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         union {
  40                 unsigned long           expires;
  41                 /* point to where the dst_entry copied from */
  42                 struct dst_entry        *from;
  43         };
  44         struct dst_entry        *path;
  45         void                    *__pad0;
  46 #ifdef CONFIG_XFRM
  47         struct xfrm_state       *xfrm;
  48 #else
  49         void                    *__pad1;
  50 #endif
  51         int                     (*input)(struct sk_buff *);
  52         int                     (*output)(struct sk_buff *);
  53
  54         unsigned short          flags;
  55 #define DST_HOST                0x0001
  56 #define DST_NOXFRM              0x0002
  57 #define DST_NOPOLICY            0x0004
  58 #define DST_NOHASH              0x0008
  59 #define DST_NOCACHE             0x0010
  60 #define DST_NOCOUNT             0x0020
  61 #define DST_NOPEER              0x0040
  62 #define DST_FAKE_RTABLE         0x0080
  63 #define DST_XFRM_TUNNEL         0x0100
  64
  65         unsigned short          pending_confirm;
  66
  67         short                   error;
  68
  69         /* A non-zero value of dst->obsolete forces by-hand validation
  70          * of the route entry.  Positive values are set by the generic
  71          * dst layer to indicate that the entry has been forcefully
  72          * destroyed.
  73          *
  74          * Negative values are used by the implementation layer code to
  75          * force invocation of the dst_ops->check() method.
  76          */
  77         short                   obsolete;
  78 #define DST_OBSOLETE_NONE       0
  79 #define DST_OBSOLETE_DEAD       2
  80 #define DST_OBSOLETE_FORCE_CHK  -1
  81 #define DST_OBSOLETE_KILL       -2
  82         unsigned short          header_len;     /* more space at head required */
  83         unsigned short          trailer_len;    /* space to reserve at tail */
  84 #ifdef CONFIG_IP_ROUTE_CLASSID
  85         __u32                   tclassid;
  86 #else
  87         __u32                   __pad2;
  88 #endif
  89
  90         /*
  91          * Align __refcnt to a 64 bytes alignment
  92          * (L1_CACHE_SIZE would be too much)
  93          */
  94 #ifdef CONFIG_64BIT
  95         long                    __pad_to_align_refcnt[2];
  96 #endif
  97         /*
  98          * __refcnt wants to be on a different cache line from
  99          * input/output/ops or performance tanks badly
 100          */
 101         atomic_t                __refcnt;       /* client references    */
 102         int                     __use;
 103         unsigned long           lastuse;
 104         union {
 105                 struct dst_entry        *next;
 106                 struct rtable __rcu     *rt_next;
 107                 struct rt6_info         *rt6_next;
 108                 struct dn_route __rcu   *dn_next;
 109         };
 110 };
 111
 112 extern u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
 113 extern const u32 dst_default_metrics[];
 114
 115 #define DST_METRICS_READ_ONLY   0x1UL
 116 #define __DST_METRICS_PTR(Y)    \
 117         ((u32 *)((Y) & ~DST_METRICS_READ_ONLY))
 118 #define DST_METRICS_PTR(X)      __DST_METRICS_PTR((X)->_metrics)
 119
 120 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
 121 {
 122         return dst->_metrics & DST_METRICS_READ_ONLY;
 123 }
 124
 125 extern void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
 126
 127 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
 128 {
 129         unsigned long val = dst->_metrics;
 130         if (!(val & DST_METRICS_READ_ONLY))
 131                 __dst_destroy_metrics_generic(dst, val);
 132 }
 133
 134 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
 135 {
 136         unsigned long p = dst->_metrics;
 137
 138         BUG_ON(!p);
 139
 140         if (p & DST_METRICS_READ_ONLY)
 141                 return dst->ops->cow_metrics(dst, p);
 142         return __DST_METRICS_PTR(p);
 143 }
 144
 145 /* This may only be invoked before the entry has reached global
 146  * visibility.
 147  */
 148 static inline void dst_init_metrics(struct dst_entry *dst,
 149                                     const u32 *src_metrics,
 150                                     bool read_only)
 151 {
 152         dst->_metrics = ((unsigned long) src_metrics) |
 153                 (read_only ? DST_METRICS_READ_ONLY : 0);
 154 }
 155
 156 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
 157 {
 158         u32 *dst_metrics = dst_metrics_write_ptr(dest);
 159
 160         if (dst_metrics) {
 161                 u32 *src_metrics = DST_METRICS_PTR(src);
 162
 163                 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
 164         }
 165 }
 166
 167 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
 168 {
 169         return DST_METRICS_PTR(dst);
 170 }
 171
 172 static inline u32
 173 dst_metric_raw(const struct dst_entry *dst, const int metric)
 174 {
 175         u32 *p = DST_METRICS_PTR(dst);
 176
 177         return p[metric-1];
 178 }
 179
 180 static inline u32
 181 dst_metric(const struct dst_entry *dst, const int metric)
 182 {
 183         WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
 184                      metric == RTAX_ADVMSS ||
 185                      metric == RTAX_MTU);
 186         return dst_metric_raw(dst, metric);
 187 }
 188
 189 static inline u32
 190 dst_metric_advmss(const struct dst_entry *dst)
 191 {
 192         u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
 193
 194         if (!advmss)
 195                 advmss = dst->ops->default_advmss(dst);
 196
 197         return advmss;
 198 }
 199
 200 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
 201 {
 202         u32 *p = dst_metrics_write_ptr(dst);
 203
 204         if (p)
 205                 p[metric-1] = val;
 206 }
 207
 208 static inline u32
 209 dst_feature(const struct dst_entry *dst, u32 feature)
 210 {
 211         return dst_metric(dst, RTAX_FEATURES) & feature;
 212 }
 213
 214 static inline u32 dst_mtu(const struct dst_entry *dst)
 215 {
 216         return dst->ops->mtu(dst);
 217 }
 218
 219 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
 220 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
 221 {
 222         return msecs_to_jiffies(dst_metric(dst, metric));
 223 }
 224
 225 static inline u32
 226 dst_allfrag(const struct dst_entry *dst)
 227 {
 228         int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
 229         return ret;
 230 }
 231
 232 static inline int
 233 dst_metric_locked(const struct dst_entry *dst, int metric)
 234 {
 235         return dst_metric(dst, RTAX_LOCK) & (1<<metric);
 236 }
 237
 238 static inline void dst_hold(struct dst_entry *dst)
 239 {
 240         /*
 241          * If your kernel compilation stops here, please check
 242          * __pad_to_align_refcnt declaration in struct dst_entry
 243          */
 244         BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
 245         atomic_inc(&dst->__refcnt);
 246 }
 247
 248 static inline void dst_use(struct dst_entry *dst, unsigned long time)
 249 {
 250         dst_hold(dst);
 251         dst->__use++;
 252         dst->lastuse = time;
 253 }
 254
 255 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
 256 {
 257         dst->__use++;
 258         dst->lastuse = time;
 259 }
 260
 261 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
 262 {
 263         if (dst)
 264                 atomic_inc(&dst->__refcnt);
 265         return dst;
 266 }
 267
 268 extern void dst_release(struct dst_entry *dst);
 269
 270 static inline void refdst_drop(unsigned long refdst)
 271 {
 272         if (!(refdst & SKB_DST_NOREF))
 273                 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
 274 }
 275
 276 /**
 277  * skb_dst_drop - drops skb dst
 278  * @skb: buffer
 279  *
 280  * Drops dst reference count if a reference was taken.
 281  */
 282 static inline void skb_dst_drop(struct sk_buff *skb)
 283 {
 284         if (skb->_skb_refdst) {
 285                 refdst_drop(skb->_skb_refdst);
 286                 skb->_skb_refdst = 0UL;
 287         }
 288 }
 289
 290 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
 291 {
 292         nskb->_skb_refdst = oskb->_skb_refdst;
 293         if (!(nskb->_skb_refdst & SKB_DST_NOREF))
 294                 dst_clone(skb_dst(nskb));
 295 }
 296
 297 /**
 298  * skb_dst_force - makes sure skb dst is refcounted
 299  * @skb: buffer
 300  *
 301  * If dst is not yet refcounted, let's do it
 302  */
 303 static inline void skb_dst_force(struct sk_buff *skb)
 304 {
 305         if (skb_dst_is_noref(skb)) {
 306                 WARN_ON(!rcu_read_lock_held());
 307                 skb->_skb_refdst &= ~SKB_DST_NOREF;
 308                 dst_clone(skb_dst(skb));
 309         }
 310 }
 311
 312
 313 /**
 314  *      __skb_tunnel_rx - prepare skb for rx reinsert
 315  *      @skb: buffer
 316  *      @dev: tunnel device
 317  *
 318  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 319  *      so make some cleanups. (no accounting done)
 320  */
 321 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
 322 {
 323         skb->dev = dev;
 324
 325         /*
 326          * Clear rxhash so that we can recalulate the hash for the
 327          * encapsulated packet, unless we have already determine the hash
 328          * over the L4 4-tuple.
 329          */
 330         if (!skb->l4_rxhash)
 331                 skb->rxhash = 0;
 332         skb_set_queue_mapping(skb, 0);
 333         skb_dst_drop(skb);
 334         nf_reset(skb);
 335 }
 336
 337 /**
 338  *      skb_tunnel_rx - prepare skb for rx reinsert
 339  *      @skb: buffer
 340  *      @dev: tunnel device
 341  *
 342  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 343  *      so make some cleanups, and perform accounting.
 344  *      Note: this accounting is not SMP safe.
 345  */
 346 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
 347 {
 348         /* TODO : stats should be SMP safe */
 349         dev->stats.rx_packets++;
 350         dev->stats.rx_bytes += skb->len;
 351         __skb_tunnel_rx(skb, dev);
 352 }
 353
 354 /* Children define the path of the packet through the
 355  * Linux networking.  Thus, destinations are stackable.
 356  */
 357
 358 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
 359 {
 360         struct dst_entry *child = dst_clone(skb_dst(skb)->child);
 361
 362         skb_dst_drop(skb);
 363         return child;
 364 }
 365
 366 extern int dst_discard(struct sk_buff *skb);
 367 extern void *dst_alloc(struct dst_ops *ops, struct net_device *dev,
 368                        int initial_ref, int initial_obsolete,
 369                        unsigned short flags);
 370 extern void __dst_free(struct dst_entry *dst);
 371 extern struct dst_entry *dst_destroy(struct dst_entry *dst);
 372
 373 static inline void dst_free(struct dst_entry *dst)
 374 {
 375         if (dst->obsolete > 0)
 376                 return;
 377         if (!atomic_read(&dst->__refcnt)) {
 378                 dst = dst_destroy(dst);
 379                 if (!dst)
 380                         return;
 381         }
 382         __dst_free(dst);
 383 }
 384
 385 static inline void dst_rcu_free(struct rcu_head *head)
 386 {
 387         struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
 388         dst_free(dst);
 389 }
 390
 391 static inline void dst_confirm(struct dst_entry *dst)
 392 {
 393         dst->pending_confirm = 1;
 394 }
 395
 396 static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n,
 397                                    struct sk_buff *skb)
 398 {
 399         const struct hh_cache *hh;
 400
 401         if (dst->pending_confirm) {
 402                 unsigned long now = jiffies;
 403
 404                 dst->pending_confirm = 0;
 405                 /* avoid dirtying neighbour */
 406                 if (n->confirmed != now)
 407                         n->confirmed = now;
 408         }
 409
 410         hh = &n->hh;
 411         if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
 412                 return neigh_hh_output(hh, skb);
 413         else
 414                 return n->output(n, skb);
 415 }
 416
 417 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
 418 {
 419         return dst->ops->neigh_lookup(dst, NULL, daddr);
 420 }
 421
 422 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
 423                                                      struct sk_buff *skb)
 424 {
 425         return dst->ops->neigh_lookup(dst, skb, NULL);
 426 }
 427
 428 static inline void dst_link_failure(struct sk_buff *skb)
 429 {
 430         struct dst_entry *dst = skb_dst(skb);
 431         if (dst && dst->ops && dst->ops->link_failure)
 432                 dst->ops->link_failure(skb);
 433 }
 434
 435 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
 436 {
 437         unsigned long expires = jiffies + timeout;
 438
 439         if (expires == 0)
 440                 expires = 1;
 441
 442         if (dst->expires == 0 || time_before(expires, dst->expires))
 443                 dst->expires = expires;
 444 }
 445
 446 /* Output packet to network from transport.  */
 447 static inline int dst_output(struct sk_buff *skb)
 448 {
 449         return skb_dst(skb)->output(skb);
 450 }
 451
 452 /* Input packet from network to transport.  */
 453 static inline int dst_input(struct sk_buff *skb)
 454 {
 455         return skb_dst(skb)->input(skb);
 456 }
 457
 458 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
 459 {
 460         if (dst->obsolete)
 461                 dst = dst->ops->check(dst, cookie);
 462         return dst;
 463 }
 464
 465 extern void             dst_init(void);
 466
 467 /* Flags for xfrm_lookup flags argument. */
 468 enum {
 469         XFRM_LOOKUP_ICMP = 1 << 0,
 470 };
 471
 472 struct flowi;
 473 #ifndef CONFIG_XFRM
 474 static inline struct dst_entry *xfrm_lookup(struct net *net,
 475                                             struct dst_entry *dst_orig,
 476                                             const struct flowi *fl, struct sock *sk,
 477                                             int flags)
 478 {
 479         return dst_orig;
 480 }
 481 #else
 482 extern struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
 483                                      const struct flowi *fl, struct sock *sk,
 484                                      int flags);
 485 #endif
 486
 487 #endif /* _NET_DST_H */