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/jiffies.h>
  16 #include <net/neighbour.h>
  17 #include <asm/processor.h>
  18
  19 /*
  20  * 0 - no debugging messages
  21  * 1 - rare events and bugs (default)
  22  * 2 - trace mode.
  23  */
  24 #define RT_CACHE_DEBUG          0
  25
  26 #define DST_GC_MIN      (HZ/10)
  27 #define DST_GC_INC      (HZ/2)
  28 #define DST_GC_MAX      (120*HZ)
  29
  30 /* Each dst_entry has reference count and sits in some parent list(s).
  31  * When it is removed from parent list, it is "freed" (dst_free).
  32  * After this it enters dead state (dst->obsolete > 0) and if its refcnt
  33  * is zero, it can be destroyed immediately, otherwise it is added
  34  * to gc list and garbage collector periodically checks the refcnt.
  35  */
  36
  37 struct sk_buff;
  38
  39 struct dst_entry {
  40         struct rcu_head         rcu_head;
  41         struct dst_entry        *child;
  42         struct net_device       *dev;
  43         short                   error;
  44         short                   obsolete;
  45         int                     flags;
  46 #define DST_HOST                0x0001
  47 #define DST_NOXFRM              0x0002
  48 #define DST_NOPOLICY            0x0004
  49 #define DST_NOHASH              0x0008
  50 #define DST_NOCACHE             0x0010
  51         unsigned long           expires;
  52
  53         unsigned short          header_len;     /* more space at head required */
  54         unsigned short          trailer_len;    /* space to reserve at tail */
  55
  56         unsigned int            rate_tokens;
  57         unsigned long           rate_last;      /* rate limiting for ICMP */
  58
  59         struct dst_entry        *path;
  60
  61         struct neighbour        *neighbour;
  62         struct hh_cache         *hh;
  63 #ifdef CONFIG_XFRM
  64         struct xfrm_state       *xfrm;
  65 #else
  66         void                    *__pad1;
  67 #endif
  68         int                     (*input)(struct sk_buff*);
  69         int                     (*output)(struct sk_buff*);
  70
  71         struct  dst_ops         *ops;
  72
  73         u32                     _metrics[RTAX_MAX];
  74
  75 #ifdef CONFIG_NET_CLS_ROUTE
  76         __u32                   tclassid;
  77 #else
  78         __u32                   __pad2;
  79 #endif
  80
  81
  82         /*
  83          * Align __refcnt to a 64 bytes alignment
  84          * (L1_CACHE_SIZE would be too much)
  85          */
  86 #ifdef CONFIG_64BIT
  87         long                    __pad_to_align_refcnt[1];
  88 #endif
  89         /*
  90          * __refcnt wants to be on a different cache line from
  91          * input/output/ops or performance tanks badly
  92          */
  93         atomic_t                __refcnt;       /* client references    */
  94         int                     __use;
  95         unsigned long           lastuse;
  96         union {
  97                 struct dst_entry        *next;
  98                 struct rtable __rcu     *rt_next;
  99                 struct rt6_info         *rt6_next;
 100                 struct dn_route __rcu   *dn_next;
 101         };
 102 };
 103
 104 #ifdef __KERNEL__
 105
 106 static inline u32
 107 dst_metric_raw(const struct dst_entry *dst, const int metric)
 108 {
 109         return dst->_metrics[metric-1];
 110 }
 111
 112 static inline u32
 113 dst_metric(const struct dst_entry *dst, const int metric)
 114 {
 115         WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
 116                      metric == RTAX_ADVMSS ||
 117                      metric == RTAX_MTU);
 118         return dst_metric_raw(dst, metric);
 119 }
 120
 121 static inline u32
 122 dst_metric_advmss(const struct dst_entry *dst)
 123 {
 124         u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
 125
 126         if (!advmss)
 127                 advmss = dst->ops->default_advmss(dst);
 128
 129         return advmss;
 130 }
 131
 132 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
 133 {
 134         dst->_metrics[metric-1] = val;
 135 }
 136
 137 static inline void dst_import_metrics(struct dst_entry *dst, const u32 *src_metrics)
 138 {
 139         memcpy(dst->_metrics, src_metrics, RTAX_MAX * sizeof(u32));
 140 }
 141
 142 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
 143 {
 144         dst_import_metrics(dest, src->_metrics);
 145 }
 146
 147 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
 148 {
 149         return dst->_metrics;
 150 }
 151
 152 static inline u32
 153 dst_feature(const struct dst_entry *dst, u32 feature)
 154 {
 155         return dst_metric(dst, RTAX_FEATURES) & feature;
 156 }
 157
 158 static inline u32 dst_mtu(const struct dst_entry *dst)
 159 {
 160         u32 mtu = dst_metric_raw(dst, RTAX_MTU);
 161
 162         if (!mtu)
 163                 mtu = dst->ops->default_mtu(dst);
 164
 165         return mtu;
 166 }
 167
 168 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
 169 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
 170 {
 171         return msecs_to_jiffies(dst_metric(dst, metric));
 172 }
 173
 174 static inline void set_dst_metric_rtt(struct dst_entry *dst, int metric,
 175                                       unsigned long rtt)
 176 {
 177         dst_metric_set(dst, metric, jiffies_to_msecs(rtt));
 178 }
 179
 180 static inline u32
 181 dst_allfrag(const struct dst_entry *dst)
 182 {
 183         int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
 184         /* Yes, _exactly_. This is paranoia. */
 185         barrier();
 186         return ret;
 187 }
 188
 189 static inline int
 190 dst_metric_locked(const struct dst_entry *dst, int metric)
 191 {
 192         return dst_metric(dst, RTAX_LOCK) & (1<<metric);
 193 }
 194
 195 static inline void dst_hold(struct dst_entry * dst)
 196 {
 197         /*
 198          * If your kernel compilation stops here, please check
 199          * __pad_to_align_refcnt declaration in struct dst_entry
 200          */
 201         BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
 202         atomic_inc(&dst->__refcnt);
 203 }
 204
 205 static inline void dst_use(struct dst_entry *dst, unsigned long time)
 206 {
 207         dst_hold(dst);
 208         dst->__use++;
 209         dst->lastuse = time;
 210 }
 211
 212 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
 213 {
 214         dst->__use++;
 215         dst->lastuse = time;
 216 }
 217
 218 static inline
 219 struct dst_entry * dst_clone(struct dst_entry * dst)
 220 {
 221         if (dst)
 222                 atomic_inc(&dst->__refcnt);
 223         return dst;
 224 }
 225
 226 extern void dst_release(struct dst_entry *dst);
 227
 228 static inline void refdst_drop(unsigned long refdst)
 229 {
 230         if (!(refdst & SKB_DST_NOREF))
 231                 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
 232 }
 233
 234 /**
 235  * skb_dst_drop - drops skb dst
 236  * @skb: buffer
 237  *
 238  * Drops dst reference count if a reference was taken.
 239  */
 240 static inline void skb_dst_drop(struct sk_buff *skb)
 241 {
 242         if (skb->_skb_refdst) {
 243                 refdst_drop(skb->_skb_refdst);
 244                 skb->_skb_refdst = 0UL;
 245         }
 246 }
 247
 248 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
 249 {
 250         nskb->_skb_refdst = oskb->_skb_refdst;
 251         if (!(nskb->_skb_refdst & SKB_DST_NOREF))
 252                 dst_clone(skb_dst(nskb));
 253 }
 254
 255 /**
 256  * skb_dst_force - makes sure skb dst is refcounted
 257  * @skb: buffer
 258  *
 259  * If dst is not yet refcounted, let's do it
 260  */
 261 static inline void skb_dst_force(struct sk_buff *skb)
 262 {
 263         if (skb_dst_is_noref(skb)) {
 264                 WARN_ON(!rcu_read_lock_held());
 265                 skb->_skb_refdst &= ~SKB_DST_NOREF;
 266                 dst_clone(skb_dst(skb));
 267         }
 268 }
 269
 270
 271 /**
 272  *      __skb_tunnel_rx - prepare skb for rx reinsert
 273  *      @skb: buffer
 274  *      @dev: tunnel device
 275  *
 276  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 277  *      so make some cleanups. (no accounting done)
 278  */
 279 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
 280 {
 281         skb->dev = dev;
 282         skb->rxhash = 0;
 283         skb_set_queue_mapping(skb, 0);
 284         skb_dst_drop(skb);
 285         nf_reset(skb);
 286 }
 287
 288 /**
 289  *      skb_tunnel_rx - prepare skb for rx reinsert
 290  *      @skb: buffer
 291  *      @dev: tunnel device
 292  *
 293  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 294  *      so make some cleanups, and perform accounting.
 295  *      Note: this accounting is not SMP safe.
 296  */
 297 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
 298 {
 299         /* TODO : stats should be SMP safe */
 300         dev->stats.rx_packets++;
 301         dev->stats.rx_bytes += skb->len;
 302         __skb_tunnel_rx(skb, dev);
 303 }
 304
 305 /* Children define the path of the packet through the
 306  * Linux networking.  Thus, destinations are stackable.
 307  */
 308
 309 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
 310 {
 311         struct dst_entry *child = skb_dst(skb)->child;
 312
 313         skb_dst_drop(skb);
 314         return child;
 315 }
 316
 317 extern int dst_discard(struct sk_buff *skb);
 318 extern void * dst_alloc(struct dst_ops * ops);
 319 extern void __dst_free(struct dst_entry * dst);
 320 extern struct dst_entry *dst_destroy(struct dst_entry * dst);
 321
 322 static inline void dst_free(struct dst_entry * dst)
 323 {
 324         if (dst->obsolete > 1)
 325                 return;
 326         if (!atomic_read(&dst->__refcnt)) {
 327                 dst = dst_destroy(dst);
 328                 if (!dst)
 329                         return;
 330         }
 331         __dst_free(dst);
 332 }
 333
 334 static inline void dst_rcu_free(struct rcu_head *head)
 335 {
 336         struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
 337         dst_free(dst);
 338 }
 339
 340 static inline void dst_confirm(struct dst_entry *dst)
 341 {
 342         if (dst)
 343                 neigh_confirm(dst->neighbour);
 344 }
 345
 346 static inline void dst_link_failure(struct sk_buff *skb)
 347 {
 348         struct dst_entry *dst = skb_dst(skb);
 349         if (dst && dst->ops && dst->ops->link_failure)
 350                 dst->ops->link_failure(skb);
 351 }
 352
 353 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
 354 {
 355         unsigned long expires = jiffies + timeout;
 356
 357         if (expires == 0)
 358                 expires = 1;
 359
 360         if (dst->expires == 0 || time_before(expires, dst->expires))
 361                 dst->expires = expires;
 362 }
 363
 364 /* Output packet to network from transport.  */
 365 static inline int dst_output(struct sk_buff *skb)
 366 {
 367         return skb_dst(skb)->output(skb);
 368 }
 369
 370 /* Input packet from network to transport.  */
 371 static inline int dst_input(struct sk_buff *skb)
 372 {
 373         return skb_dst(skb)->input(skb);
 374 }
 375
 376 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
 377 {
 378         if (dst->obsolete)
 379                 dst = dst->ops->check(dst, cookie);
 380         return dst;
 381 }
 382
 383 extern void             dst_init(void);
 384
 385 /* Flags for xfrm_lookup flags argument. */
 386 enum {
 387         XFRM_LOOKUP_WAIT = 1 << 0,
 388         XFRM_LOOKUP_ICMP = 1 << 1,
 389 };
 390
 391 struct flowi;
 392 #ifndef CONFIG_XFRM
 393 static inline int xfrm_lookup(struct net *net, struct dst_entry **dst_p,
 394                               struct flowi *fl, struct sock *sk, int flags)
 395 {
 396         return 0;
 397 }
 398 static inline int __xfrm_lookup(struct net *net, struct dst_entry **dst_p,
 399                                 struct flowi *fl, struct sock *sk, int flags)
 400 {
 401         return 0;
 402 }
 403 #else
 404 extern int xfrm_lookup(struct net *net, struct dst_entry **dst_p,
 405                        struct flowi *fl, struct sock *sk, int flags);
 406 extern int __xfrm_lookup(struct net *net, struct dst_entry **dst_p,
 407                          struct flowi *fl, struct sock *sk, int flags);
 408 #endif
 409 #endif
 410
 411 #endif /* _NET_DST_H */