]>
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 | |
111 | ||
68a5e3dd DM |
112 | #define RT_FL_TOS(oldflp4) \ |
113 | ((u32)(oldflp4->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK))) | |
1da177e4 LT |
114 | |
115 | #define IP_MAX_MTU 0xFFF0 | |
116 | ||
117 | #define RT_GC_TIMEOUT (300*HZ) | |
118 | ||
1da177e4 | 119 | static int ip_rt_max_size; |
817bc4db SH |
120 | static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT; |
121 | static int ip_rt_gc_interval __read_mostly = 60 * HZ; | |
122 | static int ip_rt_gc_min_interval __read_mostly = HZ / 2; | |
123 | static int ip_rt_redirect_number __read_mostly = 9; | |
124 | static int ip_rt_redirect_load __read_mostly = HZ / 50; | |
125 | static int ip_rt_redirect_silence __read_mostly = ((HZ / 50) << (9 + 1)); | |
126 | static int ip_rt_error_cost __read_mostly = HZ; | |
127 | static int ip_rt_error_burst __read_mostly = 5 * HZ; | |
128 | static int ip_rt_gc_elasticity __read_mostly = 8; | |
129 | static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ; | |
130 | static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20; | |
131 | static int ip_rt_min_advmss __read_mostly = 256; | |
1080d709 | 132 | static int rt_chain_length_max __read_mostly = 20; |
1da177e4 | 133 | |
1da177e4 LT |
134 | /* |
135 | * Interface to generic destination cache. | |
136 | */ | |
137 | ||
138 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie); | |
0dbaee3b | 139 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst); |
d33e4553 | 140 | static unsigned int ipv4_default_mtu(const struct dst_entry *dst); |
1da177e4 | 141 | static void ipv4_dst_destroy(struct dst_entry *dst); |
1da177e4 LT |
142 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst); |
143 | static void ipv4_link_failure(struct sk_buff *skb); | |
144 | static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu); | |
569d3645 | 145 | static int rt_garbage_collect(struct dst_ops *ops); |
1da177e4 | 146 | |
72cdd1d9 ED |
147 | static void ipv4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, |
148 | int how) | |
149 | { | |
150 | } | |
1da177e4 | 151 | |
62fa8a84 DM |
152 | static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old) |
153 | { | |
06582540 DM |
154 | struct rtable *rt = (struct rtable *) dst; |
155 | struct inet_peer *peer; | |
156 | u32 *p = NULL; | |
157 | ||
158 | if (!rt->peer) | |
159 | rt_bind_peer(rt, 1); | |
62fa8a84 | 160 | |
06582540 DM |
161 | peer = rt->peer; |
162 | if (peer) { | |
62fa8a84 DM |
163 | u32 *old_p = __DST_METRICS_PTR(old); |
164 | unsigned long prev, new; | |
165 | ||
06582540 DM |
166 | p = peer->metrics; |
167 | if (inet_metrics_new(peer)) | |
168 | memcpy(p, old_p, sizeof(u32) * RTAX_MAX); | |
62fa8a84 DM |
169 | |
170 | new = (unsigned long) p; | |
171 | prev = cmpxchg(&dst->_metrics, old, new); | |
172 | ||
173 | if (prev != old) { | |
62fa8a84 DM |
174 | p = __DST_METRICS_PTR(prev); |
175 | if (prev & DST_METRICS_READ_ONLY) | |
176 | p = NULL; | |
177 | } else { | |
62fa8a84 DM |
178 | if (rt->fi) { |
179 | fib_info_put(rt->fi); | |
180 | rt->fi = NULL; | |
181 | } | |
182 | } | |
183 | } | |
184 | return p; | |
185 | } | |
186 | ||
1da177e4 LT |
187 | static struct dst_ops ipv4_dst_ops = { |
188 | .family = AF_INET, | |
09640e63 | 189 | .protocol = cpu_to_be16(ETH_P_IP), |
1da177e4 LT |
190 | .gc = rt_garbage_collect, |
191 | .check = ipv4_dst_check, | |
0dbaee3b | 192 | .default_advmss = ipv4_default_advmss, |
d33e4553 | 193 | .default_mtu = ipv4_default_mtu, |
62fa8a84 | 194 | .cow_metrics = ipv4_cow_metrics, |
1da177e4 LT |
195 | .destroy = ipv4_dst_destroy, |
196 | .ifdown = ipv4_dst_ifdown, | |
197 | .negative_advice = ipv4_negative_advice, | |
198 | .link_failure = ipv4_link_failure, | |
199 | .update_pmtu = ip_rt_update_pmtu, | |
1ac06e03 | 200 | .local_out = __ip_local_out, |
1da177e4 LT |
201 | }; |
202 | ||
203 | #define ECN_OR_COST(class) TC_PRIO_##class | |
204 | ||
4839c52b | 205 | const __u8 ip_tos2prio[16] = { |
1da177e4 | 206 | TC_PRIO_BESTEFFORT, |
4a2b9c37 | 207 | ECN_OR_COST(BESTEFFORT), |
1da177e4 LT |
208 | TC_PRIO_BESTEFFORT, |
209 | ECN_OR_COST(BESTEFFORT), | |
210 | TC_PRIO_BULK, | |
211 | ECN_OR_COST(BULK), | |
212 | TC_PRIO_BULK, | |
213 | ECN_OR_COST(BULK), | |
214 | TC_PRIO_INTERACTIVE, | |
215 | ECN_OR_COST(INTERACTIVE), | |
216 | TC_PRIO_INTERACTIVE, | |
217 | ECN_OR_COST(INTERACTIVE), | |
218 | TC_PRIO_INTERACTIVE_BULK, | |
219 | ECN_OR_COST(INTERACTIVE_BULK), | |
220 | TC_PRIO_INTERACTIVE_BULK, | |
221 | ECN_OR_COST(INTERACTIVE_BULK) | |
222 | }; | |
223 | ||
224 | ||
225 | /* | |
226 | * Route cache. | |
227 | */ | |
228 | ||
229 | /* The locking scheme is rather straight forward: | |
230 | * | |
231 | * 1) Read-Copy Update protects the buckets of the central route hash. | |
232 | * 2) Only writers remove entries, and they hold the lock | |
233 | * as they look at rtable reference counts. | |
234 | * 3) Only readers acquire references to rtable entries, | |
235 | * they do so with atomic increments and with the | |
236 | * lock held. | |
237 | */ | |
238 | ||
239 | struct rt_hash_bucket { | |
1c31720a | 240 | struct rtable __rcu *chain; |
22c047cc | 241 | }; |
1080d709 | 242 | |
8a25d5de IM |
243 | #if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK) || \ |
244 | defined(CONFIG_PROVE_LOCKING) | |
22c047cc ED |
245 | /* |
246 | * Instead of using one spinlock for each rt_hash_bucket, we use a table of spinlocks | |
247 | * The size of this table is a power of two and depends on the number of CPUS. | |
62051200 | 248 | * (on lockdep we have a quite big spinlock_t, so keep the size down there) |
22c047cc | 249 | */ |
62051200 IM |
250 | #ifdef CONFIG_LOCKDEP |
251 | # define RT_HASH_LOCK_SZ 256 | |
22c047cc | 252 | #else |
62051200 IM |
253 | # if NR_CPUS >= 32 |
254 | # define RT_HASH_LOCK_SZ 4096 | |
255 | # elif NR_CPUS >= 16 | |
256 | # define RT_HASH_LOCK_SZ 2048 | |
257 | # elif NR_CPUS >= 8 | |
258 | # define RT_HASH_LOCK_SZ 1024 | |
259 | # elif NR_CPUS >= 4 | |
260 | # define RT_HASH_LOCK_SZ 512 | |
261 | # else | |
262 | # define RT_HASH_LOCK_SZ 256 | |
263 | # endif | |
22c047cc ED |
264 | #endif |
265 | ||
266 | static spinlock_t *rt_hash_locks; | |
267 | # define rt_hash_lock_addr(slot) &rt_hash_locks[(slot) & (RT_HASH_LOCK_SZ - 1)] | |
1ff1cc20 PE |
268 | |
269 | static __init void rt_hash_lock_init(void) | |
270 | { | |
271 | int i; | |
272 | ||
273 | rt_hash_locks = kmalloc(sizeof(spinlock_t) * RT_HASH_LOCK_SZ, | |
274 | GFP_KERNEL); | |
275 | if (!rt_hash_locks) | |
276 | panic("IP: failed to allocate rt_hash_locks\n"); | |
277 | ||
278 | for (i = 0; i < RT_HASH_LOCK_SZ; i++) | |
279 | spin_lock_init(&rt_hash_locks[i]); | |
280 | } | |
22c047cc ED |
281 | #else |
282 | # define rt_hash_lock_addr(slot) NULL | |
1ff1cc20 PE |
283 | |
284 | static inline void rt_hash_lock_init(void) | |
285 | { | |
286 | } | |
22c047cc | 287 | #endif |
1da177e4 | 288 | |
817bc4db SH |
289 | static struct rt_hash_bucket *rt_hash_table __read_mostly; |
290 | static unsigned rt_hash_mask __read_mostly; | |
291 | static unsigned int rt_hash_log __read_mostly; | |
1da177e4 | 292 | |
2f970d83 | 293 | static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat); |
27f39c73 | 294 | #define RT_CACHE_STAT_INC(field) __this_cpu_inc(rt_cache_stat.field) |
1da177e4 | 295 | |
b00180de | 296 | static inline unsigned int rt_hash(__be32 daddr, __be32 saddr, int idx, |
0eae88f3 | 297 | int genid) |
1da177e4 | 298 | { |
0eae88f3 | 299 | return jhash_3words((__force u32)daddr, (__force u32)saddr, |
b00180de | 300 | idx, genid) |
29e75252 | 301 | & rt_hash_mask; |
1da177e4 LT |
302 | } |
303 | ||
e84f84f2 DL |
304 | static inline int rt_genid(struct net *net) |
305 | { | |
306 | return atomic_read(&net->ipv4.rt_genid); | |
307 | } | |
308 | ||
1da177e4 LT |
309 | #ifdef CONFIG_PROC_FS |
310 | struct rt_cache_iter_state { | |
a75e936f | 311 | struct seq_net_private p; |
1da177e4 | 312 | int bucket; |
29e75252 | 313 | int genid; |
1da177e4 LT |
314 | }; |
315 | ||
1218854a | 316 | static struct rtable *rt_cache_get_first(struct seq_file *seq) |
1da177e4 | 317 | { |
1218854a | 318 | struct rt_cache_iter_state *st = seq->private; |
1da177e4 | 319 | struct rtable *r = NULL; |
1da177e4 LT |
320 | |
321 | for (st->bucket = rt_hash_mask; st->bucket >= 0; --st->bucket) { | |
1c31720a | 322 | if (!rcu_dereference_raw(rt_hash_table[st->bucket].chain)) |
a6272665 | 323 | continue; |
1da177e4 | 324 | rcu_read_lock_bh(); |
a898def2 | 325 | r = rcu_dereference_bh(rt_hash_table[st->bucket].chain); |
29e75252 | 326 | while (r) { |
d8d1f30b | 327 | if (dev_net(r->dst.dev) == seq_file_net(seq) && |
a75e936f | 328 | r->rt_genid == st->genid) |
29e75252 | 329 | return r; |
d8d1f30b | 330 | r = rcu_dereference_bh(r->dst.rt_next); |
29e75252 | 331 | } |
1da177e4 LT |
332 | rcu_read_unlock_bh(); |
333 | } | |
29e75252 | 334 | return r; |
1da177e4 LT |
335 | } |
336 | ||
1218854a | 337 | static struct rtable *__rt_cache_get_next(struct seq_file *seq, |
642d6318 | 338 | struct rtable *r) |
1da177e4 | 339 | { |
1218854a | 340 | struct rt_cache_iter_state *st = seq->private; |
a6272665 | 341 | |
1c31720a | 342 | r = rcu_dereference_bh(r->dst.rt_next); |
1da177e4 LT |
343 | while (!r) { |
344 | rcu_read_unlock_bh(); | |
a6272665 ED |
345 | do { |
346 | if (--st->bucket < 0) | |
347 | return NULL; | |
1c31720a | 348 | } while (!rcu_dereference_raw(rt_hash_table[st->bucket].chain)); |
1da177e4 | 349 | rcu_read_lock_bh(); |
1c31720a | 350 | r = rcu_dereference_bh(rt_hash_table[st->bucket].chain); |
1da177e4 | 351 | } |
1c31720a | 352 | return r; |
1da177e4 LT |
353 | } |
354 | ||
1218854a | 355 | static struct rtable *rt_cache_get_next(struct seq_file *seq, |
642d6318 DL |
356 | struct rtable *r) |
357 | { | |
1218854a YH |
358 | struct rt_cache_iter_state *st = seq->private; |
359 | while ((r = __rt_cache_get_next(seq, r)) != NULL) { | |
d8d1f30b | 360 | if (dev_net(r->dst.dev) != seq_file_net(seq)) |
a75e936f | 361 | continue; |
642d6318 DL |
362 | if (r->rt_genid == st->genid) |
363 | break; | |
364 | } | |
365 | return r; | |
366 | } | |
367 | ||
1218854a | 368 | static struct rtable *rt_cache_get_idx(struct seq_file *seq, loff_t pos) |
1da177e4 | 369 | { |
1218854a | 370 | struct rtable *r = rt_cache_get_first(seq); |
1da177e4 LT |
371 | |
372 | if (r) | |
1218854a | 373 | while (pos && (r = rt_cache_get_next(seq, r))) |
1da177e4 LT |
374 | --pos; |
375 | return pos ? NULL : r; | |
376 | } | |
377 | ||
378 | static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos) | |
379 | { | |
29e75252 | 380 | struct rt_cache_iter_state *st = seq->private; |
29e75252 | 381 | if (*pos) |
1218854a | 382 | return rt_cache_get_idx(seq, *pos - 1); |
e84f84f2 | 383 | st->genid = rt_genid(seq_file_net(seq)); |
29e75252 | 384 | return SEQ_START_TOKEN; |
1da177e4 LT |
385 | } |
386 | ||
387 | static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
388 | { | |
29e75252 | 389 | struct rtable *r; |
1da177e4 LT |
390 | |
391 | if (v == SEQ_START_TOKEN) | |
1218854a | 392 | r = rt_cache_get_first(seq); |
1da177e4 | 393 | else |
1218854a | 394 | r = rt_cache_get_next(seq, v); |
1da177e4 LT |
395 | ++*pos; |
396 | return r; | |
397 | } | |
398 | ||
399 | static void rt_cache_seq_stop(struct seq_file *seq, void *v) | |
400 | { | |
401 | if (v && v != SEQ_START_TOKEN) | |
402 | rcu_read_unlock_bh(); | |
403 | } | |
404 | ||
405 | static int rt_cache_seq_show(struct seq_file *seq, void *v) | |
406 | { | |
407 | if (v == SEQ_START_TOKEN) | |
408 | seq_printf(seq, "%-127s\n", | |
409 | "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t" | |
410 | "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t" | |
411 | "HHUptod\tSpecDst"); | |
412 | else { | |
413 | struct rtable *r = v; | |
5e659e4c | 414 | int len; |
1da177e4 | 415 | |
0eae88f3 ED |
416 | seq_printf(seq, "%s\t%08X\t%08X\t%8X\t%d\t%u\t%d\t" |
417 | "%08X\t%d\t%u\t%u\t%02X\t%d\t%1d\t%08X%n", | |
d8d1f30b | 418 | r->dst.dev ? r->dst.dev->name : "*", |
0eae88f3 ED |
419 | (__force u32)r->rt_dst, |
420 | (__force u32)r->rt_gateway, | |
d8d1f30b CG |
421 | r->rt_flags, atomic_read(&r->dst.__refcnt), |
422 | r->dst.__use, 0, (__force u32)r->rt_src, | |
0dbaee3b | 423 | dst_metric_advmss(&r->dst) + 40, |
d8d1f30b CG |
424 | dst_metric(&r->dst, RTAX_WINDOW), |
425 | (int)((dst_metric(&r->dst, RTAX_RTT) >> 3) + | |
426 | dst_metric(&r->dst, RTAX_RTTVAR)), | |
5e2b61f7 | 427 | r->rt_tos, |
d8d1f30b CG |
428 | r->dst.hh ? atomic_read(&r->dst.hh->hh_refcnt) : -1, |
429 | r->dst.hh ? (r->dst.hh->hh_output == | |
1da177e4 | 430 | dev_queue_xmit) : 0, |
5e659e4c PE |
431 | r->rt_spec_dst, &len); |
432 | ||
433 | seq_printf(seq, "%*s\n", 127 - len, ""); | |
e905a9ed YH |
434 | } |
435 | return 0; | |
1da177e4 LT |
436 | } |
437 | ||
f690808e | 438 | static const struct seq_operations rt_cache_seq_ops = { |
1da177e4 LT |
439 | .start = rt_cache_seq_start, |
440 | .next = rt_cache_seq_next, | |
441 | .stop = rt_cache_seq_stop, | |
442 | .show = rt_cache_seq_show, | |
443 | }; | |
444 | ||
445 | static int rt_cache_seq_open(struct inode *inode, struct file *file) | |
446 | { | |
a75e936f | 447 | return seq_open_net(inode, file, &rt_cache_seq_ops, |
cf7732e4 | 448 | sizeof(struct rt_cache_iter_state)); |
1da177e4 LT |
449 | } |
450 | ||
9a32144e | 451 | static const struct file_operations rt_cache_seq_fops = { |
1da177e4 LT |
452 | .owner = THIS_MODULE, |
453 | .open = rt_cache_seq_open, | |
454 | .read = seq_read, | |
455 | .llseek = seq_lseek, | |
a75e936f | 456 | .release = seq_release_net, |
1da177e4 LT |
457 | }; |
458 | ||
459 | ||
460 | static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos) | |
461 | { | |
462 | int cpu; | |
463 | ||
464 | if (*pos == 0) | |
465 | return SEQ_START_TOKEN; | |
466 | ||
0f23174a | 467 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
468 | if (!cpu_possible(cpu)) |
469 | continue; | |
470 | *pos = cpu+1; | |
2f970d83 | 471 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
472 | } |
473 | return NULL; | |
474 | } | |
475 | ||
476 | static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
477 | { | |
478 | int cpu; | |
479 | ||
0f23174a | 480 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
481 | if (!cpu_possible(cpu)) |
482 | continue; | |
483 | *pos = cpu+1; | |
2f970d83 | 484 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
485 | } |
486 | return NULL; | |
e905a9ed | 487 | |
1da177e4 LT |
488 | } |
489 | ||
490 | static void rt_cpu_seq_stop(struct seq_file *seq, void *v) | |
491 | { | |
492 | ||
493 | } | |
494 | ||
495 | static int rt_cpu_seq_show(struct seq_file *seq, void *v) | |
496 | { | |
497 | struct rt_cache_stat *st = v; | |
498 | ||
499 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 500 | 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 |
501 | return 0; |
502 | } | |
e905a9ed | 503 | |
1da177e4 LT |
504 | seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x " |
505 | " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n", | |
fc66f95c | 506 | dst_entries_get_slow(&ipv4_dst_ops), |
1da177e4 LT |
507 | st->in_hit, |
508 | st->in_slow_tot, | |
509 | st->in_slow_mc, | |
510 | st->in_no_route, | |
511 | st->in_brd, | |
512 | st->in_martian_dst, | |
513 | st->in_martian_src, | |
514 | ||
515 | st->out_hit, | |
516 | st->out_slow_tot, | |
e905a9ed | 517 | st->out_slow_mc, |
1da177e4 LT |
518 | |
519 | st->gc_total, | |
520 | st->gc_ignored, | |
521 | st->gc_goal_miss, | |
522 | st->gc_dst_overflow, | |
523 | st->in_hlist_search, | |
524 | st->out_hlist_search | |
525 | ); | |
526 | return 0; | |
527 | } | |
528 | ||
f690808e | 529 | static const struct seq_operations rt_cpu_seq_ops = { |
1da177e4 LT |
530 | .start = rt_cpu_seq_start, |
531 | .next = rt_cpu_seq_next, | |
532 | .stop = rt_cpu_seq_stop, | |
533 | .show = rt_cpu_seq_show, | |
534 | }; | |
535 | ||
536 | ||
537 | static int rt_cpu_seq_open(struct inode *inode, struct file *file) | |
538 | { | |
539 | return seq_open(file, &rt_cpu_seq_ops); | |
540 | } | |
541 | ||
9a32144e | 542 | static const struct file_operations rt_cpu_seq_fops = { |
1da177e4 LT |
543 | .owner = THIS_MODULE, |
544 | .open = rt_cpu_seq_open, | |
545 | .read = seq_read, | |
546 | .llseek = seq_lseek, | |
547 | .release = seq_release, | |
548 | }; | |
549 | ||
c7066f70 | 550 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 551 | static int rt_acct_proc_show(struct seq_file *m, void *v) |
78c686e9 | 552 | { |
a661c419 AD |
553 | struct ip_rt_acct *dst, *src; |
554 | unsigned int i, j; | |
555 | ||
556 | dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL); | |
557 | if (!dst) | |
558 | return -ENOMEM; | |
559 | ||
560 | for_each_possible_cpu(i) { | |
561 | src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i); | |
562 | for (j = 0; j < 256; j++) { | |
563 | dst[j].o_bytes += src[j].o_bytes; | |
564 | dst[j].o_packets += src[j].o_packets; | |
565 | dst[j].i_bytes += src[j].i_bytes; | |
566 | dst[j].i_packets += src[j].i_packets; | |
567 | } | |
78c686e9 PE |
568 | } |
569 | ||
a661c419 AD |
570 | seq_write(m, dst, 256 * sizeof(struct ip_rt_acct)); |
571 | kfree(dst); | |
572 | return 0; | |
573 | } | |
78c686e9 | 574 | |
a661c419 AD |
575 | static int rt_acct_proc_open(struct inode *inode, struct file *file) |
576 | { | |
577 | return single_open(file, rt_acct_proc_show, NULL); | |
78c686e9 | 578 | } |
a661c419 AD |
579 | |
580 | static const struct file_operations rt_acct_proc_fops = { | |
581 | .owner = THIS_MODULE, | |
582 | .open = rt_acct_proc_open, | |
583 | .read = seq_read, | |
584 | .llseek = seq_lseek, | |
585 | .release = single_release, | |
586 | }; | |
78c686e9 | 587 | #endif |
107f1634 | 588 | |
73b38711 | 589 | static int __net_init ip_rt_do_proc_init(struct net *net) |
107f1634 PE |
590 | { |
591 | struct proc_dir_entry *pde; | |
592 | ||
593 | pde = proc_net_fops_create(net, "rt_cache", S_IRUGO, | |
594 | &rt_cache_seq_fops); | |
595 | if (!pde) | |
596 | goto err1; | |
597 | ||
77020720 WC |
598 | pde = proc_create("rt_cache", S_IRUGO, |
599 | net->proc_net_stat, &rt_cpu_seq_fops); | |
107f1634 PE |
600 | if (!pde) |
601 | goto err2; | |
602 | ||
c7066f70 | 603 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 604 | pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops); |
107f1634 PE |
605 | if (!pde) |
606 | goto err3; | |
607 | #endif | |
608 | return 0; | |
609 | ||
c7066f70 | 610 | #ifdef CONFIG_IP_ROUTE_CLASSID |
107f1634 PE |
611 | err3: |
612 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
613 | #endif | |
614 | err2: | |
615 | remove_proc_entry("rt_cache", net->proc_net); | |
616 | err1: | |
617 | return -ENOMEM; | |
618 | } | |
73b38711 DL |
619 | |
620 | static void __net_exit ip_rt_do_proc_exit(struct net *net) | |
621 | { | |
622 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
623 | remove_proc_entry("rt_cache", net->proc_net); | |
c7066f70 | 624 | #ifdef CONFIG_IP_ROUTE_CLASSID |
73b38711 | 625 | remove_proc_entry("rt_acct", net->proc_net); |
0a931acf | 626 | #endif |
73b38711 DL |
627 | } |
628 | ||
629 | static struct pernet_operations ip_rt_proc_ops __net_initdata = { | |
630 | .init = ip_rt_do_proc_init, | |
631 | .exit = ip_rt_do_proc_exit, | |
632 | }; | |
633 | ||
634 | static int __init ip_rt_proc_init(void) | |
635 | { | |
636 | return register_pernet_subsys(&ip_rt_proc_ops); | |
637 | } | |
638 | ||
107f1634 | 639 | #else |
73b38711 | 640 | static inline int ip_rt_proc_init(void) |
107f1634 PE |
641 | { |
642 | return 0; | |
643 | } | |
1da177e4 | 644 | #endif /* CONFIG_PROC_FS */ |
e905a9ed | 645 | |
5969f71d | 646 | static inline void rt_free(struct rtable *rt) |
1da177e4 | 647 | { |
d8d1f30b | 648 | call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free); |
1da177e4 LT |
649 | } |
650 | ||
5969f71d | 651 | static inline void rt_drop(struct rtable *rt) |
1da177e4 | 652 | { |
1da177e4 | 653 | ip_rt_put(rt); |
d8d1f30b | 654 | call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free); |
1da177e4 LT |
655 | } |
656 | ||
5969f71d | 657 | static inline int rt_fast_clean(struct rtable *rth) |
1da177e4 LT |
658 | { |
659 | /* Kill broadcast/multicast entries very aggresively, if they | |
660 | collide in hash table with more useful entries */ | |
661 | return (rth->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && | |
c7537967 | 662 | rt_is_input_route(rth) && rth->dst.rt_next; |
1da177e4 LT |
663 | } |
664 | ||
5969f71d | 665 | static inline int rt_valuable(struct rtable *rth) |
1da177e4 LT |
666 | { |
667 | return (rth->rt_flags & (RTCF_REDIRECTED | RTCF_NOTIFY)) || | |
2c8cec5c | 668 | (rth->peer && rth->peer->pmtu_expires); |
1da177e4 LT |
669 | } |
670 | ||
671 | static int rt_may_expire(struct rtable *rth, unsigned long tmo1, unsigned long tmo2) | |
672 | { | |
673 | unsigned long age; | |
674 | int ret = 0; | |
675 | ||
d8d1f30b | 676 | if (atomic_read(&rth->dst.__refcnt)) |
1da177e4 LT |
677 | goto out; |
678 | ||
d8d1f30b | 679 | age = jiffies - rth->dst.lastuse; |
1da177e4 LT |
680 | if ((age <= tmo1 && !rt_fast_clean(rth)) || |
681 | (age <= tmo2 && rt_valuable(rth))) | |
682 | goto out; | |
683 | ret = 1; | |
684 | out: return ret; | |
685 | } | |
686 | ||
687 | /* Bits of score are: | |
688 | * 31: very valuable | |
689 | * 30: not quite useless | |
690 | * 29..0: usage counter | |
691 | */ | |
692 | static inline u32 rt_score(struct rtable *rt) | |
693 | { | |
d8d1f30b | 694 | u32 score = jiffies - rt->dst.lastuse; |
1da177e4 LT |
695 | |
696 | score = ~score & ~(3<<30); | |
697 | ||
698 | if (rt_valuable(rt)) | |
699 | score |= (1<<31); | |
700 | ||
c7537967 | 701 | if (rt_is_output_route(rt) || |
1da177e4 LT |
702 | !(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL))) |
703 | score |= (1<<30); | |
704 | ||
705 | return score; | |
706 | } | |
707 | ||
1080d709 NH |
708 | static inline bool rt_caching(const struct net *net) |
709 | { | |
710 | return net->ipv4.current_rt_cache_rebuild_count <= | |
711 | net->ipv4.sysctl_rt_cache_rebuild_count; | |
712 | } | |
713 | ||
5e2b61f7 DM |
714 | static inline bool compare_hash_inputs(const struct rtable *rt1, |
715 | const struct rtable *rt2) | |
1080d709 | 716 | { |
5e2b61f7 DM |
717 | return ((((__force u32)rt1->rt_key_dst ^ (__force u32)rt2->rt_key_dst) | |
718 | ((__force u32)rt1->rt_key_src ^ (__force u32)rt2->rt_key_src) | | |
719 | (rt1->rt_iif ^ rt2->rt_iif)) == 0); | |
1080d709 NH |
720 | } |
721 | ||
5e2b61f7 | 722 | static inline int compare_keys(struct rtable *rt1, struct rtable *rt2) |
1da177e4 | 723 | { |
5e2b61f7 DM |
724 | return (((__force u32)rt1->rt_key_dst ^ (__force u32)rt2->rt_key_dst) | |
725 | ((__force u32)rt1->rt_key_src ^ (__force u32)rt2->rt_key_src) | | |
726 | (rt1->rt_mark ^ rt2->rt_mark) | | |
727 | (rt1->rt_tos ^ rt2->rt_tos) | | |
728 | (rt1->rt_oif ^ rt2->rt_oif) | | |
729 | (rt1->rt_iif ^ rt2->rt_iif)) == 0; | |
1da177e4 LT |
730 | } |
731 | ||
b5921910 DL |
732 | static inline int compare_netns(struct rtable *rt1, struct rtable *rt2) |
733 | { | |
d8d1f30b | 734 | return net_eq(dev_net(rt1->dst.dev), dev_net(rt2->dst.dev)); |
b5921910 DL |
735 | } |
736 | ||
e84f84f2 DL |
737 | static inline int rt_is_expired(struct rtable *rth) |
738 | { | |
d8d1f30b | 739 | return rth->rt_genid != rt_genid(dev_net(rth->dst.dev)); |
e84f84f2 DL |
740 | } |
741 | ||
beb659bd ED |
742 | /* |
743 | * Perform a full scan of hash table and free all entries. | |
744 | * Can be called by a softirq or a process. | |
745 | * In the later case, we want to be reschedule if necessary | |
746 | */ | |
6561a3b1 | 747 | static void rt_do_flush(struct net *net, int process_context) |
beb659bd ED |
748 | { |
749 | unsigned int i; | |
750 | struct rtable *rth, *next; | |
751 | ||
752 | for (i = 0; i <= rt_hash_mask; i++) { | |
6561a3b1 DM |
753 | struct rtable __rcu **pprev; |
754 | struct rtable *list; | |
755 | ||
beb659bd ED |
756 | if (process_context && need_resched()) |
757 | cond_resched(); | |
1c31720a | 758 | rth = rcu_dereference_raw(rt_hash_table[i].chain); |
beb659bd ED |
759 | if (!rth) |
760 | continue; | |
761 | ||
762 | spin_lock_bh(rt_hash_lock_addr(i)); | |
32cb5b4e | 763 | |
6561a3b1 DM |
764 | list = NULL; |
765 | pprev = &rt_hash_table[i].chain; | |
766 | rth = rcu_dereference_protected(*pprev, | |
1c31720a | 767 | lockdep_is_held(rt_hash_lock_addr(i))); |
32cb5b4e | 768 | |
6561a3b1 DM |
769 | while (rth) { |
770 | next = rcu_dereference_protected(rth->dst.rt_next, | |
1c31720a | 771 | lockdep_is_held(rt_hash_lock_addr(i))); |
6561a3b1 DM |
772 | |
773 | if (!net || | |
774 | net_eq(dev_net(rth->dst.dev), net)) { | |
775 | rcu_assign_pointer(*pprev, next); | |
776 | rcu_assign_pointer(rth->dst.rt_next, list); | |
777 | list = rth; | |
32cb5b4e | 778 | } else { |
6561a3b1 | 779 | pprev = &rth->dst.rt_next; |
32cb5b4e | 780 | } |
6561a3b1 | 781 | rth = next; |
32cb5b4e | 782 | } |
6561a3b1 | 783 | |
beb659bd ED |
784 | spin_unlock_bh(rt_hash_lock_addr(i)); |
785 | ||
6561a3b1 DM |
786 | for (; list; list = next) { |
787 | next = rcu_dereference_protected(list->dst.rt_next, 1); | |
788 | rt_free(list); | |
beb659bd ED |
789 | } |
790 | } | |
791 | } | |
792 | ||
1080d709 NH |
793 | /* |
794 | * While freeing expired entries, we compute average chain length | |
795 | * and standard deviation, using fixed-point arithmetic. | |
796 | * This to have an estimation of rt_chain_length_max | |
797 | * rt_chain_length_max = max(elasticity, AVG + 4*SD) | |
798 | * We use 3 bits for frational part, and 29 (or 61) for magnitude. | |
799 | */ | |
800 | ||
801 | #define FRACT_BITS 3 | |
802 | #define ONE (1UL << FRACT_BITS) | |
803 | ||
98376387 ED |
804 | /* |
805 | * Given a hash chain and an item in this hash chain, | |
806 | * find if a previous entry has the same hash_inputs | |
807 | * (but differs on tos, mark or oif) | |
808 | * Returns 0 if an alias is found. | |
809 | * Returns ONE if rth has no alias before itself. | |
810 | */ | |
811 | static int has_noalias(const struct rtable *head, const struct rtable *rth) | |
812 | { | |
813 | const struct rtable *aux = head; | |
814 | ||
815 | while (aux != rth) { | |
5e2b61f7 | 816 | if (compare_hash_inputs(aux, rth)) |
98376387 | 817 | return 0; |
1c31720a | 818 | aux = rcu_dereference_protected(aux->dst.rt_next, 1); |
98376387 ED |
819 | } |
820 | return ONE; | |
821 | } | |
822 | ||
29e75252 | 823 | /* |
25985edc | 824 | * Perturbation of rt_genid by a small quantity [1..256] |
29e75252 ED |
825 | * Using 8 bits of shuffling ensure we can call rt_cache_invalidate() |
826 | * many times (2^24) without giving recent rt_genid. | |
827 | * Jenkins hash is strong enough that litle changes of rt_genid are OK. | |
1da177e4 | 828 | */ |
86c657f6 | 829 | static void rt_cache_invalidate(struct net *net) |
1da177e4 | 830 | { |
29e75252 | 831 | unsigned char shuffle; |
1da177e4 | 832 | |
29e75252 | 833 | get_random_bytes(&shuffle, sizeof(shuffle)); |
e84f84f2 | 834 | atomic_add(shuffle + 1U, &net->ipv4.rt_genid); |
1da177e4 LT |
835 | } |
836 | ||
29e75252 ED |
837 | /* |
838 | * delay < 0 : invalidate cache (fast : entries will be deleted later) | |
839 | * delay >= 0 : invalidate & flush cache (can be long) | |
840 | */ | |
76e6ebfb | 841 | void rt_cache_flush(struct net *net, int delay) |
1da177e4 | 842 | { |
86c657f6 | 843 | rt_cache_invalidate(net); |
29e75252 | 844 | if (delay >= 0) |
6561a3b1 | 845 | rt_do_flush(net, !in_softirq()); |
1da177e4 LT |
846 | } |
847 | ||
a5ee1551 | 848 | /* Flush previous cache invalidated entries from the cache */ |
6561a3b1 | 849 | void rt_cache_flush_batch(struct net *net) |
a5ee1551 | 850 | { |
6561a3b1 | 851 | rt_do_flush(net, !in_softirq()); |
a5ee1551 EB |
852 | } |
853 | ||
1080d709 NH |
854 | static void rt_emergency_hash_rebuild(struct net *net) |
855 | { | |
3ee94372 | 856 | if (net_ratelimit()) |
1080d709 | 857 | printk(KERN_WARNING "Route hash chain too long!\n"); |
3ee94372 | 858 | rt_cache_invalidate(net); |
1080d709 NH |
859 | } |
860 | ||
1da177e4 LT |
861 | /* |
862 | Short description of GC goals. | |
863 | ||
864 | We want to build algorithm, which will keep routing cache | |
865 | at some equilibrium point, when number of aged off entries | |
866 | is kept approximately equal to newly generated ones. | |
867 | ||
868 | Current expiration strength is variable "expire". | |
869 | We try to adjust it dynamically, so that if networking | |
870 | is idle expires is large enough to keep enough of warm entries, | |
871 | and when load increases it reduces to limit cache size. | |
872 | */ | |
873 | ||
569d3645 | 874 | static int rt_garbage_collect(struct dst_ops *ops) |
1da177e4 LT |
875 | { |
876 | static unsigned long expire = RT_GC_TIMEOUT; | |
877 | static unsigned long last_gc; | |
878 | static int rover; | |
879 | static int equilibrium; | |
1c31720a ED |
880 | struct rtable *rth; |
881 | struct rtable __rcu **rthp; | |
1da177e4 LT |
882 | unsigned long now = jiffies; |
883 | int goal; | |
fc66f95c | 884 | int entries = dst_entries_get_fast(&ipv4_dst_ops); |
1da177e4 LT |
885 | |
886 | /* | |
887 | * Garbage collection is pretty expensive, | |
888 | * do not make it too frequently. | |
889 | */ | |
890 | ||
891 | RT_CACHE_STAT_INC(gc_total); | |
892 | ||
893 | if (now - last_gc < ip_rt_gc_min_interval && | |
fc66f95c | 894 | entries < ip_rt_max_size) { |
1da177e4 LT |
895 | RT_CACHE_STAT_INC(gc_ignored); |
896 | goto out; | |
897 | } | |
898 | ||
fc66f95c | 899 | entries = dst_entries_get_slow(&ipv4_dst_ops); |
1da177e4 | 900 | /* Calculate number of entries, which we want to expire now. */ |
fc66f95c | 901 | goal = entries - (ip_rt_gc_elasticity << rt_hash_log); |
1da177e4 LT |
902 | if (goal <= 0) { |
903 | if (equilibrium < ipv4_dst_ops.gc_thresh) | |
904 | equilibrium = ipv4_dst_ops.gc_thresh; | |
fc66f95c | 905 | goal = entries - equilibrium; |
1da177e4 | 906 | if (goal > 0) { |
b790cedd | 907 | equilibrium += min_t(unsigned int, goal >> 1, rt_hash_mask + 1); |
fc66f95c | 908 | goal = entries - equilibrium; |
1da177e4 LT |
909 | } |
910 | } else { | |
911 | /* We are in dangerous area. Try to reduce cache really | |
912 | * aggressively. | |
913 | */ | |
b790cedd | 914 | goal = max_t(unsigned int, goal >> 1, rt_hash_mask + 1); |
fc66f95c | 915 | equilibrium = entries - goal; |
1da177e4 LT |
916 | } |
917 | ||
918 | if (now - last_gc >= ip_rt_gc_min_interval) | |
919 | last_gc = now; | |
920 | ||
921 | if (goal <= 0) { | |
922 | equilibrium += goal; | |
923 | goto work_done; | |
924 | } | |
925 | ||
926 | do { | |
927 | int i, k; | |
928 | ||
929 | for (i = rt_hash_mask, k = rover; i >= 0; i--) { | |
930 | unsigned long tmo = expire; | |
931 | ||
932 | k = (k + 1) & rt_hash_mask; | |
933 | rthp = &rt_hash_table[k].chain; | |
22c047cc | 934 | spin_lock_bh(rt_hash_lock_addr(k)); |
1c31720a ED |
935 | while ((rth = rcu_dereference_protected(*rthp, |
936 | lockdep_is_held(rt_hash_lock_addr(k)))) != NULL) { | |
e84f84f2 | 937 | if (!rt_is_expired(rth) && |
29e75252 | 938 | !rt_may_expire(rth, tmo, expire)) { |
1da177e4 | 939 | tmo >>= 1; |
d8d1f30b | 940 | rthp = &rth->dst.rt_next; |
1da177e4 LT |
941 | continue; |
942 | } | |
d8d1f30b | 943 | *rthp = rth->dst.rt_next; |
1da177e4 LT |
944 | rt_free(rth); |
945 | goal--; | |
1da177e4 | 946 | } |
22c047cc | 947 | spin_unlock_bh(rt_hash_lock_addr(k)); |
1da177e4 LT |
948 | if (goal <= 0) |
949 | break; | |
950 | } | |
951 | rover = k; | |
952 | ||
953 | if (goal <= 0) | |
954 | goto work_done; | |
955 | ||
956 | /* Goal is not achieved. We stop process if: | |
957 | ||
958 | - if expire reduced to zero. Otherwise, expire is halfed. | |
959 | - if table is not full. | |
960 | - if we are called from interrupt. | |
961 | - jiffies check is just fallback/debug loop breaker. | |
962 | We will not spin here for long time in any case. | |
963 | */ | |
964 | ||
965 | RT_CACHE_STAT_INC(gc_goal_miss); | |
966 | ||
967 | if (expire == 0) | |
968 | break; | |
969 | ||
970 | expire >>= 1; | |
971 | #if RT_CACHE_DEBUG >= 2 | |
972 | printk(KERN_DEBUG "expire>> %u %d %d %d\n", expire, | |
fc66f95c | 973 | dst_entries_get_fast(&ipv4_dst_ops), goal, i); |
1da177e4 LT |
974 | #endif |
975 | ||
fc66f95c | 976 | if (dst_entries_get_fast(&ipv4_dst_ops) < ip_rt_max_size) |
1da177e4 LT |
977 | goto out; |
978 | } while (!in_softirq() && time_before_eq(jiffies, now)); | |
979 | ||
fc66f95c ED |
980 | if (dst_entries_get_fast(&ipv4_dst_ops) < ip_rt_max_size) |
981 | goto out; | |
982 | if (dst_entries_get_slow(&ipv4_dst_ops) < ip_rt_max_size) | |
1da177e4 LT |
983 | goto out; |
984 | if (net_ratelimit()) | |
985 | printk(KERN_WARNING "dst cache overflow\n"); | |
986 | RT_CACHE_STAT_INC(gc_dst_overflow); | |
987 | return 1; | |
988 | ||
989 | work_done: | |
990 | expire += ip_rt_gc_min_interval; | |
991 | if (expire > ip_rt_gc_timeout || | |
fc66f95c ED |
992 | dst_entries_get_fast(&ipv4_dst_ops) < ipv4_dst_ops.gc_thresh || |
993 | dst_entries_get_slow(&ipv4_dst_ops) < ipv4_dst_ops.gc_thresh) | |
1da177e4 LT |
994 | expire = ip_rt_gc_timeout; |
995 | #if RT_CACHE_DEBUG >= 2 | |
996 | printk(KERN_DEBUG "expire++ %u %d %d %d\n", expire, | |
fc66f95c | 997 | dst_entries_get_fast(&ipv4_dst_ops), goal, rover); |
1da177e4 LT |
998 | #endif |
999 | out: return 0; | |
1000 | } | |
1001 | ||
98376387 ED |
1002 | /* |
1003 | * Returns number of entries in a hash chain that have different hash_inputs | |
1004 | */ | |
1005 | static int slow_chain_length(const struct rtable *head) | |
1006 | { | |
1007 | int length = 0; | |
1008 | const struct rtable *rth = head; | |
1009 | ||
1010 | while (rth) { | |
1011 | length += has_noalias(head, rth); | |
1c31720a | 1012 | rth = rcu_dereference_protected(rth->dst.rt_next, 1); |
98376387 ED |
1013 | } |
1014 | return length >> FRACT_BITS; | |
1015 | } | |
1016 | ||
b23dd4fe DM |
1017 | static struct rtable *rt_intern_hash(unsigned hash, struct rtable *rt, |
1018 | struct sk_buff *skb, int ifindex) | |
1da177e4 | 1019 | { |
1c31720a ED |
1020 | struct rtable *rth, *cand; |
1021 | struct rtable __rcu **rthp, **candp; | |
1da177e4 | 1022 | unsigned long now; |
1da177e4 LT |
1023 | u32 min_score; |
1024 | int chain_length; | |
1025 | int attempts = !in_softirq(); | |
1026 | ||
1027 | restart: | |
1028 | chain_length = 0; | |
1029 | min_score = ~(u32)0; | |
1030 | cand = NULL; | |
1031 | candp = NULL; | |
1032 | now = jiffies; | |
1033 | ||
d8d1f30b | 1034 | if (!rt_caching(dev_net(rt->dst.dev))) { |
73e42897 NH |
1035 | /* |
1036 | * If we're not caching, just tell the caller we | |
1037 | * were successful and don't touch the route. The | |
1038 | * caller hold the sole reference to the cache entry, and | |
1039 | * it will be released when the caller is done with it. | |
1040 | * If we drop it here, the callers have no way to resolve routes | |
1041 | * when we're not caching. Instead, just point *rp at rt, so | |
1042 | * the caller gets a single use out of the route | |
b6280b47 NH |
1043 | * Note that we do rt_free on this new route entry, so that |
1044 | * once its refcount hits zero, we are still able to reap it | |
1045 | * (Thanks Alexey) | |
27b75c95 ED |
1046 | * Note: To avoid expensive rcu stuff for this uncached dst, |
1047 | * we set DST_NOCACHE so that dst_release() can free dst without | |
1048 | * waiting a grace period. | |
73e42897 | 1049 | */ |
b6280b47 | 1050 | |
c7d4426a | 1051 | rt->dst.flags |= DST_NOCACHE; |
c7537967 | 1052 | if (rt->rt_type == RTN_UNICAST || rt_is_output_route(rt)) { |
d8d1f30b | 1053 | int err = arp_bind_neighbour(&rt->dst); |
b6280b47 NH |
1054 | if (err) { |
1055 | if (net_ratelimit()) | |
1056 | printk(KERN_WARNING | |
1057 | "Neighbour table failure & not caching routes.\n"); | |
27b75c95 | 1058 | ip_rt_put(rt); |
b23dd4fe | 1059 | return ERR_PTR(err); |
b6280b47 NH |
1060 | } |
1061 | } | |
1062 | ||
b6280b47 | 1063 | goto skip_hashing; |
1080d709 NH |
1064 | } |
1065 | ||
1da177e4 LT |
1066 | rthp = &rt_hash_table[hash].chain; |
1067 | ||
22c047cc | 1068 | spin_lock_bh(rt_hash_lock_addr(hash)); |
1c31720a ED |
1069 | while ((rth = rcu_dereference_protected(*rthp, |
1070 | lockdep_is_held(rt_hash_lock_addr(hash)))) != NULL) { | |
e84f84f2 | 1071 | if (rt_is_expired(rth)) { |
d8d1f30b | 1072 | *rthp = rth->dst.rt_next; |
29e75252 ED |
1073 | rt_free(rth); |
1074 | continue; | |
1075 | } | |
5e2b61f7 | 1076 | if (compare_keys(rth, rt) && compare_netns(rth, rt)) { |
1da177e4 | 1077 | /* Put it first */ |
d8d1f30b | 1078 | *rthp = rth->dst.rt_next; |
1da177e4 LT |
1079 | /* |
1080 | * Since lookup is lockfree, the deletion | |
1081 | * must be visible to another weakly ordered CPU before | |
1082 | * the insertion at the start of the hash chain. | |
1083 | */ | |
d8d1f30b | 1084 | rcu_assign_pointer(rth->dst.rt_next, |
1da177e4 LT |
1085 | rt_hash_table[hash].chain); |
1086 | /* | |
1087 | * Since lookup is lockfree, the update writes | |
1088 | * must be ordered for consistency on SMP. | |
1089 | */ | |
1090 | rcu_assign_pointer(rt_hash_table[hash].chain, rth); | |
1091 | ||
d8d1f30b | 1092 | dst_use(&rth->dst, now); |
22c047cc | 1093 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1094 | |
1095 | rt_drop(rt); | |
b23dd4fe | 1096 | if (skb) |
d8d1f30b | 1097 | skb_dst_set(skb, &rth->dst); |
b23dd4fe | 1098 | return rth; |
1da177e4 LT |
1099 | } |
1100 | ||
d8d1f30b | 1101 | if (!atomic_read(&rth->dst.__refcnt)) { |
1da177e4 LT |
1102 | u32 score = rt_score(rth); |
1103 | ||
1104 | if (score <= min_score) { | |
1105 | cand = rth; | |
1106 | candp = rthp; | |
1107 | min_score = score; | |
1108 | } | |
1109 | } | |
1110 | ||
1111 | chain_length++; | |
1112 | ||
d8d1f30b | 1113 | rthp = &rth->dst.rt_next; |
1da177e4 LT |
1114 | } |
1115 | ||
1116 | if (cand) { | |
1117 | /* ip_rt_gc_elasticity used to be average length of chain | |
1118 | * length, when exceeded gc becomes really aggressive. | |
1119 | * | |
1120 | * The second limit is less certain. At the moment it allows | |
1121 | * only 2 entries per bucket. We will see. | |
1122 | */ | |
1123 | if (chain_length > ip_rt_gc_elasticity) { | |
d8d1f30b | 1124 | *candp = cand->dst.rt_next; |
1da177e4 LT |
1125 | rt_free(cand); |
1126 | } | |
1080d709 | 1127 | } else { |
98376387 ED |
1128 | if (chain_length > rt_chain_length_max && |
1129 | slow_chain_length(rt_hash_table[hash].chain) > rt_chain_length_max) { | |
d8d1f30b | 1130 | struct net *net = dev_net(rt->dst.dev); |
1080d709 | 1131 | int num = ++net->ipv4.current_rt_cache_rebuild_count; |
b35ecb5d | 1132 | if (!rt_caching(net)) { |
1080d709 | 1133 | printk(KERN_WARNING "%s: %d rebuilds is over limit, route caching disabled\n", |
d8d1f30b | 1134 | rt->dst.dev->name, num); |
1080d709 | 1135 | } |
b35ecb5d | 1136 | rt_emergency_hash_rebuild(net); |
6a2bad70 PE |
1137 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1138 | ||
5e2b61f7 | 1139 | hash = rt_hash(rt->rt_key_dst, rt->rt_key_src, |
6a2bad70 PE |
1140 | ifindex, rt_genid(net)); |
1141 | goto restart; | |
1080d709 | 1142 | } |
1da177e4 LT |
1143 | } |
1144 | ||
1145 | /* Try to bind route to arp only if it is output | |
1146 | route or unicast forwarding path. | |
1147 | */ | |
c7537967 | 1148 | if (rt->rt_type == RTN_UNICAST || rt_is_output_route(rt)) { |
d8d1f30b | 1149 | int err = arp_bind_neighbour(&rt->dst); |
1da177e4 | 1150 | if (err) { |
22c047cc | 1151 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1152 | |
1153 | if (err != -ENOBUFS) { | |
1154 | rt_drop(rt); | |
b23dd4fe | 1155 | return ERR_PTR(err); |
1da177e4 LT |
1156 | } |
1157 | ||
1158 | /* Neighbour tables are full and nothing | |
1159 | can be released. Try to shrink route cache, | |
1160 | it is most likely it holds some neighbour records. | |
1161 | */ | |
1162 | if (attempts-- > 0) { | |
1163 | int saved_elasticity = ip_rt_gc_elasticity; | |
1164 | int saved_int = ip_rt_gc_min_interval; | |
1165 | ip_rt_gc_elasticity = 1; | |
1166 | ip_rt_gc_min_interval = 0; | |
569d3645 | 1167 | rt_garbage_collect(&ipv4_dst_ops); |
1da177e4 LT |
1168 | ip_rt_gc_min_interval = saved_int; |
1169 | ip_rt_gc_elasticity = saved_elasticity; | |
1170 | goto restart; | |
1171 | } | |
1172 | ||
1173 | if (net_ratelimit()) | |
7e1b33e5 | 1174 | printk(KERN_WARNING "ipv4: Neighbour table overflow.\n"); |
1da177e4 | 1175 | rt_drop(rt); |
b23dd4fe | 1176 | return ERR_PTR(-ENOBUFS); |
1da177e4 LT |
1177 | } |
1178 | } | |
1179 | ||
d8d1f30b | 1180 | rt->dst.rt_next = rt_hash_table[hash].chain; |
1080d709 | 1181 | |
1da177e4 | 1182 | #if RT_CACHE_DEBUG >= 2 |
d8d1f30b | 1183 | if (rt->dst.rt_next) { |
1da177e4 | 1184 | struct rtable *trt; |
b6280b47 NH |
1185 | printk(KERN_DEBUG "rt_cache @%02x: %pI4", |
1186 | hash, &rt->rt_dst); | |
d8d1f30b | 1187 | for (trt = rt->dst.rt_next; trt; trt = trt->dst.rt_next) |
673d57e7 | 1188 | printk(" . %pI4", &trt->rt_dst); |
1da177e4 LT |
1189 | printk("\n"); |
1190 | } | |
1191 | #endif | |
00269b54 ED |
1192 | /* |
1193 | * Since lookup is lockfree, we must make sure | |
25985edc | 1194 | * previous writes to rt are committed to memory |
00269b54 ED |
1195 | * before making rt visible to other CPUS. |
1196 | */ | |
1ddbcb00 | 1197 | rcu_assign_pointer(rt_hash_table[hash].chain, rt); |
1080d709 | 1198 | |
22c047cc | 1199 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
73e42897 | 1200 | |
b6280b47 | 1201 | skip_hashing: |
b23dd4fe | 1202 | if (skb) |
d8d1f30b | 1203 | skb_dst_set(skb, &rt->dst); |
b23dd4fe | 1204 | return rt; |
1da177e4 LT |
1205 | } |
1206 | ||
6431cbc2 DM |
1207 | static atomic_t __rt_peer_genid = ATOMIC_INIT(0); |
1208 | ||
1209 | static u32 rt_peer_genid(void) | |
1210 | { | |
1211 | return atomic_read(&__rt_peer_genid); | |
1212 | } | |
1213 | ||
1da177e4 LT |
1214 | void rt_bind_peer(struct rtable *rt, int create) |
1215 | { | |
1da177e4 LT |
1216 | struct inet_peer *peer; |
1217 | ||
b534ecf1 | 1218 | peer = inet_getpeer_v4(rt->rt_dst, create); |
1da177e4 | 1219 | |
49e8ab03 | 1220 | if (peer && cmpxchg(&rt->peer, NULL, peer) != NULL) |
1da177e4 | 1221 | inet_putpeer(peer); |
6431cbc2 DM |
1222 | else |
1223 | rt->rt_peer_genid = rt_peer_genid(); | |
1da177e4 LT |
1224 | } |
1225 | ||
1226 | /* | |
1227 | * Peer allocation may fail only in serious out-of-memory conditions. However | |
1228 | * we still can generate some output. | |
1229 | * Random ID selection looks a bit dangerous because we have no chances to | |
1230 | * select ID being unique in a reasonable period of time. | |
1231 | * But broken packet identifier may be better than no packet at all. | |
1232 | */ | |
1233 | static void ip_select_fb_ident(struct iphdr *iph) | |
1234 | { | |
1235 | static DEFINE_SPINLOCK(ip_fb_id_lock); | |
1236 | static u32 ip_fallback_id; | |
1237 | u32 salt; | |
1238 | ||
1239 | spin_lock_bh(&ip_fb_id_lock); | |
e448515c | 1240 | salt = secure_ip_id((__force __be32)ip_fallback_id ^ iph->daddr); |
1da177e4 LT |
1241 | iph->id = htons(salt & 0xFFFF); |
1242 | ip_fallback_id = salt; | |
1243 | spin_unlock_bh(&ip_fb_id_lock); | |
1244 | } | |
1245 | ||
1246 | void __ip_select_ident(struct iphdr *iph, struct dst_entry *dst, int more) | |
1247 | { | |
1248 | struct rtable *rt = (struct rtable *) dst; | |
1249 | ||
1250 | if (rt) { | |
1251 | if (rt->peer == NULL) | |
1252 | rt_bind_peer(rt, 1); | |
1253 | ||
1254 | /* If peer is attached to destination, it is never detached, | |
1255 | so that we need not to grab a lock to dereference it. | |
1256 | */ | |
1257 | if (rt->peer) { | |
1258 | iph->id = htons(inet_getid(rt->peer, more)); | |
1259 | return; | |
1260 | } | |
1261 | } else | |
e905a9ed | 1262 | printk(KERN_DEBUG "rt_bind_peer(0) @%p\n", |
9c2b3328 | 1263 | __builtin_return_address(0)); |
1da177e4 LT |
1264 | |
1265 | ip_select_fb_ident(iph); | |
1266 | } | |
4bc2f18b | 1267 | EXPORT_SYMBOL(__ip_select_ident); |
1da177e4 LT |
1268 | |
1269 | static void rt_del(unsigned hash, struct rtable *rt) | |
1270 | { | |
1c31720a ED |
1271 | struct rtable __rcu **rthp; |
1272 | struct rtable *aux; | |
1da177e4 | 1273 | |
29e75252 | 1274 | rthp = &rt_hash_table[hash].chain; |
22c047cc | 1275 | spin_lock_bh(rt_hash_lock_addr(hash)); |
1da177e4 | 1276 | ip_rt_put(rt); |
1c31720a ED |
1277 | while ((aux = rcu_dereference_protected(*rthp, |
1278 | lockdep_is_held(rt_hash_lock_addr(hash)))) != NULL) { | |
e84f84f2 | 1279 | if (aux == rt || rt_is_expired(aux)) { |
d8d1f30b | 1280 | *rthp = aux->dst.rt_next; |
29e75252 ED |
1281 | rt_free(aux); |
1282 | continue; | |
1da177e4 | 1283 | } |
d8d1f30b | 1284 | rthp = &aux->dst.rt_next; |
29e75252 | 1285 | } |
22c047cc | 1286 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1287 | } |
1288 | ||
ed7865a4 | 1289 | /* called in rcu_read_lock() section */ |
f7655229 AV |
1290 | void ip_rt_redirect(__be32 old_gw, __be32 daddr, __be32 new_gw, |
1291 | __be32 saddr, struct net_device *dev) | |
1da177e4 | 1292 | { |
ed7865a4 | 1293 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
f39925db | 1294 | struct inet_peer *peer; |
317805b8 | 1295 | struct net *net; |
1da177e4 | 1296 | |
1da177e4 LT |
1297 | if (!in_dev) |
1298 | return; | |
1299 | ||
c346dca1 | 1300 | net = dev_net(dev); |
9d4fb27d JP |
1301 | if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) || |
1302 | ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) || | |
1303 | ipv4_is_zeronet(new_gw)) | |
1da177e4 LT |
1304 | goto reject_redirect; |
1305 | ||
1306 | if (!IN_DEV_SHARED_MEDIA(in_dev)) { | |
1307 | if (!inet_addr_onlink(in_dev, new_gw, old_gw)) | |
1308 | goto reject_redirect; | |
1309 | if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev)) | |
1310 | goto reject_redirect; | |
1311 | } else { | |
317805b8 | 1312 | if (inet_addr_type(net, new_gw) != RTN_UNICAST) |
1da177e4 LT |
1313 | goto reject_redirect; |
1314 | } | |
1315 | ||
f39925db DM |
1316 | peer = inet_getpeer_v4(daddr, 1); |
1317 | if (peer) { | |
1318 | peer->redirect_learned.a4 = new_gw; | |
e905a9ed | 1319 | |
f39925db | 1320 | inet_putpeer(peer); |
1da177e4 | 1321 | |
f39925db | 1322 | atomic_inc(&__rt_peer_genid); |
1da177e4 | 1323 | } |
1da177e4 LT |
1324 | return; |
1325 | ||
1326 | reject_redirect: | |
1327 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1328 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) | |
673d57e7 HH |
1329 | printk(KERN_INFO "Redirect from %pI4 on %s about %pI4 ignored.\n" |
1330 | " Advised path = %pI4 -> %pI4\n", | |
1331 | &old_gw, dev->name, &new_gw, | |
1332 | &saddr, &daddr); | |
1da177e4 | 1333 | #endif |
ed7865a4 | 1334 | ; |
1da177e4 LT |
1335 | } |
1336 | ||
1337 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst) | |
1338 | { | |
ee6b9673 | 1339 | struct rtable *rt = (struct rtable *)dst; |
1da177e4 LT |
1340 | struct dst_entry *ret = dst; |
1341 | ||
1342 | if (rt) { | |
d11a4dc1 | 1343 | if (dst->obsolete > 0) { |
1da177e4 LT |
1344 | ip_rt_put(rt); |
1345 | ret = NULL; | |
2c8cec5c | 1346 | } else if (rt->rt_flags & RTCF_REDIRECTED) { |
5e2b61f7 DM |
1347 | unsigned hash = rt_hash(rt->rt_key_dst, rt->rt_key_src, |
1348 | rt->rt_oif, | |
e84f84f2 | 1349 | rt_genid(dev_net(dst->dev))); |
1da177e4 | 1350 | #if RT_CACHE_DEBUG >= 1 |
673d57e7 | 1351 | printk(KERN_DEBUG "ipv4_negative_advice: redirect to %pI4/%02x dropped\n", |
5e2b61f7 | 1352 | &rt->rt_dst, rt->rt_tos); |
1da177e4 LT |
1353 | #endif |
1354 | rt_del(hash, rt); | |
1355 | ret = NULL; | |
2c8cec5c DM |
1356 | } else if (rt->peer && |
1357 | rt->peer->pmtu_expires && | |
1358 | time_after_eq(jiffies, rt->peer->pmtu_expires)) { | |
1359 | unsigned long orig = rt->peer->pmtu_expires; | |
1360 | ||
1361 | if (cmpxchg(&rt->peer->pmtu_expires, orig, 0) == orig) | |
1362 | dst_metric_set(dst, RTAX_MTU, | |
1363 | rt->peer->pmtu_orig); | |
1da177e4 LT |
1364 | } |
1365 | } | |
1366 | return ret; | |
1367 | } | |
1368 | ||
1369 | /* | |
1370 | * Algorithm: | |
1371 | * 1. The first ip_rt_redirect_number redirects are sent | |
1372 | * with exponential backoff, then we stop sending them at all, | |
1373 | * assuming that the host ignores our redirects. | |
1374 | * 2. If we did not see packets requiring redirects | |
1375 | * during ip_rt_redirect_silence, we assume that the host | |
1376 | * forgot redirected route and start to send redirects again. | |
1377 | * | |
1378 | * This algorithm is much cheaper and more intelligent than dumb load limiting | |
1379 | * in icmp.c. | |
1380 | * | |
1381 | * NOTE. Do not forget to inhibit load limiting for redirects (redundant) | |
1382 | * and "frag. need" (breaks PMTU discovery) in icmp.c. | |
1383 | */ | |
1384 | ||
1385 | void ip_rt_send_redirect(struct sk_buff *skb) | |
1386 | { | |
511c3f92 | 1387 | struct rtable *rt = skb_rtable(skb); |
30038fc6 | 1388 | struct in_device *in_dev; |
92d86829 | 1389 | struct inet_peer *peer; |
30038fc6 | 1390 | int log_martians; |
1da177e4 | 1391 | |
30038fc6 | 1392 | rcu_read_lock(); |
d8d1f30b | 1393 | in_dev = __in_dev_get_rcu(rt->dst.dev); |
30038fc6 ED |
1394 | if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) { |
1395 | rcu_read_unlock(); | |
1da177e4 | 1396 | return; |
30038fc6 ED |
1397 | } |
1398 | log_martians = IN_DEV_LOG_MARTIANS(in_dev); | |
1399 | rcu_read_unlock(); | |
1da177e4 | 1400 | |
92d86829 DM |
1401 | if (!rt->peer) |
1402 | rt_bind_peer(rt, 1); | |
1403 | peer = rt->peer; | |
1404 | if (!peer) { | |
1405 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway); | |
1406 | return; | |
1407 | } | |
1408 | ||
1da177e4 LT |
1409 | /* No redirected packets during ip_rt_redirect_silence; |
1410 | * reset the algorithm. | |
1411 | */ | |
92d86829 DM |
1412 | if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence)) |
1413 | peer->rate_tokens = 0; | |
1da177e4 LT |
1414 | |
1415 | /* Too many ignored redirects; do not send anything | |
d8d1f30b | 1416 | * set dst.rate_last to the last seen redirected packet. |
1da177e4 | 1417 | */ |
92d86829 DM |
1418 | if (peer->rate_tokens >= ip_rt_redirect_number) { |
1419 | peer->rate_last = jiffies; | |
30038fc6 | 1420 | return; |
1da177e4 LT |
1421 | } |
1422 | ||
1423 | /* Check for load limit; set rate_last to the latest sent | |
1424 | * redirect. | |
1425 | */ | |
92d86829 | 1426 | if (peer->rate_tokens == 0 || |
14fb8a76 | 1427 | time_after(jiffies, |
92d86829 DM |
1428 | (peer->rate_last + |
1429 | (ip_rt_redirect_load << peer->rate_tokens)))) { | |
1da177e4 | 1430 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway); |
92d86829 DM |
1431 | peer->rate_last = jiffies; |
1432 | ++peer->rate_tokens; | |
1da177e4 | 1433 | #ifdef CONFIG_IP_ROUTE_VERBOSE |
30038fc6 | 1434 | if (log_martians && |
92d86829 | 1435 | peer->rate_tokens == ip_rt_redirect_number && |
1da177e4 | 1436 | net_ratelimit()) |
673d57e7 HH |
1437 | printk(KERN_WARNING "host %pI4/if%d ignores redirects for %pI4 to %pI4.\n", |
1438 | &rt->rt_src, rt->rt_iif, | |
1439 | &rt->rt_dst, &rt->rt_gateway); | |
1da177e4 LT |
1440 | #endif |
1441 | } | |
1da177e4 LT |
1442 | } |
1443 | ||
1444 | static int ip_error(struct sk_buff *skb) | |
1445 | { | |
511c3f92 | 1446 | struct rtable *rt = skb_rtable(skb); |
92d86829 | 1447 | struct inet_peer *peer; |
1da177e4 | 1448 | unsigned long now; |
92d86829 | 1449 | bool send; |
1da177e4 LT |
1450 | int code; |
1451 | ||
d8d1f30b | 1452 | switch (rt->dst.error) { |
1da177e4 LT |
1453 | case EINVAL: |
1454 | default: | |
1455 | goto out; | |
1456 | case EHOSTUNREACH: | |
1457 | code = ICMP_HOST_UNREACH; | |
1458 | break; | |
1459 | case ENETUNREACH: | |
1460 | code = ICMP_NET_UNREACH; | |
d8d1f30b | 1461 | IP_INC_STATS_BH(dev_net(rt->dst.dev), |
7c73a6fa | 1462 | IPSTATS_MIB_INNOROUTES); |
1da177e4 LT |
1463 | break; |
1464 | case EACCES: | |
1465 | code = ICMP_PKT_FILTERED; | |
1466 | break; | |
1467 | } | |
1468 | ||
92d86829 DM |
1469 | if (!rt->peer) |
1470 | rt_bind_peer(rt, 1); | |
1471 | peer = rt->peer; | |
1472 | ||
1473 | send = true; | |
1474 | if (peer) { | |
1475 | now = jiffies; | |
1476 | peer->rate_tokens += now - peer->rate_last; | |
1477 | if (peer->rate_tokens > ip_rt_error_burst) | |
1478 | peer->rate_tokens = ip_rt_error_burst; | |
1479 | peer->rate_last = now; | |
1480 | if (peer->rate_tokens >= ip_rt_error_cost) | |
1481 | peer->rate_tokens -= ip_rt_error_cost; | |
1482 | else | |
1483 | send = false; | |
1da177e4 | 1484 | } |
92d86829 DM |
1485 | if (send) |
1486 | icmp_send(skb, ICMP_DEST_UNREACH, code, 0); | |
1da177e4 LT |
1487 | |
1488 | out: kfree_skb(skb); | |
1489 | return 0; | |
e905a9ed | 1490 | } |
1da177e4 LT |
1491 | |
1492 | /* | |
1493 | * The last two values are not from the RFC but | |
1494 | * are needed for AMPRnet AX.25 paths. | |
1495 | */ | |
1496 | ||
9b5b5cff | 1497 | static const unsigned short mtu_plateau[] = |
1da177e4 LT |
1498 | {32000, 17914, 8166, 4352, 2002, 1492, 576, 296, 216, 128 }; |
1499 | ||
5969f71d | 1500 | static inline unsigned short guess_mtu(unsigned short old_mtu) |
1da177e4 LT |
1501 | { |
1502 | int i; | |
e905a9ed | 1503 | |
1da177e4 LT |
1504 | for (i = 0; i < ARRAY_SIZE(mtu_plateau); i++) |
1505 | if (old_mtu > mtu_plateau[i]) | |
1506 | return mtu_plateau[i]; | |
1507 | return 68; | |
1508 | } | |
1509 | ||
b71d1d42 | 1510 | unsigned short ip_rt_frag_needed(struct net *net, const struct iphdr *iph, |
0010e465 TT |
1511 | unsigned short new_mtu, |
1512 | struct net_device *dev) | |
1da177e4 | 1513 | { |
1da177e4 | 1514 | unsigned short old_mtu = ntohs(iph->tot_len); |
1da177e4 | 1515 | unsigned short est_mtu = 0; |
2c8cec5c | 1516 | struct inet_peer *peer; |
1da177e4 | 1517 | |
2c8cec5c DM |
1518 | peer = inet_getpeer_v4(iph->daddr, 1); |
1519 | if (peer) { | |
1520 | unsigned short mtu = new_mtu; | |
1da177e4 | 1521 | |
2c8cec5c DM |
1522 | if (new_mtu < 68 || new_mtu >= old_mtu) { |
1523 | /* BSD 4.2 derived systems incorrectly adjust | |
1524 | * tot_len by the IP header length, and report | |
1525 | * a zero MTU in the ICMP message. | |
1526 | */ | |
1527 | if (mtu == 0 && | |
1528 | old_mtu >= 68 + (iph->ihl << 2)) | |
1529 | old_mtu -= iph->ihl << 2; | |
1530 | mtu = guess_mtu(old_mtu); | |
1531 | } | |
0010e465 | 1532 | |
2c8cec5c DM |
1533 | if (mtu < ip_rt_min_pmtu) |
1534 | mtu = ip_rt_min_pmtu; | |
1535 | if (!peer->pmtu_expires || mtu < peer->pmtu_learned) { | |
46af3180 HS |
1536 | unsigned long pmtu_expires; |
1537 | ||
1538 | pmtu_expires = jiffies + ip_rt_mtu_expires; | |
1539 | if (!pmtu_expires) | |
1540 | pmtu_expires = 1UL; | |
1541 | ||
2c8cec5c DM |
1542 | est_mtu = mtu; |
1543 | peer->pmtu_learned = mtu; | |
46af3180 | 1544 | peer->pmtu_expires = pmtu_expires; |
2c8cec5c | 1545 | } |
1da177e4 | 1546 | |
2c8cec5c | 1547 | inet_putpeer(peer); |
1da177e4 | 1548 | |
2c8cec5c | 1549 | atomic_inc(&__rt_peer_genid); |
1da177e4 LT |
1550 | } |
1551 | return est_mtu ? : new_mtu; | |
1552 | } | |
1553 | ||
2c8cec5c DM |
1554 | static void check_peer_pmtu(struct dst_entry *dst, struct inet_peer *peer) |
1555 | { | |
1556 | unsigned long expires = peer->pmtu_expires; | |
1557 | ||
46af3180 | 1558 | if (time_before(jiffies, expires)) { |
2c8cec5c DM |
1559 | u32 orig_dst_mtu = dst_mtu(dst); |
1560 | if (peer->pmtu_learned < orig_dst_mtu) { | |
1561 | if (!peer->pmtu_orig) | |
1562 | peer->pmtu_orig = dst_metric_raw(dst, RTAX_MTU); | |
1563 | dst_metric_set(dst, RTAX_MTU, peer->pmtu_learned); | |
1564 | } | |
1565 | } else if (cmpxchg(&peer->pmtu_expires, expires, 0) == expires) | |
1566 | dst_metric_set(dst, RTAX_MTU, peer->pmtu_orig); | |
1567 | } | |
1568 | ||
1da177e4 LT |
1569 | static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu) |
1570 | { | |
2c8cec5c DM |
1571 | struct rtable *rt = (struct rtable *) dst; |
1572 | struct inet_peer *peer; | |
1573 | ||
1574 | dst_confirm(dst); | |
1575 | ||
1576 | if (!rt->peer) | |
1577 | rt_bind_peer(rt, 1); | |
1578 | peer = rt->peer; | |
1579 | if (peer) { | |
1580 | if (mtu < ip_rt_min_pmtu) | |
1da177e4 | 1581 | mtu = ip_rt_min_pmtu; |
2c8cec5c | 1582 | if (!peer->pmtu_expires || mtu < peer->pmtu_learned) { |
46af3180 HS |
1583 | unsigned long pmtu_expires; |
1584 | ||
1585 | pmtu_expires = jiffies + ip_rt_mtu_expires; | |
1586 | if (!pmtu_expires) | |
1587 | pmtu_expires = 1UL; | |
1588 | ||
2c8cec5c | 1589 | peer->pmtu_learned = mtu; |
46af3180 | 1590 | peer->pmtu_expires = pmtu_expires; |
2c8cec5c DM |
1591 | |
1592 | atomic_inc(&__rt_peer_genid); | |
1593 | rt->rt_peer_genid = rt_peer_genid(); | |
1da177e4 | 1594 | } |
46af3180 | 1595 | check_peer_pmtu(dst, peer); |
1da177e4 LT |
1596 | } |
1597 | } | |
1598 | ||
f39925db DM |
1599 | static int check_peer_redir(struct dst_entry *dst, struct inet_peer *peer) |
1600 | { | |
1601 | struct rtable *rt = (struct rtable *) dst; | |
1602 | __be32 orig_gw = rt->rt_gateway; | |
1603 | ||
1604 | dst_confirm(&rt->dst); | |
1605 | ||
1606 | neigh_release(rt->dst.neighbour); | |
1607 | rt->dst.neighbour = NULL; | |
1608 | ||
1609 | rt->rt_gateway = peer->redirect_learned.a4; | |
1610 | if (arp_bind_neighbour(&rt->dst) || | |
1611 | !(rt->dst.neighbour->nud_state & NUD_VALID)) { | |
1612 | if (rt->dst.neighbour) | |
1613 | neigh_event_send(rt->dst.neighbour, NULL); | |
1614 | rt->rt_gateway = orig_gw; | |
1615 | return -EAGAIN; | |
1616 | } else { | |
1617 | rt->rt_flags |= RTCF_REDIRECTED; | |
1618 | call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, | |
1619 | rt->dst.neighbour); | |
1620 | } | |
1621 | return 0; | |
1622 | } | |
1623 | ||
1da177e4 LT |
1624 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie) |
1625 | { | |
6431cbc2 DM |
1626 | struct rtable *rt = (struct rtable *) dst; |
1627 | ||
1628 | if (rt_is_expired(rt)) | |
d11a4dc1 | 1629 | return NULL; |
6431cbc2 | 1630 | if (rt->rt_peer_genid != rt_peer_genid()) { |
2c8cec5c DM |
1631 | struct inet_peer *peer; |
1632 | ||
6431cbc2 DM |
1633 | if (!rt->peer) |
1634 | rt_bind_peer(rt, 0); | |
1635 | ||
2c8cec5c DM |
1636 | peer = rt->peer; |
1637 | if (peer && peer->pmtu_expires) | |
1638 | check_peer_pmtu(dst, peer); | |
1639 | ||
f39925db DM |
1640 | if (peer && peer->redirect_learned.a4 && |
1641 | peer->redirect_learned.a4 != rt->rt_gateway) { | |
1642 | if (check_peer_redir(dst, peer)) | |
1643 | return NULL; | |
1644 | } | |
1645 | ||
6431cbc2 DM |
1646 | rt->rt_peer_genid = rt_peer_genid(); |
1647 | } | |
d11a4dc1 | 1648 | return dst; |
1da177e4 LT |
1649 | } |
1650 | ||
1651 | static void ipv4_dst_destroy(struct dst_entry *dst) | |
1652 | { | |
1653 | struct rtable *rt = (struct rtable *) dst; | |
1654 | struct inet_peer *peer = rt->peer; | |
1da177e4 | 1655 | |
62fa8a84 DM |
1656 | if (rt->fi) { |
1657 | fib_info_put(rt->fi); | |
1658 | rt->fi = NULL; | |
1659 | } | |
1da177e4 LT |
1660 | if (peer) { |
1661 | rt->peer = NULL; | |
1662 | inet_putpeer(peer); | |
1663 | } | |
1da177e4 LT |
1664 | } |
1665 | ||
1da177e4 LT |
1666 | |
1667 | static void ipv4_link_failure(struct sk_buff *skb) | |
1668 | { | |
1669 | struct rtable *rt; | |
1670 | ||
1671 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0); | |
1672 | ||
511c3f92 | 1673 | rt = skb_rtable(skb); |
2c8cec5c DM |
1674 | if (rt && |
1675 | rt->peer && | |
1676 | rt->peer->pmtu_expires) { | |
1677 | unsigned long orig = rt->peer->pmtu_expires; | |
1678 | ||
1679 | if (cmpxchg(&rt->peer->pmtu_expires, orig, 0) == orig) | |
1680 | dst_metric_set(&rt->dst, RTAX_MTU, rt->peer->pmtu_orig); | |
1681 | } | |
1da177e4 LT |
1682 | } |
1683 | ||
1684 | static int ip_rt_bug(struct sk_buff *skb) | |
1685 | { | |
673d57e7 HH |
1686 | printk(KERN_DEBUG "ip_rt_bug: %pI4 -> %pI4, %s\n", |
1687 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
1da177e4 LT |
1688 | skb->dev ? skb->dev->name : "?"); |
1689 | kfree_skb(skb); | |
1690 | return 0; | |
1691 | } | |
1692 | ||
1693 | /* | |
1694 | We do not cache source address of outgoing interface, | |
1695 | because it is used only by IP RR, TS and SRR options, | |
1696 | so that it out of fast path. | |
1697 | ||
1698 | BTW remember: "addr" is allowed to be not aligned | |
1699 | in IP options! | |
1700 | */ | |
1701 | ||
1702 | void ip_rt_get_source(u8 *addr, struct rtable *rt) | |
1703 | { | |
a61ced5d | 1704 | __be32 src; |
1da177e4 LT |
1705 | struct fib_result res; |
1706 | ||
c7537967 | 1707 | if (rt_is_output_route(rt)) |
1da177e4 | 1708 | src = rt->rt_src; |
ebc0ffae | 1709 | else { |
68a5e3dd DM |
1710 | struct flowi4 fl4 = { |
1711 | .daddr = rt->rt_key_dst, | |
1712 | .saddr = rt->rt_key_src, | |
1713 | .flowi4_tos = rt->rt_tos, | |
1714 | .flowi4_oif = rt->rt_oif, | |
1715 | .flowi4_iif = rt->rt_iif, | |
1716 | .flowi4_mark = rt->rt_mark, | |
5e2b61f7 DM |
1717 | }; |
1718 | ||
ebc0ffae | 1719 | rcu_read_lock(); |
68a5e3dd | 1720 | if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res) == 0) |
436c3b66 | 1721 | src = FIB_RES_PREFSRC(dev_net(rt->dst.dev), res); |
ebc0ffae ED |
1722 | else |
1723 | src = inet_select_addr(rt->dst.dev, rt->rt_gateway, | |
1da177e4 | 1724 | RT_SCOPE_UNIVERSE); |
ebc0ffae ED |
1725 | rcu_read_unlock(); |
1726 | } | |
1da177e4 LT |
1727 | memcpy(addr, &src, 4); |
1728 | } | |
1729 | ||
c7066f70 | 1730 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 LT |
1731 | static void set_class_tag(struct rtable *rt, u32 tag) |
1732 | { | |
d8d1f30b CG |
1733 | if (!(rt->dst.tclassid & 0xFFFF)) |
1734 | rt->dst.tclassid |= tag & 0xFFFF; | |
1735 | if (!(rt->dst.tclassid & 0xFFFF0000)) | |
1736 | rt->dst.tclassid |= tag & 0xFFFF0000; | |
1da177e4 LT |
1737 | } |
1738 | #endif | |
1739 | ||
0dbaee3b DM |
1740 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst) |
1741 | { | |
1742 | unsigned int advmss = dst_metric_raw(dst, RTAX_ADVMSS); | |
1743 | ||
1744 | if (advmss == 0) { | |
1745 | advmss = max_t(unsigned int, dst->dev->mtu - 40, | |
1746 | ip_rt_min_advmss); | |
1747 | if (advmss > 65535 - 40) | |
1748 | advmss = 65535 - 40; | |
1749 | } | |
1750 | return advmss; | |
1751 | } | |
1752 | ||
d33e4553 DM |
1753 | static unsigned int ipv4_default_mtu(const struct dst_entry *dst) |
1754 | { | |
1755 | unsigned int mtu = dst->dev->mtu; | |
1756 | ||
1757 | if (unlikely(dst_metric_locked(dst, RTAX_MTU))) { | |
1758 | const struct rtable *rt = (const struct rtable *) dst; | |
1759 | ||
1760 | if (rt->rt_gateway != rt->rt_dst && mtu > 576) | |
1761 | mtu = 576; | |
1762 | } | |
1763 | ||
1764 | if (mtu > IP_MAX_MTU) | |
1765 | mtu = IP_MAX_MTU; | |
1766 | ||
1767 | return mtu; | |
1768 | } | |
1769 | ||
68a5e3dd | 1770 | static void rt_init_metrics(struct rtable *rt, const struct flowi4 *oldflp4, |
5e2b61f7 | 1771 | struct fib_info *fi) |
a4daad6b | 1772 | { |
0131ba45 DM |
1773 | struct inet_peer *peer; |
1774 | int create = 0; | |
a4daad6b | 1775 | |
0131ba45 DM |
1776 | /* If a peer entry exists for this destination, we must hook |
1777 | * it up in order to get at cached metrics. | |
1778 | */ | |
68a5e3dd | 1779 | if (oldflp4 && (oldflp4->flowi4_flags & FLOWI_FLAG_PRECOW_METRICS)) |
0131ba45 DM |
1780 | create = 1; |
1781 | ||
3c0afdca | 1782 | rt->peer = peer = inet_getpeer_v4(rt->rt_dst, create); |
0131ba45 | 1783 | if (peer) { |
3c0afdca | 1784 | rt->rt_peer_genid = rt_peer_genid(); |
a4daad6b DM |
1785 | if (inet_metrics_new(peer)) |
1786 | memcpy(peer->metrics, fi->fib_metrics, | |
1787 | sizeof(u32) * RTAX_MAX); | |
1788 | dst_init_metrics(&rt->dst, peer->metrics, false); | |
2c8cec5c DM |
1789 | |
1790 | if (peer->pmtu_expires) | |
1791 | check_peer_pmtu(&rt->dst, peer); | |
f39925db DM |
1792 | if (peer->redirect_learned.a4 && |
1793 | peer->redirect_learned.a4 != rt->rt_gateway) { | |
1794 | rt->rt_gateway = peer->redirect_learned.a4; | |
1795 | rt->rt_flags |= RTCF_REDIRECTED; | |
1796 | } | |
0131ba45 DM |
1797 | } else { |
1798 | if (fi->fib_metrics != (u32 *) dst_default_metrics) { | |
1799 | rt->fi = fi; | |
1800 | atomic_inc(&fi->fib_clntref); | |
1801 | } | |
1802 | dst_init_metrics(&rt->dst, fi->fib_metrics, true); | |
a4daad6b DM |
1803 | } |
1804 | } | |
1805 | ||
68a5e3dd | 1806 | static void rt_set_nexthop(struct rtable *rt, const struct flowi4 *oldflp4, |
5e2b61f7 | 1807 | const struct fib_result *res, |
982721f3 | 1808 | struct fib_info *fi, u16 type, u32 itag) |
1da177e4 | 1809 | { |
defb3519 | 1810 | struct dst_entry *dst = &rt->dst; |
1da177e4 LT |
1811 | |
1812 | if (fi) { | |
1813 | if (FIB_RES_GW(*res) && | |
1814 | FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) | |
1815 | rt->rt_gateway = FIB_RES_GW(*res); | |
68a5e3dd | 1816 | rt_init_metrics(rt, oldflp4, fi); |
c7066f70 | 1817 | #ifdef CONFIG_IP_ROUTE_CLASSID |
defb3519 | 1818 | dst->tclassid = FIB_RES_NH(*res).nh_tclassid; |
1da177e4 | 1819 | #endif |
d33e4553 | 1820 | } |
defb3519 | 1821 | |
defb3519 DM |
1822 | if (dst_mtu(dst) > IP_MAX_MTU) |
1823 | dst_metric_set(dst, RTAX_MTU, IP_MAX_MTU); | |
0dbaee3b | 1824 | if (dst_metric_raw(dst, RTAX_ADVMSS) > 65535 - 40) |
defb3519 | 1825 | dst_metric_set(dst, RTAX_ADVMSS, 65535 - 40); |
1da177e4 | 1826 | |
c7066f70 | 1827 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 LT |
1828 | #ifdef CONFIG_IP_MULTIPLE_TABLES |
1829 | set_class_tag(rt, fib_rules_tclass(res)); | |
1830 | #endif | |
1831 | set_class_tag(rt, itag); | |
1832 | #endif | |
982721f3 | 1833 | rt->rt_type = type; |
1da177e4 LT |
1834 | } |
1835 | ||
0c4dcd58 DM |
1836 | static struct rtable *rt_dst_alloc(bool nopolicy, bool noxfrm) |
1837 | { | |
3c7bd1a1 | 1838 | struct rtable *rt = dst_alloc(&ipv4_dst_ops, 1); |
0c4dcd58 DM |
1839 | if (rt) { |
1840 | rt->dst.obsolete = -1; | |
1841 | ||
0c4dcd58 DM |
1842 | rt->dst.flags = DST_HOST | |
1843 | (nopolicy ? DST_NOPOLICY : 0) | | |
1844 | (noxfrm ? DST_NOXFRM : 0); | |
1845 | } | |
1846 | return rt; | |
1847 | } | |
1848 | ||
96d36220 | 1849 | /* called in rcu_read_lock() section */ |
9e12bb22 | 1850 | static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
1851 | u8 tos, struct net_device *dev, int our) |
1852 | { | |
96d36220 | 1853 | unsigned int hash; |
1da177e4 | 1854 | struct rtable *rth; |
a61ced5d | 1855 | __be32 spec_dst; |
96d36220 | 1856 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1857 | u32 itag = 0; |
b5f7e755 | 1858 | int err; |
1da177e4 LT |
1859 | |
1860 | /* Primary sanity checks. */ | |
1861 | ||
1862 | if (in_dev == NULL) | |
1863 | return -EINVAL; | |
1864 | ||
1e637c74 | 1865 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
f97c1e0c | 1866 | ipv4_is_loopback(saddr) || skb->protocol != htons(ETH_P_IP)) |
1da177e4 LT |
1867 | goto e_inval; |
1868 | ||
f97c1e0c JP |
1869 | if (ipv4_is_zeronet(saddr)) { |
1870 | if (!ipv4_is_local_multicast(daddr)) | |
1da177e4 LT |
1871 | goto e_inval; |
1872 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); | |
b5f7e755 | 1873 | } else { |
5c04c819 MS |
1874 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, &spec_dst, |
1875 | &itag); | |
b5f7e755 ED |
1876 | if (err < 0) |
1877 | goto e_err; | |
1878 | } | |
0c4dcd58 | 1879 | rth = rt_dst_alloc(IN_DEV_CONF_GET(in_dev, NOPOLICY), false); |
1da177e4 LT |
1880 | if (!rth) |
1881 | goto e_nobufs; | |
1882 | ||
d8d1f30b | 1883 | rth->dst.output = ip_rt_bug; |
1da177e4 | 1884 | |
5e2b61f7 | 1885 | rth->rt_key_dst = daddr; |
1da177e4 | 1886 | rth->rt_dst = daddr; |
5e2b61f7 DM |
1887 | rth->rt_tos = tos; |
1888 | rth->rt_mark = skb->mark; | |
1889 | rth->rt_key_src = saddr; | |
1da177e4 | 1890 | rth->rt_src = saddr; |
c7066f70 | 1891 | #ifdef CONFIG_IP_ROUTE_CLASSID |
d8d1f30b | 1892 | rth->dst.tclassid = itag; |
1da177e4 | 1893 | #endif |
1b86a58f | 1894 | rth->rt_route_iif = dev->ifindex; |
5e2b61f7 | 1895 | rth->rt_iif = dev->ifindex; |
d8d1f30b CG |
1896 | rth->dst.dev = init_net.loopback_dev; |
1897 | dev_hold(rth->dst.dev); | |
5e2b61f7 | 1898 | rth->rt_oif = 0; |
1da177e4 LT |
1899 | rth->rt_gateway = daddr; |
1900 | rth->rt_spec_dst= spec_dst; | |
e84f84f2 | 1901 | rth->rt_genid = rt_genid(dev_net(dev)); |
1da177e4 | 1902 | rth->rt_flags = RTCF_MULTICAST; |
29e75252 | 1903 | rth->rt_type = RTN_MULTICAST; |
1da177e4 | 1904 | if (our) { |
d8d1f30b | 1905 | rth->dst.input= ip_local_deliver; |
1da177e4 LT |
1906 | rth->rt_flags |= RTCF_LOCAL; |
1907 | } | |
1908 | ||
1909 | #ifdef CONFIG_IP_MROUTE | |
f97c1e0c | 1910 | if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev)) |
d8d1f30b | 1911 | rth->dst.input = ip_mr_input; |
1da177e4 LT |
1912 | #endif |
1913 | RT_CACHE_STAT_INC(in_slow_mc); | |
1914 | ||
e84f84f2 | 1915 | hash = rt_hash(daddr, saddr, dev->ifindex, rt_genid(dev_net(dev))); |
b23dd4fe DM |
1916 | rth = rt_intern_hash(hash, rth, skb, dev->ifindex); |
1917 | err = 0; | |
1918 | if (IS_ERR(rth)) | |
1919 | err = PTR_ERR(rth); | |
1da177e4 LT |
1920 | |
1921 | e_nobufs: | |
1da177e4 | 1922 | return -ENOBUFS; |
1da177e4 | 1923 | e_inval: |
96d36220 | 1924 | return -EINVAL; |
b5f7e755 | 1925 | e_err: |
b5f7e755 | 1926 | return err; |
1da177e4 LT |
1927 | } |
1928 | ||
1929 | ||
1930 | static void ip_handle_martian_source(struct net_device *dev, | |
1931 | struct in_device *in_dev, | |
1932 | struct sk_buff *skb, | |
9e12bb22 AV |
1933 | __be32 daddr, |
1934 | __be32 saddr) | |
1da177e4 LT |
1935 | { |
1936 | RT_CACHE_STAT_INC(in_martian_src); | |
1937 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1938 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) { | |
1939 | /* | |
1940 | * RFC1812 recommendation, if source is martian, | |
1941 | * the only hint is MAC header. | |
1942 | */ | |
673d57e7 HH |
1943 | printk(KERN_WARNING "martian source %pI4 from %pI4, on dev %s\n", |
1944 | &daddr, &saddr, dev->name); | |
98e399f8 | 1945 | if (dev->hard_header_len && skb_mac_header_was_set(skb)) { |
1da177e4 | 1946 | int i; |
98e399f8 | 1947 | const unsigned char *p = skb_mac_header(skb); |
1da177e4 LT |
1948 | printk(KERN_WARNING "ll header: "); |
1949 | for (i = 0; i < dev->hard_header_len; i++, p++) { | |
1950 | printk("%02x", *p); | |
1951 | if (i < (dev->hard_header_len - 1)) | |
1952 | printk(":"); | |
1953 | } | |
1954 | printk("\n"); | |
1955 | } | |
1956 | } | |
1957 | #endif | |
1958 | } | |
1959 | ||
47360228 | 1960 | /* called in rcu_read_lock() section */ |
5969f71d | 1961 | static int __mkroute_input(struct sk_buff *skb, |
982721f3 | 1962 | const struct fib_result *res, |
5969f71d SH |
1963 | struct in_device *in_dev, |
1964 | __be32 daddr, __be32 saddr, u32 tos, | |
1965 | struct rtable **result) | |
1da177e4 | 1966 | { |
1da177e4 LT |
1967 | struct rtable *rth; |
1968 | int err; | |
1969 | struct in_device *out_dev; | |
47360228 | 1970 | unsigned int flags = 0; |
d9c9df8c AV |
1971 | __be32 spec_dst; |
1972 | u32 itag; | |
1da177e4 LT |
1973 | |
1974 | /* get a working reference to the output device */ | |
47360228 | 1975 | out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res)); |
1da177e4 LT |
1976 | if (out_dev == NULL) { |
1977 | if (net_ratelimit()) | |
1978 | printk(KERN_CRIT "Bug in ip_route_input" \ | |
1979 | "_slow(). Please, report\n"); | |
1980 | return -EINVAL; | |
1981 | } | |
1982 | ||
1983 | ||
5c04c819 MS |
1984 | err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res), |
1985 | in_dev->dev, &spec_dst, &itag); | |
1da177e4 | 1986 | if (err < 0) { |
e905a9ed | 1987 | ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr, |
1da177e4 | 1988 | saddr); |
e905a9ed | 1989 | |
1da177e4 LT |
1990 | goto cleanup; |
1991 | } | |
1992 | ||
1993 | if (err) | |
1994 | flags |= RTCF_DIRECTSRC; | |
1995 | ||
51b77cae | 1996 | if (out_dev == in_dev && err && |
1da177e4 LT |
1997 | (IN_DEV_SHARED_MEDIA(out_dev) || |
1998 | inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) | |
1999 | flags |= RTCF_DOREDIRECT; | |
2000 | ||
2001 | if (skb->protocol != htons(ETH_P_IP)) { | |
2002 | /* Not IP (i.e. ARP). Do not create route, if it is | |
2003 | * invalid for proxy arp. DNAT routes are always valid. | |
65324144 JDB |
2004 | * |
2005 | * Proxy arp feature have been extended to allow, ARP | |
2006 | * replies back to the same interface, to support | |
2007 | * Private VLAN switch technologies. See arp.c. | |
1da177e4 | 2008 | */ |
65324144 JDB |
2009 | if (out_dev == in_dev && |
2010 | IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) { | |
1da177e4 LT |
2011 | err = -EINVAL; |
2012 | goto cleanup; | |
2013 | } | |
2014 | } | |
2015 | ||
0c4dcd58 DM |
2016 | rth = rt_dst_alloc(IN_DEV_CONF_GET(in_dev, NOPOLICY), |
2017 | IN_DEV_CONF_GET(out_dev, NOXFRM)); | |
1da177e4 LT |
2018 | if (!rth) { |
2019 | err = -ENOBUFS; | |
2020 | goto cleanup; | |
2021 | } | |
2022 | ||
5e2b61f7 | 2023 | rth->rt_key_dst = daddr; |
1da177e4 | 2024 | rth->rt_dst = daddr; |
5e2b61f7 DM |
2025 | rth->rt_tos = tos; |
2026 | rth->rt_mark = skb->mark; | |
2027 | rth->rt_key_src = saddr; | |
1da177e4 LT |
2028 | rth->rt_src = saddr; |
2029 | rth->rt_gateway = daddr; | |
1b86a58f | 2030 | rth->rt_route_iif = in_dev->dev->ifindex; |
5e2b61f7 | 2031 | rth->rt_iif = in_dev->dev->ifindex; |
d8d1f30b CG |
2032 | rth->dst.dev = (out_dev)->dev; |
2033 | dev_hold(rth->dst.dev); | |
5e2b61f7 | 2034 | rth->rt_oif = 0; |
1da177e4 LT |
2035 | rth->rt_spec_dst= spec_dst; |
2036 | ||
d8d1f30b CG |
2037 | rth->dst.input = ip_forward; |
2038 | rth->dst.output = ip_output; | |
2039 | rth->rt_genid = rt_genid(dev_net(rth->dst.dev)); | |
1da177e4 | 2040 | |
5e2b61f7 | 2041 | rt_set_nexthop(rth, NULL, res, res->fi, res->type, itag); |
1da177e4 LT |
2042 | |
2043 | rth->rt_flags = flags; | |
2044 | ||
2045 | *result = rth; | |
2046 | err = 0; | |
2047 | cleanup: | |
1da177e4 | 2048 | return err; |
e905a9ed | 2049 | } |
1da177e4 | 2050 | |
5969f71d SH |
2051 | static int ip_mkroute_input(struct sk_buff *skb, |
2052 | struct fib_result *res, | |
68a5e3dd | 2053 | const struct flowi4 *fl4, |
5969f71d SH |
2054 | struct in_device *in_dev, |
2055 | __be32 daddr, __be32 saddr, u32 tos) | |
1da177e4 | 2056 | { |
7abaa27c | 2057 | struct rtable* rth = NULL; |
1da177e4 LT |
2058 | int err; |
2059 | unsigned hash; | |
2060 | ||
2061 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
ff3fccb3 | 2062 | if (res->fi && res->fi->fib_nhs > 1) |
1b7fe593 | 2063 | fib_select_multipath(res); |
1da177e4 LT |
2064 | #endif |
2065 | ||
2066 | /* create a routing cache entry */ | |
2067 | err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth); | |
2068 | if (err) | |
2069 | return err; | |
1da177e4 LT |
2070 | |
2071 | /* put it into the cache */ | |
68a5e3dd | 2072 | hash = rt_hash(daddr, saddr, fl4->flowi4_iif, |
d8d1f30b | 2073 | rt_genid(dev_net(rth->dst.dev))); |
68a5e3dd | 2074 | rth = rt_intern_hash(hash, rth, skb, fl4->flowi4_iif); |
b23dd4fe DM |
2075 | if (IS_ERR(rth)) |
2076 | return PTR_ERR(rth); | |
2077 | return 0; | |
1da177e4 LT |
2078 | } |
2079 | ||
1da177e4 LT |
2080 | /* |
2081 | * NOTE. We drop all the packets that has local source | |
2082 | * addresses, because every properly looped back packet | |
2083 | * must have correct destination already attached by output routine. | |
2084 | * | |
2085 | * Such approach solves two big problems: | |
2086 | * 1. Not simplex devices are handled properly. | |
2087 | * 2. IP spoofing attempts are filtered with 100% of guarantee. | |
ebc0ffae | 2088 | * called with rcu_read_lock() |
1da177e4 LT |
2089 | */ |
2090 | ||
9e12bb22 | 2091 | static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
2092 | u8 tos, struct net_device *dev) |
2093 | { | |
2094 | struct fib_result res; | |
96d36220 | 2095 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
68a5e3dd | 2096 | struct flowi4 fl4; |
1da177e4 LT |
2097 | unsigned flags = 0; |
2098 | u32 itag = 0; | |
2099 | struct rtable * rth; | |
2100 | unsigned hash; | |
9e12bb22 | 2101 | __be32 spec_dst; |
1da177e4 | 2102 | int err = -EINVAL; |
c346dca1 | 2103 | struct net * net = dev_net(dev); |
1da177e4 LT |
2104 | |
2105 | /* IP on this device is disabled. */ | |
2106 | ||
2107 | if (!in_dev) | |
2108 | goto out; | |
2109 | ||
2110 | /* Check for the most weird martians, which can be not detected | |
2111 | by fib_lookup. | |
2112 | */ | |
2113 | ||
1e637c74 | 2114 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
f97c1e0c | 2115 | ipv4_is_loopback(saddr)) |
1da177e4 LT |
2116 | goto martian_source; |
2117 | ||
27a954bd | 2118 | if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0)) |
1da177e4 LT |
2119 | goto brd_input; |
2120 | ||
2121 | /* Accept zero addresses only to limited broadcast; | |
2122 | * I even do not know to fix it or not. Waiting for complains :-) | |
2123 | */ | |
f97c1e0c | 2124 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
2125 | goto martian_source; |
2126 | ||
27a954bd | 2127 | if (ipv4_is_zeronet(daddr) || ipv4_is_loopback(daddr)) |
1da177e4 LT |
2128 | goto martian_destination; |
2129 | ||
2130 | /* | |
2131 | * Now we are ready to route packet. | |
2132 | */ | |
68a5e3dd DM |
2133 | fl4.flowi4_oif = 0; |
2134 | fl4.flowi4_iif = dev->ifindex; | |
2135 | fl4.flowi4_mark = skb->mark; | |
2136 | fl4.flowi4_tos = tos; | |
2137 | fl4.flowi4_scope = RT_SCOPE_UNIVERSE; | |
2138 | fl4.daddr = daddr; | |
2139 | fl4.saddr = saddr; | |
2140 | err = fib_lookup(net, &fl4, &res); | |
ebc0ffae | 2141 | if (err != 0) { |
1da177e4 | 2142 | if (!IN_DEV_FORWARD(in_dev)) |
2c2910a4 | 2143 | goto e_hostunreach; |
1da177e4 LT |
2144 | goto no_route; |
2145 | } | |
1da177e4 LT |
2146 | |
2147 | RT_CACHE_STAT_INC(in_slow_tot); | |
2148 | ||
2149 | if (res.type == RTN_BROADCAST) | |
2150 | goto brd_input; | |
2151 | ||
2152 | if (res.type == RTN_LOCAL) { | |
5c04c819 | 2153 | err = fib_validate_source(skb, saddr, daddr, tos, |
ebc0ffae | 2154 | net->loopback_dev->ifindex, |
5c04c819 | 2155 | dev, &spec_dst, &itag); |
b5f7e755 ED |
2156 | if (err < 0) |
2157 | goto martian_source_keep_err; | |
2158 | if (err) | |
1da177e4 LT |
2159 | flags |= RTCF_DIRECTSRC; |
2160 | spec_dst = daddr; | |
2161 | goto local_input; | |
2162 | } | |
2163 | ||
2164 | if (!IN_DEV_FORWARD(in_dev)) | |
2c2910a4 | 2165 | goto e_hostunreach; |
1da177e4 LT |
2166 | if (res.type != RTN_UNICAST) |
2167 | goto martian_destination; | |
2168 | ||
68a5e3dd | 2169 | err = ip_mkroute_input(skb, &res, &fl4, in_dev, daddr, saddr, tos); |
1da177e4 LT |
2170 | out: return err; |
2171 | ||
2172 | brd_input: | |
2173 | if (skb->protocol != htons(ETH_P_IP)) | |
2174 | goto e_inval; | |
2175 | ||
f97c1e0c | 2176 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
2177 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
2178 | else { | |
5c04c819 MS |
2179 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, &spec_dst, |
2180 | &itag); | |
1da177e4 | 2181 | if (err < 0) |
b5f7e755 | 2182 | goto martian_source_keep_err; |
1da177e4 LT |
2183 | if (err) |
2184 | flags |= RTCF_DIRECTSRC; | |
2185 | } | |
2186 | flags |= RTCF_BROADCAST; | |
2187 | res.type = RTN_BROADCAST; | |
2188 | RT_CACHE_STAT_INC(in_brd); | |
2189 | ||
2190 | local_input: | |
0c4dcd58 | 2191 | rth = rt_dst_alloc(IN_DEV_CONF_GET(in_dev, NOPOLICY), false); |
1da177e4 LT |
2192 | if (!rth) |
2193 | goto e_nobufs; | |
2194 | ||
d8d1f30b | 2195 | rth->dst.output= ip_rt_bug; |
e84f84f2 | 2196 | rth->rt_genid = rt_genid(net); |
1da177e4 | 2197 | |
5e2b61f7 | 2198 | rth->rt_key_dst = daddr; |
1da177e4 | 2199 | rth->rt_dst = daddr; |
5e2b61f7 DM |
2200 | rth->rt_tos = tos; |
2201 | rth->rt_mark = skb->mark; | |
2202 | rth->rt_key_src = saddr; | |
1da177e4 | 2203 | rth->rt_src = saddr; |
c7066f70 | 2204 | #ifdef CONFIG_IP_ROUTE_CLASSID |
d8d1f30b | 2205 | rth->dst.tclassid = itag; |
1da177e4 | 2206 | #endif |
1b86a58f | 2207 | rth->rt_route_iif = dev->ifindex; |
5e2b61f7 | 2208 | rth->rt_iif = dev->ifindex; |
d8d1f30b CG |
2209 | rth->dst.dev = net->loopback_dev; |
2210 | dev_hold(rth->dst.dev); | |
1da177e4 LT |
2211 | rth->rt_gateway = daddr; |
2212 | rth->rt_spec_dst= spec_dst; | |
d8d1f30b | 2213 | rth->dst.input= ip_local_deliver; |
1da177e4 LT |
2214 | rth->rt_flags = flags|RTCF_LOCAL; |
2215 | if (res.type == RTN_UNREACHABLE) { | |
d8d1f30b CG |
2216 | rth->dst.input= ip_error; |
2217 | rth->dst.error= -err; | |
1da177e4 LT |
2218 | rth->rt_flags &= ~RTCF_LOCAL; |
2219 | } | |
2220 | rth->rt_type = res.type; | |
68a5e3dd DM |
2221 | hash = rt_hash(daddr, saddr, fl4.flowi4_iif, rt_genid(net)); |
2222 | rth = rt_intern_hash(hash, rth, skb, fl4.flowi4_iif); | |
b23dd4fe DM |
2223 | err = 0; |
2224 | if (IS_ERR(rth)) | |
2225 | err = PTR_ERR(rth); | |
ebc0ffae | 2226 | goto out; |
1da177e4 LT |
2227 | |
2228 | no_route: | |
2229 | RT_CACHE_STAT_INC(in_no_route); | |
2230 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE); | |
2231 | res.type = RTN_UNREACHABLE; | |
7f53878d MC |
2232 | if (err == -ESRCH) |
2233 | err = -ENETUNREACH; | |
1da177e4 LT |
2234 | goto local_input; |
2235 | ||
2236 | /* | |
2237 | * Do not cache martian addresses: they should be logged (RFC1812) | |
2238 | */ | |
2239 | martian_destination: | |
2240 | RT_CACHE_STAT_INC(in_martian_dst); | |
2241 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
2242 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) | |
673d57e7 HH |
2243 | printk(KERN_WARNING "martian destination %pI4 from %pI4, dev %s\n", |
2244 | &daddr, &saddr, dev->name); | |
1da177e4 | 2245 | #endif |
2c2910a4 DE |
2246 | |
2247 | e_hostunreach: | |
e905a9ed | 2248 | err = -EHOSTUNREACH; |
ebc0ffae | 2249 | goto out; |
2c2910a4 | 2250 | |
1da177e4 LT |
2251 | e_inval: |
2252 | err = -EINVAL; | |
ebc0ffae | 2253 | goto out; |
1da177e4 LT |
2254 | |
2255 | e_nobufs: | |
2256 | err = -ENOBUFS; | |
ebc0ffae | 2257 | goto out; |
1da177e4 LT |
2258 | |
2259 | martian_source: | |
b5f7e755 ED |
2260 | err = -EINVAL; |
2261 | martian_source_keep_err: | |
1da177e4 | 2262 | ip_handle_martian_source(dev, in_dev, skb, daddr, saddr); |
ebc0ffae | 2263 | goto out; |
1da177e4 LT |
2264 | } |
2265 | ||
407eadd9 ED |
2266 | int ip_route_input_common(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
2267 | u8 tos, struct net_device *dev, bool noref) | |
1da177e4 LT |
2268 | { |
2269 | struct rtable * rth; | |
2270 | unsigned hash; | |
2271 | int iif = dev->ifindex; | |
b5921910 | 2272 | struct net *net; |
96d36220 | 2273 | int res; |
1da177e4 | 2274 | |
c346dca1 | 2275 | net = dev_net(dev); |
1080d709 | 2276 | |
96d36220 ED |
2277 | rcu_read_lock(); |
2278 | ||
1080d709 NH |
2279 | if (!rt_caching(net)) |
2280 | goto skip_cache; | |
2281 | ||
1da177e4 | 2282 | tos &= IPTOS_RT_MASK; |
e84f84f2 | 2283 | hash = rt_hash(daddr, saddr, iif, rt_genid(net)); |
1da177e4 | 2284 | |
1da177e4 | 2285 | for (rth = rcu_dereference(rt_hash_table[hash].chain); rth; |
d8d1f30b | 2286 | rth = rcu_dereference(rth->dst.rt_next)) { |
5e2b61f7 DM |
2287 | if ((((__force u32)rth->rt_key_dst ^ (__force u32)daddr) | |
2288 | ((__force u32)rth->rt_key_src ^ (__force u32)saddr) | | |
2289 | (rth->rt_iif ^ iif) | | |
2290 | rth->rt_oif | | |
2291 | (rth->rt_tos ^ tos)) == 0 && | |
2292 | rth->rt_mark == skb->mark && | |
d8d1f30b | 2293 | net_eq(dev_net(rth->dst.dev), net) && |
e84f84f2 | 2294 | !rt_is_expired(rth)) { |
407eadd9 | 2295 | if (noref) { |
d8d1f30b CG |
2296 | dst_use_noref(&rth->dst, jiffies); |
2297 | skb_dst_set_noref(skb, &rth->dst); | |
407eadd9 | 2298 | } else { |
d8d1f30b CG |
2299 | dst_use(&rth->dst, jiffies); |
2300 | skb_dst_set(skb, &rth->dst); | |
407eadd9 | 2301 | } |
1da177e4 LT |
2302 | RT_CACHE_STAT_INC(in_hit); |
2303 | rcu_read_unlock(); | |
1da177e4 LT |
2304 | return 0; |
2305 | } | |
2306 | RT_CACHE_STAT_INC(in_hlist_search); | |
2307 | } | |
1da177e4 | 2308 | |
1080d709 | 2309 | skip_cache: |
1da177e4 LT |
2310 | /* Multicast recognition logic is moved from route cache to here. |
2311 | The problem was that too many Ethernet cards have broken/missing | |
2312 | hardware multicast filters :-( As result the host on multicasting | |
2313 | network acquires a lot of useless route cache entries, sort of | |
2314 | SDR messages from all the world. Now we try to get rid of them. | |
2315 | Really, provided software IP multicast filter is organized | |
2316 | reasonably (at least, hashed), it does not result in a slowdown | |
2317 | comparing with route cache reject entries. | |
2318 | Note, that multicast routers are not affected, because | |
2319 | route cache entry is created eventually. | |
2320 | */ | |
f97c1e0c | 2321 | if (ipv4_is_multicast(daddr)) { |
96d36220 | 2322 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 2323 | |
96d36220 | 2324 | if (in_dev) { |
dbdd9a52 DM |
2325 | int our = ip_check_mc_rcu(in_dev, daddr, saddr, |
2326 | ip_hdr(skb)->protocol); | |
1da177e4 LT |
2327 | if (our |
2328 | #ifdef CONFIG_IP_MROUTE | |
9d4fb27d JP |
2329 | || |
2330 | (!ipv4_is_local_multicast(daddr) && | |
2331 | IN_DEV_MFORWARD(in_dev)) | |
1da177e4 | 2332 | #endif |
9d4fb27d | 2333 | ) { |
96d36220 ED |
2334 | int res = ip_route_input_mc(skb, daddr, saddr, |
2335 | tos, dev, our); | |
1da177e4 | 2336 | rcu_read_unlock(); |
96d36220 | 2337 | return res; |
1da177e4 LT |
2338 | } |
2339 | } | |
2340 | rcu_read_unlock(); | |
2341 | return -EINVAL; | |
2342 | } | |
96d36220 ED |
2343 | res = ip_route_input_slow(skb, daddr, saddr, tos, dev); |
2344 | rcu_read_unlock(); | |
2345 | return res; | |
1da177e4 | 2346 | } |
407eadd9 | 2347 | EXPORT_SYMBOL(ip_route_input_common); |
1da177e4 | 2348 | |
ebc0ffae | 2349 | /* called with rcu_read_lock() */ |
982721f3 | 2350 | static struct rtable *__mkroute_output(const struct fib_result *res, |
68a5e3dd DM |
2351 | const struct flowi4 *fl4, |
2352 | const struct flowi4 *oldflp4, | |
5ada5527 DM |
2353 | struct net_device *dev_out, |
2354 | unsigned int flags) | |
1da177e4 | 2355 | { |
982721f3 | 2356 | struct fib_info *fi = res->fi; |
68a5e3dd | 2357 | u32 tos = RT_FL_TOS(oldflp4); |
5ada5527 | 2358 | struct in_device *in_dev; |
982721f3 | 2359 | u16 type = res->type; |
5ada5527 | 2360 | struct rtable *rth; |
1da177e4 | 2361 | |
68a5e3dd | 2362 | if (ipv4_is_loopback(fl4->saddr) && !(dev_out->flags & IFF_LOOPBACK)) |
5ada5527 | 2363 | return ERR_PTR(-EINVAL); |
1da177e4 | 2364 | |
68a5e3dd | 2365 | if (ipv4_is_lbcast(fl4->daddr)) |
982721f3 | 2366 | type = RTN_BROADCAST; |
68a5e3dd | 2367 | else if (ipv4_is_multicast(fl4->daddr)) |
982721f3 | 2368 | type = RTN_MULTICAST; |
68a5e3dd | 2369 | else if (ipv4_is_zeronet(fl4->daddr)) |
5ada5527 | 2370 | return ERR_PTR(-EINVAL); |
1da177e4 LT |
2371 | |
2372 | if (dev_out->flags & IFF_LOOPBACK) | |
2373 | flags |= RTCF_LOCAL; | |
2374 | ||
dd28d1a0 | 2375 | in_dev = __in_dev_get_rcu(dev_out); |
ebc0ffae | 2376 | if (!in_dev) |
5ada5527 | 2377 | return ERR_PTR(-EINVAL); |
ebc0ffae | 2378 | |
982721f3 | 2379 | if (type == RTN_BROADCAST) { |
1da177e4 | 2380 | flags |= RTCF_BROADCAST | RTCF_LOCAL; |
982721f3 DM |
2381 | fi = NULL; |
2382 | } else if (type == RTN_MULTICAST) { | |
dd28d1a0 | 2383 | flags |= RTCF_MULTICAST | RTCF_LOCAL; |
68a5e3dd DM |
2384 | if (!ip_check_mc_rcu(in_dev, oldflp4->daddr, oldflp4->saddr, |
2385 | oldflp4->flowi4_proto)) | |
1da177e4 LT |
2386 | flags &= ~RTCF_LOCAL; |
2387 | /* If multicast route do not exist use | |
dd28d1a0 ED |
2388 | * default one, but do not gateway in this case. |
2389 | * Yes, it is hack. | |
1da177e4 | 2390 | */ |
982721f3 DM |
2391 | if (fi && res->prefixlen < 4) |
2392 | fi = NULL; | |
1da177e4 LT |
2393 | } |
2394 | ||
0c4dcd58 DM |
2395 | rth = rt_dst_alloc(IN_DEV_CONF_GET(in_dev, NOPOLICY), |
2396 | IN_DEV_CONF_GET(in_dev, NOXFRM)); | |
8391d07b | 2397 | if (!rth) |
5ada5527 | 2398 | return ERR_PTR(-ENOBUFS); |
8391d07b | 2399 | |
68a5e3dd | 2400 | rth->rt_key_dst = oldflp4->daddr; |
5e2b61f7 | 2401 | rth->rt_tos = tos; |
68a5e3dd DM |
2402 | rth->rt_key_src = oldflp4->saddr; |
2403 | rth->rt_oif = oldflp4->flowi4_oif; | |
2404 | rth->rt_mark = oldflp4->flowi4_mark; | |
2405 | rth->rt_dst = fl4->daddr; | |
2406 | rth->rt_src = fl4->saddr; | |
1b86a58f OH |
2407 | rth->rt_route_iif = 0; |
2408 | rth->rt_iif = oldflp4->flowi4_oif ? : dev_out->ifindex; | |
e905a9ed | 2409 | /* get references to the devices that are to be hold by the routing |
1da177e4 | 2410 | cache entry */ |
d8d1f30b | 2411 | rth->dst.dev = dev_out; |
1da177e4 | 2412 | dev_hold(dev_out); |
68a5e3dd DM |
2413 | rth->rt_gateway = fl4->daddr; |
2414 | rth->rt_spec_dst= fl4->saddr; | |
1da177e4 | 2415 | |
d8d1f30b | 2416 | rth->dst.output=ip_output; |
e84f84f2 | 2417 | rth->rt_genid = rt_genid(dev_net(dev_out)); |
1da177e4 LT |
2418 | |
2419 | RT_CACHE_STAT_INC(out_slow_tot); | |
2420 | ||
2421 | if (flags & RTCF_LOCAL) { | |
d8d1f30b | 2422 | rth->dst.input = ip_local_deliver; |
68a5e3dd | 2423 | rth->rt_spec_dst = fl4->daddr; |
1da177e4 LT |
2424 | } |
2425 | if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { | |
68a5e3dd | 2426 | rth->rt_spec_dst = fl4->saddr; |
e905a9ed | 2427 | if (flags & RTCF_LOCAL && |
1da177e4 | 2428 | !(dev_out->flags & IFF_LOOPBACK)) { |
d8d1f30b | 2429 | rth->dst.output = ip_mc_output; |
1da177e4 LT |
2430 | RT_CACHE_STAT_INC(out_slow_mc); |
2431 | } | |
2432 | #ifdef CONFIG_IP_MROUTE | |
982721f3 | 2433 | if (type == RTN_MULTICAST) { |
1da177e4 | 2434 | if (IN_DEV_MFORWARD(in_dev) && |
68a5e3dd | 2435 | !ipv4_is_local_multicast(oldflp4->daddr)) { |
d8d1f30b CG |
2436 | rth->dst.input = ip_mr_input; |
2437 | rth->dst.output = ip_mc_output; | |
1da177e4 LT |
2438 | } |
2439 | } | |
2440 | #endif | |
2441 | } | |
2442 | ||
68a5e3dd | 2443 | rt_set_nexthop(rth, oldflp4, res, fi, type, 0); |
1da177e4 LT |
2444 | |
2445 | rth->rt_flags = flags; | |
5ada5527 | 2446 | return rth; |
1da177e4 LT |
2447 | } |
2448 | ||
1da177e4 LT |
2449 | /* |
2450 | * Major route resolver routine. | |
0197aa38 | 2451 | * called with rcu_read_lock(); |
1da177e4 LT |
2452 | */ |
2453 | ||
b23dd4fe | 2454 | static struct rtable *ip_route_output_slow(struct net *net, |
68a5e3dd | 2455 | const struct flowi4 *oldflp4) |
1da177e4 | 2456 | { |
68a5e3dd DM |
2457 | u32 tos = RT_FL_TOS(oldflp4); |
2458 | struct flowi4 fl4; | |
1da177e4 | 2459 | struct fib_result res; |
0197aa38 | 2460 | unsigned int flags = 0; |
1da177e4 | 2461 | struct net_device *dev_out = NULL; |
5ada5527 | 2462 | struct rtable *rth; |
1da177e4 LT |
2463 | |
2464 | res.fi = NULL; | |
2465 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
2466 | res.r = NULL; | |
2467 | #endif | |
2468 | ||
68a5e3dd DM |
2469 | fl4.flowi4_oif = oldflp4->flowi4_oif; |
2470 | fl4.flowi4_iif = net->loopback_dev->ifindex; | |
2471 | fl4.flowi4_mark = oldflp4->flowi4_mark; | |
2472 | fl4.daddr = oldflp4->daddr; | |
2473 | fl4.saddr = oldflp4->saddr; | |
2474 | fl4.flowi4_tos = tos & IPTOS_RT_MASK; | |
2475 | fl4.flowi4_scope = ((tos & RTO_ONLINK) ? | |
44713b67 DM |
2476 | RT_SCOPE_LINK : RT_SCOPE_UNIVERSE); |
2477 | ||
010c2708 | 2478 | rcu_read_lock(); |
68a5e3dd | 2479 | if (oldflp4->saddr) { |
b23dd4fe | 2480 | rth = ERR_PTR(-EINVAL); |
68a5e3dd DM |
2481 | if (ipv4_is_multicast(oldflp4->saddr) || |
2482 | ipv4_is_lbcast(oldflp4->saddr) || | |
2483 | ipv4_is_zeronet(oldflp4->saddr)) | |
1da177e4 LT |
2484 | goto out; |
2485 | ||
1da177e4 LT |
2486 | /* I removed check for oif == dev_out->oif here. |
2487 | It was wrong for two reasons: | |
1ab35276 DL |
2488 | 1. ip_dev_find(net, saddr) can return wrong iface, if saddr |
2489 | is assigned to multiple interfaces. | |
1da177e4 LT |
2490 | 2. Moreover, we are allowed to send packets with saddr |
2491 | of another iface. --ANK | |
2492 | */ | |
2493 | ||
68a5e3dd DM |
2494 | if (oldflp4->flowi4_oif == 0 && |
2495 | (ipv4_is_multicast(oldflp4->daddr) || | |
2496 | ipv4_is_lbcast(oldflp4->daddr))) { | |
a210d01a | 2497 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
68a5e3dd | 2498 | dev_out = __ip_dev_find(net, oldflp4->saddr, false); |
a210d01a JA |
2499 | if (dev_out == NULL) |
2500 | goto out; | |
2501 | ||
1da177e4 LT |
2502 | /* Special hack: user can direct multicasts |
2503 | and limited broadcast via necessary interface | |
2504 | without fiddling with IP_MULTICAST_IF or IP_PKTINFO. | |
2505 | This hack is not just for fun, it allows | |
2506 | vic,vat and friends to work. | |
2507 | They bind socket to loopback, set ttl to zero | |
2508 | and expect that it will work. | |
2509 | From the viewpoint of routing cache they are broken, | |
2510 | because we are not allowed to build multicast path | |
2511 | with loopback source addr (look, routing cache | |
2512 | cannot know, that ttl is zero, so that packet | |
2513 | will not leave this host and route is valid). | |
2514 | Luckily, this hack is good workaround. | |
2515 | */ | |
2516 | ||
68a5e3dd | 2517 | fl4.flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2518 | goto make_route; |
2519 | } | |
a210d01a | 2520 | |
68a5e3dd | 2521 | if (!(oldflp4->flowi4_flags & FLOWI_FLAG_ANYSRC)) { |
a210d01a | 2522 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
68a5e3dd | 2523 | if (!__ip_dev_find(net, oldflp4->saddr, false)) |
a210d01a | 2524 | goto out; |
a210d01a | 2525 | } |
1da177e4 LT |
2526 | } |
2527 | ||
2528 | ||
68a5e3dd DM |
2529 | if (oldflp4->flowi4_oif) { |
2530 | dev_out = dev_get_by_index_rcu(net, oldflp4->flowi4_oif); | |
b23dd4fe | 2531 | rth = ERR_PTR(-ENODEV); |
1da177e4 LT |
2532 | if (dev_out == NULL) |
2533 | goto out; | |
e5ed6399 HX |
2534 | |
2535 | /* RACE: Check return value of inet_select_addr instead. */ | |
fc75fc83 | 2536 | if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) { |
b23dd4fe | 2537 | rth = ERR_PTR(-ENETUNREACH); |
fc75fc83 ED |
2538 | goto out; |
2539 | } | |
68a5e3dd DM |
2540 | if (ipv4_is_local_multicast(oldflp4->daddr) || |
2541 | ipv4_is_lbcast(oldflp4->daddr)) { | |
2542 | if (!fl4.saddr) | |
2543 | fl4.saddr = inet_select_addr(dev_out, 0, | |
2544 | RT_SCOPE_LINK); | |
1da177e4 LT |
2545 | goto make_route; |
2546 | } | |
68a5e3dd DM |
2547 | if (!fl4.saddr) { |
2548 | if (ipv4_is_multicast(oldflp4->daddr)) | |
2549 | fl4.saddr = inet_select_addr(dev_out, 0, | |
2550 | fl4.flowi4_scope); | |
2551 | else if (!oldflp4->daddr) | |
2552 | fl4.saddr = inet_select_addr(dev_out, 0, | |
2553 | RT_SCOPE_HOST); | |
1da177e4 LT |
2554 | } |
2555 | } | |
2556 | ||
68a5e3dd DM |
2557 | if (!fl4.daddr) { |
2558 | fl4.daddr = fl4.saddr; | |
2559 | if (!fl4.daddr) | |
2560 | fl4.daddr = fl4.saddr = htonl(INADDR_LOOPBACK); | |
b40afd0e | 2561 | dev_out = net->loopback_dev; |
68a5e3dd | 2562 | fl4.flowi4_oif = net->loopback_dev->ifindex; |
1da177e4 LT |
2563 | res.type = RTN_LOCAL; |
2564 | flags |= RTCF_LOCAL; | |
2565 | goto make_route; | |
2566 | } | |
2567 | ||
68a5e3dd | 2568 | if (fib_lookup(net, &fl4, &res)) { |
1da177e4 | 2569 | res.fi = NULL; |
68a5e3dd | 2570 | if (oldflp4->flowi4_oif) { |
1da177e4 LT |
2571 | /* Apparently, routing tables are wrong. Assume, |
2572 | that the destination is on link. | |
2573 | ||
2574 | WHY? DW. | |
2575 | Because we are allowed to send to iface | |
2576 | even if it has NO routes and NO assigned | |
2577 | addresses. When oif is specified, routing | |
2578 | tables are looked up with only one purpose: | |
2579 | to catch if destination is gatewayed, rather than | |
2580 | direct. Moreover, if MSG_DONTROUTE is set, | |
2581 | we send packet, ignoring both routing tables | |
2582 | and ifaddr state. --ANK | |
2583 | ||
2584 | ||
2585 | We could make it even if oif is unknown, | |
2586 | likely IPv6, but we do not. | |
2587 | */ | |
2588 | ||
68a5e3dd DM |
2589 | if (fl4.saddr == 0) |
2590 | fl4.saddr = inet_select_addr(dev_out, 0, | |
2591 | RT_SCOPE_LINK); | |
1da177e4 LT |
2592 | res.type = RTN_UNICAST; |
2593 | goto make_route; | |
2594 | } | |
b23dd4fe | 2595 | rth = ERR_PTR(-ENETUNREACH); |
1da177e4 LT |
2596 | goto out; |
2597 | } | |
1da177e4 LT |
2598 | |
2599 | if (res.type == RTN_LOCAL) { | |
68a5e3dd | 2600 | if (!fl4.saddr) { |
9fc3bbb4 | 2601 | if (res.fi->fib_prefsrc) |
68a5e3dd | 2602 | fl4.saddr = res.fi->fib_prefsrc; |
9fc3bbb4 | 2603 | else |
68a5e3dd | 2604 | fl4.saddr = fl4.daddr; |
9fc3bbb4 | 2605 | } |
b40afd0e | 2606 | dev_out = net->loopback_dev; |
68a5e3dd | 2607 | fl4.flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2608 | res.fi = NULL; |
2609 | flags |= RTCF_LOCAL; | |
2610 | goto make_route; | |
2611 | } | |
2612 | ||
2613 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
68a5e3dd | 2614 | if (res.fi->fib_nhs > 1 && fl4.flowi4_oif == 0) |
1b7fe593 | 2615 | fib_select_multipath(&res); |
1da177e4 LT |
2616 | else |
2617 | #endif | |
21d8c49e DM |
2618 | if (!res.prefixlen && |
2619 | res.table->tb_num_default > 1 && | |
2620 | res.type == RTN_UNICAST && !fl4.flowi4_oif) | |
0c838ff1 | 2621 | fib_select_default(&res); |
1da177e4 | 2622 | |
68a5e3dd | 2623 | if (!fl4.saddr) |
436c3b66 | 2624 | fl4.saddr = FIB_RES_PREFSRC(net, res); |
1da177e4 | 2625 | |
1da177e4 | 2626 | dev_out = FIB_RES_DEV(res); |
68a5e3dd | 2627 | fl4.flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2628 | |
2629 | ||
2630 | make_route: | |
68a5e3dd | 2631 | rth = __mkroute_output(&res, &fl4, oldflp4, dev_out, flags); |
b23dd4fe | 2632 | if (!IS_ERR(rth)) { |
5ada5527 DM |
2633 | unsigned int hash; |
2634 | ||
68a5e3dd | 2635 | hash = rt_hash(oldflp4->daddr, oldflp4->saddr, oldflp4->flowi4_oif, |
5ada5527 | 2636 | rt_genid(dev_net(dev_out))); |
68a5e3dd | 2637 | rth = rt_intern_hash(hash, rth, NULL, oldflp4->flowi4_oif); |
5ada5527 | 2638 | } |
1da177e4 | 2639 | |
010c2708 DM |
2640 | out: |
2641 | rcu_read_unlock(); | |
b23dd4fe | 2642 | return rth; |
1da177e4 LT |
2643 | } |
2644 | ||
9d6ec938 | 2645 | struct rtable *__ip_route_output_key(struct net *net, const struct flowi4 *flp4) |
1da177e4 | 2646 | { |
1da177e4 | 2647 | struct rtable *rth; |
010c2708 | 2648 | unsigned int hash; |
1da177e4 | 2649 | |
1080d709 NH |
2650 | if (!rt_caching(net)) |
2651 | goto slow_output; | |
2652 | ||
9d6ec938 | 2653 | hash = rt_hash(flp4->daddr, flp4->saddr, flp4->flowi4_oif, rt_genid(net)); |
1da177e4 LT |
2654 | |
2655 | rcu_read_lock_bh(); | |
a898def2 | 2656 | for (rth = rcu_dereference_bh(rt_hash_table[hash].chain); rth; |
d8d1f30b | 2657 | rth = rcu_dereference_bh(rth->dst.rt_next)) { |
9d6ec938 DM |
2658 | if (rth->rt_key_dst == flp4->daddr && |
2659 | rth->rt_key_src == flp4->saddr && | |
c7537967 | 2660 | rt_is_output_route(rth) && |
9d6ec938 DM |
2661 | rth->rt_oif == flp4->flowi4_oif && |
2662 | rth->rt_mark == flp4->flowi4_mark && | |
2663 | !((rth->rt_tos ^ flp4->flowi4_tos) & | |
b5921910 | 2664 | (IPTOS_RT_MASK | RTO_ONLINK)) && |
d8d1f30b | 2665 | net_eq(dev_net(rth->dst.dev), net) && |
e84f84f2 | 2666 | !rt_is_expired(rth)) { |
d8d1f30b | 2667 | dst_use(&rth->dst, jiffies); |
1da177e4 LT |
2668 | RT_CACHE_STAT_INC(out_hit); |
2669 | rcu_read_unlock_bh(); | |
b23dd4fe | 2670 | return rth; |
1da177e4 LT |
2671 | } |
2672 | RT_CACHE_STAT_INC(out_hlist_search); | |
2673 | } | |
2674 | rcu_read_unlock_bh(); | |
2675 | ||
1080d709 | 2676 | slow_output: |
9d6ec938 | 2677 | return ip_route_output_slow(net, flp4); |
1da177e4 | 2678 | } |
d8c97a94 ACM |
2679 | EXPORT_SYMBOL_GPL(__ip_route_output_key); |
2680 | ||
ae2688d5 JW |
2681 | static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie) |
2682 | { | |
2683 | return NULL; | |
2684 | } | |
2685 | ||
ec831ea7 RD |
2686 | static unsigned int ipv4_blackhole_default_mtu(const struct dst_entry *dst) |
2687 | { | |
2688 | return 0; | |
2689 | } | |
2690 | ||
14e50e57 DM |
2691 | static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu) |
2692 | { | |
2693 | } | |
2694 | ||
2695 | static struct dst_ops ipv4_dst_blackhole_ops = { | |
2696 | .family = AF_INET, | |
09640e63 | 2697 | .protocol = cpu_to_be16(ETH_P_IP), |
14e50e57 | 2698 | .destroy = ipv4_dst_destroy, |
ae2688d5 | 2699 | .check = ipv4_blackhole_dst_check, |
ec831ea7 | 2700 | .default_mtu = ipv4_blackhole_default_mtu, |
214f45c9 | 2701 | .default_advmss = ipv4_default_advmss, |
14e50e57 | 2702 | .update_pmtu = ipv4_rt_blackhole_update_pmtu, |
14e50e57 DM |
2703 | }; |
2704 | ||
2774c131 | 2705 | struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig) |
14e50e57 | 2706 | { |
2774c131 DM |
2707 | struct rtable *rt = dst_alloc(&ipv4_dst_blackhole_ops, 1); |
2708 | struct rtable *ort = (struct rtable *) dst_orig; | |
14e50e57 DM |
2709 | |
2710 | if (rt) { | |
d8d1f30b | 2711 | struct dst_entry *new = &rt->dst; |
14e50e57 | 2712 | |
14e50e57 | 2713 | new->__use = 1; |
352e512c HX |
2714 | new->input = dst_discard; |
2715 | new->output = dst_discard; | |
defb3519 | 2716 | dst_copy_metrics(new, &ort->dst); |
14e50e57 | 2717 | |
d8d1f30b | 2718 | new->dev = ort->dst.dev; |
14e50e57 DM |
2719 | if (new->dev) |
2720 | dev_hold(new->dev); | |
2721 | ||
5e2b61f7 DM |
2722 | rt->rt_key_dst = ort->rt_key_dst; |
2723 | rt->rt_key_src = ort->rt_key_src; | |
2724 | rt->rt_tos = ort->rt_tos; | |
1b86a58f | 2725 | rt->rt_route_iif = ort->rt_route_iif; |
5e2b61f7 DM |
2726 | rt->rt_iif = ort->rt_iif; |
2727 | rt->rt_oif = ort->rt_oif; | |
2728 | rt->rt_mark = ort->rt_mark; | |
14e50e57 | 2729 | |
e84f84f2 | 2730 | rt->rt_genid = rt_genid(net); |
14e50e57 DM |
2731 | rt->rt_flags = ort->rt_flags; |
2732 | rt->rt_type = ort->rt_type; | |
2733 | rt->rt_dst = ort->rt_dst; | |
2734 | rt->rt_src = ort->rt_src; | |
14e50e57 DM |
2735 | rt->rt_gateway = ort->rt_gateway; |
2736 | rt->rt_spec_dst = ort->rt_spec_dst; | |
2737 | rt->peer = ort->peer; | |
2738 | if (rt->peer) | |
2739 | atomic_inc(&rt->peer->refcnt); | |
62fa8a84 DM |
2740 | rt->fi = ort->fi; |
2741 | if (rt->fi) | |
2742 | atomic_inc(&rt->fi->fib_clntref); | |
14e50e57 DM |
2743 | |
2744 | dst_free(new); | |
2745 | } | |
2746 | ||
2774c131 DM |
2747 | dst_release(dst_orig); |
2748 | ||
2749 | return rt ? &rt->dst : ERR_PTR(-ENOMEM); | |
14e50e57 DM |
2750 | } |
2751 | ||
9d6ec938 | 2752 | struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4, |
b23dd4fe | 2753 | struct sock *sk) |
1da177e4 | 2754 | { |
9d6ec938 | 2755 | struct rtable *rt = __ip_route_output_key(net, flp4); |
1da177e4 | 2756 | |
b23dd4fe DM |
2757 | if (IS_ERR(rt)) |
2758 | return rt; | |
1da177e4 | 2759 | |
9d6ec938 DM |
2760 | if (flp4->flowi4_proto) { |
2761 | if (!flp4->saddr) | |
2762 | flp4->saddr = rt->rt_src; | |
2763 | if (!flp4->daddr) | |
2764 | flp4->daddr = rt->rt_dst; | |
2765 | rt = (struct rtable *) xfrm_lookup(net, &rt->dst, | |
2766 | flowi4_to_flowi(flp4), | |
2767 | sk, 0); | |
1da177e4 LT |
2768 | } |
2769 | ||
b23dd4fe | 2770 | return rt; |
1da177e4 | 2771 | } |
d8c97a94 ACM |
2772 | EXPORT_SYMBOL_GPL(ip_route_output_flow); |
2773 | ||
4feb88e5 BT |
2774 | static int rt_fill_info(struct net *net, |
2775 | struct sk_buff *skb, u32 pid, u32 seq, int event, | |
b6544c0b | 2776 | int nowait, unsigned int flags) |
1da177e4 | 2777 | { |
511c3f92 | 2778 | struct rtable *rt = skb_rtable(skb); |
1da177e4 | 2779 | struct rtmsg *r; |
be403ea1 | 2780 | struct nlmsghdr *nlh; |
e3703b3d TG |
2781 | long expires; |
2782 | u32 id = 0, ts = 0, tsage = 0, error; | |
be403ea1 TG |
2783 | |
2784 | nlh = nlmsg_put(skb, pid, seq, event, sizeof(*r), flags); | |
2785 | if (nlh == NULL) | |
26932566 | 2786 | return -EMSGSIZE; |
be403ea1 TG |
2787 | |
2788 | r = nlmsg_data(nlh); | |
1da177e4 LT |
2789 | r->rtm_family = AF_INET; |
2790 | r->rtm_dst_len = 32; | |
2791 | r->rtm_src_len = 0; | |
5e2b61f7 | 2792 | r->rtm_tos = rt->rt_tos; |
1da177e4 | 2793 | r->rtm_table = RT_TABLE_MAIN; |
be403ea1 | 2794 | NLA_PUT_U32(skb, RTA_TABLE, RT_TABLE_MAIN); |
1da177e4 LT |
2795 | r->rtm_type = rt->rt_type; |
2796 | r->rtm_scope = RT_SCOPE_UNIVERSE; | |
2797 | r->rtm_protocol = RTPROT_UNSPEC; | |
2798 | r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED; | |
2799 | if (rt->rt_flags & RTCF_NOTIFY) | |
2800 | r->rtm_flags |= RTM_F_NOTIFY; | |
be403ea1 | 2801 | |
17fb2c64 | 2802 | NLA_PUT_BE32(skb, RTA_DST, rt->rt_dst); |
be403ea1 | 2803 | |
5e2b61f7 | 2804 | if (rt->rt_key_src) { |
1da177e4 | 2805 | r->rtm_src_len = 32; |
5e2b61f7 | 2806 | NLA_PUT_BE32(skb, RTA_SRC, rt->rt_key_src); |
1da177e4 | 2807 | } |
d8d1f30b CG |
2808 | if (rt->dst.dev) |
2809 | NLA_PUT_U32(skb, RTA_OIF, rt->dst.dev->ifindex); | |
c7066f70 | 2810 | #ifdef CONFIG_IP_ROUTE_CLASSID |
d8d1f30b CG |
2811 | if (rt->dst.tclassid) |
2812 | NLA_PUT_U32(skb, RTA_FLOW, rt->dst.tclassid); | |
1da177e4 | 2813 | #endif |
c7537967 | 2814 | if (rt_is_input_route(rt)) |
17fb2c64 | 2815 | NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_spec_dst); |
5e2b61f7 | 2816 | else if (rt->rt_src != rt->rt_key_src) |
17fb2c64 | 2817 | NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_src); |
be403ea1 | 2818 | |
1da177e4 | 2819 | if (rt->rt_dst != rt->rt_gateway) |
17fb2c64 | 2820 | NLA_PUT_BE32(skb, RTA_GATEWAY, rt->rt_gateway); |
be403ea1 | 2821 | |
defb3519 | 2822 | if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0) |
be403ea1 TG |
2823 | goto nla_put_failure; |
2824 | ||
5e2b61f7 DM |
2825 | if (rt->rt_mark) |
2826 | NLA_PUT_BE32(skb, RTA_MARK, rt->rt_mark); | |
963bfeee | 2827 | |
d8d1f30b | 2828 | error = rt->dst.error; |
2c8cec5c DM |
2829 | expires = (rt->peer && rt->peer->pmtu_expires) ? |
2830 | rt->peer->pmtu_expires - jiffies : 0; | |
1da177e4 | 2831 | if (rt->peer) { |
317fe0e6 | 2832 | inet_peer_refcheck(rt->peer); |
2c1409a0 | 2833 | id = atomic_read(&rt->peer->ip_id_count) & 0xffff; |
1da177e4 | 2834 | if (rt->peer->tcp_ts_stamp) { |
e3703b3d | 2835 | ts = rt->peer->tcp_ts; |
9d729f72 | 2836 | tsage = get_seconds() - rt->peer->tcp_ts_stamp; |
1da177e4 LT |
2837 | } |
2838 | } | |
be403ea1 | 2839 | |
c7537967 | 2840 | if (rt_is_input_route(rt)) { |
1da177e4 | 2841 | #ifdef CONFIG_IP_MROUTE |
e448515c | 2842 | __be32 dst = rt->rt_dst; |
1da177e4 | 2843 | |
f97c1e0c | 2844 | if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) && |
4feb88e5 BT |
2845 | IPV4_DEVCONF_ALL(net, MC_FORWARDING)) { |
2846 | int err = ipmr_get_route(net, skb, r, nowait); | |
1da177e4 LT |
2847 | if (err <= 0) { |
2848 | if (!nowait) { | |
2849 | if (err == 0) | |
2850 | return 0; | |
be403ea1 | 2851 | goto nla_put_failure; |
1da177e4 LT |
2852 | } else { |
2853 | if (err == -EMSGSIZE) | |
be403ea1 | 2854 | goto nla_put_failure; |
e3703b3d | 2855 | error = err; |
1da177e4 LT |
2856 | } |
2857 | } | |
2858 | } else | |
2859 | #endif | |
5e2b61f7 | 2860 | NLA_PUT_U32(skb, RTA_IIF, rt->rt_iif); |
1da177e4 LT |
2861 | } |
2862 | ||
d8d1f30b | 2863 | if (rtnl_put_cacheinfo(skb, &rt->dst, id, ts, tsage, |
e3703b3d TG |
2864 | expires, error) < 0) |
2865 | goto nla_put_failure; | |
be403ea1 TG |
2866 | |
2867 | return nlmsg_end(skb, nlh); | |
1da177e4 | 2868 | |
be403ea1 | 2869 | nla_put_failure: |
26932566 PM |
2870 | nlmsg_cancel(skb, nlh); |
2871 | return -EMSGSIZE; | |
1da177e4 LT |
2872 | } |
2873 | ||
63f3444f | 2874 | static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg) |
1da177e4 | 2875 | { |
3b1e0a65 | 2876 | struct net *net = sock_net(in_skb->sk); |
d889ce3b TG |
2877 | struct rtmsg *rtm; |
2878 | struct nlattr *tb[RTA_MAX+1]; | |
1da177e4 | 2879 | struct rtable *rt = NULL; |
9e12bb22 AV |
2880 | __be32 dst = 0; |
2881 | __be32 src = 0; | |
2882 | u32 iif; | |
d889ce3b | 2883 | int err; |
963bfeee | 2884 | int mark; |
1da177e4 LT |
2885 | struct sk_buff *skb; |
2886 | ||
d889ce3b TG |
2887 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy); |
2888 | if (err < 0) | |
2889 | goto errout; | |
2890 | ||
2891 | rtm = nlmsg_data(nlh); | |
2892 | ||
1da177e4 | 2893 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
d889ce3b TG |
2894 | if (skb == NULL) { |
2895 | err = -ENOBUFS; | |
2896 | goto errout; | |
2897 | } | |
1da177e4 LT |
2898 | |
2899 | /* Reserve room for dummy headers, this skb can pass | |
2900 | through good chunk of routing engine. | |
2901 | */ | |
459a98ed | 2902 | skb_reset_mac_header(skb); |
c1d2bbe1 | 2903 | skb_reset_network_header(skb); |
d2c962b8 SH |
2904 | |
2905 | /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */ | |
eddc9ec5 | 2906 | ip_hdr(skb)->protocol = IPPROTO_ICMP; |
1da177e4 LT |
2907 | skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr)); |
2908 | ||
17fb2c64 AV |
2909 | src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0; |
2910 | dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0; | |
d889ce3b | 2911 | iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0; |
963bfeee | 2912 | mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0; |
1da177e4 LT |
2913 | |
2914 | if (iif) { | |
d889ce3b TG |
2915 | struct net_device *dev; |
2916 | ||
1937504d | 2917 | dev = __dev_get_by_index(net, iif); |
d889ce3b TG |
2918 | if (dev == NULL) { |
2919 | err = -ENODEV; | |
2920 | goto errout_free; | |
2921 | } | |
2922 | ||
1da177e4 LT |
2923 | skb->protocol = htons(ETH_P_IP); |
2924 | skb->dev = dev; | |
963bfeee | 2925 | skb->mark = mark; |
1da177e4 LT |
2926 | local_bh_disable(); |
2927 | err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev); | |
2928 | local_bh_enable(); | |
d889ce3b | 2929 | |
511c3f92 | 2930 | rt = skb_rtable(skb); |
d8d1f30b CG |
2931 | if (err == 0 && rt->dst.error) |
2932 | err = -rt->dst.error; | |
1da177e4 | 2933 | } else { |
68a5e3dd DM |
2934 | struct flowi4 fl4 = { |
2935 | .daddr = dst, | |
2936 | .saddr = src, | |
2937 | .flowi4_tos = rtm->rtm_tos, | |
2938 | .flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0, | |
2939 | .flowi4_mark = mark, | |
d889ce3b | 2940 | }; |
9d6ec938 | 2941 | rt = ip_route_output_key(net, &fl4); |
b23dd4fe DM |
2942 | |
2943 | err = 0; | |
2944 | if (IS_ERR(rt)) | |
2945 | err = PTR_ERR(rt); | |
1da177e4 | 2946 | } |
d889ce3b | 2947 | |
1da177e4 | 2948 | if (err) |
d889ce3b | 2949 | goto errout_free; |
1da177e4 | 2950 | |
d8d1f30b | 2951 | skb_dst_set(skb, &rt->dst); |
1da177e4 LT |
2952 | if (rtm->rtm_flags & RTM_F_NOTIFY) |
2953 | rt->rt_flags |= RTCF_NOTIFY; | |
2954 | ||
4feb88e5 | 2955 | err = rt_fill_info(net, skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, |
1937504d | 2956 | RTM_NEWROUTE, 0, 0); |
d889ce3b TG |
2957 | if (err <= 0) |
2958 | goto errout_free; | |
1da177e4 | 2959 | |
1937504d | 2960 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid); |
d889ce3b | 2961 | errout: |
2942e900 | 2962 | return err; |
1da177e4 | 2963 | |
d889ce3b | 2964 | errout_free: |
1da177e4 | 2965 | kfree_skb(skb); |
d889ce3b | 2966 | goto errout; |
1da177e4 LT |
2967 | } |
2968 | ||
2969 | int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb) | |
2970 | { | |
2971 | struct rtable *rt; | |
2972 | int h, s_h; | |
2973 | int idx, s_idx; | |
1937504d DL |
2974 | struct net *net; |
2975 | ||
3b1e0a65 | 2976 | net = sock_net(skb->sk); |
1da177e4 LT |
2977 | |
2978 | s_h = cb->args[0]; | |
d8c92830 ED |
2979 | if (s_h < 0) |
2980 | s_h = 0; | |
1da177e4 | 2981 | s_idx = idx = cb->args[1]; |
a6272665 ED |
2982 | for (h = s_h; h <= rt_hash_mask; h++, s_idx = 0) { |
2983 | if (!rt_hash_table[h].chain) | |
2984 | continue; | |
1da177e4 | 2985 | rcu_read_lock_bh(); |
a898def2 | 2986 | for (rt = rcu_dereference_bh(rt_hash_table[h].chain), idx = 0; rt; |
d8d1f30b CG |
2987 | rt = rcu_dereference_bh(rt->dst.rt_next), idx++) { |
2988 | if (!net_eq(dev_net(rt->dst.dev), net) || idx < s_idx) | |
1da177e4 | 2989 | continue; |
e84f84f2 | 2990 | if (rt_is_expired(rt)) |
29e75252 | 2991 | continue; |
d8d1f30b | 2992 | skb_dst_set_noref(skb, &rt->dst); |
4feb88e5 | 2993 | if (rt_fill_info(net, skb, NETLINK_CB(cb->skb).pid, |
e905a9ed | 2994 | cb->nlh->nlmsg_seq, RTM_NEWROUTE, |
b6544c0b | 2995 | 1, NLM_F_MULTI) <= 0) { |
adf30907 | 2996 | skb_dst_drop(skb); |
1da177e4 LT |
2997 | rcu_read_unlock_bh(); |
2998 | goto done; | |
2999 | } | |
adf30907 | 3000 | skb_dst_drop(skb); |
1da177e4 LT |
3001 | } |
3002 | rcu_read_unlock_bh(); | |
3003 | } | |
3004 | ||
3005 | done: | |
3006 | cb->args[0] = h; | |
3007 | cb->args[1] = idx; | |
3008 | return skb->len; | |
3009 | } | |
3010 | ||
3011 | void ip_rt_multicast_event(struct in_device *in_dev) | |
3012 | { | |
76e6ebfb | 3013 | rt_cache_flush(dev_net(in_dev->dev), 0); |
1da177e4 LT |
3014 | } |
3015 | ||
3016 | #ifdef CONFIG_SYSCTL | |
81c684d1 | 3017 | static int ipv4_sysctl_rtcache_flush(ctl_table *__ctl, int write, |
8d65af78 | 3018 | void __user *buffer, |
1da177e4 LT |
3019 | size_t *lenp, loff_t *ppos) |
3020 | { | |
3021 | if (write) { | |
639e104f | 3022 | int flush_delay; |
81c684d1 | 3023 | ctl_table ctl; |
39a23e75 | 3024 | struct net *net; |
639e104f | 3025 | |
81c684d1 DL |
3026 | memcpy(&ctl, __ctl, sizeof(ctl)); |
3027 | ctl.data = &flush_delay; | |
8d65af78 | 3028 | proc_dointvec(&ctl, write, buffer, lenp, ppos); |
639e104f | 3029 | |
81c684d1 | 3030 | net = (struct net *)__ctl->extra1; |
39a23e75 | 3031 | rt_cache_flush(net, flush_delay); |
1da177e4 | 3032 | return 0; |
e905a9ed | 3033 | } |
1da177e4 LT |
3034 | |
3035 | return -EINVAL; | |
3036 | } | |
3037 | ||
eeb61f71 | 3038 | static ctl_table ipv4_route_table[] = { |
1da177e4 | 3039 | { |
1da177e4 LT |
3040 | .procname = "gc_thresh", |
3041 | .data = &ipv4_dst_ops.gc_thresh, | |
3042 | .maxlen = sizeof(int), | |
3043 | .mode = 0644, | |
6d9f239a | 3044 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3045 | }, |
3046 | { | |
1da177e4 LT |
3047 | .procname = "max_size", |
3048 | .data = &ip_rt_max_size, | |
3049 | .maxlen = sizeof(int), | |
3050 | .mode = 0644, | |
6d9f239a | 3051 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3052 | }, |
3053 | { | |
3054 | /* Deprecated. Use gc_min_interval_ms */ | |
e905a9ed | 3055 | |
1da177e4 LT |
3056 | .procname = "gc_min_interval", |
3057 | .data = &ip_rt_gc_min_interval, | |
3058 | .maxlen = sizeof(int), | |
3059 | .mode = 0644, | |
6d9f239a | 3060 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3061 | }, |
3062 | { | |
1da177e4 LT |
3063 | .procname = "gc_min_interval_ms", |
3064 | .data = &ip_rt_gc_min_interval, | |
3065 | .maxlen = sizeof(int), | |
3066 | .mode = 0644, | |
6d9f239a | 3067 | .proc_handler = proc_dointvec_ms_jiffies, |
1da177e4 LT |
3068 | }, |
3069 | { | |
1da177e4 LT |
3070 | .procname = "gc_timeout", |
3071 | .data = &ip_rt_gc_timeout, | |
3072 | .maxlen = sizeof(int), | |
3073 | .mode = 0644, | |
6d9f239a | 3074 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3075 | }, |
3076 | { | |
1da177e4 LT |
3077 | .procname = "gc_interval", |
3078 | .data = &ip_rt_gc_interval, | |
3079 | .maxlen = sizeof(int), | |
3080 | .mode = 0644, | |
6d9f239a | 3081 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3082 | }, |
3083 | { | |
1da177e4 LT |
3084 | .procname = "redirect_load", |
3085 | .data = &ip_rt_redirect_load, | |
3086 | .maxlen = sizeof(int), | |
3087 | .mode = 0644, | |
6d9f239a | 3088 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3089 | }, |
3090 | { | |
1da177e4 LT |
3091 | .procname = "redirect_number", |
3092 | .data = &ip_rt_redirect_number, | |
3093 | .maxlen = sizeof(int), | |
3094 | .mode = 0644, | |
6d9f239a | 3095 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3096 | }, |
3097 | { | |
1da177e4 LT |
3098 | .procname = "redirect_silence", |
3099 | .data = &ip_rt_redirect_silence, | |
3100 | .maxlen = sizeof(int), | |
3101 | .mode = 0644, | |
6d9f239a | 3102 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3103 | }, |
3104 | { | |
1da177e4 LT |
3105 | .procname = "error_cost", |
3106 | .data = &ip_rt_error_cost, | |
3107 | .maxlen = sizeof(int), | |
3108 | .mode = 0644, | |
6d9f239a | 3109 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3110 | }, |
3111 | { | |
1da177e4 LT |
3112 | .procname = "error_burst", |
3113 | .data = &ip_rt_error_burst, | |
3114 | .maxlen = sizeof(int), | |
3115 | .mode = 0644, | |
6d9f239a | 3116 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3117 | }, |
3118 | { | |
1da177e4 LT |
3119 | .procname = "gc_elasticity", |
3120 | .data = &ip_rt_gc_elasticity, | |
3121 | .maxlen = sizeof(int), | |
3122 | .mode = 0644, | |
6d9f239a | 3123 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3124 | }, |
3125 | { | |
1da177e4 LT |
3126 | .procname = "mtu_expires", |
3127 | .data = &ip_rt_mtu_expires, | |
3128 | .maxlen = sizeof(int), | |
3129 | .mode = 0644, | |
6d9f239a | 3130 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3131 | }, |
3132 | { | |
1da177e4 LT |
3133 | .procname = "min_pmtu", |
3134 | .data = &ip_rt_min_pmtu, | |
3135 | .maxlen = sizeof(int), | |
3136 | .mode = 0644, | |
6d9f239a | 3137 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3138 | }, |
3139 | { | |
1da177e4 LT |
3140 | .procname = "min_adv_mss", |
3141 | .data = &ip_rt_min_advmss, | |
3142 | .maxlen = sizeof(int), | |
3143 | .mode = 0644, | |
6d9f239a | 3144 | .proc_handler = proc_dointvec, |
1da177e4 | 3145 | }, |
f8572d8f | 3146 | { } |
1da177e4 | 3147 | }; |
39a23e75 | 3148 | |
2f4520d3 AV |
3149 | static struct ctl_table empty[1]; |
3150 | ||
3151 | static struct ctl_table ipv4_skeleton[] = | |
3152 | { | |
f8572d8f | 3153 | { .procname = "route", |
d994af0d | 3154 | .mode = 0555, .child = ipv4_route_table}, |
f8572d8f | 3155 | { .procname = "neigh", |
d994af0d | 3156 | .mode = 0555, .child = empty}, |
2f4520d3 AV |
3157 | { } |
3158 | }; | |
3159 | ||
3160 | static __net_initdata struct ctl_path ipv4_path[] = { | |
f8572d8f EB |
3161 | { .procname = "net", }, |
3162 | { .procname = "ipv4", }, | |
39a23e75 DL |
3163 | { }, |
3164 | }; | |
3165 | ||
39a23e75 DL |
3166 | static struct ctl_table ipv4_route_flush_table[] = { |
3167 | { | |
39a23e75 DL |
3168 | .procname = "flush", |
3169 | .maxlen = sizeof(int), | |
3170 | .mode = 0200, | |
6d9f239a | 3171 | .proc_handler = ipv4_sysctl_rtcache_flush, |
39a23e75 | 3172 | }, |
f8572d8f | 3173 | { }, |
39a23e75 DL |
3174 | }; |
3175 | ||
2f4520d3 | 3176 | static __net_initdata struct ctl_path ipv4_route_path[] = { |
f8572d8f EB |
3177 | { .procname = "net", }, |
3178 | { .procname = "ipv4", }, | |
3179 | { .procname = "route", }, | |
2f4520d3 AV |
3180 | { }, |
3181 | }; | |
3182 | ||
39a23e75 DL |
3183 | static __net_init int sysctl_route_net_init(struct net *net) |
3184 | { | |
3185 | struct ctl_table *tbl; | |
3186 | ||
3187 | tbl = ipv4_route_flush_table; | |
09ad9bc7 | 3188 | if (!net_eq(net, &init_net)) { |
39a23e75 DL |
3189 | tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL); |
3190 | if (tbl == NULL) | |
3191 | goto err_dup; | |
3192 | } | |
3193 | tbl[0].extra1 = net; | |
3194 | ||
3195 | net->ipv4.route_hdr = | |
3196 | register_net_sysctl_table(net, ipv4_route_path, tbl); | |
3197 | if (net->ipv4.route_hdr == NULL) | |
3198 | goto err_reg; | |
3199 | return 0; | |
3200 | ||
3201 | err_reg: | |
3202 | if (tbl != ipv4_route_flush_table) | |
3203 | kfree(tbl); | |
3204 | err_dup: | |
3205 | return -ENOMEM; | |
3206 | } | |
3207 | ||
3208 | static __net_exit void sysctl_route_net_exit(struct net *net) | |
3209 | { | |
3210 | struct ctl_table *tbl; | |
3211 | ||
3212 | tbl = net->ipv4.route_hdr->ctl_table_arg; | |
3213 | unregister_net_sysctl_table(net->ipv4.route_hdr); | |
3214 | BUG_ON(tbl == ipv4_route_flush_table); | |
3215 | kfree(tbl); | |
3216 | } | |
3217 | ||
3218 | static __net_initdata struct pernet_operations sysctl_route_ops = { | |
3219 | .init = sysctl_route_net_init, | |
3220 | .exit = sysctl_route_net_exit, | |
3221 | }; | |
1da177e4 LT |
3222 | #endif |
3223 | ||
3ee94372 | 3224 | static __net_init int rt_genid_init(struct net *net) |
9f5e97e5 | 3225 | { |
3ee94372 NH |
3226 | get_random_bytes(&net->ipv4.rt_genid, |
3227 | sizeof(net->ipv4.rt_genid)); | |
436c3b66 DM |
3228 | get_random_bytes(&net->ipv4.dev_addr_genid, |
3229 | sizeof(net->ipv4.dev_addr_genid)); | |
9f5e97e5 DL |
3230 | return 0; |
3231 | } | |
3232 | ||
3ee94372 NH |
3233 | static __net_initdata struct pernet_operations rt_genid_ops = { |
3234 | .init = rt_genid_init, | |
9f5e97e5 DL |
3235 | }; |
3236 | ||
3237 | ||
c7066f70 | 3238 | #ifdef CONFIG_IP_ROUTE_CLASSID |
7d720c3e | 3239 | struct ip_rt_acct __percpu *ip_rt_acct __read_mostly; |
c7066f70 | 3240 | #endif /* CONFIG_IP_ROUTE_CLASSID */ |
1da177e4 LT |
3241 | |
3242 | static __initdata unsigned long rhash_entries; | |
3243 | static int __init set_rhash_entries(char *str) | |
3244 | { | |
3245 | if (!str) | |
3246 | return 0; | |
3247 | rhash_entries = simple_strtoul(str, &str, 0); | |
3248 | return 1; | |
3249 | } | |
3250 | __setup("rhash_entries=", set_rhash_entries); | |
3251 | ||
3252 | int __init ip_rt_init(void) | |
3253 | { | |
424c4b70 | 3254 | int rc = 0; |
1da177e4 | 3255 | |
c7066f70 | 3256 | #ifdef CONFIG_IP_ROUTE_CLASSID |
0dcec8c2 | 3257 | ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct)); |
1da177e4 LT |
3258 | if (!ip_rt_acct) |
3259 | panic("IP: failed to allocate ip_rt_acct\n"); | |
1da177e4 LT |
3260 | #endif |
3261 | ||
e5d679f3 AD |
3262 | ipv4_dst_ops.kmem_cachep = |
3263 | kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0, | |
20c2df83 | 3264 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 3265 | |
14e50e57 DM |
3266 | ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep; |
3267 | ||
fc66f95c ED |
3268 | if (dst_entries_init(&ipv4_dst_ops) < 0) |
3269 | panic("IP: failed to allocate ipv4_dst_ops counter\n"); | |
3270 | ||
3271 | if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0) | |
3272 | panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n"); | |
3273 | ||
424c4b70 ED |
3274 | rt_hash_table = (struct rt_hash_bucket *) |
3275 | alloc_large_system_hash("IP route cache", | |
3276 | sizeof(struct rt_hash_bucket), | |
3277 | rhash_entries, | |
4481374c | 3278 | (totalram_pages >= 128 * 1024) ? |
18955cfc | 3279 | 15 : 17, |
8d1502de | 3280 | 0, |
424c4b70 ED |
3281 | &rt_hash_log, |
3282 | &rt_hash_mask, | |
c9503e0f | 3283 | rhash_entries ? 0 : 512 * 1024); |
22c047cc ED |
3284 | memset(rt_hash_table, 0, (rt_hash_mask + 1) * sizeof(struct rt_hash_bucket)); |
3285 | rt_hash_lock_init(); | |
1da177e4 LT |
3286 | |
3287 | ipv4_dst_ops.gc_thresh = (rt_hash_mask + 1); | |
3288 | ip_rt_max_size = (rt_hash_mask + 1) * 16; | |
3289 | ||
1da177e4 LT |
3290 | devinet_init(); |
3291 | ip_fib_init(); | |
3292 | ||
73b38711 | 3293 | if (ip_rt_proc_init()) |
107f1634 | 3294 | printk(KERN_ERR "Unable to create route proc files\n"); |
1da177e4 LT |
3295 | #ifdef CONFIG_XFRM |
3296 | xfrm_init(); | |
a33bc5c1 | 3297 | xfrm4_init(ip_rt_max_size); |
1da177e4 | 3298 | #endif |
63f3444f TG |
3299 | rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL); |
3300 | ||
39a23e75 DL |
3301 | #ifdef CONFIG_SYSCTL |
3302 | register_pernet_subsys(&sysctl_route_ops); | |
3303 | #endif | |
3ee94372 | 3304 | register_pernet_subsys(&rt_genid_ops); |
1da177e4 LT |
3305 | return rc; |
3306 | } | |
3307 | ||
a1bc6eb4 | 3308 | #ifdef CONFIG_SYSCTL |
eeb61f71 AV |
3309 | /* |
3310 | * We really need to sanitize the damn ipv4 init order, then all | |
3311 | * this nonsense will go away. | |
3312 | */ | |
3313 | void __init ip_static_sysctl_init(void) | |
3314 | { | |
2f4520d3 | 3315 | register_sysctl_paths(ipv4_path, ipv4_skeleton); |
eeb61f71 | 3316 | } |
a1bc6eb4 | 3317 | #endif |