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         return dst_metric_raw(dst, metric);
 118 }
 119
 120 static inline u32
 121 dst_metric_advmss(const struct dst_entry *dst)
 122 {
 123         u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
 124
 125         if (!advmss)
 126                 advmss = dst->ops->default_advmss(dst);
 127
 128         return advmss;
 129 }
 130
 131 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
 132 {
 133         dst->_metrics[metric-1] = val;
 134 }
 135
 136 static inline void dst_import_metrics(struct dst_entry *dst, const u32 *src_metrics)
 137 {
 138         memcpy(dst->_metrics, src_metrics, RTAX_MAX * sizeof(u32));
 139 }
 140
 141 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
 142 {
 143         dst_import_metrics(dest, src->_metrics);
 144 }
 145
 146 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
 147 {
 148         return dst->_metrics;
 149 }
 150
 151 static inline u32
 152 dst_feature(const struct dst_entry *dst, u32 feature)
 153 {
 154         return dst_metric(dst, RTAX_FEATURES) & feature;
 155 }
 156
 157 static inline u32 dst_mtu(const struct dst_entry *dst)
 158 {
 159         u32 mtu = dst_metric(dst, RTAX_MTU);
 160         /*
 161          * Alexey put it here, so ask him about it :)
 162          */
 163         barrier();
 164         return mtu;
 165 }
 166
 167 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
 168 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
 169 {
 170         return msecs_to_jiffies(dst_metric(dst, metric));
 171 }
 172
 173 static inline void set_dst_metric_rtt(struct dst_entry *dst, int metric,
 174                                       unsigned long rtt)
 175 {
 176         dst_metric_set(dst, metric, jiffies_to_msecs(rtt));
 177 }
 178
 179 static inline u32
 180 dst_allfrag(const struct dst_entry *dst)
 181 {
 182         int ret = dst_feature(dst,  RTAX_FEATURE_ALLFRAG);
 183         /* Yes, _exactly_. This is paranoia. */
 184         barrier();
 185         return ret;
 186 }
 187
 188 static inline int
 189 dst_metric_locked(struct dst_entry *dst, int metric)
 190 {
 191         return dst_metric(dst, RTAX_LOCK) & (1<<metric);
 192 }
 193
 194 static inline void dst_hold(struct dst_entry * dst)
 195 {
 196         /*
 197          * If your kernel compilation stops here, please check
 198          * __pad_to_align_refcnt declaration in struct dst_entry
 199          */
 200         BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
 201         atomic_inc(&dst->__refcnt);
 202 }
 203
 204 static inline void dst_use(struct dst_entry *dst, unsigned long time)
 205 {
 206         dst_hold(dst);
 207         dst->__use++;
 208         dst->lastuse = time;
 209 }
 210
 211 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
 212 {
 213         dst->__use++;
 214         dst->lastuse = time;
 215 }
 216
 217 static inline
 218 struct dst_entry * dst_clone(struct dst_entry * dst)
 219 {
 220         if (dst)
 221                 atomic_inc(&dst->__refcnt);
 222         return dst;
 223 }
 224
 225 extern void dst_release(struct dst_entry *dst);
 226
 227 static inline void refdst_drop(unsigned long refdst)
 228 {
 229         if (!(refdst & SKB_DST_NOREF))
 230                 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
 231 }
 232
 233 /**
 234  * skb_dst_drop - drops skb dst
 235  * @skb: buffer
 236  *
 237  * Drops dst reference count if a reference was taken.
 238  */
 239 static inline void skb_dst_drop(struct sk_buff *skb)
 240 {
 241         if (skb->_skb_refdst) {
 242                 refdst_drop(skb->_skb_refdst);
 243                 skb->_skb_refdst = 0UL;
 244         }
 245 }
 246
 247 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
 248 {
 249         nskb->_skb_refdst = oskb->_skb_refdst;
 250         if (!(nskb->_skb_refdst & SKB_DST_NOREF))
 251                 dst_clone(skb_dst(nskb));
 252 }
 253
 254 /**
 255  * skb_dst_force - makes sure skb dst is refcounted
 256  * @skb: buffer
 257  *
 258  * If dst is not yet refcounted, let's do it
 259  */
 260 static inline void skb_dst_force(struct sk_buff *skb)
 261 {
 262         if (skb_dst_is_noref(skb)) {
 263                 WARN_ON(!rcu_read_lock_held());
 264                 skb->_skb_refdst &= ~SKB_DST_NOREF;
 265                 dst_clone(skb_dst(skb));
 266         }
 267 }
 268
 269
 270 /**
 271  *      __skb_tunnel_rx - prepare skb for rx reinsert
 272  *      @skb: buffer
 273  *      @dev: tunnel device
 274  *
 275  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 276  *      so make some cleanups. (no accounting done)
 277  */
 278 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
 279 {
 280         skb->dev = dev;
 281         skb->rxhash = 0;
 282         skb_set_queue_mapping(skb, 0);
 283         skb_dst_drop(skb);
 284         nf_reset(skb);
 285 }
 286
 287 /**
 288  *      skb_tunnel_rx - prepare skb for rx reinsert
 289  *      @skb: buffer
 290  *      @dev: tunnel device
 291  *
 292  *      After decapsulation, packet is going to re-enter (netif_rx()) our stack,
 293  *      so make some cleanups, and perform accounting.
 294  *      Note: this accounting is not SMP safe.
 295  */
 296 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
 297 {
 298         /* TODO : stats should be SMP safe */
 299         dev->stats.rx_packets++;
 300         dev->stats.rx_bytes += skb->len;
 301         __skb_tunnel_rx(skb, dev);
 302 }
 303
 304 /* Children define the path of the packet through the
 305  * Linux networking.  Thus, destinations are stackable.
 306  */
 307
 308 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
 309 {
 310         struct dst_entry *child = skb_dst(skb)->child;
 311
 312         skb_dst_drop(skb);
 313         return child;
 314 }
 315
 316 extern int dst_discard(struct sk_buff *skb);
 317 extern void * dst_alloc(struct dst_ops * ops);
 318 extern void __dst_free(struct dst_entry * dst);
 319 extern struct dst_entry *dst_destroy(struct dst_entry * dst);
 320
 321 static inline void dst_free(struct dst_entry * dst)
 322 {
 323         if (dst->obsolete > 1)
 324                 return;
 325         if (!atomic_read(&dst->__refcnt)) {
 326                 dst = dst_destroy(dst);
 327                 if (!dst)
 328                         return;
 329         }
 330         __dst_free(dst);
 331 }
 332
 333 static inline void dst_rcu_free(struct rcu_head *head)
 334 {
 335         struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
 336         dst_free(dst);
 337 }
 338
 339 static inline void dst_confirm(struct dst_entry *dst)
 340 {
 341         if (dst)
 342                 neigh_confirm(dst->neighbour);
 343 }
 344
 345 static inline void dst_link_failure(struct sk_buff *skb)
 346 {
 347         struct dst_entry *dst = skb_dst(skb);
 348         if (dst && dst->ops && dst->ops->link_failure)
 349                 dst->ops->link_failure(skb);
 350 }
 351
 352 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
 353 {
 354         unsigned long expires = jiffies + timeout;
 355
 356         if (expires == 0)
 357                 expires = 1;
 358
 359         if (dst->expires == 0 || time_before(expires, dst->expires))
 360                 dst->expires = expires;
 361 }
 362
 363 /* Output packet to network from transport.  */
 364 static inline int dst_output(struct sk_buff *skb)
 365 {
 366         return skb_dst(skb)->output(skb);
 367 }
 368
 369 /* Input packet from network to transport.  */
 370 static inline int dst_input(struct sk_buff *skb)
 371 {
 372         return skb_dst(skb)->input(skb);
 373 }
 374
 375 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
 376 {
 377         if (dst->obsolete)
 378                 dst = dst->ops->check(dst, cookie);
 379         return dst;
 380 }
 381
 382 extern void             dst_init(void);
 383
 384 /* Flags for xfrm_lookup flags argument. */
 385 enum {
 386         XFRM_LOOKUP_WAIT = 1 << 0,
 387         XFRM_LOOKUP_ICMP = 1 << 1,
 388 };
 389
 390 struct flowi;
 391 #ifndef CONFIG_XFRM
 392 static inline int xfrm_lookup(struct net *net, struct dst_entry **dst_p,
 393                               struct flowi *fl, struct sock *sk, int flags)
 394 {
 395         return 0;
 396 }
 397 static inline int __xfrm_lookup(struct net *net, struct dst_entry **dst_p,
 398                                 struct flowi *fl, struct sock *sk, int flags)
 399 {
 400         return 0;
 401 }
 402 #else
 403 extern int xfrm_lookup(struct net *net, struct dst_entry **dst_p,
 404                        struct flowi *fl, struct sock *sk, int flags);
 405 extern int __xfrm_lookup(struct net *net, struct dst_entry **dst_p,
 406                          struct flowi *fl, struct sock *sk, int flags);
 407 #endif
 408 #endif
 409
 410 #endif /* _NET_DST_H */