Avoid beyond bounds copy while caching ACL
[zen-stable.git] / include / net / dst.h
blob344c8dd028745857bdbe43efd85f386c0a64eade
1 /*
2 * net/dst.h Protocol independent destination cache definitions.
4 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
6 */
8 #ifndef _NET_DST_H
9 #define _NET_DST_H
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>
19 #define DST_GC_MIN (HZ/10)
20 #define DST_GC_INC (HZ/2)
21 #define DST_GC_MAX (120*HZ)
23 /* Each dst_entry has reference count and sits in some parent list(s).
24 * When it is removed from parent list, it is "freed" (dst_free).
25 * After this it enters dead state (dst->obsolete > 0) and if its refcnt
26 * is zero, it can be destroyed immediately, otherwise it is added
27 * to gc list and garbage collector periodically checks the refcnt.
30 struct sk_buff;
32 struct dst_entry {
33 struct rcu_head rcu_head;
34 struct dst_entry *child;
35 struct net_device *dev;
36 struct dst_ops *ops;
37 unsigned long _metrics;
38 unsigned long expires;
39 struct dst_entry *path;
40 struct neighbour __rcu *_neighbour;
41 #ifdef CONFIG_XFRM
42 struct xfrm_state *xfrm;
43 #else
44 void *__pad1;
45 #endif
46 int (*input)(struct sk_buff*);
47 int (*output)(struct sk_buff*);
49 int flags;
50 #define DST_HOST 0x0001
51 #define DST_NOXFRM 0x0002
52 #define DST_NOPOLICY 0x0004
53 #define DST_NOHASH 0x0008
54 #define DST_NOCACHE 0x0010
55 #define DST_NOCOUNT 0x0020
56 #define DST_NOPEER 0x0040
58 short error;
59 short obsolete;
60 unsigned short header_len; /* more space at head required */
61 unsigned short trailer_len; /* space to reserve at tail */
62 #ifdef CONFIG_IP_ROUTE_CLASSID
63 __u32 tclassid;
64 #else
65 __u32 __pad2;
66 #endif
69 * Align __refcnt to a 64 bytes alignment
70 * (L1_CACHE_SIZE would be too much)
72 #ifdef CONFIG_64BIT
73 long __pad_to_align_refcnt[2];
74 #endif
76 * __refcnt wants to be on a different cache line from
77 * input/output/ops or performance tanks badly
79 atomic_t __refcnt; /* client references */
80 int __use;
81 unsigned long lastuse;
82 union {
83 struct dst_entry *next;
84 struct rtable __rcu *rt_next;
85 struct rt6_info *rt6_next;
86 struct dn_route __rcu *dn_next;
90 static inline struct neighbour *dst_get_neighbour_noref(struct dst_entry *dst)
92 return rcu_dereference(dst->_neighbour);
95 static inline struct neighbour *dst_get_neighbour_noref_raw(struct dst_entry *dst)
97 return rcu_dereference_raw(dst->_neighbour);
100 static inline void dst_set_neighbour(struct dst_entry *dst, struct neighbour *neigh)
102 rcu_assign_pointer(dst->_neighbour, neigh);
105 extern u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old);
106 extern const u32 dst_default_metrics[RTAX_MAX];
108 #define DST_METRICS_READ_ONLY 0x1UL
109 #define __DST_METRICS_PTR(Y) \
110 ((u32 *)((Y) & ~DST_METRICS_READ_ONLY))
111 #define DST_METRICS_PTR(X) __DST_METRICS_PTR((X)->_metrics)
113 static inline bool dst_metrics_read_only(const struct dst_entry *dst)
115 return dst->_metrics & DST_METRICS_READ_ONLY;
118 extern void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old);
120 static inline void dst_destroy_metrics_generic(struct dst_entry *dst)
122 unsigned long val = dst->_metrics;
123 if (!(val & DST_METRICS_READ_ONLY))
124 __dst_destroy_metrics_generic(dst, val);
127 static inline u32 *dst_metrics_write_ptr(struct dst_entry *dst)
129 unsigned long p = dst->_metrics;
131 BUG_ON(!p);
133 if (p & DST_METRICS_READ_ONLY)
134 return dst->ops->cow_metrics(dst, p);
135 return __DST_METRICS_PTR(p);
138 /* This may only be invoked before the entry has reached global
139 * visibility.
141 static inline void dst_init_metrics(struct dst_entry *dst,
142 const u32 *src_metrics,
143 bool read_only)
145 dst->_metrics = ((unsigned long) src_metrics) |
146 (read_only ? DST_METRICS_READ_ONLY : 0);
149 static inline void dst_copy_metrics(struct dst_entry *dest, const struct dst_entry *src)
151 u32 *dst_metrics = dst_metrics_write_ptr(dest);
153 if (dst_metrics) {
154 u32 *src_metrics = DST_METRICS_PTR(src);
156 memcpy(dst_metrics, src_metrics, RTAX_MAX * sizeof(u32));
160 static inline u32 *dst_metrics_ptr(struct dst_entry *dst)
162 return DST_METRICS_PTR(dst);
165 static inline u32
166 dst_metric_raw(const struct dst_entry *dst, const int metric)
168 u32 *p = DST_METRICS_PTR(dst);
170 return p[metric-1];
173 static inline u32
174 dst_metric(const struct dst_entry *dst, const int metric)
176 WARN_ON_ONCE(metric == RTAX_HOPLIMIT ||
177 metric == RTAX_ADVMSS ||
178 metric == RTAX_MTU);
179 return dst_metric_raw(dst, metric);
182 static inline u32
183 dst_metric_advmss(const struct dst_entry *dst)
185 u32 advmss = dst_metric_raw(dst, RTAX_ADVMSS);
187 if (!advmss)
188 advmss = dst->ops->default_advmss(dst);
190 return advmss;
193 static inline void dst_metric_set(struct dst_entry *dst, int metric, u32 val)
195 u32 *p = dst_metrics_write_ptr(dst);
197 if (p)
198 p[metric-1] = val;
201 static inline u32
202 dst_feature(const struct dst_entry *dst, u32 feature)
204 return dst_metric(dst, RTAX_FEATURES) & feature;
207 static inline u32 dst_mtu(const struct dst_entry *dst)
209 return dst->ops->mtu(dst);
212 /* RTT metrics are stored in milliseconds for user ABI, but used as jiffies */
213 static inline unsigned long dst_metric_rtt(const struct dst_entry *dst, int metric)
215 return msecs_to_jiffies(dst_metric(dst, metric));
218 static inline void set_dst_metric_rtt(struct dst_entry *dst, int metric,
219 unsigned long rtt)
221 dst_metric_set(dst, metric, jiffies_to_msecs(rtt));
224 static inline u32
225 dst_allfrag(const struct dst_entry *dst)
227 int ret = dst_feature(dst, RTAX_FEATURE_ALLFRAG);
228 return ret;
231 static inline int
232 dst_metric_locked(const struct dst_entry *dst, int metric)
234 return dst_metric(dst, RTAX_LOCK) & (1<<metric);
237 static inline void dst_hold(struct dst_entry * dst)
240 * If your kernel compilation stops here, please check
241 * __pad_to_align_refcnt declaration in struct dst_entry
243 BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
244 atomic_inc(&dst->__refcnt);
247 static inline void dst_use(struct dst_entry *dst, unsigned long time)
249 dst_hold(dst);
250 dst->__use++;
251 dst->lastuse = time;
254 static inline void dst_use_noref(struct dst_entry *dst, unsigned long time)
256 dst->__use++;
257 dst->lastuse = time;
260 static inline
261 struct dst_entry * dst_clone(struct dst_entry * dst)
263 if (dst)
264 atomic_inc(&dst->__refcnt);
265 return dst;
268 extern void dst_release(struct dst_entry *dst);
270 static inline void refdst_drop(unsigned long refdst)
272 if (!(refdst & SKB_DST_NOREF))
273 dst_release((struct dst_entry *)(refdst & SKB_DST_PTRMASK));
277 * skb_dst_drop - drops skb dst
278 * @skb: buffer
280 * Drops dst reference count if a reference was taken.
282 static inline void skb_dst_drop(struct sk_buff *skb)
284 if (skb->_skb_refdst) {
285 refdst_drop(skb->_skb_refdst);
286 skb->_skb_refdst = 0UL;
290 static inline void skb_dst_copy(struct sk_buff *nskb, const struct sk_buff *oskb)
292 nskb->_skb_refdst = oskb->_skb_refdst;
293 if (!(nskb->_skb_refdst & SKB_DST_NOREF))
294 dst_clone(skb_dst(nskb));
298 * skb_dst_force - makes sure skb dst is refcounted
299 * @skb: buffer
301 * If dst is not yet refcounted, let's do it
303 static inline void skb_dst_force(struct sk_buff *skb)
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));
314 * __skb_tunnel_rx - prepare skb for rx reinsert
315 * @skb: buffer
316 * @dev: tunnel device
318 * After decapsulation, packet is going to re-enter (netif_rx()) our stack,
319 * so make some cleanups. (no accounting done)
321 static inline void __skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
323 skb->dev = dev;
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.
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);
338 * skb_tunnel_rx - prepare skb for rx reinsert
339 * @skb: buffer
340 * @dev: tunnel device
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.
346 static inline void skb_tunnel_rx(struct sk_buff *skb, struct net_device *dev)
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);
354 /* Children define the path of the packet through the
355 * Linux networking. Thus, destinations are stackable.
358 static inline struct dst_entry *skb_dst_pop(struct sk_buff *skb)
360 struct dst_entry *child = dst_clone(skb_dst(skb)->child);
362 skb_dst_drop(skb);
363 return child;
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, int flags);
369 extern void __dst_free(struct dst_entry * dst);
370 extern struct dst_entry *dst_destroy(struct dst_entry * dst);
372 static inline void dst_free(struct dst_entry * dst)
374 if (dst->obsolete > 1)
375 return;
376 if (!atomic_read(&dst->__refcnt)) {
377 dst = dst_destroy(dst);
378 if (!dst)
379 return;
381 __dst_free(dst);
384 static inline void dst_rcu_free(struct rcu_head *head)
386 struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
387 dst_free(dst);
390 static inline void dst_confirm(struct dst_entry *dst)
392 if (dst) {
393 struct neighbour *n;
395 rcu_read_lock();
396 n = dst_get_neighbour_noref(dst);
397 neigh_confirm(n);
398 rcu_read_unlock();
402 static inline struct neighbour *dst_neigh_lookup(const struct dst_entry *dst, const void *daddr)
404 return dst->ops->neigh_lookup(dst, daddr);
407 static inline void dst_link_failure(struct sk_buff *skb)
409 struct dst_entry *dst = skb_dst(skb);
410 if (dst && dst->ops && dst->ops->link_failure)
411 dst->ops->link_failure(skb);
414 static inline void dst_set_expires(struct dst_entry *dst, int timeout)
416 unsigned long expires = jiffies + timeout;
418 if (expires == 0)
419 expires = 1;
421 if (dst->expires == 0 || time_before(expires, dst->expires))
422 dst->expires = expires;
425 /* Output packet to network from transport. */
426 static inline int dst_output(struct sk_buff *skb)
428 return skb_dst(skb)->output(skb);
431 /* Input packet from network to transport. */
432 static inline int dst_input(struct sk_buff *skb)
434 return skb_dst(skb)->input(skb);
437 static inline struct dst_entry *dst_check(struct dst_entry *dst, u32 cookie)
439 if (dst->obsolete)
440 dst = dst->ops->check(dst, cookie);
441 return dst;
444 extern void dst_init(void);
446 /* Flags for xfrm_lookup flags argument. */
447 enum {
448 XFRM_LOOKUP_ICMP = 1 << 0,
451 struct flowi;
452 #ifndef CONFIG_XFRM
453 static inline struct dst_entry *xfrm_lookup(struct net *net,
454 struct dst_entry *dst_orig,
455 const struct flowi *fl, struct sock *sk,
456 int flags)
458 return dst_orig;
460 #else
461 extern struct dst_entry *xfrm_lookup(struct net *net, struct dst_entry *dst_orig,
462 const struct flowi *fl, struct sock *sk,
463 int flags);
464 #endif
466 #endif /* _NET_DST_H */