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 rcu_head         rcu_head;
  35         struct dst_entry        *child;
  36         struct net_device       *dev;
  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 sk_buff *);
  49
  50         unsigned short          flags;
  51 #define DST_HOST                0x0001
  52 #define DST_NOXFRM              0x0002
  53 #define DST_NOPOLICY            0x0004
  54 #define DST_NOHASH              0x0008
  55 #define DST_NOCACHE             0x0010
  56 #define DST_NOCOUNT             0x0020
  57 #define DST_NOPEER              0x0040
  58 #define DST_FAKE_RTABLE         0x0080
  59 #define DST_XFRM_TUNNEL         0x0100
  60 #define DST_XFRM_QUEUE          0x0200
  61
  62         unsigned short          pending_confirm;
  63
  64         short                   error;
  65
  66         /* A non-zero value of dst->obsolete forces by-hand validation
  67          * of the route entry.  Positive values are set by the generic
  68          * dst layer to indicate that the entry has been forcefully
  69          * destroyed.
  70          *
  71          * Negative values are used by the implementation layer code to
  72          * force invocation of the dst_ops->check() method.
  73          */
  74         short                   obsolete;
  75 #define DST_OBSOLETE_NONE       0
  76 #define DST_OBSOLETE_DEAD       2
  77 #define DST_OBSOLETE_FORCE_CHK  -1
  78 #define DST_OBSOLETE_KILL       -2
  79         unsigned short          header_len;     /* more space at head required */
  80         unsigned short          trailer_len;    /* space to reserve at tail */
  81 #ifdef CONFIG_IP_ROUTE_CLASSID
  82         __u32                   tclassid;
  83 #else
  84         __u32                   __pad2;
  85 #endif
  86
  87         /*
  88          * Align __refcnt to a 64 bytes alignment
  89          * (L1_CACHE_SIZE would be too much)
  90          */
  91 #ifdef CONFIG_64BIT
  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         union {
 102                 struct dst_entry        *next;
 103                 struct rtable __rcu     *rt_next;
 104                 struct rt6_info         *rt6_next;
 105                 struct dn_route __rcu   *dn_next;
 106         };
 107 };
 108
 109 extern u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
 110 extern const u32 dst_default_metrics[];
 111
 112 #define DST_METRICS_READ_ONLY   0x1UL
 113 #define __DST_METRICS_PTR(Y)    \
 114         ((u32 *)((Y) & ~DST_METRICS_READ_ONLY))
 115 #define DST_METRICS_PTR(X)      __DST_METRICS_PTR((X)->_metrics)
 116
 117 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
 118 {
 119         return dst->_metrics & DST_METRICS_READ_ONLY;
 120 }
 121
 122 extern void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
 123
 124 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
 125 {
 126         unsigned long val = dst->_metrics;
 127         if (!(val & DST_METRICS_READ_ONLY))
 128                 __dst_destroy_metrics_generic(dst, val);
 129 }
 130
 131 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
 132 {
 133         unsigned long p = dst->_metrics;
 134
 135         BUG_ON(!p);
 136
 137         if (p & DST_METRICS_READ_ONLY)
 138                 return dst->ops->cow_metrics(dst, p);
 139         return __DST_METRICS_PTR(p);
 140 }
 141
 142 /* This may only be invoked before the entry has reached global
 143  * visibility.
 144  */
 145 static inline void dst_init_metrics(struct dst_entry *dst,
 146                                     const u32 *src_metrics,
 147                                     bool read_only)
 148 {
 149         dst->_metrics = ((unsigned long) src_metrics) |
 150                 (read_only ? DST_METRICS_READ_ONLY : 0);
 151 }
 152
 153 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
 154 {
 155         u32 *dst_metrics = dst_metrics_write_ptr(dest);
 156
 157         if (dst_metrics) {
 158                 u32 *src_metrics = DST_METRICS_PTR(src);
 159
 160                 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
 161         }
 162 }
 163
 164 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
 165 {
 166         return DST_METRICS_PTR(dst);
 167 }
 168
 169 static inline u32
 170 dst_metric_raw(const struct dst_entry *dst, const int metric)
 171 {
 172         u32 *p = DST_METRICS_PTR(dst);
 173
 174         return p[metric-1];
 175 }
 176
 177 static inline u32
 178 dst_metric(const struct dst_entry *dst, const int metric)
 179 {
 180         WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
 181                      metric == RTAX_ADVMSS ||
 182                      metric == RTAX_MTU);
 183         return dst_metric_raw(dst, metric);
 184 }
 185
 186 static inline u32
 187 dst_metric_advmss(const struct dst_entry *dst)
 188 {
 189         u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
 190
 191         if (!advmss)
 192                 advmss = dst->ops->default_advmss(dst);
 193
 194         return advmss;
 195 }
 196
 197 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
 198 {
 199         u32 *p = dst_metrics_write_ptr(dst);
 200
 201         if (p)
 202                 p[metric-1] = val;
 203 }
 204
 205 static inline u32
 206 dst_feature(const struct dst_entry *dst, u32 feature)
 207 {
 208         return dst_metric(dst, RTAX_FEATURES) & feature;
 209 }
 210
 211 static inline u32 dst_mtu(const struct dst_entry *dst)
 212 {
 213         return dst->ops->mtu(dst);
 214 }
 215
 216 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
 217 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
 218 {
 219         return msecs_to_jiffies(dst_metric(dst, metric));
 220 }
 221
 222 static inline u32
 223 dst_allfrag(const struct dst_entry *dst)
 224 {
 225         int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
 226         return ret;
 227 }
 228
 229 static inline int
 230 dst_metric_locked(const struct dst_entry *dst, int metric)
 231 {
 232         return dst_metric(dst, RTAX_LOCK) & (1<<metric);
 233 }
 234
 235 static inline void dst_hold(struct dst_entry *dst)
 236 {
 237         /*
 238          * If your kernel compilation stops here, please check
 239          * __pad_to_align_refcnt declaration in struct dst_entry
 240          */
 241         BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
 242         atomic_inc(&dst->__refcnt);
 243 }
 244
 245 static inline void dst_use(struct dst_entry *dst, unsigned long time)
 246 {
 247         dst_hold(dst);
 248         dst->__use++;
 249         dst->lastuse = time;
 250 }
 251
 252 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
 253 {
 254         dst->__use++;
 255         dst->lastuse = time;
 256 }
 257
 258 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
 259 {
 260         if (dst)
 261                 atomic_inc(&dst->__refcnt);
 262         return dst;
 263 }
 264
 265 extern void dst_release(struct dst_entry *dst);
 266
 267 static inline void refdst_drop(unsigned long refdst)
 268 {
 269         if (!(refdst & SKB_DST_NOREF))
 270                 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
 271 }
 272
 273 /**
 274  * skb_dst_drop - drops skb dst
 275  * @skb: buffer
 276  *
 277  * Drops dst reference count if a reference was taken.
 278  */
 279 static inline void skb_dst_drop(struct sk_buff *skb)
 280 {
 281         if (skb->_skb_refdst) {
 282                 refdst_drop(skb->_skb_refdst);
 283                 skb->_skb_refdst = 0UL;
 284         }
 285 }
 286
 287 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
 288 {
 289         nskb->_skb_refdst = oskb->_skb_refdst;
 290         if (!(nskb->_skb_refdst & SKB_DST_NOREF))
 291                 dst_clone(skb_dst(nskb));
 292 }
 293
 294 /**
 295  * skb_dst_force - makes sure skb dst is refcounted
 296  * @skb: buffer
 297  *
 298  * If dst is not yet refcounted, let's do it
 299  */
 300 static inline void skb_dst_force(struct sk_buff *skb)
 301 {
 302         if (skb_dst_is_noref(skb)) {
 303                 WARN_ON(!rcu_read_lock_held());
 304                 skb->_skb_refdst &= ~SKB_DST_NOREF;
 305                 dst_clone(skb_dst(skb));
 306         }
 307 }
 308
 309
 310 /**
 311  *      __skb_tunnel_rx - prepare skb for rx reinsert
 312  *      @skb: buffer
 313  *      @dev: tunnel device
 314  *
 315  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 316  *      so make some cleanups. (no accounting done)
 317  */
 318 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
 319 {
 320         skb->dev = dev;
 321
 322         /*
 323          * Clear rxhash so that we can recalulate the hash for the
 324          * encapsulated packet, unless we have already determine the hash
 325          * over the L4 4-tuple.
 326          */
 327         if (!skb->l4_rxhash)
 328                 skb->rxhash = 0;
 329         skb_set_queue_mapping(skb, 0);
 330         skb_dst_drop(skb);
 331         nf_reset(skb);
 332 }
 333
 334 /**
 335  *      skb_tunnel_rx - prepare skb for rx reinsert
 336  *      @skb: buffer
 337  *      @dev: tunnel device
 338  *
 339  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 340  *      so make some cleanups, and perform accounting.
 341  *      Note: this accounting is not SMP safe.
 342  */
 343 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
 344 {
 345         /* TODO : stats should be SMP safe */
 346         dev->stats.rx_packets++;
 347         dev->stats.rx_bytes += skb->len;
 348         __skb_tunnel_rx(skb, dev);
 349 }
 350
 351 /* Children define the path of the packet through the
 352  * Linux networking.  Thus, destinations are stackable.
 353  */
 354
 355 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
 356 {
 357         struct dst_entry *child = dst_clone(skb_dst(skb)->child);
 358
 359         skb_dst_drop(skb);
 360         return child;
 361 }
 362
 363 extern int dst_discard(struct sk_buff *skb);
 364 extern void *dst_alloc(struct dst_ops *ops, struct net_device *dev,
 365                        int initial_ref, int initial_obsolete,
 366                        unsigned short flags);
 367 extern void __dst_free(struct dst_entry *dst);
 368 extern struct dst_entry *dst_destroy(struct dst_entry *dst);
 369
 370 static inline void dst_free(struct dst_entry *dst)
 371 {
 372         if (dst->obsolete > 0)
 373                 return;
 374         if (!atomic_read(&dst->__refcnt)) {
 375                 dst = dst_destroy(dst);
 376                 if (!dst)
 377                         return;
 378         }
 379         __dst_free(dst);
 380 }
 381
 382 static inline void dst_rcu_free(struct rcu_head *head)
 383 {
 384         struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
 385         dst_free(dst);
 386 }
 387
 388 static inline void dst_confirm(struct dst_entry *dst)
 389 {
 390         dst->pending_confirm = 1;
 391 }
 392
 393 static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n,
 394                                    struct sk_buff *skb)
 395 {
 396         const struct hh_cache *hh;
 397
 398         if (dst->pending_confirm) {
 399                 unsigned long now = jiffies;
 400
 401                 dst->pending_confirm = 0;
 402                 /* avoid dirtying neighbour */
 403                 if (n->confirmed != now)
 404                         n->confirmed = now;
 405         }
 406
 407         hh = &n->hh;
 408         if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
 409                 return neigh_hh_output(hh, skb);
 410         else
 411                 return n->output(n, skb);
 412 }
 413
 414 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
 415 {
 416         struct neighbour *n = dst->ops->neigh_lookup(dst, NULL, daddr);
 417         return IS_ERR(n) ? NULL : n;
 418 }
 419
 420 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
 421                                                      struct sk_buff *skb)
 422 {
 423         struct neighbour *n =  dst->ops->neigh_lookup(dst, skb, NULL);
 424         return IS_ERR(n) ? NULL : n;
 425 }
 426
 427 static inline void dst_link_failure(struct sk_buff *skb)
 428 {
 429         struct dst_entry *dst = skb_dst(skb);
 430         if (dst && dst->ops && dst->ops->link_failure)
 431                 dst->ops->link_failure(skb);
 432 }
 433
 434 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
 435 {
 436         unsigned long expires = jiffies + timeout;
 437
 438         if (expires == 0)
 439                 expires = 1;
 440
 441         if (dst->expires == 0 || time_before(expires, dst->expires))
 442                 dst->expires = expires;
 443 }
 444
 445 /* Output packet to network from transport.  */
 446 static inline int dst_output(struct sk_buff *skb)
 447 {
 448         return skb_dst(skb)->output(skb);
 449 }
 450
 451 /* Input packet from network to transport.  */
 452 static inline int dst_input(struct sk_buff *skb)
 453 {
 454         return skb_dst(skb)->input(skb);
 455 }
 456
 457 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
 458 {
 459         if (dst->obsolete)
 460                 dst = dst->ops->check(dst, cookie);
 461         return dst;
 462 }
 463
 464 extern void             dst_init(void);
 465
 466 /* Flags for xfrm_lookup flags argument. */
 467 enum {
 468         XFRM_LOOKUP_ICMP = 1 << 0,
 469 };
 470
 471 struct flowi;
 472 #ifndef CONFIG_XFRM
 473 static inline struct dst_entry *xfrm_lookup(struct net *net,
 474                                             struct dst_entry *dst_orig,
 475                                             const struct flowi *fl, struct sock *sk,
 476                                             int flags)
 477 {
 478         return dst_orig;
 479 }
 480 #else
 481 extern struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
 482                                      const struct flowi *fl, struct sock *sk,
 483                                      int flags);
 484 #endif
 485
 486 #endif /* _NET_DST_H */