include/net/dst.h

   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 net_device       *dev;
  35         struct rcu_head         rcu_head;
  36         struct dst_entry        *child;
  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_NOCOUNT             0x0008
  55 #define DST_FAKE_RTABLE         0x0010
  56 #define DST_XFRM_TUNNEL         0x0020
  57 #define DST_XFRM_QUEUE          0x0040
  58 #define DST_METADATA            0x0080
  59
  60         short                   error;
  61
  62         /* A non-zero value of dst->obsolete forces by-hand validation
  63          * of the route entry.  Positive values are set by the generic
  64          * dst layer to indicate that the entry has been forcefully
  65          * destroyed.
  66          *
  67          * Negative values are used by the implementation layer code to
  68          * force invocation of the dst_ops->check() method.
  69          */
  70         short                   obsolete;
  71 #define DST_OBSOLETE_NONE       0
  72 #define DST_OBSOLETE_DEAD       2
  73 #define DST_OBSOLETE_FORCE_CHK  -1
  74 #define DST_OBSOLETE_KILL       -2
  75         unsigned short          header_len;     /* more space at head required */
  76         unsigned short          trailer_len;    /* space to reserve at tail */
  77         unsigned short          __pad3;
  78
  79 #ifdef CONFIG_IP_ROUTE_CLASSID
  80         __u32                   tclassid;
  81 #else
  82         __u32                   __pad2;
  83 #endif
  84
  85 #ifdef CONFIG_64BIT
  86         /*
  87          * Align __refcnt to a 64 bytes alignment
  88          * (L1_CACHE_SIZE would be too much)
  89          */
  90         long                    __pad_to_align_refcnt[2];
  91 #endif
  92         /*
  93          * __refcnt wants to be on a different cache line from
  94          * input/output/ops or performance tanks badly
  95          */
  96         atomic_t                __refcnt;       /* client references    */
  97         int                     __use;
  98         unsigned long           lastuse;
  99         struct lwtunnel_state   *lwtstate;
 100         union {
 101                 struct dst_entry        *next;
 102                 struct rtable __rcu     *rt_next;
 103                 struct rt6_info         *rt6_next;
 104                 struct dn_route __rcu   *dn_next;
 105         };
 106 };
 107
 108 struct dst_metrics {
 109         u32             metrics[RTAX_MAX];
 110         atomic_t        refcnt;
 111 };
 112 extern const struct dst_metrics dst_default_metrics;
 113
 114 u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
 115
 116 #define DST_METRICS_READ_ONLY           0x1UL
 117 #define DST_METRICS_REFCOUNTED          0x2UL
 118 #define DST_METRICS_FLAGS               0x3UL
 119 #define __DST_METRICS_PTR(Y)    \
 120         ((u32 *)((Y) & ~DST_METRICS_FLAGS))
 121 #define DST_METRICS_PTR(X)      __DST_METRICS_PTR((X)->_metrics)
 122
 123 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
 124 {
 125         return dst->_metrics & DST_METRICS_READ_ONLY;
 126 }
 127
 128 void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
 129
 130 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
 131 {
 132         unsigned long val = dst->_metrics;
 133         if (!(val & DST_METRICS_READ_ONLY))
 134                 __dst_destroy_metrics_generic(dst, val);
 135 }
 136
 137 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
 138 {
 139         unsigned long p = dst->_metrics;
 140
 141         BUG_ON(!p);
 142
 143         if (p & DST_METRICS_READ_ONLY)
 144                 return dst->ops->cow_metrics(dst, p);
 145         return __DST_METRICS_PTR(p);
 146 }
 147
 148 /* This may only be invoked before the entry has reached global
 149  * visibility.
 150  */
 151 static inline void dst_init_metrics(struct dst_entry *dst,
 152                                     const u32 *src_metrics,
 153                                     bool read_only)
 154 {
 155         dst->_metrics = ((unsigned long) src_metrics) |
 156                 (read_only ? DST_METRICS_READ_ONLY : 0);
 157 }
 158
 159 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
 160 {
 161         u32 *dst_metrics = dst_metrics_write_ptr(dest);
 162
 163         if (dst_metrics) {
 164                 u32 *src_metrics = DST_METRICS_PTR(src);
 165
 166                 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
 167         }
 168 }
 169
 170 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
 171 {
 172         return DST_METRICS_PTR(dst);
 173 }
 174
 175 static inline u32
 176 dst_metric_raw(const struct dst_entry *dst, const int metric)
 177 {
 178         u32 *p = DST_METRICS_PTR(dst);
 179
 180         return p[metric-1];
 181 }
 182
 183 static inline u32
 184 dst_metric(const struct dst_entry *dst, const int metric)
 185 {
 186         WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
 187                      metric == RTAX_ADVMSS ||
 188                      metric == RTAX_MTU);
 189         return dst_metric_raw(dst, metric);
 190 }
 191
 192 static inline u32
 193 dst_metric_advmss(const struct dst_entry *dst)
 194 {
 195         u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
 196
 197         if (!advmss)
 198                 advmss = dst->ops->default_advmss(dst);
 199
 200         return advmss;
 201 }
 202
 203 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
 204 {
 205         u32 *p = dst_metrics_write_ptr(dst);
 206
 207         if (p)
 208                 p[metric-1] = val;
 209 }
 210
 211 /* Kernel-internal feature bits that are unallocated in user space. */
 212 #define DST_FEATURE_ECN_CA      (1 << 31)
 213
 214 #define DST_FEATURE_MASK        (DST_FEATURE_ECN_CA)
 215 #define DST_FEATURE_ECN_MASK    (DST_FEATURE_ECN_CA | RTAX_FEATURE_ECN)
 216
 217 static inline u32
 218 dst_feature(const struct dst_entry *dst, u32 feature)
 219 {
 220         return dst_metric(dst, RTAX_FEATURES) & feature;
 221 }
 222
 223 static inline u32 dst_mtu(const struct dst_entry *dst)
 224 {
 225         return dst->ops->mtu(dst);
 226 }
 227
 228 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
 229 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
 230 {
 231         return msecs_to_jiffies(dst_metric(dst, metric));
 232 }
 233
 234 static inline u32
 235 dst_allfrag(const struct dst_entry *dst)
 236 {
 237         int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
 238         return ret;
 239 }
 240
 241 static inline int
 242 dst_metric_locked(const struct dst_entry *dst, int metric)
 243 {
 244         return dst_metric(dst, RTAX_LOCK) & (1<<metric);
 245 }
 246
 247 static inline void dst_hold(struct dst_entry *dst)
 248 {
 249         /*
 250          * If your kernel compilation stops here, please check
 251          * __pad_to_align_refcnt declaration in struct dst_entry
 252          */
 253         BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
 254         WARN_ON(atomic_inc_not_zero(&dst->__refcnt) == 0);
 255 }
 256
 257 static inline void dst_use(struct dst_entry *dst, unsigned long time)
 258 {
 259         dst_hold(dst);
 260         dst->__use++;
 261         dst->lastuse = time;
 262 }
 263
 264 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
 265 {
 266         dst->__use++;
 267         dst->lastuse = time;
 268 }
 269
 270 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
 271 {
 272         if (dst)
 273                 atomic_inc(&dst->__refcnt);
 274         return dst;
 275 }
 276
 277 void dst_release(struct dst_entry *dst);
 278
 279 void dst_release_immediate(struct dst_entry *dst);
 280
 281 static inline void refdst_drop(unsigned long refdst)
 282 {
 283         if (!(refdst & SKB_DST_NOREF))
 284                 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
 285 }
 286
 287 /**
 288  * skb_dst_drop - drops skb dst
 289  * @skb: buffer
 290  *
 291  * Drops dst reference count if a reference was taken.
 292  */
 293 static inline void skb_dst_drop(struct sk_buff *skb)
 294 {
 295         if (skb->_skb_refdst) {
 296                 refdst_drop(skb->_skb_refdst);
 297                 skb->_skb_refdst = 0UL;
 298         }
 299 }
 300
 301 static inline void __skb_dst_copy(struct sk_buff *nskb, unsigned long refdst)
 302 {
 303         nskb->_skb_refdst = refdst;
 304         if (!(nskb->_skb_refdst & SKB_DST_NOREF))
 305                 dst_clone(skb_dst(nskb));
 306 }
 307
 308 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
 309 {
 310         __skb_dst_copy(nskb, oskb->_skb_refdst);
 311 }
 312
 313 /**
 314  * skb_dst_force - makes sure skb dst is refcounted
 315  * @skb: buffer
 316  *
 317  * If dst is not yet refcounted, let's do it
 318  */
 319 static inline void skb_dst_force(struct sk_buff *skb)
 320 {
 321         if (skb_dst_is_noref(skb)) {
 322                 WARN_ON(!rcu_read_lock_held());
 323                 skb->_skb_refdst &= ~SKB_DST_NOREF;
 324                 dst_clone(skb_dst(skb));
 325         }
 326 }
 327
 328 /**
 329  * dst_hold_safe - Take a reference on a dst if possible
 330  * @dst: pointer to dst entry
 331  *
 332  * This helper returns false if it could not safely
 333  * take a reference on a dst.
 334  */
 335 static inline bool dst_hold_safe(struct dst_entry *dst)
 336 {
 337         return atomic_inc_not_zero(&dst->__refcnt);
 338 }
 339
 340 /**
 341  * skb_dst_force_safe - makes sure skb dst is refcounted
 342  * @skb: buffer
 343  *
 344  * If dst is not yet refcounted and not destroyed, grab a ref on it.
 345  */
 346 static inline void skb_dst_force_safe(struct sk_buff *skb)
 347 {
 348         if (skb_dst_is_noref(skb)) {
 349                 struct dst_entry *dst = skb_dst(skb);
 350
 351                 if (!dst_hold_safe(dst))
 352                         dst = NULL;
 353
 354                 skb->_skb_refdst = (unsigned long)dst;
 355         }
 356 }
 357
 358
 359 /**
 360  *      __skb_tunnel_rx - prepare skb for rx reinsert
 361  *      @skb: buffer
 362  *      @dev: tunnel device
 363  *      @net: netns for packet i/o
 364  *
 365  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 366  *      so make some cleanups. (no accounting done)
 367  */
 368 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
 369                                    struct net *net)
 370 {
 371         skb->dev = dev;
 372
 373         /*
 374          * Clear hash so that we can recalulate the hash for the
 375          * encapsulated packet, unless we have already determine the hash
 376          * over the L4 4-tuple.
 377          */
 378         skb_clear_hash_if_not_l4(skb);
 379         skb_set_queue_mapping(skb, 0);
 380         skb_scrub_packet(skb, !net_eq(net, dev_net(dev)));
 381 }
 382
 383 /**
 384  *      skb_tunnel_rx - prepare skb for rx reinsert
 385  *      @skb: buffer
 386  *      @dev: tunnel device
 387  *
 388  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 389  *      so make some cleanups, and perform accounting.
 390  *      Note: this accounting is not SMP safe.
 391  */
 392 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev,
 393                                  struct net *net)
 394 {
 395         /* TODO : stats should be SMP safe */
 396         dev->stats.rx_packets++;
 397         dev->stats.rx_bytes += skb->len;
 398         __skb_tunnel_rx(skb, dev, net);
 399 }
 400
 401 static inline u32 dst_tclassid(const struct sk_buff *skb)
 402 {
 403 #ifdef CONFIG_IP_ROUTE_CLASSID
 404         const struct dst_entry *dst;
 405
 406         dst = skb_dst(skb);
 407         if (dst)
 408                 return dst->tclassid;
 409 #endif
 410         return 0;
 411 }
 412
 413 int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
 414 static inline int dst_discard(struct sk_buff *skb)
 415 {
 416         return dst_discard_out(&init_net, skb->sk, skb);
 417 }
 418 void *dst_alloc(struct dst_ops *ops, struct net_device *dev, int initial_ref,
 419                 int initial_obsolete, unsigned short flags);
 420 void dst_init(struct dst_entry *dst, struct dst_ops *ops,
 421               struct net_device *dev, int initial_ref, int initial_obsolete,
 422               unsigned short flags);
 423 struct dst_entry *dst_destroy(struct dst_entry *dst);
 424 void dst_dev_put(struct dst_entry *dst);
 425
 426 static inline void dst_confirm(struct dst_entry *dst)
 427 {
 428 }
 429
 430 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
 431 {
 432         struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
 433         return IS_ERR(n) ? NULL : n;
 434 }
 435
 436 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
 437                                                      struct sk_buff *skb)
 438 {
 439         struct neighbour *n =  dst->ops->neigh_lookup(dst, skb, NULL);
 440         return IS_ERR(n) ? NULL : n;
 441 }
 442
 443 static inline void dst_confirm_neigh(const struct dst_entry *dst,
 444                                      const void *daddr)
 445 {
 446         if (dst->ops->confirm_neigh)
 447                 dst->ops->confirm_neigh(dst, daddr);
 448 }
 449
 450 static inline void dst_link_failure(struct sk_buff *skb)
 451 {
 452         struct dst_entry *dst = skb_dst(skb);
 453         if (dst && dst->ops && dst->ops->link_failure)
 454                 dst->ops->link_failure(skb);
 455 }
 456
 457 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
 458 {
 459         unsigned long expires = jiffies + timeout;
 460
 461         if (expires == 0)
 462                 expires = 1;
 463
 464         if (dst->expires == 0 || time_before(expires, dst->expires))
 465                 dst->expires = expires;
 466 }
 467
 468 /* Output packet to network from transport.  */
 469 static inline int dst_output(struct net *net, struct sock *sk, struct sk_buff *skb)
 470 {
 471         return skb_dst(skb)->output(net, sk, skb);
 472 }
 473
 474 /* Input packet from network to transport.  */
 475 static inline int dst_input(struct sk_buff *skb)
 476 {
 477         return skb_dst(skb)->input(skb);
 478 }
 479
 480 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
 481 {
 482         if (dst->obsolete)
 483                 dst = dst->ops->check(dst, cookie);
 484         return dst;
 485 }
 486
 487 /* Flags for xfrm_lookup flags argument. */
 488 enum {
 489         XFRM_LOOKUP_ICMP = 1 << 0,
 490         XFRM_LOOKUP_QUEUE = 1 << 1,
 491         XFRM_LOOKUP_KEEP_DST_REF = 1 << 2,
 492 };
 493
 494 struct flowi;
 495 #ifndef CONFIG_XFRM
 496 static inline struct dst_entry *xfrm_lookup(struct net *net,
 497                                             struct dst_entry *dst_orig,
 498                                             const struct flowi *fl,
 499                                             const struct sock *sk,
 500                                             int flags)
 501 {
 502         return dst_orig;
 503 }
 504
 505 static inline struct dst_entry *xfrm_lookup_route(struct net *net,
 506                                                   struct dst_entry *dst_orig,
 507                                                   const struct flowi *fl,
 508                                                   const struct sock *sk,
 509                                                   int flags)
 510 {
 511         return dst_orig;
 512 }
 513
 514 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
 515 {
 516         return NULL;
 517 }
 518
 519 #else
 520 struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
 521                               const struct flowi *fl, const struct sock *sk,
 522                               int flags);
 523
 524 struct dst_entry *xfrm_lookup_route(struct net *net, struct dst_entry *dst_orig,
 525                                     const struct flowi *fl, const struct sock *sk,
 526                                     int flags);
 527
 528 /* skb attached with this dst needs transformation if dst->xfrm is valid */
 529 static inline struct xfrm_state *dst_xfrm(const struct dst_entry *dst)
 530 {
 531         return dst->xfrm;
 532 }
 533 #endif
 534
 535 #endif /* _NET_DST_H */