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 #define DST_XFRM_QUEUE          0x0200
  65
  66         unsigned short          pending_confirm;
  67
  68         short                   error;
  69
  70         /* A non-zero value of dst->obsolete forces by-hand validation
  71          * of the route entry.  Positive values are set by the generic
  72          * dst layer to indicate that the entry has been forcefully
  73          * destroyed.
  74          *
  75          * Negative values are used by the implementation layer code to
  76          * force invocation of the dst_ops->check() method.
  77          */
  78         short                   obsolete;
  79 #define DST_OBSOLETE_NONE       0
  80 #define DST_OBSOLETE_DEAD       2
  81 #define DST_OBSOLETE_FORCE_CHK  -1
  82 #define DST_OBSOLETE_KILL       -2
  83         unsigned short          header_len;     /* more space at head required */
  84         unsigned short          trailer_len;    /* space to reserve at tail */
  85 #ifdef CONFIG_IP_ROUTE_CLASSID
  86         __u32                   tclassid;
  87 #else
  88         __u32                   __pad2;
  89 #endif
  90
  91         /*
  92          * Align __refcnt to a 64 bytes alignment
  93          * (L1_CACHE_SIZE would be too much)
  94          */
  95 #ifdef CONFIG_64BIT
  96         long                    __pad_to_align_refcnt[2];
  97 #endif
  98         /*
  99          * __refcnt wants to be on a different cache line from
 100          * input/output/ops or performance tanks badly
 101          */
 102         atomic_t                __refcnt;       /* client references    */
 103         int                     __use;
 104         unsigned long           lastuse;
 105         union {
 106                 struct dst_entry        *next;
 107                 struct rtable __rcu     *rt_next;
 108                 struct rt6_info         *rt6_next;
 109                 struct dn_route __rcu   *dn_next;
 110         };
 111 };
 112
 113 extern u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
 114 extern const u32 dst_default_metrics[];
 115
 116 #define DST_METRICS_READ_ONLY   0x1UL
 117 #define __DST_METRICS_PTR(Y)    \
 118         ((u32 *)((Y) & ~DST_METRICS_READ_ONLY))
 119 #define DST_METRICS_PTR(X)      __DST_METRICS_PTR((X)->_metrics)
 120
 121 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
 122 {
 123         return dst->_metrics & DST_METRICS_READ_ONLY;
 124 }
 125
 126 extern void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
 127
 128 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
 129 {
 130         unsigned long val = dst->_metrics;
 131         if (!(val & DST_METRICS_READ_ONLY))
 132                 __dst_destroy_metrics_generic(dst, val);
 133 }
 134
 135 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
 136 {
 137         unsigned long p = dst->_metrics;
 138
 139         BUG_ON(!p);
 140
 141         if (p & DST_METRICS_READ_ONLY)
 142                 return dst->ops->cow_metrics(dst, p);
 143         return __DST_METRICS_PTR(p);
 144 }
 145
 146 /* This may only be invoked before the entry has reached global
 147  * visibility.
 148  */
 149 static inline void dst_init_metrics(struct dst_entry *dst,
 150                                     const u32 *src_metrics,
 151                                     bool read_only)
 152 {
 153         dst->_metrics = ((unsigned long) src_metrics) |
 154                 (read_only ? DST_METRICS_READ_ONLY : 0);
 155 }
 156
 157 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
 158 {
 159         u32 *dst_metrics = dst_metrics_write_ptr(dest);
 160
 161         if (dst_metrics) {
 162                 u32 *src_metrics = DST_METRICS_PTR(src);
 163
 164                 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
 165         }
 166 }
 167
 168 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
 169 {
 170         return DST_METRICS_PTR(dst);
 171 }
 172
 173 static inline u32
 174 dst_metric_raw(const struct dst_entry *dst, const int metric)
 175 {
 176         u32 *p = DST_METRICS_PTR(dst);
 177
 178         return p[metric-1];
 179 }
 180
 181 static inline u32
 182 dst_metric(const struct dst_entry *dst, const int metric)
 183 {
 184         WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
 185                      metric == RTAX_ADVMSS ||
 186                      metric == RTAX_MTU);
 187         return dst_metric_raw(dst, metric);
 188 }
 189
 190 static inline u32
 191 dst_metric_advmss(const struct dst_entry *dst)
 192 {
 193         u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
 194
 195         if (!advmss)
 196                 advmss = dst->ops->default_advmss(dst);
 197
 198         return advmss;
 199 }
 200
 201 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
 202 {
 203         u32 *p = dst_metrics_write_ptr(dst);
 204
 205         if (p)
 206                 p[metric-1] = val;
 207 }
 208
 209 static inline u32
 210 dst_feature(const struct dst_entry *dst, u32 feature)
 211 {
 212         return dst_metric(dst, RTAX_FEATURES) & feature;
 213 }
 214
 215 static inline u32 dst_mtu(const struct dst_entry *dst)
 216 {
 217         return dst->ops->mtu(dst);
 218 }
 219
 220 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
 221 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
 222 {
 223         return msecs_to_jiffies(dst_metric(dst, metric));
 224 }
 225
 226 static inline u32
 227 dst_allfrag(const struct dst_entry *dst)
 228 {
 229         int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
 230         return ret;
 231 }
 232
 233 static inline int
 234 dst_metric_locked(const struct dst_entry *dst, int metric)
 235 {
 236         return dst_metric(dst, RTAX_LOCK) & (1<<metric);
 237 }
 238
 239 static inline void dst_hold(struct dst_entry *dst)
 240 {
 241         /*
 242          * If your kernel compilation stops here, please check
 243          * __pad_to_align_refcnt declaration in struct dst_entry
 244          */
 245         BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
 246         atomic_inc(&dst->__refcnt);
 247 }
 248
 249 static inline void dst_use(struct dst_entry *dst, unsigned long time)
 250 {
 251         dst_hold(dst);
 252         dst->__use++;
 253         dst->lastuse = time;
 254 }
 255
 256 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
 257 {
 258         dst->__use++;
 259         dst->lastuse = time;
 260 }
 261
 262 static inline struct dst_entry *dst_clone(struct dst_entry *dst)
 263 {
 264         if (dst)
 265                 atomic_inc(&dst->__refcnt);
 266         return dst;
 267 }
 268
 269 extern void dst_release(struct dst_entry *dst);
 270
 271 static inline void refdst_drop(unsigned long refdst)
 272 {
 273         if (!(refdst & SKB_DST_NOREF))
 274                 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
 275 }
 276
 277 /**
 278  * skb_dst_drop - drops skb dst
 279  * @skb: buffer
 280  *
 281  * Drops dst reference count if a reference was taken.
 282  */
 283 static inline void skb_dst_drop(struct sk_buff *skb)
 284 {
 285         if (skb->_skb_refdst) {
 286                 refdst_drop(skb->_skb_refdst);
 287                 skb->_skb_refdst = 0UL;
 288         }
 289 }
 290
 291 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
 292 {
 293         nskb->_skb_refdst = oskb->_skb_refdst;
 294         if (!(nskb->_skb_refdst & SKB_DST_NOREF))
 295                 dst_clone(skb_dst(nskb));
 296 }
 297
 298 /**
 299  * skb_dst_force - makes sure skb dst is refcounted
 300  * @skb: buffer
 301  *
 302  * If dst is not yet refcounted, let's do it
 303  */
 304 static inline void skb_dst_force(struct sk_buff *skb)
 305 {
 306         if (skb_dst_is_noref(skb)) {
 307                 WARN_ON(!rcu_read_lock_held());
 308                 skb->_skb_refdst &= ~SKB_DST_NOREF;
 309                 dst_clone(skb_dst(skb));
 310         }
 311 }
 312
 313
 314 /**
 315  *      __skb_tunnel_rx - prepare skb for rx reinsert
 316  *      @skb: buffer
 317  *      @dev: tunnel device
 318  *
 319  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 320  *      so make some cleanups. (no accounting done)
 321  */
 322 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
 323 {
 324         skb->dev = dev;
 325
 326         /*
 327          * Clear rxhash so that we can recalulate the hash for the
 328          * encapsulated packet, unless we have already determine the hash
 329          * over the L4 4-tuple.
 330          */
 331         if (!skb->l4_rxhash)
 332                 skb->rxhash = 0;
 333         skb_set_queue_mapping(skb, 0);
 334         skb_dst_drop(skb);
 335         nf_reset(skb);
 336 }
 337
 338 /**
 339  *      skb_tunnel_rx - prepare skb for rx reinsert
 340  *      @skb: buffer
 341  *      @dev: tunnel device
 342  *
 343  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 344  *      so make some cleanups, and perform accounting.
 345  *      Note: this accounting is not SMP safe.
 346  */
 347 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
 348 {
 349         /* TODO : stats should be SMP safe */
 350         dev->stats.rx_packets++;
 351         dev->stats.rx_bytes += skb->len;
 352         __skb_tunnel_rx(skb, dev);
 353 }
 354
 355 /* Children define the path of the packet through the
 356  * Linux networking.  Thus, destinations are stackable.
 357  */
 358
 359 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
 360 {
 361         struct dst_entry *child = dst_clone(skb_dst(skb)->child);
 362
 363         skb_dst_drop(skb);
 364         return child;
 365 }
 366
 367 extern int dst_discard(struct sk_buff *skb);
 368 extern void *dst_alloc(struct dst_ops *ops, struct net_device *dev,
 369                        int initial_ref, int initial_obsolete,
 370                        unsigned short flags);
 371 extern void __dst_free(struct dst_entry *dst);
 372 extern struct dst_entry *dst_destroy(struct dst_entry *dst);
 373
 374 static inline void dst_free(struct dst_entry *dst)
 375 {
 376         if (dst->obsolete > 0)
 377                 return;
 378         if (!atomic_read(&dst->__refcnt)) {
 379                 dst = dst_destroy(dst);
 380                 if (!dst)
 381                         return;
 382         }
 383         __dst_free(dst);
 384 }
 385
 386 static inline void dst_rcu_free(struct rcu_head *head)
 387 {
 388         struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
 389         dst_free(dst);
 390 }
 391
 392 static inline void dst_confirm(struct dst_entry *dst)
 393 {
 394         dst->pending_confirm = 1;
 395 }
 396
 397 static inline int dst_neigh_output(struct dst_entry *dst, struct neighbour *n,
 398                                    struct sk_buff *skb)
 399 {
 400         const struct hh_cache *hh;
 401
 402         if (dst->pending_confirm) {
 403                 unsigned long now = jiffies;
 404
 405                 dst->pending_confirm = 0;
 406                 /* avoid dirtying neighbour */
 407                 if (n->confirmed != now)
 408                         n->confirmed = now;
 409         }
 410
 411         hh = &n->hh;
 412         if ((n->nud_state & NUD_CONNECTED) && hh->hh_len)
 413                 return neigh_hh_output(hh, skb);
 414         else
 415                 return n->output(n, skb);
 416 }
 417
 418 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
 419 {
 420         return dst->ops->neigh_lookup(dst, NULL, daddr);
 421 }
 422
 423 static inline struct neighbour *dst_neigh_lookup_skb(const struct dst_entry *dst,
 424                                                      struct sk_buff *skb)
 425 {
 426         return dst->ops->neigh_lookup(dst, skb, NULL);
 427 }
 428
 429 static inline void dst_link_failure(struct sk_buff *skb)
 430 {
 431         struct dst_entry *dst = skb_dst(skb);
 432         if (dst && dst->ops && dst->ops->link_failure)
 433                 dst->ops->link_failure(skb);
 434 }
 435
 436 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
 437 {
 438         unsigned long expires = jiffies + timeout;
 439
 440         if (expires == 0)
 441                 expires = 1;
 442
 443         if (dst->expires == 0 || time_before(expires, dst->expires))
 444                 dst->expires = expires;
 445 }
 446
 447 /* Output packet to network from transport.  */
 448 static inline int dst_output(struct sk_buff *skb)
 449 {
 450         return skb_dst(skb)->output(skb);
 451 }
 452
 453 /* Input packet from network to transport.  */
 454 static inline int dst_input(struct sk_buff *skb)
 455 {
 456         return skb_dst(skb)->input(skb);
 457 }
 458
 459 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
 460 {
 461         if (dst->obsolete)
 462                 dst = dst->ops->check(dst, cookie);
 463         return dst;
 464 }
 465
 466 extern void             dst_init(void);
 467
 468 /* Flags for xfrm_lookup flags argument. */
 469 enum {
 470         XFRM_LOOKUP_ICMP = 1 << 0,
 471 };
 472
 473 struct flowi;
 474 #ifndef CONFIG_XFRM
 475 static inline struct dst_entry *xfrm_lookup(struct net *net,
 476                                             struct dst_entry *dst_orig,
 477                                             const struct flowi *fl, struct sock *sk,
 478                                             int flags)
 479 {
 480         return dst_orig;
 481 }
 482 #else
 483 extern struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
 484                                      const struct flowi *fl, struct sock *sk,
 485                                      int flags);
 486 #endif
 487
 488 #endif /* _NET_DST_H */