]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * ROUTE - implementation of the IP router. | |
7 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * Alan Cox, <gw4pts@gw4pts.ampr.org> | |
11 | * Linus Torvalds, <Linus.Torvalds@helsinki.fi> | |
12 | * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> | |
13 | * | |
14 | * Fixes: | |
15 | * Alan Cox : Verify area fixes. | |
16 | * Alan Cox : cli() protects routing changes | |
17 | * Rui Oliveira : ICMP routing table updates | |
18 | * (rco@di.uminho.pt) Routing table insertion and update | |
19 | * Linus Torvalds : Rewrote bits to be sensible | |
20 | * Alan Cox : Added BSD route gw semantics | |
e905a9ed | 21 | * Alan Cox : Super /proc >4K |
1da177e4 LT |
22 | * Alan Cox : MTU in route table |
23 | * Alan Cox : MSS actually. Also added the window | |
24 | * clamper. | |
25 | * Sam Lantinga : Fixed route matching in rt_del() | |
26 | * Alan Cox : Routing cache support. | |
27 | * Alan Cox : Removed compatibility cruft. | |
28 | * Alan Cox : RTF_REJECT support. | |
29 | * Alan Cox : TCP irtt support. | |
30 | * Jonathan Naylor : Added Metric support. | |
31 | * Miquel van Smoorenburg : BSD API fixes. | |
32 | * Miquel van Smoorenburg : Metrics. | |
33 | * Alan Cox : Use __u32 properly | |
34 | * Alan Cox : Aligned routing errors more closely with BSD | |
35 | * our system is still very different. | |
36 | * Alan Cox : Faster /proc handling | |
37 | * Alexey Kuznetsov : Massive rework to support tree based routing, | |
38 | * routing caches and better behaviour. | |
e905a9ed | 39 | * |
1da177e4 LT |
40 | * Olaf Erb : irtt wasn't being copied right. |
41 | * Bjorn Ekwall : Kerneld route support. | |
42 | * Alan Cox : Multicast fixed (I hope) | |
43 | * Pavel Krauz : Limited broadcast fixed | |
44 | * Mike McLagan : Routing by source | |
45 | * Alexey Kuznetsov : End of old history. Split to fib.c and | |
46 | * route.c and rewritten from scratch. | |
47 | * Andi Kleen : Load-limit warning messages. | |
48 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. | |
49 | * Vitaly E. Lavrov : Race condition in ip_route_input_slow. | |
50 | * Tobias Ringstrom : Uninitialized res.type in ip_route_output_slow. | |
51 | * Vladimir V. Ivanov : IP rule info (flowid) is really useful. | |
52 | * Marc Boucher : routing by fwmark | |
53 | * Robert Olsson : Added rt_cache statistics | |
54 | * Arnaldo C. Melo : Convert proc stuff to seq_file | |
bb1d23b0 | 55 | * Eric Dumazet : hashed spinlocks and rt_check_expire() fixes. |
cef2685e IS |
56 | * Ilia Sotnikov : Ignore TOS on PMTUD and Redirect |
57 | * Ilia Sotnikov : Removed TOS from hash calculations | |
1da177e4 LT |
58 | * |
59 | * This program is free software; you can redistribute it and/or | |
60 | * modify it under the terms of the GNU General Public License | |
61 | * as published by the Free Software Foundation; either version | |
62 | * 2 of the License, or (at your option) any later version. | |
63 | */ | |
64 | ||
1da177e4 LT |
65 | #include <linux/module.h> |
66 | #include <asm/uaccess.h> | |
67 | #include <asm/system.h> | |
68 | #include <linux/bitops.h> | |
69 | #include <linux/types.h> | |
70 | #include <linux/kernel.h> | |
1da177e4 | 71 | #include <linux/mm.h> |
424c4b70 | 72 | #include <linux/bootmem.h> |
1da177e4 LT |
73 | #include <linux/string.h> |
74 | #include <linux/socket.h> | |
75 | #include <linux/sockios.h> | |
76 | #include <linux/errno.h> | |
77 | #include <linux/in.h> | |
78 | #include <linux/inet.h> | |
79 | #include <linux/netdevice.h> | |
80 | #include <linux/proc_fs.h> | |
81 | #include <linux/init.h> | |
39c90ece | 82 | #include <linux/workqueue.h> |
1da177e4 | 83 | #include <linux/skbuff.h> |
1da177e4 LT |
84 | #include <linux/inetdevice.h> |
85 | #include <linux/igmp.h> | |
86 | #include <linux/pkt_sched.h> | |
87 | #include <linux/mroute.h> | |
88 | #include <linux/netfilter_ipv4.h> | |
89 | #include <linux/random.h> | |
90 | #include <linux/jhash.h> | |
91 | #include <linux/rcupdate.h> | |
92 | #include <linux/times.h> | |
5a0e3ad6 | 93 | #include <linux/slab.h> |
352e512c | 94 | #include <net/dst.h> |
457c4cbc | 95 | #include <net/net_namespace.h> |
1da177e4 LT |
96 | #include <net/protocol.h> |
97 | #include <net/ip.h> | |
98 | #include <net/route.h> | |
99 | #include <net/inetpeer.h> | |
100 | #include <net/sock.h> | |
101 | #include <net/ip_fib.h> | |
102 | #include <net/arp.h> | |
103 | #include <net/tcp.h> | |
104 | #include <net/icmp.h> | |
105 | #include <net/xfrm.h> | |
8d71740c | 106 | #include <net/netevent.h> |
63f3444f | 107 | #include <net/rtnetlink.h> |
1da177e4 LT |
108 | #ifdef CONFIG_SYSCTL |
109 | #include <linux/sysctl.h> | |
110 | #endif | |
3769cffb | 111 | #include <net/atmclip.h> |
6e5714ea | 112 | #include <net/secure_seq.h> |
1da177e4 | 113 | |
68a5e3dd DM |
114 | #define RT_FL_TOS(oldflp4) \ |
115 | ((u32)(oldflp4->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK))) | |
1da177e4 LT |
116 | |
117 | #define IP_MAX_MTU 0xFFF0 | |
118 | ||
119 | #define RT_GC_TIMEOUT (300*HZ) | |
120 | ||
1da177e4 | 121 | static int ip_rt_max_size; |
817bc4db | 122 | static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT; |
817bc4db SH |
123 | static int ip_rt_gc_min_interval __read_mostly = HZ / 2; |
124 | static int ip_rt_redirect_number __read_mostly = 9; | |
125 | static int ip_rt_redirect_load __read_mostly = HZ / 50; | |
126 | static int ip_rt_redirect_silence __read_mostly = ((HZ / 50) << (9 + 1)); | |
127 | static int ip_rt_error_cost __read_mostly = HZ; | |
128 | static int ip_rt_error_burst __read_mostly = 5 * HZ; | |
129 | static int ip_rt_gc_elasticity __read_mostly = 8; | |
130 | static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ; | |
131 | static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20; | |
132 | static int ip_rt_min_advmss __read_mostly = 256; | |
1080d709 | 133 | static int rt_chain_length_max __read_mostly = 20; |
1da177e4 | 134 | |
1da177e4 LT |
135 | /* |
136 | * Interface to generic destination cache. | |
137 | */ | |
138 | ||
139 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie); | |
0dbaee3b | 140 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst); |
d33e4553 | 141 | static unsigned int ipv4_default_mtu(const struct dst_entry *dst); |
1da177e4 | 142 | static void ipv4_dst_destroy(struct dst_entry *dst); |
1da177e4 LT |
143 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst); |
144 | static void ipv4_link_failure(struct sk_buff *skb); | |
145 | static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu); | |
569d3645 | 146 | static int rt_garbage_collect(struct dst_ops *ops); |
1da177e4 | 147 | |
72cdd1d9 ED |
148 | static void ipv4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, |
149 | int how) | |
150 | { | |
151 | } | |
1da177e4 | 152 | |
62fa8a84 DM |
153 | static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old) |
154 | { | |
06582540 DM |
155 | struct rtable *rt = (struct rtable *) dst; |
156 | struct inet_peer *peer; | |
157 | u32 *p = NULL; | |
158 | ||
159 | if (!rt->peer) | |
a48eff12 | 160 | rt_bind_peer(rt, rt->rt_dst, 1); |
62fa8a84 | 161 | |
06582540 DM |
162 | peer = rt->peer; |
163 | if (peer) { | |
62fa8a84 DM |
164 | u32 *old_p = __DST_METRICS_PTR(old); |
165 | unsigned long prev, new; | |
166 | ||
06582540 DM |
167 | p = peer->metrics; |
168 | if (inet_metrics_new(peer)) | |
169 | memcpy(p, old_p, sizeof(u32) * RTAX_MAX); | |
62fa8a84 DM |
170 | |
171 | new = (unsigned long) p; | |
172 | prev = cmpxchg(&dst->_metrics, old, new); | |
173 | ||
174 | if (prev != old) { | |
62fa8a84 DM |
175 | p = __DST_METRICS_PTR(prev); |
176 | if (prev & DST_METRICS_READ_ONLY) | |
177 | p = NULL; | |
178 | } else { | |
62fa8a84 DM |
179 | if (rt->fi) { |
180 | fib_info_put(rt->fi); | |
181 | rt->fi = NULL; | |
182 | } | |
183 | } | |
184 | } | |
185 | return p; | |
186 | } | |
187 | ||
d3aaeb38 DM |
188 | static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, const void *daddr); |
189 | ||
1da177e4 LT |
190 | static struct dst_ops ipv4_dst_ops = { |
191 | .family = AF_INET, | |
09640e63 | 192 | .protocol = cpu_to_be16(ETH_P_IP), |
1da177e4 LT |
193 | .gc = rt_garbage_collect, |
194 | .check = ipv4_dst_check, | |
0dbaee3b | 195 | .default_advmss = ipv4_default_advmss, |
d33e4553 | 196 | .default_mtu = ipv4_default_mtu, |
62fa8a84 | 197 | .cow_metrics = ipv4_cow_metrics, |
1da177e4 LT |
198 | .destroy = ipv4_dst_destroy, |
199 | .ifdown = ipv4_dst_ifdown, | |
200 | .negative_advice = ipv4_negative_advice, | |
201 | .link_failure = ipv4_link_failure, | |
202 | .update_pmtu = ip_rt_update_pmtu, | |
1ac06e03 | 203 | .local_out = __ip_local_out, |
d3aaeb38 | 204 | .neigh_lookup = ipv4_neigh_lookup, |
1da177e4 LT |
205 | }; |
206 | ||
207 | #define ECN_OR_COST(class) TC_PRIO_##class | |
208 | ||
4839c52b | 209 | const __u8 ip_tos2prio[16] = { |
1da177e4 | 210 | TC_PRIO_BESTEFFORT, |
4a2b9c37 | 211 | ECN_OR_COST(BESTEFFORT), |
1da177e4 LT |
212 | TC_PRIO_BESTEFFORT, |
213 | ECN_OR_COST(BESTEFFORT), | |
214 | TC_PRIO_BULK, | |
215 | ECN_OR_COST(BULK), | |
216 | TC_PRIO_BULK, | |
217 | ECN_OR_COST(BULK), | |
218 | TC_PRIO_INTERACTIVE, | |
219 | ECN_OR_COST(INTERACTIVE), | |
220 | TC_PRIO_INTERACTIVE, | |
221 | ECN_OR_COST(INTERACTIVE), | |
222 | TC_PRIO_INTERACTIVE_BULK, | |
223 | ECN_OR_COST(INTERACTIVE_BULK), | |
224 | TC_PRIO_INTERACTIVE_BULK, | |
225 | ECN_OR_COST(INTERACTIVE_BULK) | |
226 | }; | |
227 | ||
228 | ||
229 | /* | |
230 | * Route cache. | |
231 | */ | |
232 | ||
233 | /* The locking scheme is rather straight forward: | |
234 | * | |
235 | * 1) Read-Copy Update protects the buckets of the central route hash. | |
236 | * 2) Only writers remove entries, and they hold the lock | |
237 | * as they look at rtable reference counts. | |
238 | * 3) Only readers acquire references to rtable entries, | |
239 | * they do so with atomic increments and with the | |
240 | * lock held. | |
241 | */ | |
242 | ||
243 | struct rt_hash_bucket { | |
1c31720a | 244 | struct rtable __rcu *chain; |
22c047cc | 245 | }; |
1080d709 | 246 | |
8a25d5de IM |
247 | #if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK) || \ |
248 | defined(CONFIG_PROVE_LOCKING) | |
22c047cc ED |
249 | /* |
250 | * Instead of using one spinlock for each rt_hash_bucket, we use a table of spinlocks | |
251 | * The size of this table is a power of two and depends on the number of CPUS. | |
62051200 | 252 | * (on lockdep we have a quite big spinlock_t, so keep the size down there) |
22c047cc | 253 | */ |
62051200 IM |
254 | #ifdef CONFIG_LOCKDEP |
255 | # define RT_HASH_LOCK_SZ 256 | |
22c047cc | 256 | #else |
62051200 IM |
257 | # if NR_CPUS >= 32 |
258 | # define RT_HASH_LOCK_SZ 4096 | |
259 | # elif NR_CPUS >= 16 | |
260 | # define RT_HASH_LOCK_SZ 2048 | |
261 | # elif NR_CPUS >= 8 | |
262 | # define RT_HASH_LOCK_SZ 1024 | |
263 | # elif NR_CPUS >= 4 | |
264 | # define RT_HASH_LOCK_SZ 512 | |
265 | # else | |
266 | # define RT_HASH_LOCK_SZ 256 | |
267 | # endif | |
22c047cc ED |
268 | #endif |
269 | ||
270 | static spinlock_t *rt_hash_locks; | |
271 | # define rt_hash_lock_addr(slot) &rt_hash_locks[(slot) & (RT_HASH_LOCK_SZ - 1)] | |
1ff1cc20 PE |
272 | |
273 | static __init void rt_hash_lock_init(void) | |
274 | { | |
275 | int i; | |
276 | ||
277 | rt_hash_locks = kmalloc(sizeof(spinlock_t) * RT_HASH_LOCK_SZ, | |
278 | GFP_KERNEL); | |
279 | if (!rt_hash_locks) | |
280 | panic("IP: failed to allocate rt_hash_locks\n"); | |
281 | ||
282 | for (i = 0; i < RT_HASH_LOCK_SZ; i++) | |
283 | spin_lock_init(&rt_hash_locks[i]); | |
284 | } | |
22c047cc ED |
285 | #else |
286 | # define rt_hash_lock_addr(slot) NULL | |
1ff1cc20 PE |
287 | |
288 | static inline void rt_hash_lock_init(void) | |
289 | { | |
290 | } | |
22c047cc | 291 | #endif |
1da177e4 | 292 | |
817bc4db SH |
293 | static struct rt_hash_bucket *rt_hash_table __read_mostly; |
294 | static unsigned rt_hash_mask __read_mostly; | |
295 | static unsigned int rt_hash_log __read_mostly; | |
1da177e4 | 296 | |
2f970d83 | 297 | static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat); |
27f39c73 | 298 | #define RT_CACHE_STAT_INC(field) __this_cpu_inc(rt_cache_stat.field) |
1da177e4 | 299 | |
b00180de | 300 | static inline unsigned int rt_hash(__be32 daddr, __be32 saddr, int idx, |
0eae88f3 | 301 | int genid) |
1da177e4 | 302 | { |
0eae88f3 | 303 | return jhash_3words((__force u32)daddr, (__force u32)saddr, |
b00180de | 304 | idx, genid) |
29e75252 | 305 | & rt_hash_mask; |
1da177e4 LT |
306 | } |
307 | ||
e84f84f2 DL |
308 | static inline int rt_genid(struct net *net) |
309 | { | |
310 | return atomic_read(&net->ipv4.rt_genid); | |
311 | } | |
312 | ||
1da177e4 LT |
313 | #ifdef CONFIG_PROC_FS |
314 | struct rt_cache_iter_state { | |
a75e936f | 315 | struct seq_net_private p; |
1da177e4 | 316 | int bucket; |
29e75252 | 317 | int genid; |
1da177e4 LT |
318 | }; |
319 | ||
1218854a | 320 | static struct rtable *rt_cache_get_first(struct seq_file *seq) |
1da177e4 | 321 | { |
1218854a | 322 | struct rt_cache_iter_state *st = seq->private; |
1da177e4 | 323 | struct rtable *r = NULL; |
1da177e4 LT |
324 | |
325 | for (st->bucket = rt_hash_mask; st->bucket >= 0; --st->bucket) { | |
33d480ce | 326 | if (!rcu_access_pointer(rt_hash_table[st->bucket].chain)) |
a6272665 | 327 | continue; |
1da177e4 | 328 | rcu_read_lock_bh(); |
a898def2 | 329 | r = rcu_dereference_bh(rt_hash_table[st->bucket].chain); |
29e75252 | 330 | while (r) { |
d8d1f30b | 331 | if (dev_net(r->dst.dev) == seq_file_net(seq) && |
a75e936f | 332 | r->rt_genid == st->genid) |
29e75252 | 333 | return r; |
d8d1f30b | 334 | r = rcu_dereference_bh(r->dst.rt_next); |
29e75252 | 335 | } |
1da177e4 LT |
336 | rcu_read_unlock_bh(); |
337 | } | |
29e75252 | 338 | return r; |
1da177e4 LT |
339 | } |
340 | ||
1218854a | 341 | static struct rtable *__rt_cache_get_next(struct seq_file *seq, |
642d6318 | 342 | struct rtable *r) |
1da177e4 | 343 | { |
1218854a | 344 | struct rt_cache_iter_state *st = seq->private; |
a6272665 | 345 | |
1c31720a | 346 | r = rcu_dereference_bh(r->dst.rt_next); |
1da177e4 LT |
347 | while (!r) { |
348 | rcu_read_unlock_bh(); | |
a6272665 ED |
349 | do { |
350 | if (--st->bucket < 0) | |
351 | return NULL; | |
33d480ce | 352 | } while (!rcu_access_pointer(rt_hash_table[st->bucket].chain)); |
1da177e4 | 353 | rcu_read_lock_bh(); |
1c31720a | 354 | r = rcu_dereference_bh(rt_hash_table[st->bucket].chain); |
1da177e4 | 355 | } |
1c31720a | 356 | return r; |
1da177e4 LT |
357 | } |
358 | ||
1218854a | 359 | static struct rtable *rt_cache_get_next(struct seq_file *seq, |
642d6318 DL |
360 | struct rtable *r) |
361 | { | |
1218854a YH |
362 | struct rt_cache_iter_state *st = seq->private; |
363 | while ((r = __rt_cache_get_next(seq, r)) != NULL) { | |
d8d1f30b | 364 | if (dev_net(r->dst.dev) != seq_file_net(seq)) |
a75e936f | 365 | continue; |
642d6318 DL |
366 | if (r->rt_genid == st->genid) |
367 | break; | |
368 | } | |
369 | return r; | |
370 | } | |
371 | ||
1218854a | 372 | static struct rtable *rt_cache_get_idx(struct seq_file *seq, loff_t pos) |
1da177e4 | 373 | { |
1218854a | 374 | struct rtable *r = rt_cache_get_first(seq); |
1da177e4 LT |
375 | |
376 | if (r) | |
1218854a | 377 | while (pos && (r = rt_cache_get_next(seq, r))) |
1da177e4 LT |
378 | --pos; |
379 | return pos ? NULL : r; | |
380 | } | |
381 | ||
382 | static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos) | |
383 | { | |
29e75252 | 384 | struct rt_cache_iter_state *st = seq->private; |
29e75252 | 385 | if (*pos) |
1218854a | 386 | return rt_cache_get_idx(seq, *pos - 1); |
e84f84f2 | 387 | st->genid = rt_genid(seq_file_net(seq)); |
29e75252 | 388 | return SEQ_START_TOKEN; |
1da177e4 LT |
389 | } |
390 | ||
391 | static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
392 | { | |
29e75252 | 393 | struct rtable *r; |
1da177e4 LT |
394 | |
395 | if (v == SEQ_START_TOKEN) | |
1218854a | 396 | r = rt_cache_get_first(seq); |
1da177e4 | 397 | else |
1218854a | 398 | r = rt_cache_get_next(seq, v); |
1da177e4 LT |
399 | ++*pos; |
400 | return r; | |
401 | } | |
402 | ||
403 | static void rt_cache_seq_stop(struct seq_file *seq, void *v) | |
404 | { | |
405 | if (v && v != SEQ_START_TOKEN) | |
406 | rcu_read_unlock_bh(); | |
407 | } | |
408 | ||
409 | static int rt_cache_seq_show(struct seq_file *seq, void *v) | |
410 | { | |
411 | if (v == SEQ_START_TOKEN) | |
412 | seq_printf(seq, "%-127s\n", | |
413 | "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t" | |
414 | "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t" | |
415 | "HHUptod\tSpecDst"); | |
416 | else { | |
417 | struct rtable *r = v; | |
69cce1d1 | 418 | struct neighbour *n; |
5e659e4c | 419 | int len; |
1da177e4 | 420 | |
69cce1d1 | 421 | n = dst_get_neighbour(&r->dst); |
0eae88f3 ED |
422 | seq_printf(seq, "%s\t%08X\t%08X\t%8X\t%d\t%u\t%d\t" |
423 | "%08X\t%d\t%u\t%u\t%02X\t%d\t%1d\t%08X%n", | |
d8d1f30b | 424 | r->dst.dev ? r->dst.dev->name : "*", |
0eae88f3 ED |
425 | (__force u32)r->rt_dst, |
426 | (__force u32)r->rt_gateway, | |
d8d1f30b CG |
427 | r->rt_flags, atomic_read(&r->dst.__refcnt), |
428 | r->dst.__use, 0, (__force u32)r->rt_src, | |
0dbaee3b | 429 | dst_metric_advmss(&r->dst) + 40, |
d8d1f30b CG |
430 | dst_metric(&r->dst, RTAX_WINDOW), |
431 | (int)((dst_metric(&r->dst, RTAX_RTT) >> 3) + | |
432 | dst_metric(&r->dst, RTAX_RTTVAR)), | |
475949d8 | 433 | r->rt_key_tos, |
f6b72b62 | 434 | -1, |
69cce1d1 | 435 | (n && (n->nud_state & NUD_CONNECTED)) ? 1 : 0, |
5e659e4c PE |
436 | r->rt_spec_dst, &len); |
437 | ||
438 | seq_printf(seq, "%*s\n", 127 - len, ""); | |
e905a9ed YH |
439 | } |
440 | return 0; | |
1da177e4 LT |
441 | } |
442 | ||
f690808e | 443 | static const struct seq_operations rt_cache_seq_ops = { |
1da177e4 LT |
444 | .start = rt_cache_seq_start, |
445 | .next = rt_cache_seq_next, | |
446 | .stop = rt_cache_seq_stop, | |
447 | .show = rt_cache_seq_show, | |
448 | }; | |
449 | ||
450 | static int rt_cache_seq_open(struct inode *inode, struct file *file) | |
451 | { | |
a75e936f | 452 | return seq_open_net(inode, file, &rt_cache_seq_ops, |
cf7732e4 | 453 | sizeof(struct rt_cache_iter_state)); |
1da177e4 LT |
454 | } |
455 | ||
9a32144e | 456 | static const struct file_operations rt_cache_seq_fops = { |
1da177e4 LT |
457 | .owner = THIS_MODULE, |
458 | .open = rt_cache_seq_open, | |
459 | .read = seq_read, | |
460 | .llseek = seq_lseek, | |
a75e936f | 461 | .release = seq_release_net, |
1da177e4 LT |
462 | }; |
463 | ||
464 | ||
465 | static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos) | |
466 | { | |
467 | int cpu; | |
468 | ||
469 | if (*pos == 0) | |
470 | return SEQ_START_TOKEN; | |
471 | ||
0f23174a | 472 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
473 | if (!cpu_possible(cpu)) |
474 | continue; | |
475 | *pos = cpu+1; | |
2f970d83 | 476 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
477 | } |
478 | return NULL; | |
479 | } | |
480 | ||
481 | static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
482 | { | |
483 | int cpu; | |
484 | ||
0f23174a | 485 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
486 | if (!cpu_possible(cpu)) |
487 | continue; | |
488 | *pos = cpu+1; | |
2f970d83 | 489 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
490 | } |
491 | return NULL; | |
e905a9ed | 492 | |
1da177e4 LT |
493 | } |
494 | ||
495 | static void rt_cpu_seq_stop(struct seq_file *seq, void *v) | |
496 | { | |
497 | ||
498 | } | |
499 | ||
500 | static int rt_cpu_seq_show(struct seq_file *seq, void *v) | |
501 | { | |
502 | struct rt_cache_stat *st = v; | |
503 | ||
504 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 505 | seq_printf(seq, "entries in_hit in_slow_tot in_slow_mc in_no_route in_brd in_martian_dst in_martian_src out_hit out_slow_tot out_slow_mc gc_total gc_ignored gc_goal_miss gc_dst_overflow in_hlist_search out_hlist_search\n"); |
1da177e4 LT |
506 | return 0; |
507 | } | |
e905a9ed | 508 | |
1da177e4 LT |
509 | seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x " |
510 | " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n", | |
fc66f95c | 511 | dst_entries_get_slow(&ipv4_dst_ops), |
1da177e4 LT |
512 | st->in_hit, |
513 | st->in_slow_tot, | |
514 | st->in_slow_mc, | |
515 | st->in_no_route, | |
516 | st->in_brd, | |
517 | st->in_martian_dst, | |
518 | st->in_martian_src, | |
519 | ||
520 | st->out_hit, | |
521 | st->out_slow_tot, | |
e905a9ed | 522 | st->out_slow_mc, |
1da177e4 LT |
523 | |
524 | st->gc_total, | |
525 | st->gc_ignored, | |
526 | st->gc_goal_miss, | |
527 | st->gc_dst_overflow, | |
528 | st->in_hlist_search, | |
529 | st->out_hlist_search | |
530 | ); | |
531 | return 0; | |
532 | } | |
533 | ||
f690808e | 534 | static const struct seq_operations rt_cpu_seq_ops = { |
1da177e4 LT |
535 | .start = rt_cpu_seq_start, |
536 | .next = rt_cpu_seq_next, | |
537 | .stop = rt_cpu_seq_stop, | |
538 | .show = rt_cpu_seq_show, | |
539 | }; | |
540 | ||
541 | ||
542 | static int rt_cpu_seq_open(struct inode *inode, struct file *file) | |
543 | { | |
544 | return seq_open(file, &rt_cpu_seq_ops); | |
545 | } | |
546 | ||
9a32144e | 547 | static const struct file_operations rt_cpu_seq_fops = { |
1da177e4 LT |
548 | .owner = THIS_MODULE, |
549 | .open = rt_cpu_seq_open, | |
550 | .read = seq_read, | |
551 | .llseek = seq_lseek, | |
552 | .release = seq_release, | |
553 | }; | |
554 | ||
c7066f70 | 555 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 556 | static int rt_acct_proc_show(struct seq_file *m, void *v) |
78c686e9 | 557 | { |
a661c419 AD |
558 | struct ip_rt_acct *dst, *src; |
559 | unsigned int i, j; | |
560 | ||
561 | dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL); | |
562 | if (!dst) | |
563 | return -ENOMEM; | |
564 | ||
565 | for_each_possible_cpu(i) { | |
566 | src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i); | |
567 | for (j = 0; j < 256; j++) { | |
568 | dst[j].o_bytes += src[j].o_bytes; | |
569 | dst[j].o_packets += src[j].o_packets; | |
570 | dst[j].i_bytes += src[j].i_bytes; | |
571 | dst[j].i_packets += src[j].i_packets; | |
572 | } | |
78c686e9 PE |
573 | } |
574 | ||
a661c419 AD |
575 | seq_write(m, dst, 256 * sizeof(struct ip_rt_acct)); |
576 | kfree(dst); | |
577 | return 0; | |
578 | } | |
78c686e9 | 579 | |
a661c419 AD |
580 | static int rt_acct_proc_open(struct inode *inode, struct file *file) |
581 | { | |
582 | return single_open(file, rt_acct_proc_show, NULL); | |
78c686e9 | 583 | } |
a661c419 AD |
584 | |
585 | static const struct file_operations rt_acct_proc_fops = { | |
586 | .owner = THIS_MODULE, | |
587 | .open = rt_acct_proc_open, | |
588 | .read = seq_read, | |
589 | .llseek = seq_lseek, | |
590 | .release = single_release, | |
591 | }; | |
78c686e9 | 592 | #endif |
107f1634 | 593 | |
73b38711 | 594 | static int __net_init ip_rt_do_proc_init(struct net *net) |
107f1634 PE |
595 | { |
596 | struct proc_dir_entry *pde; | |
597 | ||
598 | pde = proc_net_fops_create(net, "rt_cache", S_IRUGO, | |
599 | &rt_cache_seq_fops); | |
600 | if (!pde) | |
601 | goto err1; | |
602 | ||
77020720 WC |
603 | pde = proc_create("rt_cache", S_IRUGO, |
604 | net->proc_net_stat, &rt_cpu_seq_fops); | |
107f1634 PE |
605 | if (!pde) |
606 | goto err2; | |
607 | ||
c7066f70 | 608 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 609 | pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops); |
107f1634 PE |
610 | if (!pde) |
611 | goto err3; | |
612 | #endif | |
613 | return 0; | |
614 | ||
c7066f70 | 615 | #ifdef CONFIG_IP_ROUTE_CLASSID |
107f1634 PE |
616 | err3: |
617 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
618 | #endif | |
619 | err2: | |
620 | remove_proc_entry("rt_cache", net->proc_net); | |
621 | err1: | |
622 | return -ENOMEM; | |
623 | } | |
73b38711 DL |
624 | |
625 | static void __net_exit ip_rt_do_proc_exit(struct net *net) | |
626 | { | |
627 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
628 | remove_proc_entry("rt_cache", net->proc_net); | |
c7066f70 | 629 | #ifdef CONFIG_IP_ROUTE_CLASSID |
73b38711 | 630 | remove_proc_entry("rt_acct", net->proc_net); |
0a931acf | 631 | #endif |
73b38711 DL |
632 | } |
633 | ||
634 | static struct pernet_operations ip_rt_proc_ops __net_initdata = { | |
635 | .init = ip_rt_do_proc_init, | |
636 | .exit = ip_rt_do_proc_exit, | |
637 | }; | |
638 | ||
639 | static int __init ip_rt_proc_init(void) | |
640 | { | |
641 | return register_pernet_subsys(&ip_rt_proc_ops); | |
642 | } | |
643 | ||
107f1634 | 644 | #else |
73b38711 | 645 | static inline int ip_rt_proc_init(void) |
107f1634 PE |
646 | { |
647 | return 0; | |
648 | } | |
1da177e4 | 649 | #endif /* CONFIG_PROC_FS */ |
e905a9ed | 650 | |
5969f71d | 651 | static inline void rt_free(struct rtable *rt) |
1da177e4 | 652 | { |
d8d1f30b | 653 | call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free); |
1da177e4 LT |
654 | } |
655 | ||
5969f71d | 656 | static inline void rt_drop(struct rtable *rt) |
1da177e4 | 657 | { |
1da177e4 | 658 | ip_rt_put(rt); |
d8d1f30b | 659 | call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free); |
1da177e4 LT |
660 | } |
661 | ||
5969f71d | 662 | static inline int rt_fast_clean(struct rtable *rth) |
1da177e4 LT |
663 | { |
664 | /* Kill broadcast/multicast entries very aggresively, if they | |
665 | collide in hash table with more useful entries */ | |
666 | return (rth->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && | |
c7537967 | 667 | rt_is_input_route(rth) && rth->dst.rt_next; |
1da177e4 LT |
668 | } |
669 | ||
5969f71d | 670 | static inline int rt_valuable(struct rtable *rth) |
1da177e4 LT |
671 | { |
672 | return (rth->rt_flags & (RTCF_REDIRECTED | RTCF_NOTIFY)) || | |
2c8cec5c | 673 | (rth->peer && rth->peer->pmtu_expires); |
1da177e4 LT |
674 | } |
675 | ||
676 | static int rt_may_expire(struct rtable *rth, unsigned long tmo1, unsigned long tmo2) | |
677 | { | |
678 | unsigned long age; | |
679 | int ret = 0; | |
680 | ||
d8d1f30b | 681 | if (atomic_read(&rth->dst.__refcnt)) |
1da177e4 LT |
682 | goto out; |
683 | ||
d8d1f30b | 684 | age = jiffies - rth->dst.lastuse; |
1da177e4 LT |
685 | if ((age <= tmo1 && !rt_fast_clean(rth)) || |
686 | (age <= tmo2 && rt_valuable(rth))) | |
687 | goto out; | |
688 | ret = 1; | |
689 | out: return ret; | |
690 | } | |
691 | ||
692 | /* Bits of score are: | |
693 | * 31: very valuable | |
694 | * 30: not quite useless | |
695 | * 29..0: usage counter | |
696 | */ | |
697 | static inline u32 rt_score(struct rtable *rt) | |
698 | { | |
d8d1f30b | 699 | u32 score = jiffies - rt->dst.lastuse; |
1da177e4 LT |
700 | |
701 | score = ~score & ~(3<<30); | |
702 | ||
703 | if (rt_valuable(rt)) | |
704 | score |= (1<<31); | |
705 | ||
c7537967 | 706 | if (rt_is_output_route(rt) || |
1da177e4 LT |
707 | !(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL))) |
708 | score |= (1<<30); | |
709 | ||
710 | return score; | |
711 | } | |
712 | ||
1080d709 NH |
713 | static inline bool rt_caching(const struct net *net) |
714 | { | |
715 | return net->ipv4.current_rt_cache_rebuild_count <= | |
716 | net->ipv4.sysctl_rt_cache_rebuild_count; | |
717 | } | |
718 | ||
5e2b61f7 DM |
719 | static inline bool compare_hash_inputs(const struct rtable *rt1, |
720 | const struct rtable *rt2) | |
1080d709 | 721 | { |
5e2b61f7 DM |
722 | return ((((__force u32)rt1->rt_key_dst ^ (__force u32)rt2->rt_key_dst) | |
723 | ((__force u32)rt1->rt_key_src ^ (__force u32)rt2->rt_key_src) | | |
97a80410 | 724 | (rt1->rt_route_iif ^ rt2->rt_route_iif)) == 0); |
1080d709 NH |
725 | } |
726 | ||
5e2b61f7 | 727 | static inline int compare_keys(struct rtable *rt1, struct rtable *rt2) |
1da177e4 | 728 | { |
5e2b61f7 DM |
729 | return (((__force u32)rt1->rt_key_dst ^ (__force u32)rt2->rt_key_dst) | |
730 | ((__force u32)rt1->rt_key_src ^ (__force u32)rt2->rt_key_src) | | |
731 | (rt1->rt_mark ^ rt2->rt_mark) | | |
475949d8 | 732 | (rt1->rt_key_tos ^ rt2->rt_key_tos) | |
d547f727 | 733 | (rt1->rt_route_iif ^ rt2->rt_route_iif) | |
97a80410 | 734 | (rt1->rt_oif ^ rt2->rt_oif)) == 0; |
1da177e4 LT |
735 | } |
736 | ||
b5921910 DL |
737 | static inline int compare_netns(struct rtable *rt1, struct rtable *rt2) |
738 | { | |
d8d1f30b | 739 | return net_eq(dev_net(rt1->dst.dev), dev_net(rt2->dst.dev)); |
b5921910 DL |
740 | } |
741 | ||
e84f84f2 DL |
742 | static inline int rt_is_expired(struct rtable *rth) |
743 | { | |
d8d1f30b | 744 | return rth->rt_genid != rt_genid(dev_net(rth->dst.dev)); |
e84f84f2 DL |
745 | } |
746 | ||
beb659bd ED |
747 | /* |
748 | * Perform a full scan of hash table and free all entries. | |
749 | * Can be called by a softirq or a process. | |
750 | * In the later case, we want to be reschedule if necessary | |
751 | */ | |
6561a3b1 | 752 | static void rt_do_flush(struct net *net, int process_context) |
beb659bd ED |
753 | { |
754 | unsigned int i; | |
755 | struct rtable *rth, *next; | |
756 | ||
757 | for (i = 0; i <= rt_hash_mask; i++) { | |
6561a3b1 DM |
758 | struct rtable __rcu **pprev; |
759 | struct rtable *list; | |
760 | ||
beb659bd ED |
761 | if (process_context && need_resched()) |
762 | cond_resched(); | |
33d480ce | 763 | rth = rcu_access_pointer(rt_hash_table[i].chain); |
beb659bd ED |
764 | if (!rth) |
765 | continue; | |
766 | ||
767 | spin_lock_bh(rt_hash_lock_addr(i)); | |
32cb5b4e | 768 | |
6561a3b1 DM |
769 | list = NULL; |
770 | pprev = &rt_hash_table[i].chain; | |
771 | rth = rcu_dereference_protected(*pprev, | |
1c31720a | 772 | lockdep_is_held(rt_hash_lock_addr(i))); |
32cb5b4e | 773 | |
6561a3b1 DM |
774 | while (rth) { |
775 | next = rcu_dereference_protected(rth->dst.rt_next, | |
1c31720a | 776 | lockdep_is_held(rt_hash_lock_addr(i))); |
6561a3b1 DM |
777 | |
778 | if (!net || | |
779 | net_eq(dev_net(rth->dst.dev), net)) { | |
780 | rcu_assign_pointer(*pprev, next); | |
781 | rcu_assign_pointer(rth->dst.rt_next, list); | |
782 | list = rth; | |
32cb5b4e | 783 | } else { |
6561a3b1 | 784 | pprev = &rth->dst.rt_next; |
32cb5b4e | 785 | } |
6561a3b1 | 786 | rth = next; |
32cb5b4e | 787 | } |
6561a3b1 | 788 | |
beb659bd ED |
789 | spin_unlock_bh(rt_hash_lock_addr(i)); |
790 | ||
6561a3b1 DM |
791 | for (; list; list = next) { |
792 | next = rcu_dereference_protected(list->dst.rt_next, 1); | |
793 | rt_free(list); | |
beb659bd ED |
794 | } |
795 | } | |
796 | } | |
797 | ||
1080d709 NH |
798 | /* |
799 | * While freeing expired entries, we compute average chain length | |
800 | * and standard deviation, using fixed-point arithmetic. | |
801 | * This to have an estimation of rt_chain_length_max | |
802 | * rt_chain_length_max = max(elasticity, AVG + 4*SD) | |
803 | * We use 3 bits for frational part, and 29 (or 61) for magnitude. | |
804 | */ | |
805 | ||
806 | #define FRACT_BITS 3 | |
807 | #define ONE (1UL << FRACT_BITS) | |
808 | ||
98376387 ED |
809 | /* |
810 | * Given a hash chain and an item in this hash chain, | |
811 | * find if a previous entry has the same hash_inputs | |
812 | * (but differs on tos, mark or oif) | |
813 | * Returns 0 if an alias is found. | |
814 | * Returns ONE if rth has no alias before itself. | |
815 | */ | |
816 | static int has_noalias(const struct rtable *head, const struct rtable *rth) | |
817 | { | |
818 | const struct rtable *aux = head; | |
819 | ||
820 | while (aux != rth) { | |
5e2b61f7 | 821 | if (compare_hash_inputs(aux, rth)) |
98376387 | 822 | return 0; |
1c31720a | 823 | aux = rcu_dereference_protected(aux->dst.rt_next, 1); |
98376387 ED |
824 | } |
825 | return ONE; | |
826 | } | |
827 | ||
29e75252 | 828 | /* |
25985edc | 829 | * Perturbation of rt_genid by a small quantity [1..256] |
29e75252 ED |
830 | * Using 8 bits of shuffling ensure we can call rt_cache_invalidate() |
831 | * many times (2^24) without giving recent rt_genid. | |
832 | * Jenkins hash is strong enough that litle changes of rt_genid are OK. | |
1da177e4 | 833 | */ |
86c657f6 | 834 | static void rt_cache_invalidate(struct net *net) |
1da177e4 | 835 | { |
29e75252 | 836 | unsigned char shuffle; |
1da177e4 | 837 | |
29e75252 | 838 | get_random_bytes(&shuffle, sizeof(shuffle)); |
e84f84f2 | 839 | atomic_add(shuffle + 1U, &net->ipv4.rt_genid); |
1da177e4 LT |
840 | } |
841 | ||
29e75252 ED |
842 | /* |
843 | * delay < 0 : invalidate cache (fast : entries will be deleted later) | |
844 | * delay >= 0 : invalidate & flush cache (can be long) | |
845 | */ | |
76e6ebfb | 846 | void rt_cache_flush(struct net *net, int delay) |
1da177e4 | 847 | { |
86c657f6 | 848 | rt_cache_invalidate(net); |
29e75252 | 849 | if (delay >= 0) |
6561a3b1 | 850 | rt_do_flush(net, !in_softirq()); |
1da177e4 LT |
851 | } |
852 | ||
a5ee1551 | 853 | /* Flush previous cache invalidated entries from the cache */ |
6561a3b1 | 854 | void rt_cache_flush_batch(struct net *net) |
a5ee1551 | 855 | { |
6561a3b1 | 856 | rt_do_flush(net, !in_softirq()); |
a5ee1551 EB |
857 | } |
858 | ||
1080d709 NH |
859 | static void rt_emergency_hash_rebuild(struct net *net) |
860 | { | |
3ee94372 | 861 | if (net_ratelimit()) |
1080d709 | 862 | printk(KERN_WARNING "Route hash chain too long!\n"); |
3ee94372 | 863 | rt_cache_invalidate(net); |
1080d709 NH |
864 | } |
865 | ||
1da177e4 LT |
866 | /* |
867 | Short description of GC goals. | |
868 | ||
869 | We want to build algorithm, which will keep routing cache | |
870 | at some equilibrium point, when number of aged off entries | |
871 | is kept approximately equal to newly generated ones. | |
872 | ||
873 | Current expiration strength is variable "expire". | |
874 | We try to adjust it dynamically, so that if networking | |
875 | is idle expires is large enough to keep enough of warm entries, | |
876 | and when load increases it reduces to limit cache size. | |
877 | */ | |
878 | ||
569d3645 | 879 | static int rt_garbage_collect(struct dst_ops *ops) |
1da177e4 LT |
880 | { |
881 | static unsigned long expire = RT_GC_TIMEOUT; | |
882 | static unsigned long last_gc; | |
883 | static int rover; | |
884 | static int equilibrium; | |
1c31720a ED |
885 | struct rtable *rth; |
886 | struct rtable __rcu **rthp; | |
1da177e4 LT |
887 | unsigned long now = jiffies; |
888 | int goal; | |
fc66f95c | 889 | int entries = dst_entries_get_fast(&ipv4_dst_ops); |
1da177e4 LT |
890 | |
891 | /* | |
892 | * Garbage collection is pretty expensive, | |
893 | * do not make it too frequently. | |
894 | */ | |
895 | ||
896 | RT_CACHE_STAT_INC(gc_total); | |
897 | ||
898 | if (now - last_gc < ip_rt_gc_min_interval && | |
fc66f95c | 899 | entries < ip_rt_max_size) { |
1da177e4 LT |
900 | RT_CACHE_STAT_INC(gc_ignored); |
901 | goto out; | |
902 | } | |
903 | ||
fc66f95c | 904 | entries = dst_entries_get_slow(&ipv4_dst_ops); |
1da177e4 | 905 | /* Calculate number of entries, which we want to expire now. */ |
fc66f95c | 906 | goal = entries - (ip_rt_gc_elasticity << rt_hash_log); |
1da177e4 LT |
907 | if (goal <= 0) { |
908 | if (equilibrium < ipv4_dst_ops.gc_thresh) | |
909 | equilibrium = ipv4_dst_ops.gc_thresh; | |
fc66f95c | 910 | goal = entries - equilibrium; |
1da177e4 | 911 | if (goal > 0) { |
b790cedd | 912 | equilibrium += min_t(unsigned int, goal >> 1, rt_hash_mask + 1); |
fc66f95c | 913 | goal = entries - equilibrium; |
1da177e4 LT |
914 | } |
915 | } else { | |
916 | /* We are in dangerous area. Try to reduce cache really | |
917 | * aggressively. | |
918 | */ | |
b790cedd | 919 | goal = max_t(unsigned int, goal >> 1, rt_hash_mask + 1); |
fc66f95c | 920 | equilibrium = entries - goal; |
1da177e4 LT |
921 | } |
922 | ||
923 | if (now - last_gc >= ip_rt_gc_min_interval) | |
924 | last_gc = now; | |
925 | ||
926 | if (goal <= 0) { | |
927 | equilibrium += goal; | |
928 | goto work_done; | |
929 | } | |
930 | ||
931 | do { | |
932 | int i, k; | |
933 | ||
934 | for (i = rt_hash_mask, k = rover; i >= 0; i--) { | |
935 | unsigned long tmo = expire; | |
936 | ||
937 | k = (k + 1) & rt_hash_mask; | |
938 | rthp = &rt_hash_table[k].chain; | |
22c047cc | 939 | spin_lock_bh(rt_hash_lock_addr(k)); |
1c31720a ED |
940 | while ((rth = rcu_dereference_protected(*rthp, |
941 | lockdep_is_held(rt_hash_lock_addr(k)))) != NULL) { | |
e84f84f2 | 942 | if (!rt_is_expired(rth) && |
29e75252 | 943 | !rt_may_expire(rth, tmo, expire)) { |
1da177e4 | 944 | tmo >>= 1; |
d8d1f30b | 945 | rthp = &rth->dst.rt_next; |
1da177e4 LT |
946 | continue; |
947 | } | |
d8d1f30b | 948 | *rthp = rth->dst.rt_next; |
1da177e4 LT |
949 | rt_free(rth); |
950 | goal--; | |
1da177e4 | 951 | } |
22c047cc | 952 | spin_unlock_bh(rt_hash_lock_addr(k)); |
1da177e4 LT |
953 | if (goal <= 0) |
954 | break; | |
955 | } | |
956 | rover = k; | |
957 | ||
958 | if (goal <= 0) | |
959 | goto work_done; | |
960 | ||
961 | /* Goal is not achieved. We stop process if: | |
962 | ||
963 | - if expire reduced to zero. Otherwise, expire is halfed. | |
964 | - if table is not full. | |
965 | - if we are called from interrupt. | |
966 | - jiffies check is just fallback/debug loop breaker. | |
967 | We will not spin here for long time in any case. | |
968 | */ | |
969 | ||
970 | RT_CACHE_STAT_INC(gc_goal_miss); | |
971 | ||
972 | if (expire == 0) | |
973 | break; | |
974 | ||
975 | expire >>= 1; | |
1da177e4 | 976 | |
fc66f95c | 977 | if (dst_entries_get_fast(&ipv4_dst_ops) < ip_rt_max_size) |
1da177e4 LT |
978 | goto out; |
979 | } while (!in_softirq() && time_before_eq(jiffies, now)); | |
980 | ||
fc66f95c ED |
981 | if (dst_entries_get_fast(&ipv4_dst_ops) < ip_rt_max_size) |
982 | goto out; | |
983 | if (dst_entries_get_slow(&ipv4_dst_ops) < ip_rt_max_size) | |
1da177e4 LT |
984 | goto out; |
985 | if (net_ratelimit()) | |
986 | printk(KERN_WARNING "dst cache overflow\n"); | |
987 | RT_CACHE_STAT_INC(gc_dst_overflow); | |
988 | return 1; | |
989 | ||
990 | work_done: | |
991 | expire += ip_rt_gc_min_interval; | |
992 | if (expire > ip_rt_gc_timeout || | |
fc66f95c ED |
993 | dst_entries_get_fast(&ipv4_dst_ops) < ipv4_dst_ops.gc_thresh || |
994 | dst_entries_get_slow(&ipv4_dst_ops) < ipv4_dst_ops.gc_thresh) | |
1da177e4 | 995 | expire = ip_rt_gc_timeout; |
1da177e4 LT |
996 | out: return 0; |
997 | } | |
998 | ||
98376387 ED |
999 | /* |
1000 | * Returns number of entries in a hash chain that have different hash_inputs | |
1001 | */ | |
1002 | static int slow_chain_length(const struct rtable *head) | |
1003 | { | |
1004 | int length = 0; | |
1005 | const struct rtable *rth = head; | |
1006 | ||
1007 | while (rth) { | |
1008 | length += has_noalias(head, rth); | |
1c31720a | 1009 | rth = rcu_dereference_protected(rth->dst.rt_next, 1); |
98376387 ED |
1010 | } |
1011 | return length >> FRACT_BITS; | |
1012 | } | |
1013 | ||
d3aaeb38 | 1014 | static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, const void *daddr) |
3769cffb | 1015 | { |
3769cffb | 1016 | struct neigh_table *tbl = &arp_tbl; |
d3aaeb38 DM |
1017 | static const __be32 inaddr_any = 0; |
1018 | struct net_device *dev = dst->dev; | |
1019 | const __be32 *pkey = daddr; | |
3769cffb DM |
1020 | struct neighbour *n; |
1021 | ||
1022 | #if defined(CONFIG_ATM_CLIP) || defined(CONFIG_ATM_CLIP_MODULE) | |
1023 | if (dev->type == ARPHRD_ATM) | |
1024 | tbl = clip_tbl_hook; | |
1025 | #endif | |
3769cffb | 1026 | if (dev->flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) |
d3aaeb38 DM |
1027 | pkey = &inaddr_any; |
1028 | ||
1029 | n = __ipv4_neigh_lookup(tbl, dev, *(__force u32 *)pkey); | |
1030 | if (n) | |
1031 | return n; | |
1032 | return neigh_create(tbl, pkey, dev); | |
1033 | } | |
1034 | ||
1035 | static int rt_bind_neighbour(struct rtable *rt) | |
1036 | { | |
1037 | struct neighbour *n = ipv4_neigh_lookup(&rt->dst, &rt->rt_gateway); | |
3769cffb DM |
1038 | if (IS_ERR(n)) |
1039 | return PTR_ERR(n); | |
69cce1d1 | 1040 | dst_set_neighbour(&rt->dst, n); |
3769cffb DM |
1041 | |
1042 | return 0; | |
1043 | } | |
1044 | ||
b23dd4fe DM |
1045 | static struct rtable *rt_intern_hash(unsigned hash, struct rtable *rt, |
1046 | struct sk_buff *skb, int ifindex) | |
1da177e4 | 1047 | { |
1c31720a ED |
1048 | struct rtable *rth, *cand; |
1049 | struct rtable __rcu **rthp, **candp; | |
1da177e4 | 1050 | unsigned long now; |
1da177e4 LT |
1051 | u32 min_score; |
1052 | int chain_length; | |
1053 | int attempts = !in_softirq(); | |
1054 | ||
1055 | restart: | |
1056 | chain_length = 0; | |
1057 | min_score = ~(u32)0; | |
1058 | cand = NULL; | |
1059 | candp = NULL; | |
1060 | now = jiffies; | |
1061 | ||
d8d1f30b | 1062 | if (!rt_caching(dev_net(rt->dst.dev))) { |
73e42897 NH |
1063 | /* |
1064 | * If we're not caching, just tell the caller we | |
1065 | * were successful and don't touch the route. The | |
1066 | * caller hold the sole reference to the cache entry, and | |
1067 | * it will be released when the caller is done with it. | |
1068 | * If we drop it here, the callers have no way to resolve routes | |
1069 | * when we're not caching. Instead, just point *rp at rt, so | |
1070 | * the caller gets a single use out of the route | |
b6280b47 NH |
1071 | * Note that we do rt_free on this new route entry, so that |
1072 | * once its refcount hits zero, we are still able to reap it | |
1073 | * (Thanks Alexey) | |
27b75c95 ED |
1074 | * Note: To avoid expensive rcu stuff for this uncached dst, |
1075 | * we set DST_NOCACHE so that dst_release() can free dst without | |
1076 | * waiting a grace period. | |
73e42897 | 1077 | */ |
b6280b47 | 1078 | |
c7d4426a | 1079 | rt->dst.flags |= DST_NOCACHE; |
c7537967 | 1080 | if (rt->rt_type == RTN_UNICAST || rt_is_output_route(rt)) { |
3769cffb | 1081 | int err = rt_bind_neighbour(rt); |
b6280b47 NH |
1082 | if (err) { |
1083 | if (net_ratelimit()) | |
1084 | printk(KERN_WARNING | |
1085 | "Neighbour table failure & not caching routes.\n"); | |
27b75c95 | 1086 | ip_rt_put(rt); |
b23dd4fe | 1087 | return ERR_PTR(err); |
b6280b47 NH |
1088 | } |
1089 | } | |
1090 | ||
b6280b47 | 1091 | goto skip_hashing; |
1080d709 NH |
1092 | } |
1093 | ||
1da177e4 LT |
1094 | rthp = &rt_hash_table[hash].chain; |
1095 | ||
22c047cc | 1096 | spin_lock_bh(rt_hash_lock_addr(hash)); |
1c31720a ED |
1097 | while ((rth = rcu_dereference_protected(*rthp, |
1098 | lockdep_is_held(rt_hash_lock_addr(hash)))) != NULL) { | |
e84f84f2 | 1099 | if (rt_is_expired(rth)) { |
d8d1f30b | 1100 | *rthp = rth->dst.rt_next; |
29e75252 ED |
1101 | rt_free(rth); |
1102 | continue; | |
1103 | } | |
5e2b61f7 | 1104 | if (compare_keys(rth, rt) && compare_netns(rth, rt)) { |
1da177e4 | 1105 | /* Put it first */ |
d8d1f30b | 1106 | *rthp = rth->dst.rt_next; |
1da177e4 LT |
1107 | /* |
1108 | * Since lookup is lockfree, the deletion | |
1109 | * must be visible to another weakly ordered CPU before | |
1110 | * the insertion at the start of the hash chain. | |
1111 | */ | |
d8d1f30b | 1112 | rcu_assign_pointer(rth->dst.rt_next, |
1da177e4 LT |
1113 | rt_hash_table[hash].chain); |
1114 | /* | |
1115 | * Since lookup is lockfree, the update writes | |
1116 | * must be ordered for consistency on SMP. | |
1117 | */ | |
1118 | rcu_assign_pointer(rt_hash_table[hash].chain, rth); | |
1119 | ||
d8d1f30b | 1120 | dst_use(&rth->dst, now); |
22c047cc | 1121 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1122 | |
1123 | rt_drop(rt); | |
b23dd4fe | 1124 | if (skb) |
d8d1f30b | 1125 | skb_dst_set(skb, &rth->dst); |
b23dd4fe | 1126 | return rth; |
1da177e4 LT |
1127 | } |
1128 | ||
d8d1f30b | 1129 | if (!atomic_read(&rth->dst.__refcnt)) { |
1da177e4 LT |
1130 | u32 score = rt_score(rth); |
1131 | ||
1132 | if (score <= min_score) { | |
1133 | cand = rth; | |
1134 | candp = rthp; | |
1135 | min_score = score; | |
1136 | } | |
1137 | } | |
1138 | ||
1139 | chain_length++; | |
1140 | ||
d8d1f30b | 1141 | rthp = &rth->dst.rt_next; |
1da177e4 LT |
1142 | } |
1143 | ||
1144 | if (cand) { | |
1145 | /* ip_rt_gc_elasticity used to be average length of chain | |
1146 | * length, when exceeded gc becomes really aggressive. | |
1147 | * | |
1148 | * The second limit is less certain. At the moment it allows | |
1149 | * only 2 entries per bucket. We will see. | |
1150 | */ | |
1151 | if (chain_length > ip_rt_gc_elasticity) { | |
d8d1f30b | 1152 | *candp = cand->dst.rt_next; |
1da177e4 LT |
1153 | rt_free(cand); |
1154 | } | |
1080d709 | 1155 | } else { |
98376387 ED |
1156 | if (chain_length > rt_chain_length_max && |
1157 | slow_chain_length(rt_hash_table[hash].chain) > rt_chain_length_max) { | |
d8d1f30b | 1158 | struct net *net = dev_net(rt->dst.dev); |
1080d709 | 1159 | int num = ++net->ipv4.current_rt_cache_rebuild_count; |
b35ecb5d | 1160 | if (!rt_caching(net)) { |
1080d709 | 1161 | printk(KERN_WARNING "%s: %d rebuilds is over limit, route caching disabled\n", |
d8d1f30b | 1162 | rt->dst.dev->name, num); |
1080d709 | 1163 | } |
b35ecb5d | 1164 | rt_emergency_hash_rebuild(net); |
6a2bad70 PE |
1165 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1166 | ||
5e2b61f7 | 1167 | hash = rt_hash(rt->rt_key_dst, rt->rt_key_src, |
6a2bad70 PE |
1168 | ifindex, rt_genid(net)); |
1169 | goto restart; | |
1080d709 | 1170 | } |
1da177e4 LT |
1171 | } |
1172 | ||
1173 | /* Try to bind route to arp only if it is output | |
1174 | route or unicast forwarding path. | |
1175 | */ | |
c7537967 | 1176 | if (rt->rt_type == RTN_UNICAST || rt_is_output_route(rt)) { |
3769cffb | 1177 | int err = rt_bind_neighbour(rt); |
1da177e4 | 1178 | if (err) { |
22c047cc | 1179 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1180 | |
1181 | if (err != -ENOBUFS) { | |
1182 | rt_drop(rt); | |
b23dd4fe | 1183 | return ERR_PTR(err); |
1da177e4 LT |
1184 | } |
1185 | ||
1186 | /* Neighbour tables are full and nothing | |
1187 | can be released. Try to shrink route cache, | |
1188 | it is most likely it holds some neighbour records. | |
1189 | */ | |
1190 | if (attempts-- > 0) { | |
1191 | int saved_elasticity = ip_rt_gc_elasticity; | |
1192 | int saved_int = ip_rt_gc_min_interval; | |
1193 | ip_rt_gc_elasticity = 1; | |
1194 | ip_rt_gc_min_interval = 0; | |
569d3645 | 1195 | rt_garbage_collect(&ipv4_dst_ops); |
1da177e4 LT |
1196 | ip_rt_gc_min_interval = saved_int; |
1197 | ip_rt_gc_elasticity = saved_elasticity; | |
1198 | goto restart; | |
1199 | } | |
1200 | ||
1201 | if (net_ratelimit()) | |
7e1b33e5 | 1202 | printk(KERN_WARNING "ipv4: Neighbour table overflow.\n"); |
1da177e4 | 1203 | rt_drop(rt); |
b23dd4fe | 1204 | return ERR_PTR(-ENOBUFS); |
1da177e4 LT |
1205 | } |
1206 | } | |
1207 | ||
d8d1f30b | 1208 | rt->dst.rt_next = rt_hash_table[hash].chain; |
1080d709 | 1209 | |
00269b54 ED |
1210 | /* |
1211 | * Since lookup is lockfree, we must make sure | |
25985edc | 1212 | * previous writes to rt are committed to memory |
00269b54 ED |
1213 | * before making rt visible to other CPUS. |
1214 | */ | |
1ddbcb00 | 1215 | rcu_assign_pointer(rt_hash_table[hash].chain, rt); |
1080d709 | 1216 | |
22c047cc | 1217 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
73e42897 | 1218 | |
b6280b47 | 1219 | skip_hashing: |
b23dd4fe | 1220 | if (skb) |
d8d1f30b | 1221 | skb_dst_set(skb, &rt->dst); |
b23dd4fe | 1222 | return rt; |
1da177e4 LT |
1223 | } |
1224 | ||
6431cbc2 DM |
1225 | static atomic_t __rt_peer_genid = ATOMIC_INIT(0); |
1226 | ||
1227 | static u32 rt_peer_genid(void) | |
1228 | { | |
1229 | return atomic_read(&__rt_peer_genid); | |
1230 | } | |
1231 | ||
a48eff12 | 1232 | void rt_bind_peer(struct rtable *rt, __be32 daddr, int create) |
1da177e4 | 1233 | { |
1da177e4 LT |
1234 | struct inet_peer *peer; |
1235 | ||
a48eff12 | 1236 | peer = inet_getpeer_v4(daddr, create); |
1da177e4 | 1237 | |
49e8ab03 | 1238 | if (peer && cmpxchg(&rt->peer, NULL, peer) != NULL) |
1da177e4 | 1239 | inet_putpeer(peer); |
6431cbc2 DM |
1240 | else |
1241 | rt->rt_peer_genid = rt_peer_genid(); | |
1da177e4 LT |
1242 | } |
1243 | ||
1244 | /* | |
1245 | * Peer allocation may fail only in serious out-of-memory conditions. However | |
1246 | * we still can generate some output. | |
1247 | * Random ID selection looks a bit dangerous because we have no chances to | |
1248 | * select ID being unique in a reasonable period of time. | |
1249 | * But broken packet identifier may be better than no packet at all. | |
1250 | */ | |
1251 | static void ip_select_fb_ident(struct iphdr *iph) | |
1252 | { | |
1253 | static DEFINE_SPINLOCK(ip_fb_id_lock); | |
1254 | static u32 ip_fallback_id; | |
1255 | u32 salt; | |
1256 | ||
1257 | spin_lock_bh(&ip_fb_id_lock); | |
e448515c | 1258 | salt = secure_ip_id((__force __be32)ip_fallback_id ^ iph->daddr); |
1da177e4 LT |
1259 | iph->id = htons(salt & 0xFFFF); |
1260 | ip_fallback_id = salt; | |
1261 | spin_unlock_bh(&ip_fb_id_lock); | |
1262 | } | |
1263 | ||
1264 | void __ip_select_ident(struct iphdr *iph, struct dst_entry *dst, int more) | |
1265 | { | |
1266 | struct rtable *rt = (struct rtable *) dst; | |
1267 | ||
1268 | if (rt) { | |
1269 | if (rt->peer == NULL) | |
a48eff12 | 1270 | rt_bind_peer(rt, rt->rt_dst, 1); |
1da177e4 LT |
1271 | |
1272 | /* If peer is attached to destination, it is never detached, | |
1273 | so that we need not to grab a lock to dereference it. | |
1274 | */ | |
1275 | if (rt->peer) { | |
1276 | iph->id = htons(inet_getid(rt->peer, more)); | |
1277 | return; | |
1278 | } | |
1279 | } else | |
e905a9ed | 1280 | printk(KERN_DEBUG "rt_bind_peer(0) @%p\n", |
9c2b3328 | 1281 | __builtin_return_address(0)); |
1da177e4 LT |
1282 | |
1283 | ip_select_fb_ident(iph); | |
1284 | } | |
4bc2f18b | 1285 | EXPORT_SYMBOL(__ip_select_ident); |
1da177e4 LT |
1286 | |
1287 | static void rt_del(unsigned hash, struct rtable *rt) | |
1288 | { | |
1c31720a ED |
1289 | struct rtable __rcu **rthp; |
1290 | struct rtable *aux; | |
1da177e4 | 1291 | |
29e75252 | 1292 | rthp = &rt_hash_table[hash].chain; |
22c047cc | 1293 | spin_lock_bh(rt_hash_lock_addr(hash)); |
1da177e4 | 1294 | ip_rt_put(rt); |
1c31720a ED |
1295 | while ((aux = rcu_dereference_protected(*rthp, |
1296 | lockdep_is_held(rt_hash_lock_addr(hash)))) != NULL) { | |
e84f84f2 | 1297 | if (aux == rt || rt_is_expired(aux)) { |
d8d1f30b | 1298 | *rthp = aux->dst.rt_next; |
29e75252 ED |
1299 | rt_free(aux); |
1300 | continue; | |
1da177e4 | 1301 | } |
d8d1f30b | 1302 | rthp = &aux->dst.rt_next; |
29e75252 | 1303 | } |
22c047cc | 1304 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1305 | } |
1306 | ||
ed7865a4 | 1307 | /* called in rcu_read_lock() section */ |
f7655229 AV |
1308 | void ip_rt_redirect(__be32 old_gw, __be32 daddr, __be32 new_gw, |
1309 | __be32 saddr, struct net_device *dev) | |
1da177e4 | 1310 | { |
7cc9150e | 1311 | int s, i; |
ed7865a4 | 1312 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
7cc9150e FL |
1313 | struct rtable *rt; |
1314 | __be32 skeys[2] = { saddr, 0 }; | |
1315 | int ikeys[2] = { dev->ifindex, 0 }; | |
1316 | struct flowi4 fl4; | |
f39925db | 1317 | struct inet_peer *peer; |
317805b8 | 1318 | struct net *net; |
1da177e4 | 1319 | |
1da177e4 LT |
1320 | if (!in_dev) |
1321 | return; | |
1322 | ||
c346dca1 | 1323 | net = dev_net(dev); |
9d4fb27d JP |
1324 | if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) || |
1325 | ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) || | |
1326 | ipv4_is_zeronet(new_gw)) | |
1da177e4 LT |
1327 | goto reject_redirect; |
1328 | ||
1329 | if (!IN_DEV_SHARED_MEDIA(in_dev)) { | |
1330 | if (!inet_addr_onlink(in_dev, new_gw, old_gw)) | |
1331 | goto reject_redirect; | |
1332 | if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev)) | |
1333 | goto reject_redirect; | |
1334 | } else { | |
317805b8 | 1335 | if (inet_addr_type(net, new_gw) != RTN_UNICAST) |
1da177e4 LT |
1336 | goto reject_redirect; |
1337 | } | |
1338 | ||
7cc9150e FL |
1339 | memset(&fl4, 0, sizeof(fl4)); |
1340 | fl4.daddr = daddr; | |
1341 | for (s = 0; s < 2; s++) { | |
1342 | for (i = 0; i < 2; i++) { | |
1343 | fl4.flowi4_oif = ikeys[i]; | |
1344 | fl4.saddr = skeys[s]; | |
1345 | rt = __ip_route_output_key(net, &fl4); | |
1346 | if (IS_ERR(rt)) | |
1347 | continue; | |
1348 | ||
1349 | if (rt->dst.error || rt->dst.dev != dev || | |
1350 | rt->rt_gateway != old_gw) { | |
1351 | ip_rt_put(rt); | |
1352 | continue; | |
1353 | } | |
e905a9ed | 1354 | |
7cc9150e FL |
1355 | if (!rt->peer) |
1356 | rt_bind_peer(rt, rt->rt_dst, 1); | |
1da177e4 | 1357 | |
7cc9150e FL |
1358 | peer = rt->peer; |
1359 | if (peer) { | |
1360 | peer->redirect_learned.a4 = new_gw; | |
1361 | atomic_inc(&__rt_peer_genid); | |
1362 | } | |
1363 | ||
1364 | ip_rt_put(rt); | |
1365 | return; | |
1366 | } | |
1da177e4 | 1367 | } |
1da177e4 LT |
1368 | return; |
1369 | ||
1370 | reject_redirect: | |
1371 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1372 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) | |
673d57e7 HH |
1373 | printk(KERN_INFO "Redirect from %pI4 on %s about %pI4 ignored.\n" |
1374 | " Advised path = %pI4 -> %pI4\n", | |
1375 | &old_gw, dev->name, &new_gw, | |
1376 | &saddr, &daddr); | |
1da177e4 | 1377 | #endif |
ed7865a4 | 1378 | ; |
1da177e4 LT |
1379 | } |
1380 | ||
fe6fe792 ED |
1381 | static bool peer_pmtu_expired(struct inet_peer *peer) |
1382 | { | |
1383 | unsigned long orig = ACCESS_ONCE(peer->pmtu_expires); | |
1384 | ||
1385 | return orig && | |
1386 | time_after_eq(jiffies, orig) && | |
1387 | cmpxchg(&peer->pmtu_expires, orig, 0) == orig; | |
1388 | } | |
1389 | ||
1390 | static bool peer_pmtu_cleaned(struct inet_peer *peer) | |
1391 | { | |
1392 | unsigned long orig = ACCESS_ONCE(peer->pmtu_expires); | |
1393 | ||
1394 | return orig && | |
1395 | cmpxchg(&peer->pmtu_expires, orig, 0) == orig; | |
1396 | } | |
1397 | ||
1da177e4 LT |
1398 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst) |
1399 | { | |
ee6b9673 | 1400 | struct rtable *rt = (struct rtable *)dst; |
1da177e4 LT |
1401 | struct dst_entry *ret = dst; |
1402 | ||
1403 | if (rt) { | |
d11a4dc1 | 1404 | if (dst->obsolete > 0) { |
1da177e4 LT |
1405 | ip_rt_put(rt); |
1406 | ret = NULL; | |
2c8cec5c | 1407 | } else if (rt->rt_flags & RTCF_REDIRECTED) { |
5e2b61f7 DM |
1408 | unsigned hash = rt_hash(rt->rt_key_dst, rt->rt_key_src, |
1409 | rt->rt_oif, | |
e84f84f2 | 1410 | rt_genid(dev_net(dst->dev))); |
1da177e4 LT |
1411 | rt_del(hash, rt); |
1412 | ret = NULL; | |
fe6fe792 ED |
1413 | } else if (rt->peer && peer_pmtu_expired(rt->peer)) { |
1414 | dst_metric_set(dst, RTAX_MTU, rt->peer->pmtu_orig); | |
1da177e4 LT |
1415 | } |
1416 | } | |
1417 | return ret; | |
1418 | } | |
1419 | ||
1420 | /* | |
1421 | * Algorithm: | |
1422 | * 1. The first ip_rt_redirect_number redirects are sent | |
1423 | * with exponential backoff, then we stop sending them at all, | |
1424 | * assuming that the host ignores our redirects. | |
1425 | * 2. If we did not see packets requiring redirects | |
1426 | * during ip_rt_redirect_silence, we assume that the host | |
1427 | * forgot redirected route and start to send redirects again. | |
1428 | * | |
1429 | * This algorithm is much cheaper and more intelligent than dumb load limiting | |
1430 | * in icmp.c. | |
1431 | * | |
1432 | * NOTE. Do not forget to inhibit load limiting for redirects (redundant) | |
1433 | * and "frag. need" (breaks PMTU discovery) in icmp.c. | |
1434 | */ | |
1435 | ||
1436 | void ip_rt_send_redirect(struct sk_buff *skb) | |
1437 | { | |
511c3f92 | 1438 | struct rtable *rt = skb_rtable(skb); |
30038fc6 | 1439 | struct in_device *in_dev; |
92d86829 | 1440 | struct inet_peer *peer; |
30038fc6 | 1441 | int log_martians; |
1da177e4 | 1442 | |
30038fc6 | 1443 | rcu_read_lock(); |
d8d1f30b | 1444 | in_dev = __in_dev_get_rcu(rt->dst.dev); |
30038fc6 ED |
1445 | if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) { |
1446 | rcu_read_unlock(); | |
1da177e4 | 1447 | return; |
30038fc6 ED |
1448 | } |
1449 | log_martians = IN_DEV_LOG_MARTIANS(in_dev); | |
1450 | rcu_read_unlock(); | |
1da177e4 | 1451 | |
92d86829 | 1452 | if (!rt->peer) |
a48eff12 | 1453 | rt_bind_peer(rt, rt->rt_dst, 1); |
92d86829 DM |
1454 | peer = rt->peer; |
1455 | if (!peer) { | |
1456 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway); | |
1457 | return; | |
1458 | } | |
1459 | ||
1da177e4 LT |
1460 | /* No redirected packets during ip_rt_redirect_silence; |
1461 | * reset the algorithm. | |
1462 | */ | |
92d86829 DM |
1463 | if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence)) |
1464 | peer->rate_tokens = 0; | |
1da177e4 LT |
1465 | |
1466 | /* Too many ignored redirects; do not send anything | |
d8d1f30b | 1467 | * set dst.rate_last to the last seen redirected packet. |
1da177e4 | 1468 | */ |
92d86829 DM |
1469 | if (peer->rate_tokens >= ip_rt_redirect_number) { |
1470 | peer->rate_last = jiffies; | |
30038fc6 | 1471 | return; |
1da177e4 LT |
1472 | } |
1473 | ||
1474 | /* Check for load limit; set rate_last to the latest sent | |
1475 | * redirect. | |
1476 | */ | |
92d86829 | 1477 | if (peer->rate_tokens == 0 || |
14fb8a76 | 1478 | time_after(jiffies, |
92d86829 DM |
1479 | (peer->rate_last + |
1480 | (ip_rt_redirect_load << peer->rate_tokens)))) { | |
1da177e4 | 1481 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway); |
92d86829 DM |
1482 | peer->rate_last = jiffies; |
1483 | ++peer->rate_tokens; | |
1da177e4 | 1484 | #ifdef CONFIG_IP_ROUTE_VERBOSE |
30038fc6 | 1485 | if (log_martians && |
92d86829 | 1486 | peer->rate_tokens == ip_rt_redirect_number && |
1da177e4 | 1487 | net_ratelimit()) |
673d57e7 | 1488 | printk(KERN_WARNING "host %pI4/if%d ignores redirects for %pI4 to %pI4.\n", |
c5be24ff | 1489 | &ip_hdr(skb)->saddr, rt->rt_iif, |
673d57e7 | 1490 | &rt->rt_dst, &rt->rt_gateway); |
1da177e4 LT |
1491 | #endif |
1492 | } | |
1da177e4 LT |
1493 | } |
1494 | ||
1495 | static int ip_error(struct sk_buff *skb) | |
1496 | { | |
511c3f92 | 1497 | struct rtable *rt = skb_rtable(skb); |
92d86829 | 1498 | struct inet_peer *peer; |
1da177e4 | 1499 | unsigned long now; |
92d86829 | 1500 | bool send; |
1da177e4 LT |
1501 | int code; |
1502 | ||
d8d1f30b | 1503 | switch (rt->dst.error) { |
4500ebf8 JP |
1504 | case EINVAL: |
1505 | default: | |
1506 | goto out; | |
1507 | case EHOSTUNREACH: | |
1508 | code = ICMP_HOST_UNREACH; | |
1509 | break; | |
1510 | case ENETUNREACH: | |
1511 | code = ICMP_NET_UNREACH; | |
1512 | IP_INC_STATS_BH(dev_net(rt->dst.dev), | |
1513 | IPSTATS_MIB_INNOROUTES); | |
1514 | break; | |
1515 | case EACCES: | |
1516 | code = ICMP_PKT_FILTERED; | |
1517 | break; | |
1da177e4 LT |
1518 | } |
1519 | ||
92d86829 | 1520 | if (!rt->peer) |
a48eff12 | 1521 | rt_bind_peer(rt, rt->rt_dst, 1); |
92d86829 DM |
1522 | peer = rt->peer; |
1523 | ||
1524 | send = true; | |
1525 | if (peer) { | |
1526 | now = jiffies; | |
1527 | peer->rate_tokens += now - peer->rate_last; | |
1528 | if (peer->rate_tokens > ip_rt_error_burst) | |
1529 | peer->rate_tokens = ip_rt_error_burst; | |
1530 | peer->rate_last = now; | |
1531 | if (peer->rate_tokens >= ip_rt_error_cost) | |
1532 | peer->rate_tokens -= ip_rt_error_cost; | |
1533 | else | |
1534 | send = false; | |
1da177e4 | 1535 | } |
92d86829 DM |
1536 | if (send) |
1537 | icmp_send(skb, ICMP_DEST_UNREACH, code, 0); | |
1da177e4 LT |
1538 | |
1539 | out: kfree_skb(skb); | |
1540 | return 0; | |
e905a9ed | 1541 | } |
1da177e4 LT |
1542 | |
1543 | /* | |
1544 | * The last two values are not from the RFC but | |
1545 | * are needed for AMPRnet AX.25 paths. | |
1546 | */ | |
1547 | ||
9b5b5cff | 1548 | static const unsigned short mtu_plateau[] = |
1da177e4 LT |
1549 | {32000, 17914, 8166, 4352, 2002, 1492, 576, 296, 216, 128 }; |
1550 | ||
5969f71d | 1551 | static inline unsigned short guess_mtu(unsigned short old_mtu) |
1da177e4 LT |
1552 | { |
1553 | int i; | |
e905a9ed | 1554 | |
1da177e4 LT |
1555 | for (i = 0; i < ARRAY_SIZE(mtu_plateau); i++) |
1556 | if (old_mtu > mtu_plateau[i]) | |
1557 | return mtu_plateau[i]; | |
1558 | return 68; | |
1559 | } | |
1560 | ||
b71d1d42 | 1561 | unsigned short ip_rt_frag_needed(struct net *net, const struct iphdr *iph, |
0010e465 TT |
1562 | unsigned short new_mtu, |
1563 | struct net_device *dev) | |
1da177e4 | 1564 | { |
1da177e4 | 1565 | unsigned short old_mtu = ntohs(iph->tot_len); |
1da177e4 | 1566 | unsigned short est_mtu = 0; |
2c8cec5c | 1567 | struct inet_peer *peer; |
1da177e4 | 1568 | |
2c8cec5c DM |
1569 | peer = inet_getpeer_v4(iph->daddr, 1); |
1570 | if (peer) { | |
1571 | unsigned short mtu = new_mtu; | |
1da177e4 | 1572 | |
2c8cec5c DM |
1573 | if (new_mtu < 68 || new_mtu >= old_mtu) { |
1574 | /* BSD 4.2 derived systems incorrectly adjust | |
1575 | * tot_len by the IP header length, and report | |
1576 | * a zero MTU in the ICMP message. | |
1577 | */ | |
1578 | if (mtu == 0 && | |
1579 | old_mtu >= 68 + (iph->ihl << 2)) | |
1580 | old_mtu -= iph->ihl << 2; | |
1581 | mtu = guess_mtu(old_mtu); | |
1582 | } | |
0010e465 | 1583 | |
2c8cec5c DM |
1584 | if (mtu < ip_rt_min_pmtu) |
1585 | mtu = ip_rt_min_pmtu; | |
1586 | if (!peer->pmtu_expires || mtu < peer->pmtu_learned) { | |
46af3180 HS |
1587 | unsigned long pmtu_expires; |
1588 | ||
1589 | pmtu_expires = jiffies + ip_rt_mtu_expires; | |
1590 | if (!pmtu_expires) | |
1591 | pmtu_expires = 1UL; | |
1592 | ||
2c8cec5c DM |
1593 | est_mtu = mtu; |
1594 | peer->pmtu_learned = mtu; | |
46af3180 | 1595 | peer->pmtu_expires = pmtu_expires; |
59445b6b | 1596 | atomic_inc(&__rt_peer_genid); |
2c8cec5c | 1597 | } |
1da177e4 | 1598 | |
2c8cec5c | 1599 | inet_putpeer(peer); |
1da177e4 LT |
1600 | } |
1601 | return est_mtu ? : new_mtu; | |
1602 | } | |
1603 | ||
2c8cec5c DM |
1604 | static void check_peer_pmtu(struct dst_entry *dst, struct inet_peer *peer) |
1605 | { | |
fe6fe792 | 1606 | unsigned long expires = ACCESS_ONCE(peer->pmtu_expires); |
2c8cec5c | 1607 | |
fe6fe792 ED |
1608 | if (!expires) |
1609 | return; | |
46af3180 | 1610 | if (time_before(jiffies, expires)) { |
2c8cec5c DM |
1611 | u32 orig_dst_mtu = dst_mtu(dst); |
1612 | if (peer->pmtu_learned < orig_dst_mtu) { | |
1613 | if (!peer->pmtu_orig) | |
1614 | peer->pmtu_orig = dst_metric_raw(dst, RTAX_MTU); | |
1615 | dst_metric_set(dst, RTAX_MTU, peer->pmtu_learned); | |
1616 | } | |
1617 | } else if (cmpxchg(&peer->pmtu_expires, expires, 0) == expires) | |
1618 | dst_metric_set(dst, RTAX_MTU, peer->pmtu_orig); | |
1619 | } | |
1620 | ||
1da177e4 LT |
1621 | static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu) |
1622 | { | |
2c8cec5c DM |
1623 | struct rtable *rt = (struct rtable *) dst; |
1624 | struct inet_peer *peer; | |
1625 | ||
1626 | dst_confirm(dst); | |
1627 | ||
1628 | if (!rt->peer) | |
a48eff12 | 1629 | rt_bind_peer(rt, rt->rt_dst, 1); |
2c8cec5c DM |
1630 | peer = rt->peer; |
1631 | if (peer) { | |
fe6fe792 ED |
1632 | unsigned long pmtu_expires = ACCESS_ONCE(peer->pmtu_expires); |
1633 | ||
2c8cec5c | 1634 | if (mtu < ip_rt_min_pmtu) |
1da177e4 | 1635 | mtu = ip_rt_min_pmtu; |
fe6fe792 | 1636 | if (!pmtu_expires || mtu < peer->pmtu_learned) { |
46af3180 HS |
1637 | |
1638 | pmtu_expires = jiffies + ip_rt_mtu_expires; | |
1639 | if (!pmtu_expires) | |
1640 | pmtu_expires = 1UL; | |
1641 | ||
2c8cec5c | 1642 | peer->pmtu_learned = mtu; |
46af3180 | 1643 | peer->pmtu_expires = pmtu_expires; |
2c8cec5c DM |
1644 | |
1645 | atomic_inc(&__rt_peer_genid); | |
1646 | rt->rt_peer_genid = rt_peer_genid(); | |
1da177e4 | 1647 | } |
46af3180 | 1648 | check_peer_pmtu(dst, peer); |
1da177e4 LT |
1649 | } |
1650 | } | |
1651 | ||
f39925db DM |
1652 | static int check_peer_redir(struct dst_entry *dst, struct inet_peer *peer) |
1653 | { | |
1654 | struct rtable *rt = (struct rtable *) dst; | |
1655 | __be32 orig_gw = rt->rt_gateway; | |
f2c31e32 | 1656 | struct neighbour *n, *old_n; |
f39925db DM |
1657 | |
1658 | dst_confirm(&rt->dst); | |
1659 | ||
f39925db | 1660 | rt->rt_gateway = peer->redirect_learned.a4; |
f2c31e32 ED |
1661 | |
1662 | n = ipv4_neigh_lookup(&rt->dst, &rt->rt_gateway); | |
1663 | if (IS_ERR(n)) | |
1664 | return PTR_ERR(n); | |
1665 | old_n = xchg(&rt->dst._neighbour, n); | |
1666 | if (old_n) | |
1667 | neigh_release(old_n); | |
69cce1d1 DM |
1668 | if (!n || !(n->nud_state & NUD_VALID)) { |
1669 | if (n) | |
1670 | neigh_event_send(n, NULL); | |
f39925db DM |
1671 | rt->rt_gateway = orig_gw; |
1672 | return -EAGAIN; | |
1673 | } else { | |
1674 | rt->rt_flags |= RTCF_REDIRECTED; | |
69cce1d1 | 1675 | call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, n); |
f39925db DM |
1676 | } |
1677 | return 0; | |
1678 | } | |
1679 | ||
1da177e4 LT |
1680 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie) |
1681 | { | |
6431cbc2 DM |
1682 | struct rtable *rt = (struct rtable *) dst; |
1683 | ||
1684 | if (rt_is_expired(rt)) | |
d11a4dc1 | 1685 | return NULL; |
6431cbc2 | 1686 | if (rt->rt_peer_genid != rt_peer_genid()) { |
2c8cec5c DM |
1687 | struct inet_peer *peer; |
1688 | ||
6431cbc2 | 1689 | if (!rt->peer) |
a48eff12 | 1690 | rt_bind_peer(rt, rt->rt_dst, 0); |
6431cbc2 | 1691 | |
2c8cec5c | 1692 | peer = rt->peer; |
fe6fe792 | 1693 | if (peer) { |
2c8cec5c DM |
1694 | check_peer_pmtu(dst, peer); |
1695 | ||
fe6fe792 ED |
1696 | if (peer->redirect_learned.a4 && |
1697 | peer->redirect_learned.a4 != rt->rt_gateway) { | |
1698 | if (check_peer_redir(dst, peer)) | |
1699 | return NULL; | |
1700 | } | |
f39925db DM |
1701 | } |
1702 | ||
6431cbc2 DM |
1703 | rt->rt_peer_genid = rt_peer_genid(); |
1704 | } | |
d11a4dc1 | 1705 | return dst; |
1da177e4 LT |
1706 | } |
1707 | ||
1708 | static void ipv4_dst_destroy(struct dst_entry *dst) | |
1709 | { | |
1710 | struct rtable *rt = (struct rtable *) dst; | |
1711 | struct inet_peer *peer = rt->peer; | |
1da177e4 | 1712 | |
62fa8a84 DM |
1713 | if (rt->fi) { |
1714 | fib_info_put(rt->fi); | |
1715 | rt->fi = NULL; | |
1716 | } | |
1da177e4 LT |
1717 | if (peer) { |
1718 | rt->peer = NULL; | |
1719 | inet_putpeer(peer); | |
1720 | } | |
1da177e4 LT |
1721 | } |
1722 | ||
1da177e4 LT |
1723 | |
1724 | static void ipv4_link_failure(struct sk_buff *skb) | |
1725 | { | |
1726 | struct rtable *rt; | |
1727 | ||
1728 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0); | |
1729 | ||
511c3f92 | 1730 | rt = skb_rtable(skb); |
fe6fe792 ED |
1731 | if (rt && rt->peer && peer_pmtu_cleaned(rt->peer)) |
1732 | dst_metric_set(&rt->dst, RTAX_MTU, rt->peer->pmtu_orig); | |
1da177e4 LT |
1733 | } |
1734 | ||
1735 | static int ip_rt_bug(struct sk_buff *skb) | |
1736 | { | |
673d57e7 HH |
1737 | printk(KERN_DEBUG "ip_rt_bug: %pI4 -> %pI4, %s\n", |
1738 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
1da177e4 LT |
1739 | skb->dev ? skb->dev->name : "?"); |
1740 | kfree_skb(skb); | |
c378a9c0 | 1741 | WARN_ON(1); |
1da177e4 LT |
1742 | return 0; |
1743 | } | |
1744 | ||
1745 | /* | |
1746 | We do not cache source address of outgoing interface, | |
1747 | because it is used only by IP RR, TS and SRR options, | |
1748 | so that it out of fast path. | |
1749 | ||
1750 | BTW remember: "addr" is allowed to be not aligned | |
1751 | in IP options! | |
1752 | */ | |
1753 | ||
8e36360a | 1754 | void ip_rt_get_source(u8 *addr, struct sk_buff *skb, struct rtable *rt) |
1da177e4 | 1755 | { |
a61ced5d | 1756 | __be32 src; |
1da177e4 | 1757 | |
c7537967 | 1758 | if (rt_is_output_route(rt)) |
c5be24ff | 1759 | src = ip_hdr(skb)->saddr; |
ebc0ffae | 1760 | else { |
8e36360a DM |
1761 | struct fib_result res; |
1762 | struct flowi4 fl4; | |
1763 | struct iphdr *iph; | |
1764 | ||
1765 | iph = ip_hdr(skb); | |
1766 | ||
1767 | memset(&fl4, 0, sizeof(fl4)); | |
1768 | fl4.daddr = iph->daddr; | |
1769 | fl4.saddr = iph->saddr; | |
b0fe4a31 | 1770 | fl4.flowi4_tos = RT_TOS(iph->tos); |
8e36360a DM |
1771 | fl4.flowi4_oif = rt->dst.dev->ifindex; |
1772 | fl4.flowi4_iif = skb->dev->ifindex; | |
1773 | fl4.flowi4_mark = skb->mark; | |
5e2b61f7 | 1774 | |
ebc0ffae | 1775 | rcu_read_lock(); |
68a5e3dd | 1776 | if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res) == 0) |
436c3b66 | 1777 | src = FIB_RES_PREFSRC(dev_net(rt->dst.dev), res); |
ebc0ffae ED |
1778 | else |
1779 | src = inet_select_addr(rt->dst.dev, rt->rt_gateway, | |
1da177e4 | 1780 | RT_SCOPE_UNIVERSE); |
ebc0ffae ED |
1781 | rcu_read_unlock(); |
1782 | } | |
1da177e4 LT |
1783 | memcpy(addr, &src, 4); |
1784 | } | |
1785 | ||
c7066f70 | 1786 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 LT |
1787 | static void set_class_tag(struct rtable *rt, u32 tag) |
1788 | { | |
d8d1f30b CG |
1789 | if (!(rt->dst.tclassid & 0xFFFF)) |
1790 | rt->dst.tclassid |= tag & 0xFFFF; | |
1791 | if (!(rt->dst.tclassid & 0xFFFF0000)) | |
1792 | rt->dst.tclassid |= tag & 0xFFFF0000; | |
1da177e4 LT |
1793 | } |
1794 | #endif | |
1795 | ||
0dbaee3b DM |
1796 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst) |
1797 | { | |
1798 | unsigned int advmss = dst_metric_raw(dst, RTAX_ADVMSS); | |
1799 | ||
1800 | if (advmss == 0) { | |
1801 | advmss = max_t(unsigned int, dst->dev->mtu - 40, | |
1802 | ip_rt_min_advmss); | |
1803 | if (advmss > 65535 - 40) | |
1804 | advmss = 65535 - 40; | |
1805 | } | |
1806 | return advmss; | |
1807 | } | |
1808 | ||
d33e4553 DM |
1809 | static unsigned int ipv4_default_mtu(const struct dst_entry *dst) |
1810 | { | |
1811 | unsigned int mtu = dst->dev->mtu; | |
1812 | ||
1813 | if (unlikely(dst_metric_locked(dst, RTAX_MTU))) { | |
1814 | const struct rtable *rt = (const struct rtable *) dst; | |
1815 | ||
1816 | if (rt->rt_gateway != rt->rt_dst && mtu > 576) | |
1817 | mtu = 576; | |
1818 | } | |
1819 | ||
1820 | if (mtu > IP_MAX_MTU) | |
1821 | mtu = IP_MAX_MTU; | |
1822 | ||
1823 | return mtu; | |
1824 | } | |
1825 | ||
813b3b5d | 1826 | static void rt_init_metrics(struct rtable *rt, const struct flowi4 *fl4, |
5e2b61f7 | 1827 | struct fib_info *fi) |
a4daad6b | 1828 | { |
0131ba45 DM |
1829 | struct inet_peer *peer; |
1830 | int create = 0; | |
a4daad6b | 1831 | |
0131ba45 DM |
1832 | /* If a peer entry exists for this destination, we must hook |
1833 | * it up in order to get at cached metrics. | |
1834 | */ | |
813b3b5d | 1835 | if (fl4 && (fl4->flowi4_flags & FLOWI_FLAG_PRECOW_METRICS)) |
0131ba45 DM |
1836 | create = 1; |
1837 | ||
3c0afdca | 1838 | rt->peer = peer = inet_getpeer_v4(rt->rt_dst, create); |
0131ba45 | 1839 | if (peer) { |
3c0afdca | 1840 | rt->rt_peer_genid = rt_peer_genid(); |
a4daad6b DM |
1841 | if (inet_metrics_new(peer)) |
1842 | memcpy(peer->metrics, fi->fib_metrics, | |
1843 | sizeof(u32) * RTAX_MAX); | |
1844 | dst_init_metrics(&rt->dst, peer->metrics, false); | |
2c8cec5c | 1845 | |
fe6fe792 | 1846 | check_peer_pmtu(&rt->dst, peer); |
f39925db DM |
1847 | if (peer->redirect_learned.a4 && |
1848 | peer->redirect_learned.a4 != rt->rt_gateway) { | |
1849 | rt->rt_gateway = peer->redirect_learned.a4; | |
1850 | rt->rt_flags |= RTCF_REDIRECTED; | |
1851 | } | |
0131ba45 DM |
1852 | } else { |
1853 | if (fi->fib_metrics != (u32 *) dst_default_metrics) { | |
1854 | rt->fi = fi; | |
1855 | atomic_inc(&fi->fib_clntref); | |
1856 | } | |
1857 | dst_init_metrics(&rt->dst, fi->fib_metrics, true); | |
a4daad6b DM |
1858 | } |
1859 | } | |
1860 | ||
813b3b5d | 1861 | static void rt_set_nexthop(struct rtable *rt, const struct flowi4 *fl4, |
5e2b61f7 | 1862 | const struct fib_result *res, |
982721f3 | 1863 | struct fib_info *fi, u16 type, u32 itag) |
1da177e4 | 1864 | { |
defb3519 | 1865 | struct dst_entry *dst = &rt->dst; |
1da177e4 LT |
1866 | |
1867 | if (fi) { | |
1868 | if (FIB_RES_GW(*res) && | |
1869 | FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) | |
1870 | rt->rt_gateway = FIB_RES_GW(*res); | |
813b3b5d | 1871 | rt_init_metrics(rt, fl4, fi); |
c7066f70 | 1872 | #ifdef CONFIG_IP_ROUTE_CLASSID |
defb3519 | 1873 | dst->tclassid = FIB_RES_NH(*res).nh_tclassid; |
1da177e4 | 1874 | #endif |
d33e4553 | 1875 | } |
defb3519 | 1876 | |
defb3519 DM |
1877 | if (dst_mtu(dst) > IP_MAX_MTU) |
1878 | dst_metric_set(dst, RTAX_MTU, IP_MAX_MTU); | |
0dbaee3b | 1879 | if (dst_metric_raw(dst, RTAX_ADVMSS) > 65535 - 40) |
defb3519 | 1880 | dst_metric_set(dst, RTAX_ADVMSS, 65535 - 40); |
1da177e4 | 1881 | |
c7066f70 | 1882 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 LT |
1883 | #ifdef CONFIG_IP_MULTIPLE_TABLES |
1884 | set_class_tag(rt, fib_rules_tclass(res)); | |
1885 | #endif | |
1886 | set_class_tag(rt, itag); | |
1887 | #endif | |
1da177e4 LT |
1888 | } |
1889 | ||
5c1e6aa3 DM |
1890 | static struct rtable *rt_dst_alloc(struct net_device *dev, |
1891 | bool nopolicy, bool noxfrm) | |
0c4dcd58 | 1892 | { |
5c1e6aa3 DM |
1893 | return dst_alloc(&ipv4_dst_ops, dev, 1, -1, |
1894 | DST_HOST | | |
1895 | (nopolicy ? DST_NOPOLICY : 0) | | |
1896 | (noxfrm ? DST_NOXFRM : 0)); | |
0c4dcd58 DM |
1897 | } |
1898 | ||
96d36220 | 1899 | /* called in rcu_read_lock() section */ |
9e12bb22 | 1900 | static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
1901 | u8 tos, struct net_device *dev, int our) |
1902 | { | |
96d36220 | 1903 | unsigned int hash; |
1da177e4 | 1904 | struct rtable *rth; |
a61ced5d | 1905 | __be32 spec_dst; |
96d36220 | 1906 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1907 | u32 itag = 0; |
b5f7e755 | 1908 | int err; |
1da177e4 LT |
1909 | |
1910 | /* Primary sanity checks. */ | |
1911 | ||
1912 | if (in_dev == NULL) | |
1913 | return -EINVAL; | |
1914 | ||
1e637c74 | 1915 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
f97c1e0c | 1916 | ipv4_is_loopback(saddr) || skb->protocol != htons(ETH_P_IP)) |
1da177e4 LT |
1917 | goto e_inval; |
1918 | ||
f97c1e0c JP |
1919 | if (ipv4_is_zeronet(saddr)) { |
1920 | if (!ipv4_is_local_multicast(daddr)) | |
1da177e4 LT |
1921 | goto e_inval; |
1922 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); | |
b5f7e755 | 1923 | } else { |
5c04c819 MS |
1924 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, &spec_dst, |
1925 | &itag); | |
b5f7e755 ED |
1926 | if (err < 0) |
1927 | goto e_err; | |
1928 | } | |
5c1e6aa3 DM |
1929 | rth = rt_dst_alloc(init_net.loopback_dev, |
1930 | IN_DEV_CONF_GET(in_dev, NOPOLICY), false); | |
1da177e4 LT |
1931 | if (!rth) |
1932 | goto e_nobufs; | |
1933 | ||
cf911662 DM |
1934 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1935 | rth->dst.tclassid = itag; | |
1936 | #endif | |
d8d1f30b | 1937 | rth->dst.output = ip_rt_bug; |
1da177e4 | 1938 | |
5e2b61f7 | 1939 | rth->rt_key_dst = daddr; |
5e2b61f7 | 1940 | rth->rt_key_src = saddr; |
cf911662 DM |
1941 | rth->rt_genid = rt_genid(dev_net(dev)); |
1942 | rth->rt_flags = RTCF_MULTICAST; | |
1943 | rth->rt_type = RTN_MULTICAST; | |
475949d8 | 1944 | rth->rt_key_tos = tos; |
cf911662 | 1945 | rth->rt_dst = daddr; |
1da177e4 | 1946 | rth->rt_src = saddr; |
1b86a58f | 1947 | rth->rt_route_iif = dev->ifindex; |
5e2b61f7 | 1948 | rth->rt_iif = dev->ifindex; |
5e2b61f7 | 1949 | rth->rt_oif = 0; |
cf911662 | 1950 | rth->rt_mark = skb->mark; |
1da177e4 LT |
1951 | rth->rt_gateway = daddr; |
1952 | rth->rt_spec_dst= spec_dst; | |
cf911662 DM |
1953 | rth->rt_peer_genid = 0; |
1954 | rth->peer = NULL; | |
1955 | rth->fi = NULL; | |
1da177e4 | 1956 | if (our) { |
d8d1f30b | 1957 | rth->dst.input= ip_local_deliver; |
1da177e4 LT |
1958 | rth->rt_flags |= RTCF_LOCAL; |
1959 | } | |
1960 | ||
1961 | #ifdef CONFIG_IP_MROUTE | |
f97c1e0c | 1962 | if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev)) |
d8d1f30b | 1963 | rth->dst.input = ip_mr_input; |
1da177e4 LT |
1964 | #endif |
1965 | RT_CACHE_STAT_INC(in_slow_mc); | |
1966 | ||
e84f84f2 | 1967 | hash = rt_hash(daddr, saddr, dev->ifindex, rt_genid(dev_net(dev))); |
b23dd4fe | 1968 | rth = rt_intern_hash(hash, rth, skb, dev->ifindex); |
9aa3c94c | 1969 | return IS_ERR(rth) ? PTR_ERR(rth) : 0; |
1da177e4 LT |
1970 | |
1971 | e_nobufs: | |
1da177e4 | 1972 | return -ENOBUFS; |
1da177e4 | 1973 | e_inval: |
96d36220 | 1974 | return -EINVAL; |
b5f7e755 | 1975 | e_err: |
b5f7e755 | 1976 | return err; |
1da177e4 LT |
1977 | } |
1978 | ||
1979 | ||
1980 | static void ip_handle_martian_source(struct net_device *dev, | |
1981 | struct in_device *in_dev, | |
1982 | struct sk_buff *skb, | |
9e12bb22 AV |
1983 | __be32 daddr, |
1984 | __be32 saddr) | |
1da177e4 LT |
1985 | { |
1986 | RT_CACHE_STAT_INC(in_martian_src); | |
1987 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1988 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) { | |
1989 | /* | |
1990 | * RFC1812 recommendation, if source is martian, | |
1991 | * the only hint is MAC header. | |
1992 | */ | |
673d57e7 HH |
1993 | printk(KERN_WARNING "martian source %pI4 from %pI4, on dev %s\n", |
1994 | &daddr, &saddr, dev->name); | |
98e399f8 | 1995 | if (dev->hard_header_len && skb_mac_header_was_set(skb)) { |
1da177e4 | 1996 | int i; |
98e399f8 | 1997 | const unsigned char *p = skb_mac_header(skb); |
1da177e4 LT |
1998 | printk(KERN_WARNING "ll header: "); |
1999 | for (i = 0; i < dev->hard_header_len; i++, p++) { | |
2000 | printk("%02x", *p); | |
2001 | if (i < (dev->hard_header_len - 1)) | |
2002 | printk(":"); | |
2003 | } | |
2004 | printk("\n"); | |
2005 | } | |
2006 | } | |
2007 | #endif | |
2008 | } | |
2009 | ||
47360228 | 2010 | /* called in rcu_read_lock() section */ |
5969f71d | 2011 | static int __mkroute_input(struct sk_buff *skb, |
982721f3 | 2012 | const struct fib_result *res, |
5969f71d SH |
2013 | struct in_device *in_dev, |
2014 | __be32 daddr, __be32 saddr, u32 tos, | |
2015 | struct rtable **result) | |
1da177e4 | 2016 | { |
1da177e4 LT |
2017 | struct rtable *rth; |
2018 | int err; | |
2019 | struct in_device *out_dev; | |
47360228 | 2020 | unsigned int flags = 0; |
d9c9df8c AV |
2021 | __be32 spec_dst; |
2022 | u32 itag; | |
1da177e4 LT |
2023 | |
2024 | /* get a working reference to the output device */ | |
47360228 | 2025 | out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res)); |
1da177e4 LT |
2026 | if (out_dev == NULL) { |
2027 | if (net_ratelimit()) | |
2028 | printk(KERN_CRIT "Bug in ip_route_input" \ | |
2029 | "_slow(). Please, report\n"); | |
2030 | return -EINVAL; | |
2031 | } | |
2032 | ||
2033 | ||
5c04c819 MS |
2034 | err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res), |
2035 | in_dev->dev, &spec_dst, &itag); | |
1da177e4 | 2036 | if (err < 0) { |
e905a9ed | 2037 | ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr, |
1da177e4 | 2038 | saddr); |
e905a9ed | 2039 | |
1da177e4 LT |
2040 | goto cleanup; |
2041 | } | |
2042 | ||
2043 | if (err) | |
2044 | flags |= RTCF_DIRECTSRC; | |
2045 | ||
51b77cae | 2046 | if (out_dev == in_dev && err && |
1da177e4 LT |
2047 | (IN_DEV_SHARED_MEDIA(out_dev) || |
2048 | inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) | |
2049 | flags |= RTCF_DOREDIRECT; | |
2050 | ||
2051 | if (skb->protocol != htons(ETH_P_IP)) { | |
2052 | /* Not IP (i.e. ARP). Do not create route, if it is | |
2053 | * invalid for proxy arp. DNAT routes are always valid. | |
65324144 JDB |
2054 | * |
2055 | * Proxy arp feature have been extended to allow, ARP | |
2056 | * replies back to the same interface, to support | |
2057 | * Private VLAN switch technologies. See arp.c. | |
1da177e4 | 2058 | */ |
65324144 JDB |
2059 | if (out_dev == in_dev && |
2060 | IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) { | |
1da177e4 LT |
2061 | err = -EINVAL; |
2062 | goto cleanup; | |
2063 | } | |
2064 | } | |
2065 | ||
5c1e6aa3 DM |
2066 | rth = rt_dst_alloc(out_dev->dev, |
2067 | IN_DEV_CONF_GET(in_dev, NOPOLICY), | |
0c4dcd58 | 2068 | IN_DEV_CONF_GET(out_dev, NOXFRM)); |
1da177e4 LT |
2069 | if (!rth) { |
2070 | err = -ENOBUFS; | |
2071 | goto cleanup; | |
2072 | } | |
2073 | ||
5e2b61f7 | 2074 | rth->rt_key_dst = daddr; |
5e2b61f7 | 2075 | rth->rt_key_src = saddr; |
cf911662 DM |
2076 | rth->rt_genid = rt_genid(dev_net(rth->dst.dev)); |
2077 | rth->rt_flags = flags; | |
2078 | rth->rt_type = res->type; | |
475949d8 | 2079 | rth->rt_key_tos = tos; |
cf911662 | 2080 | rth->rt_dst = daddr; |
1da177e4 | 2081 | rth->rt_src = saddr; |
1b86a58f | 2082 | rth->rt_route_iif = in_dev->dev->ifindex; |
5e2b61f7 | 2083 | rth->rt_iif = in_dev->dev->ifindex; |
5e2b61f7 | 2084 | rth->rt_oif = 0; |
cf911662 DM |
2085 | rth->rt_mark = skb->mark; |
2086 | rth->rt_gateway = daddr; | |
1da177e4 | 2087 | rth->rt_spec_dst= spec_dst; |
cf911662 DM |
2088 | rth->rt_peer_genid = 0; |
2089 | rth->peer = NULL; | |
2090 | rth->fi = NULL; | |
1da177e4 | 2091 | |
d8d1f30b CG |
2092 | rth->dst.input = ip_forward; |
2093 | rth->dst.output = ip_output; | |
1da177e4 | 2094 | |
5e2b61f7 | 2095 | rt_set_nexthop(rth, NULL, res, res->fi, res->type, itag); |
1da177e4 | 2096 | |
1da177e4 LT |
2097 | *result = rth; |
2098 | err = 0; | |
2099 | cleanup: | |
1da177e4 | 2100 | return err; |
e905a9ed | 2101 | } |
1da177e4 | 2102 | |
5969f71d SH |
2103 | static int ip_mkroute_input(struct sk_buff *skb, |
2104 | struct fib_result *res, | |
68a5e3dd | 2105 | const struct flowi4 *fl4, |
5969f71d SH |
2106 | struct in_device *in_dev, |
2107 | __be32 daddr, __be32 saddr, u32 tos) | |
1da177e4 | 2108 | { |
7abaa27c | 2109 | struct rtable* rth = NULL; |
1da177e4 LT |
2110 | int err; |
2111 | unsigned hash; | |
2112 | ||
2113 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
ff3fccb3 | 2114 | if (res->fi && res->fi->fib_nhs > 1) |
1b7fe593 | 2115 | fib_select_multipath(res); |
1da177e4 LT |
2116 | #endif |
2117 | ||
2118 | /* create a routing cache entry */ | |
2119 | err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth); | |
2120 | if (err) | |
2121 | return err; | |
1da177e4 LT |
2122 | |
2123 | /* put it into the cache */ | |
68a5e3dd | 2124 | hash = rt_hash(daddr, saddr, fl4->flowi4_iif, |
d8d1f30b | 2125 | rt_genid(dev_net(rth->dst.dev))); |
68a5e3dd | 2126 | rth = rt_intern_hash(hash, rth, skb, fl4->flowi4_iif); |
b23dd4fe DM |
2127 | if (IS_ERR(rth)) |
2128 | return PTR_ERR(rth); | |
2129 | return 0; | |
1da177e4 LT |
2130 | } |
2131 | ||
1da177e4 LT |
2132 | /* |
2133 | * NOTE. We drop all the packets that has local source | |
2134 | * addresses, because every properly looped back packet | |
2135 | * must have correct destination already attached by output routine. | |
2136 | * | |
2137 | * Such approach solves two big problems: | |
2138 | * 1. Not simplex devices are handled properly. | |
2139 | * 2. IP spoofing attempts are filtered with 100% of guarantee. | |
ebc0ffae | 2140 | * called with rcu_read_lock() |
1da177e4 LT |
2141 | */ |
2142 | ||
9e12bb22 | 2143 | static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
2144 | u8 tos, struct net_device *dev) |
2145 | { | |
2146 | struct fib_result res; | |
96d36220 | 2147 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
68a5e3dd | 2148 | struct flowi4 fl4; |
1da177e4 LT |
2149 | unsigned flags = 0; |
2150 | u32 itag = 0; | |
2151 | struct rtable * rth; | |
2152 | unsigned hash; | |
9e12bb22 | 2153 | __be32 spec_dst; |
1da177e4 | 2154 | int err = -EINVAL; |
c346dca1 | 2155 | struct net * net = dev_net(dev); |
1da177e4 LT |
2156 | |
2157 | /* IP on this device is disabled. */ | |
2158 | ||
2159 | if (!in_dev) | |
2160 | goto out; | |
2161 | ||
2162 | /* Check for the most weird martians, which can be not detected | |
2163 | by fib_lookup. | |
2164 | */ | |
2165 | ||
1e637c74 | 2166 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
f97c1e0c | 2167 | ipv4_is_loopback(saddr)) |
1da177e4 LT |
2168 | goto martian_source; |
2169 | ||
27a954bd | 2170 | if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0)) |
1da177e4 LT |
2171 | goto brd_input; |
2172 | ||
2173 | /* Accept zero addresses only to limited broadcast; | |
2174 | * I even do not know to fix it or not. Waiting for complains :-) | |
2175 | */ | |
f97c1e0c | 2176 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
2177 | goto martian_source; |
2178 | ||
27a954bd | 2179 | if (ipv4_is_zeronet(daddr) || ipv4_is_loopback(daddr)) |
1da177e4 LT |
2180 | goto martian_destination; |
2181 | ||
2182 | /* | |
2183 | * Now we are ready to route packet. | |
2184 | */ | |
68a5e3dd DM |
2185 | fl4.flowi4_oif = 0; |
2186 | fl4.flowi4_iif = dev->ifindex; | |
2187 | fl4.flowi4_mark = skb->mark; | |
2188 | fl4.flowi4_tos = tos; | |
2189 | fl4.flowi4_scope = RT_SCOPE_UNIVERSE; | |
2190 | fl4.daddr = daddr; | |
2191 | fl4.saddr = saddr; | |
2192 | err = fib_lookup(net, &fl4, &res); | |
ebc0ffae | 2193 | if (err != 0) { |
1da177e4 | 2194 | if (!IN_DEV_FORWARD(in_dev)) |
2c2910a4 | 2195 | goto e_hostunreach; |
1da177e4 LT |
2196 | goto no_route; |
2197 | } | |
1da177e4 LT |
2198 | |
2199 | RT_CACHE_STAT_INC(in_slow_tot); | |
2200 | ||
2201 | if (res.type == RTN_BROADCAST) | |
2202 | goto brd_input; | |
2203 | ||
2204 | if (res.type == RTN_LOCAL) { | |
5c04c819 | 2205 | err = fib_validate_source(skb, saddr, daddr, tos, |
ebc0ffae | 2206 | net->loopback_dev->ifindex, |
5c04c819 | 2207 | dev, &spec_dst, &itag); |
b5f7e755 ED |
2208 | if (err < 0) |
2209 | goto martian_source_keep_err; | |
2210 | if (err) | |
1da177e4 LT |
2211 | flags |= RTCF_DIRECTSRC; |
2212 | spec_dst = daddr; | |
2213 | goto local_input; | |
2214 | } | |
2215 | ||
2216 | if (!IN_DEV_FORWARD(in_dev)) | |
2c2910a4 | 2217 | goto e_hostunreach; |
1da177e4 LT |
2218 | if (res.type != RTN_UNICAST) |
2219 | goto martian_destination; | |
2220 | ||
68a5e3dd | 2221 | err = ip_mkroute_input(skb, &res, &fl4, in_dev, daddr, saddr, tos); |
1da177e4 LT |
2222 | out: return err; |
2223 | ||
2224 | brd_input: | |
2225 | if (skb->protocol != htons(ETH_P_IP)) | |
2226 | goto e_inval; | |
2227 | ||
f97c1e0c | 2228 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
2229 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
2230 | else { | |
5c04c819 MS |
2231 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, &spec_dst, |
2232 | &itag); | |
1da177e4 | 2233 | if (err < 0) |
b5f7e755 | 2234 | goto martian_source_keep_err; |
1da177e4 LT |
2235 | if (err) |
2236 | flags |= RTCF_DIRECTSRC; | |
2237 | } | |
2238 | flags |= RTCF_BROADCAST; | |
2239 | res.type = RTN_BROADCAST; | |
2240 | RT_CACHE_STAT_INC(in_brd); | |
2241 | ||
2242 | local_input: | |
5c1e6aa3 DM |
2243 | rth = rt_dst_alloc(net->loopback_dev, |
2244 | IN_DEV_CONF_GET(in_dev, NOPOLICY), false); | |
1da177e4 LT |
2245 | if (!rth) |
2246 | goto e_nobufs; | |
2247 | ||
cf911662 | 2248 | rth->dst.input= ip_local_deliver; |
d8d1f30b | 2249 | rth->dst.output= ip_rt_bug; |
cf911662 DM |
2250 | #ifdef CONFIG_IP_ROUTE_CLASSID |
2251 | rth->dst.tclassid = itag; | |
2252 | #endif | |
1da177e4 | 2253 | |
5e2b61f7 | 2254 | rth->rt_key_dst = daddr; |
5e2b61f7 | 2255 | rth->rt_key_src = saddr; |
cf911662 DM |
2256 | rth->rt_genid = rt_genid(net); |
2257 | rth->rt_flags = flags|RTCF_LOCAL; | |
2258 | rth->rt_type = res.type; | |
475949d8 | 2259 | rth->rt_key_tos = tos; |
cf911662 | 2260 | rth->rt_dst = daddr; |
1da177e4 | 2261 | rth->rt_src = saddr; |
c7066f70 | 2262 | #ifdef CONFIG_IP_ROUTE_CLASSID |
d8d1f30b | 2263 | rth->dst.tclassid = itag; |
1da177e4 | 2264 | #endif |
1b86a58f | 2265 | rth->rt_route_iif = dev->ifindex; |
5e2b61f7 | 2266 | rth->rt_iif = dev->ifindex; |
cf911662 DM |
2267 | rth->rt_oif = 0; |
2268 | rth->rt_mark = skb->mark; | |
1da177e4 LT |
2269 | rth->rt_gateway = daddr; |
2270 | rth->rt_spec_dst= spec_dst; | |
cf911662 DM |
2271 | rth->rt_peer_genid = 0; |
2272 | rth->peer = NULL; | |
2273 | rth->fi = NULL; | |
1da177e4 | 2274 | if (res.type == RTN_UNREACHABLE) { |
d8d1f30b CG |
2275 | rth->dst.input= ip_error; |
2276 | rth->dst.error= -err; | |
1da177e4 LT |
2277 | rth->rt_flags &= ~RTCF_LOCAL; |
2278 | } | |
68a5e3dd DM |
2279 | hash = rt_hash(daddr, saddr, fl4.flowi4_iif, rt_genid(net)); |
2280 | rth = rt_intern_hash(hash, rth, skb, fl4.flowi4_iif); | |
b23dd4fe DM |
2281 | err = 0; |
2282 | if (IS_ERR(rth)) | |
2283 | err = PTR_ERR(rth); | |
ebc0ffae | 2284 | goto out; |
1da177e4 LT |
2285 | |
2286 | no_route: | |
2287 | RT_CACHE_STAT_INC(in_no_route); | |
2288 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE); | |
2289 | res.type = RTN_UNREACHABLE; | |
7f53878d MC |
2290 | if (err == -ESRCH) |
2291 | err = -ENETUNREACH; | |
1da177e4 LT |
2292 | goto local_input; |
2293 | ||
2294 | /* | |
2295 | * Do not cache martian addresses: they should be logged (RFC1812) | |
2296 | */ | |
2297 | martian_destination: | |
2298 | RT_CACHE_STAT_INC(in_martian_dst); | |
2299 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
2300 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) | |
673d57e7 HH |
2301 | printk(KERN_WARNING "martian destination %pI4 from %pI4, dev %s\n", |
2302 | &daddr, &saddr, dev->name); | |
1da177e4 | 2303 | #endif |
2c2910a4 DE |
2304 | |
2305 | e_hostunreach: | |
e905a9ed | 2306 | err = -EHOSTUNREACH; |
ebc0ffae | 2307 | goto out; |
2c2910a4 | 2308 | |
1da177e4 LT |
2309 | e_inval: |
2310 | err = -EINVAL; | |
ebc0ffae | 2311 | goto out; |
1da177e4 LT |
2312 | |
2313 | e_nobufs: | |
2314 | err = -ENOBUFS; | |
ebc0ffae | 2315 | goto out; |
1da177e4 LT |
2316 | |
2317 | martian_source: | |
b5f7e755 ED |
2318 | err = -EINVAL; |
2319 | martian_source_keep_err: | |
1da177e4 | 2320 | ip_handle_martian_source(dev, in_dev, skb, daddr, saddr); |
ebc0ffae | 2321 | goto out; |
1da177e4 LT |
2322 | } |
2323 | ||
407eadd9 ED |
2324 | int ip_route_input_common(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
2325 | u8 tos, struct net_device *dev, bool noref) | |
1da177e4 LT |
2326 | { |
2327 | struct rtable * rth; | |
2328 | unsigned hash; | |
2329 | int iif = dev->ifindex; | |
b5921910 | 2330 | struct net *net; |
96d36220 | 2331 | int res; |
1da177e4 | 2332 | |
c346dca1 | 2333 | net = dev_net(dev); |
1080d709 | 2334 | |
96d36220 ED |
2335 | rcu_read_lock(); |
2336 | ||
1080d709 NH |
2337 | if (!rt_caching(net)) |
2338 | goto skip_cache; | |
2339 | ||
1da177e4 | 2340 | tos &= IPTOS_RT_MASK; |
e84f84f2 | 2341 | hash = rt_hash(daddr, saddr, iif, rt_genid(net)); |
1da177e4 | 2342 | |
1da177e4 | 2343 | for (rth = rcu_dereference(rt_hash_table[hash].chain); rth; |
d8d1f30b | 2344 | rth = rcu_dereference(rth->dst.rt_next)) { |
5e2b61f7 DM |
2345 | if ((((__force u32)rth->rt_key_dst ^ (__force u32)daddr) | |
2346 | ((__force u32)rth->rt_key_src ^ (__force u32)saddr) | | |
97a80410 | 2347 | (rth->rt_route_iif ^ iif) | |
475949d8 | 2348 | (rth->rt_key_tos ^ tos)) == 0 && |
5e2b61f7 | 2349 | rth->rt_mark == skb->mark && |
d8d1f30b | 2350 | net_eq(dev_net(rth->dst.dev), net) && |
e84f84f2 | 2351 | !rt_is_expired(rth)) { |
407eadd9 | 2352 | if (noref) { |
d8d1f30b CG |
2353 | dst_use_noref(&rth->dst, jiffies); |
2354 | skb_dst_set_noref(skb, &rth->dst); | |
407eadd9 | 2355 | } else { |
d8d1f30b CG |
2356 | dst_use(&rth->dst, jiffies); |
2357 | skb_dst_set(skb, &rth->dst); | |
407eadd9 | 2358 | } |
1da177e4 LT |
2359 | RT_CACHE_STAT_INC(in_hit); |
2360 | rcu_read_unlock(); | |
1da177e4 LT |
2361 | return 0; |
2362 | } | |
2363 | RT_CACHE_STAT_INC(in_hlist_search); | |
2364 | } | |
1da177e4 | 2365 | |
1080d709 | 2366 | skip_cache: |
1da177e4 LT |
2367 | /* Multicast recognition logic is moved from route cache to here. |
2368 | The problem was that too many Ethernet cards have broken/missing | |
2369 | hardware multicast filters :-( As result the host on multicasting | |
2370 | network acquires a lot of useless route cache entries, sort of | |
2371 | SDR messages from all the world. Now we try to get rid of them. | |
2372 | Really, provided software IP multicast filter is organized | |
2373 | reasonably (at least, hashed), it does not result in a slowdown | |
2374 | comparing with route cache reject entries. | |
2375 | Note, that multicast routers are not affected, because | |
2376 | route cache entry is created eventually. | |
2377 | */ | |
f97c1e0c | 2378 | if (ipv4_is_multicast(daddr)) { |
96d36220 | 2379 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 2380 | |
96d36220 | 2381 | if (in_dev) { |
dbdd9a52 DM |
2382 | int our = ip_check_mc_rcu(in_dev, daddr, saddr, |
2383 | ip_hdr(skb)->protocol); | |
1da177e4 LT |
2384 | if (our |
2385 | #ifdef CONFIG_IP_MROUTE | |
9d4fb27d JP |
2386 | || |
2387 | (!ipv4_is_local_multicast(daddr) && | |
2388 | IN_DEV_MFORWARD(in_dev)) | |
1da177e4 | 2389 | #endif |
9d4fb27d | 2390 | ) { |
96d36220 ED |
2391 | int res = ip_route_input_mc(skb, daddr, saddr, |
2392 | tos, dev, our); | |
1da177e4 | 2393 | rcu_read_unlock(); |
96d36220 | 2394 | return res; |
1da177e4 LT |
2395 | } |
2396 | } | |
2397 | rcu_read_unlock(); | |
2398 | return -EINVAL; | |
2399 | } | |
96d36220 ED |
2400 | res = ip_route_input_slow(skb, daddr, saddr, tos, dev); |
2401 | rcu_read_unlock(); | |
2402 | return res; | |
1da177e4 | 2403 | } |
407eadd9 | 2404 | EXPORT_SYMBOL(ip_route_input_common); |
1da177e4 | 2405 | |
ebc0ffae | 2406 | /* called with rcu_read_lock() */ |
982721f3 | 2407 | static struct rtable *__mkroute_output(const struct fib_result *res, |
68a5e3dd | 2408 | const struct flowi4 *fl4, |
813b3b5d DM |
2409 | __be32 orig_daddr, __be32 orig_saddr, |
2410 | int orig_oif, struct net_device *dev_out, | |
5ada5527 | 2411 | unsigned int flags) |
1da177e4 | 2412 | { |
982721f3 | 2413 | struct fib_info *fi = res->fi; |
813b3b5d | 2414 | u32 tos = RT_FL_TOS(fl4); |
5ada5527 | 2415 | struct in_device *in_dev; |
982721f3 | 2416 | u16 type = res->type; |
5ada5527 | 2417 | struct rtable *rth; |
1da177e4 | 2418 | |
68a5e3dd | 2419 | if (ipv4_is_loopback(fl4->saddr) && !(dev_out->flags & IFF_LOOPBACK)) |
5ada5527 | 2420 | return ERR_PTR(-EINVAL); |
1da177e4 | 2421 | |
68a5e3dd | 2422 | if (ipv4_is_lbcast(fl4->daddr)) |
982721f3 | 2423 | type = RTN_BROADCAST; |
68a5e3dd | 2424 | else if (ipv4_is_multicast(fl4->daddr)) |
982721f3 | 2425 | type = RTN_MULTICAST; |
68a5e3dd | 2426 | else if (ipv4_is_zeronet(fl4->daddr)) |
5ada5527 | 2427 | return ERR_PTR(-EINVAL); |
1da177e4 LT |
2428 | |
2429 | if (dev_out->flags & IFF_LOOPBACK) | |
2430 | flags |= RTCF_LOCAL; | |
2431 | ||
dd28d1a0 | 2432 | in_dev = __in_dev_get_rcu(dev_out); |
ebc0ffae | 2433 | if (!in_dev) |
5ada5527 | 2434 | return ERR_PTR(-EINVAL); |
ebc0ffae | 2435 | |
982721f3 | 2436 | if (type == RTN_BROADCAST) { |
1da177e4 | 2437 | flags |= RTCF_BROADCAST | RTCF_LOCAL; |
982721f3 DM |
2438 | fi = NULL; |
2439 | } else if (type == RTN_MULTICAST) { | |
dd28d1a0 | 2440 | flags |= RTCF_MULTICAST | RTCF_LOCAL; |
813b3b5d DM |
2441 | if (!ip_check_mc_rcu(in_dev, fl4->daddr, fl4->saddr, |
2442 | fl4->flowi4_proto)) | |
1da177e4 LT |
2443 | flags &= ~RTCF_LOCAL; |
2444 | /* If multicast route do not exist use | |
dd28d1a0 ED |
2445 | * default one, but do not gateway in this case. |
2446 | * Yes, it is hack. | |
1da177e4 | 2447 | */ |
982721f3 DM |
2448 | if (fi && res->prefixlen < 4) |
2449 | fi = NULL; | |
1da177e4 LT |
2450 | } |
2451 | ||
5c1e6aa3 DM |
2452 | rth = rt_dst_alloc(dev_out, |
2453 | IN_DEV_CONF_GET(in_dev, NOPOLICY), | |
0c4dcd58 | 2454 | IN_DEV_CONF_GET(in_dev, NOXFRM)); |
8391d07b | 2455 | if (!rth) |
5ada5527 | 2456 | return ERR_PTR(-ENOBUFS); |
8391d07b | 2457 | |
cf911662 DM |
2458 | rth->dst.output = ip_output; |
2459 | ||
813b3b5d DM |
2460 | rth->rt_key_dst = orig_daddr; |
2461 | rth->rt_key_src = orig_saddr; | |
cf911662 DM |
2462 | rth->rt_genid = rt_genid(dev_net(dev_out)); |
2463 | rth->rt_flags = flags; | |
2464 | rth->rt_type = type; | |
475949d8 | 2465 | rth->rt_key_tos = tos; |
68a5e3dd DM |
2466 | rth->rt_dst = fl4->daddr; |
2467 | rth->rt_src = fl4->saddr; | |
1b86a58f | 2468 | rth->rt_route_iif = 0; |
813b3b5d DM |
2469 | rth->rt_iif = orig_oif ? : dev_out->ifindex; |
2470 | rth->rt_oif = orig_oif; | |
2471 | rth->rt_mark = fl4->flowi4_mark; | |
68a5e3dd DM |
2472 | rth->rt_gateway = fl4->daddr; |
2473 | rth->rt_spec_dst= fl4->saddr; | |
cf911662 DM |
2474 | rth->rt_peer_genid = 0; |
2475 | rth->peer = NULL; | |
2476 | rth->fi = NULL; | |
1da177e4 LT |
2477 | |
2478 | RT_CACHE_STAT_INC(out_slow_tot); | |
2479 | ||
2480 | if (flags & RTCF_LOCAL) { | |
d8d1f30b | 2481 | rth->dst.input = ip_local_deliver; |
68a5e3dd | 2482 | rth->rt_spec_dst = fl4->daddr; |
1da177e4 LT |
2483 | } |
2484 | if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { | |
68a5e3dd | 2485 | rth->rt_spec_dst = fl4->saddr; |
e905a9ed | 2486 | if (flags & RTCF_LOCAL && |
1da177e4 | 2487 | !(dev_out->flags & IFF_LOOPBACK)) { |
d8d1f30b | 2488 | rth->dst.output = ip_mc_output; |
1da177e4 LT |
2489 | RT_CACHE_STAT_INC(out_slow_mc); |
2490 | } | |
2491 | #ifdef CONFIG_IP_MROUTE | |
982721f3 | 2492 | if (type == RTN_MULTICAST) { |
1da177e4 | 2493 | if (IN_DEV_MFORWARD(in_dev) && |
813b3b5d | 2494 | !ipv4_is_local_multicast(fl4->daddr)) { |
d8d1f30b CG |
2495 | rth->dst.input = ip_mr_input; |
2496 | rth->dst.output = ip_mc_output; | |
1da177e4 LT |
2497 | } |
2498 | } | |
2499 | #endif | |
2500 | } | |
2501 | ||
813b3b5d | 2502 | rt_set_nexthop(rth, fl4, res, fi, type, 0); |
1da177e4 | 2503 | |
5ada5527 | 2504 | return rth; |
1da177e4 LT |
2505 | } |
2506 | ||
1da177e4 LT |
2507 | /* |
2508 | * Major route resolver routine. | |
0197aa38 | 2509 | * called with rcu_read_lock(); |
1da177e4 LT |
2510 | */ |
2511 | ||
813b3b5d | 2512 | static struct rtable *ip_route_output_slow(struct net *net, struct flowi4 *fl4) |
1da177e4 | 2513 | { |
1da177e4 | 2514 | struct net_device *dev_out = NULL; |
813b3b5d DM |
2515 | u32 tos = RT_FL_TOS(fl4); |
2516 | unsigned int flags = 0; | |
2517 | struct fib_result res; | |
5ada5527 | 2518 | struct rtable *rth; |
813b3b5d DM |
2519 | __be32 orig_daddr; |
2520 | __be32 orig_saddr; | |
2521 | int orig_oif; | |
1da177e4 LT |
2522 | |
2523 | res.fi = NULL; | |
2524 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
2525 | res.r = NULL; | |
2526 | #endif | |
2527 | ||
813b3b5d DM |
2528 | orig_daddr = fl4->daddr; |
2529 | orig_saddr = fl4->saddr; | |
2530 | orig_oif = fl4->flowi4_oif; | |
2531 | ||
2532 | fl4->flowi4_iif = net->loopback_dev->ifindex; | |
2533 | fl4->flowi4_tos = tos & IPTOS_RT_MASK; | |
2534 | fl4->flowi4_scope = ((tos & RTO_ONLINK) ? | |
2535 | RT_SCOPE_LINK : RT_SCOPE_UNIVERSE); | |
44713b67 | 2536 | |
010c2708 | 2537 | rcu_read_lock(); |
813b3b5d | 2538 | if (fl4->saddr) { |
b23dd4fe | 2539 | rth = ERR_PTR(-EINVAL); |
813b3b5d DM |
2540 | if (ipv4_is_multicast(fl4->saddr) || |
2541 | ipv4_is_lbcast(fl4->saddr) || | |
2542 | ipv4_is_zeronet(fl4->saddr)) | |
1da177e4 LT |
2543 | goto out; |
2544 | ||
1da177e4 LT |
2545 | /* I removed check for oif == dev_out->oif here. |
2546 | It was wrong for two reasons: | |
1ab35276 DL |
2547 | 1. ip_dev_find(net, saddr) can return wrong iface, if saddr |
2548 | is assigned to multiple interfaces. | |
1da177e4 LT |
2549 | 2. Moreover, we are allowed to send packets with saddr |
2550 | of another iface. --ANK | |
2551 | */ | |
2552 | ||
813b3b5d DM |
2553 | if (fl4->flowi4_oif == 0 && |
2554 | (ipv4_is_multicast(fl4->daddr) || | |
2555 | ipv4_is_lbcast(fl4->daddr))) { | |
a210d01a | 2556 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
813b3b5d | 2557 | dev_out = __ip_dev_find(net, fl4->saddr, false); |
a210d01a JA |
2558 | if (dev_out == NULL) |
2559 | goto out; | |
2560 | ||
1da177e4 LT |
2561 | /* Special hack: user can direct multicasts |
2562 | and limited broadcast via necessary interface | |
2563 | without fiddling with IP_MULTICAST_IF or IP_PKTINFO. | |
2564 | This hack is not just for fun, it allows | |
2565 | vic,vat and friends to work. | |
2566 | They bind socket to loopback, set ttl to zero | |
2567 | and expect that it will work. | |
2568 | From the viewpoint of routing cache they are broken, | |
2569 | because we are not allowed to build multicast path | |
2570 | with loopback source addr (look, routing cache | |
2571 | cannot know, that ttl is zero, so that packet | |
2572 | will not leave this host and route is valid). | |
2573 | Luckily, this hack is good workaround. | |
2574 | */ | |
2575 | ||
813b3b5d | 2576 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2577 | goto make_route; |
2578 | } | |
a210d01a | 2579 | |
813b3b5d | 2580 | if (!(fl4->flowi4_flags & FLOWI_FLAG_ANYSRC)) { |
a210d01a | 2581 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
813b3b5d | 2582 | if (!__ip_dev_find(net, fl4->saddr, false)) |
a210d01a | 2583 | goto out; |
a210d01a | 2584 | } |
1da177e4 LT |
2585 | } |
2586 | ||
2587 | ||
813b3b5d DM |
2588 | if (fl4->flowi4_oif) { |
2589 | dev_out = dev_get_by_index_rcu(net, fl4->flowi4_oif); | |
b23dd4fe | 2590 | rth = ERR_PTR(-ENODEV); |
1da177e4 LT |
2591 | if (dev_out == NULL) |
2592 | goto out; | |
e5ed6399 HX |
2593 | |
2594 | /* RACE: Check return value of inet_select_addr instead. */ | |
fc75fc83 | 2595 | if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) { |
b23dd4fe | 2596 | rth = ERR_PTR(-ENETUNREACH); |
fc75fc83 ED |
2597 | goto out; |
2598 | } | |
813b3b5d DM |
2599 | if (ipv4_is_local_multicast(fl4->daddr) || |
2600 | ipv4_is_lbcast(fl4->daddr)) { | |
2601 | if (!fl4->saddr) | |
2602 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2603 | RT_SCOPE_LINK); | |
1da177e4 LT |
2604 | goto make_route; |
2605 | } | |
813b3b5d DM |
2606 | if (fl4->saddr) { |
2607 | if (ipv4_is_multicast(fl4->daddr)) | |
2608 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2609 | fl4->flowi4_scope); | |
2610 | else if (!fl4->daddr) | |
2611 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2612 | RT_SCOPE_HOST); | |
1da177e4 LT |
2613 | } |
2614 | } | |
2615 | ||
813b3b5d DM |
2616 | if (!fl4->daddr) { |
2617 | fl4->daddr = fl4->saddr; | |
2618 | if (!fl4->daddr) | |
2619 | fl4->daddr = fl4->saddr = htonl(INADDR_LOOPBACK); | |
b40afd0e | 2620 | dev_out = net->loopback_dev; |
813b3b5d | 2621 | fl4->flowi4_oif = net->loopback_dev->ifindex; |
1da177e4 LT |
2622 | res.type = RTN_LOCAL; |
2623 | flags |= RTCF_LOCAL; | |
2624 | goto make_route; | |
2625 | } | |
2626 | ||
813b3b5d | 2627 | if (fib_lookup(net, fl4, &res)) { |
1da177e4 | 2628 | res.fi = NULL; |
813b3b5d | 2629 | if (fl4->flowi4_oif) { |
1da177e4 LT |
2630 | /* Apparently, routing tables are wrong. Assume, |
2631 | that the destination is on link. | |
2632 | ||
2633 | WHY? DW. | |
2634 | Because we are allowed to send to iface | |
2635 | even if it has NO routes and NO assigned | |
2636 | addresses. When oif is specified, routing | |
2637 | tables are looked up with only one purpose: | |
2638 | to catch if destination is gatewayed, rather than | |
2639 | direct. Moreover, if MSG_DONTROUTE is set, | |
2640 | we send packet, ignoring both routing tables | |
2641 | and ifaddr state. --ANK | |
2642 | ||
2643 | ||
2644 | We could make it even if oif is unknown, | |
2645 | likely IPv6, but we do not. | |
2646 | */ | |
2647 | ||
813b3b5d DM |
2648 | if (fl4->saddr == 0) |
2649 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2650 | RT_SCOPE_LINK); | |
1da177e4 LT |
2651 | res.type = RTN_UNICAST; |
2652 | goto make_route; | |
2653 | } | |
b23dd4fe | 2654 | rth = ERR_PTR(-ENETUNREACH); |
1da177e4 LT |
2655 | goto out; |
2656 | } | |
1da177e4 LT |
2657 | |
2658 | if (res.type == RTN_LOCAL) { | |
813b3b5d | 2659 | if (!fl4->saddr) { |
9fc3bbb4 | 2660 | if (res.fi->fib_prefsrc) |
813b3b5d | 2661 | fl4->saddr = res.fi->fib_prefsrc; |
9fc3bbb4 | 2662 | else |
813b3b5d | 2663 | fl4->saddr = fl4->daddr; |
9fc3bbb4 | 2664 | } |
b40afd0e | 2665 | dev_out = net->loopback_dev; |
813b3b5d | 2666 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2667 | res.fi = NULL; |
2668 | flags |= RTCF_LOCAL; | |
2669 | goto make_route; | |
2670 | } | |
2671 | ||
2672 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
813b3b5d | 2673 | if (res.fi->fib_nhs > 1 && fl4->flowi4_oif == 0) |
1b7fe593 | 2674 | fib_select_multipath(&res); |
1da177e4 LT |
2675 | else |
2676 | #endif | |
21d8c49e DM |
2677 | if (!res.prefixlen && |
2678 | res.table->tb_num_default > 1 && | |
813b3b5d | 2679 | res.type == RTN_UNICAST && !fl4->flowi4_oif) |
0c838ff1 | 2680 | fib_select_default(&res); |
1da177e4 | 2681 | |
813b3b5d DM |
2682 | if (!fl4->saddr) |
2683 | fl4->saddr = FIB_RES_PREFSRC(net, res); | |
1da177e4 | 2684 | |
1da177e4 | 2685 | dev_out = FIB_RES_DEV(res); |
813b3b5d | 2686 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2687 | |
2688 | ||
2689 | make_route: | |
813b3b5d DM |
2690 | rth = __mkroute_output(&res, fl4, orig_daddr, orig_saddr, orig_oif, |
2691 | dev_out, flags); | |
b23dd4fe | 2692 | if (!IS_ERR(rth)) { |
5ada5527 DM |
2693 | unsigned int hash; |
2694 | ||
813b3b5d | 2695 | hash = rt_hash(orig_daddr, orig_saddr, orig_oif, |
5ada5527 | 2696 | rt_genid(dev_net(dev_out))); |
813b3b5d | 2697 | rth = rt_intern_hash(hash, rth, NULL, orig_oif); |
5ada5527 | 2698 | } |
1da177e4 | 2699 | |
010c2708 DM |
2700 | out: |
2701 | rcu_read_unlock(); | |
b23dd4fe | 2702 | return rth; |
1da177e4 LT |
2703 | } |
2704 | ||
813b3b5d | 2705 | struct rtable *__ip_route_output_key(struct net *net, struct flowi4 *flp4) |
1da177e4 | 2706 | { |
1da177e4 | 2707 | struct rtable *rth; |
010c2708 | 2708 | unsigned int hash; |
1da177e4 | 2709 | |
1080d709 NH |
2710 | if (!rt_caching(net)) |
2711 | goto slow_output; | |
2712 | ||
9d6ec938 | 2713 | hash = rt_hash(flp4->daddr, flp4->saddr, flp4->flowi4_oif, rt_genid(net)); |
1da177e4 LT |
2714 | |
2715 | rcu_read_lock_bh(); | |
a898def2 | 2716 | for (rth = rcu_dereference_bh(rt_hash_table[hash].chain); rth; |
d8d1f30b | 2717 | rth = rcu_dereference_bh(rth->dst.rt_next)) { |
9d6ec938 DM |
2718 | if (rth->rt_key_dst == flp4->daddr && |
2719 | rth->rt_key_src == flp4->saddr && | |
c7537967 | 2720 | rt_is_output_route(rth) && |
9d6ec938 DM |
2721 | rth->rt_oif == flp4->flowi4_oif && |
2722 | rth->rt_mark == flp4->flowi4_mark && | |
475949d8 | 2723 | !((rth->rt_key_tos ^ flp4->flowi4_tos) & |
b5921910 | 2724 | (IPTOS_RT_MASK | RTO_ONLINK)) && |
d8d1f30b | 2725 | net_eq(dev_net(rth->dst.dev), net) && |
e84f84f2 | 2726 | !rt_is_expired(rth)) { |
d8d1f30b | 2727 | dst_use(&rth->dst, jiffies); |
1da177e4 LT |
2728 | RT_CACHE_STAT_INC(out_hit); |
2729 | rcu_read_unlock_bh(); | |
56157872 DM |
2730 | if (!flp4->saddr) |
2731 | flp4->saddr = rth->rt_src; | |
2732 | if (!flp4->daddr) | |
2733 | flp4->daddr = rth->rt_dst; | |
b23dd4fe | 2734 | return rth; |
1da177e4 LT |
2735 | } |
2736 | RT_CACHE_STAT_INC(out_hlist_search); | |
2737 | } | |
2738 | rcu_read_unlock_bh(); | |
2739 | ||
1080d709 | 2740 | slow_output: |
9d6ec938 | 2741 | return ip_route_output_slow(net, flp4); |
1da177e4 | 2742 | } |
d8c97a94 ACM |
2743 | EXPORT_SYMBOL_GPL(__ip_route_output_key); |
2744 | ||
ae2688d5 JW |
2745 | static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie) |
2746 | { | |
2747 | return NULL; | |
2748 | } | |
2749 | ||
ec831ea7 RD |
2750 | static unsigned int ipv4_blackhole_default_mtu(const struct dst_entry *dst) |
2751 | { | |
2752 | return 0; | |
2753 | } | |
2754 | ||
14e50e57 DM |
2755 | static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu) |
2756 | { | |
2757 | } | |
2758 | ||
0972ddb2 HB |
2759 | static u32 *ipv4_rt_blackhole_cow_metrics(struct dst_entry *dst, |
2760 | unsigned long old) | |
2761 | { | |
2762 | return NULL; | |
2763 | } | |
2764 | ||
14e50e57 DM |
2765 | static struct dst_ops ipv4_dst_blackhole_ops = { |
2766 | .family = AF_INET, | |
09640e63 | 2767 | .protocol = cpu_to_be16(ETH_P_IP), |
14e50e57 | 2768 | .destroy = ipv4_dst_destroy, |
ae2688d5 | 2769 | .check = ipv4_blackhole_dst_check, |
ec831ea7 | 2770 | .default_mtu = ipv4_blackhole_default_mtu, |
214f45c9 | 2771 | .default_advmss = ipv4_default_advmss, |
14e50e57 | 2772 | .update_pmtu = ipv4_rt_blackhole_update_pmtu, |
0972ddb2 | 2773 | .cow_metrics = ipv4_rt_blackhole_cow_metrics, |
d3aaeb38 | 2774 | .neigh_lookup = ipv4_neigh_lookup, |
14e50e57 DM |
2775 | }; |
2776 | ||
2774c131 | 2777 | struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig) |
14e50e57 | 2778 | { |
5c1e6aa3 | 2779 | struct rtable *rt = dst_alloc(&ipv4_dst_blackhole_ops, NULL, 1, 0, 0); |
2774c131 | 2780 | struct rtable *ort = (struct rtable *) dst_orig; |
14e50e57 DM |
2781 | |
2782 | if (rt) { | |
d8d1f30b | 2783 | struct dst_entry *new = &rt->dst; |
14e50e57 | 2784 | |
14e50e57 | 2785 | new->__use = 1; |
352e512c HX |
2786 | new->input = dst_discard; |
2787 | new->output = dst_discard; | |
defb3519 | 2788 | dst_copy_metrics(new, &ort->dst); |
14e50e57 | 2789 | |
d8d1f30b | 2790 | new->dev = ort->dst.dev; |
14e50e57 DM |
2791 | if (new->dev) |
2792 | dev_hold(new->dev); | |
2793 | ||
5e2b61f7 DM |
2794 | rt->rt_key_dst = ort->rt_key_dst; |
2795 | rt->rt_key_src = ort->rt_key_src; | |
475949d8 | 2796 | rt->rt_key_tos = ort->rt_key_tos; |
1b86a58f | 2797 | rt->rt_route_iif = ort->rt_route_iif; |
5e2b61f7 DM |
2798 | rt->rt_iif = ort->rt_iif; |
2799 | rt->rt_oif = ort->rt_oif; | |
2800 | rt->rt_mark = ort->rt_mark; | |
14e50e57 | 2801 | |
e84f84f2 | 2802 | rt->rt_genid = rt_genid(net); |
14e50e57 DM |
2803 | rt->rt_flags = ort->rt_flags; |
2804 | rt->rt_type = ort->rt_type; | |
2805 | rt->rt_dst = ort->rt_dst; | |
2806 | rt->rt_src = ort->rt_src; | |
14e50e57 DM |
2807 | rt->rt_gateway = ort->rt_gateway; |
2808 | rt->rt_spec_dst = ort->rt_spec_dst; | |
2809 | rt->peer = ort->peer; | |
2810 | if (rt->peer) | |
2811 | atomic_inc(&rt->peer->refcnt); | |
62fa8a84 DM |
2812 | rt->fi = ort->fi; |
2813 | if (rt->fi) | |
2814 | atomic_inc(&rt->fi->fib_clntref); | |
14e50e57 DM |
2815 | |
2816 | dst_free(new); | |
2817 | } | |
2818 | ||
2774c131 DM |
2819 | dst_release(dst_orig); |
2820 | ||
2821 | return rt ? &rt->dst : ERR_PTR(-ENOMEM); | |
14e50e57 DM |
2822 | } |
2823 | ||
9d6ec938 | 2824 | struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4, |
b23dd4fe | 2825 | struct sock *sk) |
1da177e4 | 2826 | { |
9d6ec938 | 2827 | struct rtable *rt = __ip_route_output_key(net, flp4); |
1da177e4 | 2828 | |
b23dd4fe DM |
2829 | if (IS_ERR(rt)) |
2830 | return rt; | |
1da177e4 | 2831 | |
56157872 | 2832 | if (flp4->flowi4_proto) |
9d6ec938 DM |
2833 | rt = (struct rtable *) xfrm_lookup(net, &rt->dst, |
2834 | flowi4_to_flowi(flp4), | |
2835 | sk, 0); | |
1da177e4 | 2836 | |
b23dd4fe | 2837 | return rt; |
1da177e4 | 2838 | } |
d8c97a94 ACM |
2839 | EXPORT_SYMBOL_GPL(ip_route_output_flow); |
2840 | ||
4feb88e5 BT |
2841 | static int rt_fill_info(struct net *net, |
2842 | struct sk_buff *skb, u32 pid, u32 seq, int event, | |
b6544c0b | 2843 | int nowait, unsigned int flags) |
1da177e4 | 2844 | { |
511c3f92 | 2845 | struct rtable *rt = skb_rtable(skb); |
1da177e4 | 2846 | struct rtmsg *r; |
be403ea1 | 2847 | struct nlmsghdr *nlh; |
2bc8ca40 | 2848 | unsigned long expires = 0; |
fe6fe792 | 2849 | const struct inet_peer *peer = rt->peer; |
e3703b3d | 2850 | u32 id = 0, ts = 0, tsage = 0, error; |
be403ea1 TG |
2851 | |
2852 | nlh = nlmsg_put(skb, pid, seq, event, sizeof(*r), flags); | |
2853 | if (nlh == NULL) | |
26932566 | 2854 | return -EMSGSIZE; |
be403ea1 TG |
2855 | |
2856 | r = nlmsg_data(nlh); | |
1da177e4 LT |
2857 | r->rtm_family = AF_INET; |
2858 | r->rtm_dst_len = 32; | |
2859 | r->rtm_src_len = 0; | |
475949d8 | 2860 | r->rtm_tos = rt->rt_key_tos; |
1da177e4 | 2861 | r->rtm_table = RT_TABLE_MAIN; |
be403ea1 | 2862 | NLA_PUT_U32(skb, RTA_TABLE, RT_TABLE_MAIN); |
1da177e4 LT |
2863 | r->rtm_type = rt->rt_type; |
2864 | r->rtm_scope = RT_SCOPE_UNIVERSE; | |
2865 | r->rtm_protocol = RTPROT_UNSPEC; | |
2866 | r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED; | |
2867 | if (rt->rt_flags & RTCF_NOTIFY) | |
2868 | r->rtm_flags |= RTM_F_NOTIFY; | |
be403ea1 | 2869 | |
17fb2c64 | 2870 | NLA_PUT_BE32(skb, RTA_DST, rt->rt_dst); |
be403ea1 | 2871 | |
5e2b61f7 | 2872 | if (rt->rt_key_src) { |
1da177e4 | 2873 | r->rtm_src_len = 32; |
5e2b61f7 | 2874 | NLA_PUT_BE32(skb, RTA_SRC, rt->rt_key_src); |
1da177e4 | 2875 | } |
d8d1f30b CG |
2876 | if (rt->dst.dev) |
2877 | NLA_PUT_U32(skb, RTA_OIF, rt->dst.dev->ifindex); | |
c7066f70 | 2878 | #ifdef CONFIG_IP_ROUTE_CLASSID |
d8d1f30b CG |
2879 | if (rt->dst.tclassid) |
2880 | NLA_PUT_U32(skb, RTA_FLOW, rt->dst.tclassid); | |
1da177e4 | 2881 | #endif |
c7537967 | 2882 | if (rt_is_input_route(rt)) |
17fb2c64 | 2883 | NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_spec_dst); |
5e2b61f7 | 2884 | else if (rt->rt_src != rt->rt_key_src) |
17fb2c64 | 2885 | NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_src); |
be403ea1 | 2886 | |
1da177e4 | 2887 | if (rt->rt_dst != rt->rt_gateway) |
17fb2c64 | 2888 | NLA_PUT_BE32(skb, RTA_GATEWAY, rt->rt_gateway); |
be403ea1 | 2889 | |
defb3519 | 2890 | if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0) |
be403ea1 TG |
2891 | goto nla_put_failure; |
2892 | ||
5e2b61f7 DM |
2893 | if (rt->rt_mark) |
2894 | NLA_PUT_BE32(skb, RTA_MARK, rt->rt_mark); | |
963bfeee | 2895 | |
d8d1f30b | 2896 | error = rt->dst.error; |
fe6fe792 | 2897 | if (peer) { |
317fe0e6 | 2898 | inet_peer_refcheck(rt->peer); |
fe6fe792 ED |
2899 | id = atomic_read(&peer->ip_id_count) & 0xffff; |
2900 | if (peer->tcp_ts_stamp) { | |
2901 | ts = peer->tcp_ts; | |
2902 | tsage = get_seconds() - peer->tcp_ts_stamp; | |
1da177e4 | 2903 | } |
fe6fe792 | 2904 | expires = ACCESS_ONCE(peer->pmtu_expires); |
2bc8ca40 SK |
2905 | if (expires) { |
2906 | if (time_before(jiffies, expires)) | |
2907 | expires -= jiffies; | |
2908 | else | |
2909 | expires = 0; | |
2910 | } | |
1da177e4 | 2911 | } |
be403ea1 | 2912 | |
c7537967 | 2913 | if (rt_is_input_route(rt)) { |
1da177e4 | 2914 | #ifdef CONFIG_IP_MROUTE |
e448515c | 2915 | __be32 dst = rt->rt_dst; |
1da177e4 | 2916 | |
f97c1e0c | 2917 | if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) && |
4feb88e5 | 2918 | IPV4_DEVCONF_ALL(net, MC_FORWARDING)) { |
9a1b9496 DM |
2919 | int err = ipmr_get_route(net, skb, |
2920 | rt->rt_src, rt->rt_dst, | |
2921 | r, nowait); | |
1da177e4 LT |
2922 | if (err <= 0) { |
2923 | if (!nowait) { | |
2924 | if (err == 0) | |
2925 | return 0; | |
be403ea1 | 2926 | goto nla_put_failure; |
1da177e4 LT |
2927 | } else { |
2928 | if (err == -EMSGSIZE) | |
be403ea1 | 2929 | goto nla_put_failure; |
e3703b3d | 2930 | error = err; |
1da177e4 LT |
2931 | } |
2932 | } | |
2933 | } else | |
2934 | #endif | |
5e2b61f7 | 2935 | NLA_PUT_U32(skb, RTA_IIF, rt->rt_iif); |
1da177e4 LT |
2936 | } |
2937 | ||
d8d1f30b | 2938 | if (rtnl_put_cacheinfo(skb, &rt->dst, id, ts, tsage, |
e3703b3d TG |
2939 | expires, error) < 0) |
2940 | goto nla_put_failure; | |
be403ea1 TG |
2941 | |
2942 | return nlmsg_end(skb, nlh); | |
1da177e4 | 2943 | |
be403ea1 | 2944 | nla_put_failure: |
26932566 PM |
2945 | nlmsg_cancel(skb, nlh); |
2946 | return -EMSGSIZE; | |
1da177e4 LT |
2947 | } |
2948 | ||
63f3444f | 2949 | static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg) |
1da177e4 | 2950 | { |
3b1e0a65 | 2951 | struct net *net = sock_net(in_skb->sk); |
d889ce3b TG |
2952 | struct rtmsg *rtm; |
2953 | struct nlattr *tb[RTA_MAX+1]; | |
1da177e4 | 2954 | struct rtable *rt = NULL; |
9e12bb22 AV |
2955 | __be32 dst = 0; |
2956 | __be32 src = 0; | |
2957 | u32 iif; | |
d889ce3b | 2958 | int err; |
963bfeee | 2959 | int mark; |
1da177e4 LT |
2960 | struct sk_buff *skb; |
2961 | ||
d889ce3b TG |
2962 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy); |
2963 | if (err < 0) | |
2964 | goto errout; | |
2965 | ||
2966 | rtm = nlmsg_data(nlh); | |
2967 | ||
1da177e4 | 2968 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
d889ce3b TG |
2969 | if (skb == NULL) { |
2970 | err = -ENOBUFS; | |
2971 | goto errout; | |
2972 | } | |
1da177e4 LT |
2973 | |
2974 | /* Reserve room for dummy headers, this skb can pass | |
2975 | through good chunk of routing engine. | |
2976 | */ | |
459a98ed | 2977 | skb_reset_mac_header(skb); |
c1d2bbe1 | 2978 | skb_reset_network_header(skb); |
d2c962b8 SH |
2979 | |
2980 | /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */ | |
eddc9ec5 | 2981 | ip_hdr(skb)->protocol = IPPROTO_ICMP; |
1da177e4 LT |
2982 | skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr)); |
2983 | ||
17fb2c64 AV |
2984 | src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0; |
2985 | dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0; | |
d889ce3b | 2986 | iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0; |
963bfeee | 2987 | mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0; |
1da177e4 LT |
2988 | |
2989 | if (iif) { | |
d889ce3b TG |
2990 | struct net_device *dev; |
2991 | ||
1937504d | 2992 | dev = __dev_get_by_index(net, iif); |
d889ce3b TG |
2993 | if (dev == NULL) { |
2994 | err = -ENODEV; | |
2995 | goto errout_free; | |
2996 | } | |
2997 | ||
1da177e4 LT |
2998 | skb->protocol = htons(ETH_P_IP); |
2999 | skb->dev = dev; | |
963bfeee | 3000 | skb->mark = mark; |
1da177e4 LT |
3001 | local_bh_disable(); |
3002 | err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev); | |
3003 | local_bh_enable(); | |
d889ce3b | 3004 | |
511c3f92 | 3005 | rt = skb_rtable(skb); |
d8d1f30b CG |
3006 | if (err == 0 && rt->dst.error) |
3007 | err = -rt->dst.error; | |
1da177e4 | 3008 | } else { |
68a5e3dd DM |
3009 | struct flowi4 fl4 = { |
3010 | .daddr = dst, | |
3011 | .saddr = src, | |
3012 | .flowi4_tos = rtm->rtm_tos, | |
3013 | .flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0, | |
3014 | .flowi4_mark = mark, | |
d889ce3b | 3015 | }; |
9d6ec938 | 3016 | rt = ip_route_output_key(net, &fl4); |
b23dd4fe DM |
3017 | |
3018 | err = 0; | |
3019 | if (IS_ERR(rt)) | |
3020 | err = PTR_ERR(rt); | |
1da177e4 | 3021 | } |
d889ce3b | 3022 | |
1da177e4 | 3023 | if (err) |
d889ce3b | 3024 | goto errout_free; |
1da177e4 | 3025 | |
d8d1f30b | 3026 | skb_dst_set(skb, &rt->dst); |
1da177e4 LT |
3027 | if (rtm->rtm_flags & RTM_F_NOTIFY) |
3028 | rt->rt_flags |= RTCF_NOTIFY; | |
3029 | ||
4feb88e5 | 3030 | err = rt_fill_info(net, skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, |
1937504d | 3031 | RTM_NEWROUTE, 0, 0); |
d889ce3b TG |
3032 | if (err <= 0) |
3033 | goto errout_free; | |
1da177e4 | 3034 | |
1937504d | 3035 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid); |
d889ce3b | 3036 | errout: |
2942e900 | 3037 | return err; |
1da177e4 | 3038 | |
d889ce3b | 3039 | errout_free: |
1da177e4 | 3040 | kfree_skb(skb); |
d889ce3b | 3041 | goto errout; |
1da177e4 LT |
3042 | } |
3043 | ||
3044 | int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb) | |
3045 | { | |
3046 | struct rtable *rt; | |
3047 | int h, s_h; | |
3048 | int idx, s_idx; | |
1937504d DL |
3049 | struct net *net; |
3050 | ||
3b1e0a65 | 3051 | net = sock_net(skb->sk); |
1da177e4 LT |
3052 | |
3053 | s_h = cb->args[0]; | |
d8c92830 ED |
3054 | if (s_h < 0) |
3055 | s_h = 0; | |
1da177e4 | 3056 | s_idx = idx = cb->args[1]; |
a6272665 ED |
3057 | for (h = s_h; h <= rt_hash_mask; h++, s_idx = 0) { |
3058 | if (!rt_hash_table[h].chain) | |
3059 | continue; | |
1da177e4 | 3060 | rcu_read_lock_bh(); |
a898def2 | 3061 | for (rt = rcu_dereference_bh(rt_hash_table[h].chain), idx = 0; rt; |
d8d1f30b CG |
3062 | rt = rcu_dereference_bh(rt->dst.rt_next), idx++) { |
3063 | if (!net_eq(dev_net(rt->dst.dev), net) || idx < s_idx) | |
1da177e4 | 3064 | continue; |
e84f84f2 | 3065 | if (rt_is_expired(rt)) |
29e75252 | 3066 | continue; |
d8d1f30b | 3067 | skb_dst_set_noref(skb, &rt->dst); |
4feb88e5 | 3068 | if (rt_fill_info(net, skb, NETLINK_CB(cb->skb).pid, |
e905a9ed | 3069 | cb->nlh->nlmsg_seq, RTM_NEWROUTE, |
b6544c0b | 3070 | 1, NLM_F_MULTI) <= 0) { |
adf30907 | 3071 | skb_dst_drop(skb); |
1da177e4 LT |
3072 | rcu_read_unlock_bh(); |
3073 | goto done; | |
3074 | } | |
adf30907 | 3075 | skb_dst_drop(skb); |
1da177e4 LT |
3076 | } |
3077 | rcu_read_unlock_bh(); | |
3078 | } | |
3079 | ||
3080 | done: | |
3081 | cb->args[0] = h; | |
3082 | cb->args[1] = idx; | |
3083 | return skb->len; | |
3084 | } | |
3085 | ||
3086 | void ip_rt_multicast_event(struct in_device *in_dev) | |
3087 | { | |
76e6ebfb | 3088 | rt_cache_flush(dev_net(in_dev->dev), 0); |
1da177e4 LT |
3089 | } |
3090 | ||
3091 | #ifdef CONFIG_SYSCTL | |
81c684d1 | 3092 | static int ipv4_sysctl_rtcache_flush(ctl_table *__ctl, int write, |
8d65af78 | 3093 | void __user *buffer, |
1da177e4 LT |
3094 | size_t *lenp, loff_t *ppos) |
3095 | { | |
3096 | if (write) { | |
639e104f | 3097 | int flush_delay; |
81c684d1 | 3098 | ctl_table ctl; |
39a23e75 | 3099 | struct net *net; |
639e104f | 3100 | |
81c684d1 DL |
3101 | memcpy(&ctl, __ctl, sizeof(ctl)); |
3102 | ctl.data = &flush_delay; | |
8d65af78 | 3103 | proc_dointvec(&ctl, write, buffer, lenp, ppos); |
639e104f | 3104 | |
81c684d1 | 3105 | net = (struct net *)__ctl->extra1; |
39a23e75 | 3106 | rt_cache_flush(net, flush_delay); |
1da177e4 | 3107 | return 0; |
e905a9ed | 3108 | } |
1da177e4 LT |
3109 | |
3110 | return -EINVAL; | |
3111 | } | |
3112 | ||
eeb61f71 | 3113 | static ctl_table ipv4_route_table[] = { |
1da177e4 | 3114 | { |
1da177e4 LT |
3115 | .procname = "gc_thresh", |
3116 | .data = &ipv4_dst_ops.gc_thresh, | |
3117 | .maxlen = sizeof(int), | |
3118 | .mode = 0644, | |
6d9f239a | 3119 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3120 | }, |
3121 | { | |
1da177e4 LT |
3122 | .procname = "max_size", |
3123 | .data = &ip_rt_max_size, | |
3124 | .maxlen = sizeof(int), | |
3125 | .mode = 0644, | |
6d9f239a | 3126 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3127 | }, |
3128 | { | |
3129 | /* Deprecated. Use gc_min_interval_ms */ | |
e905a9ed | 3130 | |
1da177e4 LT |
3131 | .procname = "gc_min_interval", |
3132 | .data = &ip_rt_gc_min_interval, | |
3133 | .maxlen = sizeof(int), | |
3134 | .mode = 0644, | |
6d9f239a | 3135 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3136 | }, |
3137 | { | |
1da177e4 LT |
3138 | .procname = "gc_min_interval_ms", |
3139 | .data = &ip_rt_gc_min_interval, | |
3140 | .maxlen = sizeof(int), | |
3141 | .mode = 0644, | |
6d9f239a | 3142 | .proc_handler = proc_dointvec_ms_jiffies, |
1da177e4 LT |
3143 | }, |
3144 | { | |
1da177e4 LT |
3145 | .procname = "gc_timeout", |
3146 | .data = &ip_rt_gc_timeout, | |
3147 | .maxlen = sizeof(int), | |
3148 | .mode = 0644, | |
6d9f239a | 3149 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 | 3150 | }, |
1da177e4 | 3151 | { |
1da177e4 LT |
3152 | .procname = "redirect_load", |
3153 | .data = &ip_rt_redirect_load, | |
3154 | .maxlen = sizeof(int), | |
3155 | .mode = 0644, | |
6d9f239a | 3156 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3157 | }, |
3158 | { | |
1da177e4 LT |
3159 | .procname = "redirect_number", |
3160 | .data = &ip_rt_redirect_number, | |
3161 | .maxlen = sizeof(int), | |
3162 | .mode = 0644, | |
6d9f239a | 3163 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3164 | }, |
3165 | { | |
1da177e4 LT |
3166 | .procname = "redirect_silence", |
3167 | .data = &ip_rt_redirect_silence, | |
3168 | .maxlen = sizeof(int), | |
3169 | .mode = 0644, | |
6d9f239a | 3170 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3171 | }, |
3172 | { | |
1da177e4 LT |
3173 | .procname = "error_cost", |
3174 | .data = &ip_rt_error_cost, | |
3175 | .maxlen = sizeof(int), | |
3176 | .mode = 0644, | |
6d9f239a | 3177 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3178 | }, |
3179 | { | |
1da177e4 LT |
3180 | .procname = "error_burst", |
3181 | .data = &ip_rt_error_burst, | |
3182 | .maxlen = sizeof(int), | |
3183 | .mode = 0644, | |
6d9f239a | 3184 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3185 | }, |
3186 | { | |
1da177e4 LT |
3187 | .procname = "gc_elasticity", |
3188 | .data = &ip_rt_gc_elasticity, | |
3189 | .maxlen = sizeof(int), | |
3190 | .mode = 0644, | |
6d9f239a | 3191 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3192 | }, |
3193 | { | |
1da177e4 LT |
3194 | .procname = "mtu_expires", |
3195 | .data = &ip_rt_mtu_expires, | |
3196 | .maxlen = sizeof(int), | |
3197 | .mode = 0644, | |
6d9f239a | 3198 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3199 | }, |
3200 | { | |
1da177e4 LT |
3201 | .procname = "min_pmtu", |
3202 | .data = &ip_rt_min_pmtu, | |
3203 | .maxlen = sizeof(int), | |
3204 | .mode = 0644, | |
6d9f239a | 3205 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3206 | }, |
3207 | { | |
1da177e4 LT |
3208 | .procname = "min_adv_mss", |
3209 | .data = &ip_rt_min_advmss, | |
3210 | .maxlen = sizeof(int), | |
3211 | .mode = 0644, | |
6d9f239a | 3212 | .proc_handler = proc_dointvec, |
1da177e4 | 3213 | }, |
f8572d8f | 3214 | { } |
1da177e4 | 3215 | }; |
39a23e75 | 3216 | |
2f4520d3 AV |
3217 | static struct ctl_table empty[1]; |
3218 | ||
3219 | static struct ctl_table ipv4_skeleton[] = | |
3220 | { | |
f8572d8f | 3221 | { .procname = "route", |
d994af0d | 3222 | .mode = 0555, .child = ipv4_route_table}, |
f8572d8f | 3223 | { .procname = "neigh", |
d994af0d | 3224 | .mode = 0555, .child = empty}, |
2f4520d3 AV |
3225 | { } |
3226 | }; | |
3227 | ||
3228 | static __net_initdata struct ctl_path ipv4_path[] = { | |
f8572d8f EB |
3229 | { .procname = "net", }, |
3230 | { .procname = "ipv4", }, | |
39a23e75 DL |
3231 | { }, |
3232 | }; | |
3233 | ||
39a23e75 DL |
3234 | static struct ctl_table ipv4_route_flush_table[] = { |
3235 | { | |
39a23e75 DL |
3236 | .procname = "flush", |
3237 | .maxlen = sizeof(int), | |
3238 | .mode = 0200, | |
6d9f239a | 3239 | .proc_handler = ipv4_sysctl_rtcache_flush, |
39a23e75 | 3240 | }, |
f8572d8f | 3241 | { }, |
39a23e75 DL |
3242 | }; |
3243 | ||
2f4520d3 | 3244 | static __net_initdata struct ctl_path ipv4_route_path[] = { |
f8572d8f EB |
3245 | { .procname = "net", }, |
3246 | { .procname = "ipv4", }, | |
3247 | { .procname = "route", }, | |
2f4520d3 AV |
3248 | { }, |
3249 | }; | |
3250 | ||
39a23e75 DL |
3251 | static __net_init int sysctl_route_net_init(struct net *net) |
3252 | { | |
3253 | struct ctl_table *tbl; | |
3254 | ||
3255 | tbl = ipv4_route_flush_table; | |
09ad9bc7 | 3256 | if (!net_eq(net, &init_net)) { |
39a23e75 DL |
3257 | tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL); |
3258 | if (tbl == NULL) | |
3259 | goto err_dup; | |
3260 | } | |
3261 | tbl[0].extra1 = net; | |
3262 | ||
3263 | net->ipv4.route_hdr = | |
3264 | register_net_sysctl_table(net, ipv4_route_path, tbl); | |
3265 | if (net->ipv4.route_hdr == NULL) | |
3266 | goto err_reg; | |
3267 | return 0; | |
3268 | ||
3269 | err_reg: | |
3270 | if (tbl != ipv4_route_flush_table) | |
3271 | kfree(tbl); | |
3272 | err_dup: | |
3273 | return -ENOMEM; | |
3274 | } | |
3275 | ||
3276 | static __net_exit void sysctl_route_net_exit(struct net *net) | |
3277 | { | |
3278 | struct ctl_table *tbl; | |
3279 | ||
3280 | tbl = net->ipv4.route_hdr->ctl_table_arg; | |
3281 | unregister_net_sysctl_table(net->ipv4.route_hdr); | |
3282 | BUG_ON(tbl == ipv4_route_flush_table); | |
3283 | kfree(tbl); | |
3284 | } | |
3285 | ||
3286 | static __net_initdata struct pernet_operations sysctl_route_ops = { | |
3287 | .init = sysctl_route_net_init, | |
3288 | .exit = sysctl_route_net_exit, | |
3289 | }; | |
1da177e4 LT |
3290 | #endif |
3291 | ||
3ee94372 | 3292 | static __net_init int rt_genid_init(struct net *net) |
9f5e97e5 | 3293 | { |
3ee94372 NH |
3294 | get_random_bytes(&net->ipv4.rt_genid, |
3295 | sizeof(net->ipv4.rt_genid)); | |
436c3b66 DM |
3296 | get_random_bytes(&net->ipv4.dev_addr_genid, |
3297 | sizeof(net->ipv4.dev_addr_genid)); | |
9f5e97e5 DL |
3298 | return 0; |
3299 | } | |
3300 | ||
3ee94372 NH |
3301 | static __net_initdata struct pernet_operations rt_genid_ops = { |
3302 | .init = rt_genid_init, | |
9f5e97e5 DL |
3303 | }; |
3304 | ||
3305 | ||
c7066f70 | 3306 | #ifdef CONFIG_IP_ROUTE_CLASSID |
7d720c3e | 3307 | struct ip_rt_acct __percpu *ip_rt_acct __read_mostly; |
c7066f70 | 3308 | #endif /* CONFIG_IP_ROUTE_CLASSID */ |
1da177e4 LT |
3309 | |
3310 | static __initdata unsigned long rhash_entries; | |
3311 | static int __init set_rhash_entries(char *str) | |
3312 | { | |
3313 | if (!str) | |
3314 | return 0; | |
3315 | rhash_entries = simple_strtoul(str, &str, 0); | |
3316 | return 1; | |
3317 | } | |
3318 | __setup("rhash_entries=", set_rhash_entries); | |
3319 | ||
3320 | int __init ip_rt_init(void) | |
3321 | { | |
424c4b70 | 3322 | int rc = 0; |
1da177e4 | 3323 | |
c7066f70 | 3324 | #ifdef CONFIG_IP_ROUTE_CLASSID |
0dcec8c2 | 3325 | ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct)); |
1da177e4 LT |
3326 | if (!ip_rt_acct) |
3327 | panic("IP: failed to allocate ip_rt_acct\n"); | |
1da177e4 LT |
3328 | #endif |
3329 | ||
e5d679f3 AD |
3330 | ipv4_dst_ops.kmem_cachep = |
3331 | kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0, | |
20c2df83 | 3332 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 3333 | |
14e50e57 DM |
3334 | ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep; |
3335 | ||
fc66f95c ED |
3336 | if (dst_entries_init(&ipv4_dst_ops) < 0) |
3337 | panic("IP: failed to allocate ipv4_dst_ops counter\n"); | |
3338 | ||
3339 | if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0) | |
3340 | panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n"); | |
3341 | ||
424c4b70 ED |
3342 | rt_hash_table = (struct rt_hash_bucket *) |
3343 | alloc_large_system_hash("IP route cache", | |
3344 | sizeof(struct rt_hash_bucket), | |
3345 | rhash_entries, | |
4481374c | 3346 | (totalram_pages >= 128 * 1024) ? |
18955cfc | 3347 | 15 : 17, |
8d1502de | 3348 | 0, |
424c4b70 ED |
3349 | &rt_hash_log, |
3350 | &rt_hash_mask, | |
c9503e0f | 3351 | rhash_entries ? 0 : 512 * 1024); |
22c047cc ED |
3352 | memset(rt_hash_table, 0, (rt_hash_mask + 1) * sizeof(struct rt_hash_bucket)); |
3353 | rt_hash_lock_init(); | |
1da177e4 LT |
3354 | |
3355 | ipv4_dst_ops.gc_thresh = (rt_hash_mask + 1); | |
3356 | ip_rt_max_size = (rt_hash_mask + 1) * 16; | |
3357 | ||
1da177e4 LT |
3358 | devinet_init(); |
3359 | ip_fib_init(); | |
3360 | ||
73b38711 | 3361 | if (ip_rt_proc_init()) |
107f1634 | 3362 | printk(KERN_ERR "Unable to create route proc files\n"); |
1da177e4 LT |
3363 | #ifdef CONFIG_XFRM |
3364 | xfrm_init(); | |
a33bc5c1 | 3365 | xfrm4_init(ip_rt_max_size); |
1da177e4 | 3366 | #endif |
c7ac8679 | 3367 | rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL, NULL); |
63f3444f | 3368 | |
39a23e75 DL |
3369 | #ifdef CONFIG_SYSCTL |
3370 | register_pernet_subsys(&sysctl_route_ops); | |
3371 | #endif | |
3ee94372 | 3372 | register_pernet_subsys(&rt_genid_ops); |
1da177e4 LT |
3373 | return rc; |
3374 | } | |
3375 | ||
a1bc6eb4 | 3376 | #ifdef CONFIG_SYSCTL |
eeb61f71 AV |
3377 | /* |
3378 | * We really need to sanitize the damn ipv4 init order, then all | |
3379 | * this nonsense will go away. | |
3380 | */ | |
3381 | void __init ip_static_sysctl_init(void) | |
3382 | { | |
2f4520d3 | 3383 | register_sysctl_paths(ipv4_path, ipv4_skeleton); |
eeb61f71 | 3384 | } |
a1bc6eb4 | 3385 | #endif |