entry *dst_orig,
 477                                                   const struct flowi *fl,
 478                                                   const struct sock *sk,
 479                                                   int flags)
 480 {
 481         return dst_orig;
 482 }
 483
 484 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
 485 {
 486         return NULL;
 487 }
 488
 489 #else
 490 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
 491                               const struct flowi *fl, const struct sock *sk,
 492                               int flags);
 493
 494 struct dst_entry *xfrm_lookup_with_ifid(struct net *net,
 495                                         struct dst_entry *dst_orig,
 496                                         const struct flowi *fl,
 497                                         const struct sock *sk, int flags,
 498                                         u32 if_id);
 499
 500 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
 501                                     const struct flowi *fl, const struct sock *sk,
 502                                     int flags);
 503
 504 /* skb attached with this dst needs transformation if dst->xfrm is valid */
 505 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
 506 {
 507         return dst->xfrm;
 508 }
 509 #endif
 510
 511 static inline void skb_dst_update_pmtu(struct sk_buff *skb, u32 mtu)
 512 {
 513         struct dst_entry *dst = skb_dst(skb);
 514
 515         if (dst && dst->ops->update_pmtu)
 516                 dst->ops->update_pmtu(dst, NULL, skb, mtu);
 517 }
 518
 519 static inline void skb_tunnel_check_pmtu(struct sk_buff *skb,
 520                                          struct dst_entry *encap_dst,
 521                                          int headroom)
 522 {
 523         u32 encap_mtu = dst_mtu(encap_dst);
 524
 525         if (skb->len > encap_mtu - headroom)
 526                 skb_dst_update_pmtu(skb, encap_mtu - headroom);
 527 }
 528
 529 #endif /* _NET_DST_H */