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 <linux/netdevice.h> 12 #include <linux/rtnetlink.h> 13 #include <linux/rcupdate.h> 14 #include <linux/jiffies.h> 15 #include <net/neighbour.h> 16 #include <asm/processor.h> 17 18 /* 19 * 0 - no debugging messages 20 * 1 - rare events and bugs (default) 21 * 2 - trace mode. 22 */ 23 #define RT_CACHE_DEBUG 0 24 25 #define DST_GC_MIN (HZ/10) 26 #define DST_GC_INC (HZ/2) 27 #define DST_GC_MAX (120*HZ) 28 29 /* Each dst_entry has reference count and sits in some parent list(s). 30 * When it is removed from parent list, it is "freed" (dst_free). 31 * After this it enters dead state (dst->obsolete > 0) and if its refcnt 32 * is zero, it can be destroyed immediately, otherwise it is added 33 * to gc list and garbage collector periodically checks the refcnt. 34 */ 35 36 struct sk_buff; 37 38 struct dst_entry 39 { 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 1 47 #define DST_NOXFRM 2 48 #define DST_NOPOLICY 4 49 #define DST_NOHASH 8 50 unsigned long expires; 51 52 unsigned short header_len; /* more space at head required */ 53 unsigned short trailer_len; /* space to reserve at tail */ 54 55 unsigned int rate_tokens; 56 unsigned long rate_last; /* rate limiting for ICMP */ 57 58 struct dst_entry *path; 59 60 struct neighbour *neighbour; 61 struct hh_cache *hh; 62 struct xfrm_state *xfrm; 63 64 int (*input)(struct sk_buff*); 65 int (*output)(struct sk_buff*); 66 67 struct dst_ops *ops; 68 69 u32 metrics[RTAX_MAX]; 70 71 #ifdef CONFIG_NET_CLS_ROUTE 72 __u32 tclassid; 73 #endif 74 75 /* 76 * __refcnt wants to be on a different cache line from 77 * input/output/ops or performance tanks badly 78 */ 79 atomic_t __refcnt; /* client references */ 80 int __use; 81 unsigned long lastuse; 82 union { 83 struct dst_entry *next; 84 struct rtable *rt_next; 85 struct rt6_info *rt6_next; 86 struct dn_route *dn_next; 87 }; 88 }; 89 90 91 struct dst_ops 92 { 93 unsigned short family; 94 __be16 protocol; 95 unsigned gc_thresh; 96 97 int (*gc)(struct dst_ops *ops); 98 struct dst_entry * (*check)(struct dst_entry *, __u32 cookie); 99 void (*destroy)(struct dst_entry *); 100 void (*ifdown)(struct dst_entry *, 101 struct net_device *dev, int how); 102 struct dst_entry * (*negative_advice)(struct dst_entry *); 103 void (*link_failure)(struct sk_buff *); 104 void (*update_pmtu)(struct dst_entry *dst, u32 mtu); 105 int (*local_out)(struct sk_buff *skb); 106 int entry_size; 107 108 atomic_t entries; 109 struct kmem_cache *kmem_cachep; 110 struct net *dst_net; 111 }; 112 113 #ifdef __KERNEL__ 114 115 static inline u32 116 dst_metric(const struct dst_entry *dst, int metric) 117 { 118 return dst->metrics[metric-1]; 119 } 120 121 static inline u32 dst_mtu(const struct dst_entry *dst) 122 { 123 u32 mtu = dst_metric(dst, RTAX_MTU); 124 /* 125 * Alexey put it here, so ask him about it :) 126 */ 127 barrier(); 128 return mtu; 129 } 130 131 static inline u32 132 dst_allfrag(const struct dst_entry *dst) 133 { 134 int ret = dst_metric(dst, RTAX_FEATURES) & RTAX_FEATURE_ALLFRAG; 135 /* Yes, _exactly_. This is paranoia. */ 136 barrier(); 137 return ret; 138 } 139 140 static inline int 141 dst_metric_locked(struct dst_entry *dst, int metric) 142 { 143 return dst_metric(dst, RTAX_LOCK) & (1<<metric); 144 } 145 146 static inline void dst_hold(struct dst_entry * dst) 147 { 148 atomic_inc(&dst->__refcnt); 149 } 150 151 static inline void dst_use(struct dst_entry *dst, unsigned long time) 152 { 153 dst_hold(dst); 154 dst->__use++; 155 dst->lastuse = time; 156 } 157 158 static inline 159 struct dst_entry * dst_clone(struct dst_entry * dst) 160 { 161 if (dst) 162 atomic_inc(&dst->__refcnt); 163 return dst; 164 } 165 166 static inline 167 void dst_release(struct dst_entry * dst) 168 { 169 if (dst) { 170 WARN_ON(atomic_read(&dst->__refcnt) < 1); 171 smp_mb__before_atomic_dec(); 172 atomic_dec(&dst->__refcnt); 173 } 174 } 175 176 /* Children define the path of the packet through the 177 * Linux networking. Thus, destinations are stackable. 178 */ 179 180 static inline struct dst_entry *dst_pop(struct dst_entry *dst) 181 { 182 struct dst_entry *child = dst_clone(dst->child); 183 184 dst_release(dst); 185 return child; 186 } 187 188 extern int dst_discard(struct sk_buff *skb); 189 extern void * dst_alloc(struct dst_ops * ops); 190 extern void __dst_free(struct dst_entry * dst); 191 extern struct dst_entry *dst_destroy(struct dst_entry * dst); 192 193 static inline void dst_free(struct dst_entry * dst) 194 { 195 if (dst->obsolete > 1) 196 return; 197 if (!atomic_read(&dst->__refcnt)) { 198 dst = dst_destroy(dst); 199 if (!dst) 200 return; 201 } 202 __dst_free(dst); 203 } 204 205 static inline void dst_rcu_free(struct rcu_head *head) 206 { 207 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head); 208 dst_free(dst); 209 } 210 211 static inline void dst_confirm(struct dst_entry *dst) 212 { 213 if (dst) 214 neigh_confirm(dst->neighbour); 215 } 216 217 static inline void dst_negative_advice(struct dst_entry **dst_p) 218 { 219 struct dst_entry * dst = *dst_p; 220 if (dst && dst->ops->negative_advice) 221 *dst_p = dst->ops->negative_advice(dst); 222 } 223 224 static inline void dst_link_failure(struct sk_buff *skb) 225 { 226 struct dst_entry * dst = skb->dst; 227 if (dst && dst->ops && dst->ops->link_failure) 228 dst->ops->link_failure(skb); 229 } 230 231 static inline void dst_set_expires(struct dst_entry *dst, int timeout) 232 { 233 unsigned long expires = jiffies + timeout; 234 235 if (expires == 0) 236 expires = 1; 237 238 if (dst->expires == 0 || time_before(expires, dst->expires)) 239 dst->expires = expires; 240 } 241 242 /* Output packet to network from transport. */ 243 static inline int dst_output(struct sk_buff *skb) 244 { 245 return skb->dst->output(skb); 246 } 247 248 /* Input packet from network to transport. */ 249 static inline int dst_input(struct sk_buff *skb) 250 { 251 int err; 252 253 for (;;) { 254 err = skb->dst->input(skb); 255 256 if (likely(err == 0)) 257 return err; 258 /* Oh, Jamal... Seems, I will not forgive you this mess. :-) */ 259 if (unlikely(err != NET_XMIT_BYPASS)) 260 return err; 261 } 262 } 263 264 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie) 265 { 266 if (dst->obsolete) 267 dst = dst->ops->check(dst, cookie); 268 return dst; 269 } 270 271 extern void dst_init(void); 272 273 /* Flags for xfrm_lookup flags argument. */ 274 enum { 275 XFRM_LOOKUP_WAIT = 1 << 0, 276 XFRM_LOOKUP_ICMP = 1 << 1, 277 }; 278 279 struct flowi; 280 #ifndef CONFIG_XFRM 281 static inline int xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl, 282 struct sock *sk, int flags) 283 { 284 return 0; 285 } 286 static inline int __xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl, 287 struct sock *sk, int flags) 288 { 289 return 0; 290 } 291 #else 292 extern int xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl, 293 struct sock *sk, int flags); 294 extern int __xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl, 295 struct sock *sk, int flags); 296 #endif 297 #endif 298 299 #endif /* _NET_DST_H */ 300