]>
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 | ||
112 | #define RT_FL_TOS(oldflp) \ | |
113 | ((u32)(oldflp->fl4_tos & (IPTOS_RT_MASK | RTO_ONLINK))) | |
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 | |
125bb8f5 ED |
134 | static struct delayed_work expires_work; |
135 | static unsigned long expires_ljiffies; | |
1da177e4 LT |
136 | |
137 | /* | |
138 | * Interface to generic destination cache. | |
139 | */ | |
140 | ||
141 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie); | |
142 | static void ipv4_dst_destroy(struct dst_entry *dst); | |
143 | static void ipv4_dst_ifdown(struct dst_entry *dst, | |
144 | struct net_device *dev, int how); | |
145 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst); | |
146 | static void ipv4_link_failure(struct sk_buff *skb); | |
147 | static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu); | |
569d3645 | 148 | static int rt_garbage_collect(struct dst_ops *ops); |
1da177e4 LT |
149 | |
150 | ||
151 | static struct dst_ops ipv4_dst_ops = { | |
152 | .family = AF_INET, | |
09640e63 | 153 | .protocol = cpu_to_be16(ETH_P_IP), |
1da177e4 LT |
154 | .gc = rt_garbage_collect, |
155 | .check = ipv4_dst_check, | |
156 | .destroy = ipv4_dst_destroy, | |
157 | .ifdown = ipv4_dst_ifdown, | |
158 | .negative_advice = ipv4_negative_advice, | |
159 | .link_failure = ipv4_link_failure, | |
160 | .update_pmtu = ip_rt_update_pmtu, | |
1ac06e03 | 161 | .local_out = __ip_local_out, |
1da177e4 LT |
162 | }; |
163 | ||
164 | #define ECN_OR_COST(class) TC_PRIO_##class | |
165 | ||
4839c52b | 166 | const __u8 ip_tos2prio[16] = { |
1da177e4 LT |
167 | TC_PRIO_BESTEFFORT, |
168 | ECN_OR_COST(FILLER), | |
169 | TC_PRIO_BESTEFFORT, | |
170 | ECN_OR_COST(BESTEFFORT), | |
171 | TC_PRIO_BULK, | |
172 | ECN_OR_COST(BULK), | |
173 | TC_PRIO_BULK, | |
174 | ECN_OR_COST(BULK), | |
175 | TC_PRIO_INTERACTIVE, | |
176 | ECN_OR_COST(INTERACTIVE), | |
177 | TC_PRIO_INTERACTIVE, | |
178 | ECN_OR_COST(INTERACTIVE), | |
179 | TC_PRIO_INTERACTIVE_BULK, | |
180 | ECN_OR_COST(INTERACTIVE_BULK), | |
181 | TC_PRIO_INTERACTIVE_BULK, | |
182 | ECN_OR_COST(INTERACTIVE_BULK) | |
183 | }; | |
184 | ||
185 | ||
186 | /* | |
187 | * Route cache. | |
188 | */ | |
189 | ||
190 | /* The locking scheme is rather straight forward: | |
191 | * | |
192 | * 1) Read-Copy Update protects the buckets of the central route hash. | |
193 | * 2) Only writers remove entries, and they hold the lock | |
194 | * as they look at rtable reference counts. | |
195 | * 3) Only readers acquire references to rtable entries, | |
196 | * they do so with atomic increments and with the | |
197 | * lock held. | |
198 | */ | |
199 | ||
200 | struct rt_hash_bucket { | |
201 | struct rtable *chain; | |
22c047cc | 202 | }; |
1080d709 | 203 | |
8a25d5de IM |
204 | #if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK) || \ |
205 | defined(CONFIG_PROVE_LOCKING) | |
22c047cc ED |
206 | /* |
207 | * Instead of using one spinlock for each rt_hash_bucket, we use a table of spinlocks | |
208 | * The size of this table is a power of two and depends on the number of CPUS. | |
62051200 | 209 | * (on lockdep we have a quite big spinlock_t, so keep the size down there) |
22c047cc | 210 | */ |
62051200 IM |
211 | #ifdef CONFIG_LOCKDEP |
212 | # define RT_HASH_LOCK_SZ 256 | |
22c047cc | 213 | #else |
62051200 IM |
214 | # if NR_CPUS >= 32 |
215 | # define RT_HASH_LOCK_SZ 4096 | |
216 | # elif NR_CPUS >= 16 | |
217 | # define RT_HASH_LOCK_SZ 2048 | |
218 | # elif NR_CPUS >= 8 | |
219 | # define RT_HASH_LOCK_SZ 1024 | |
220 | # elif NR_CPUS >= 4 | |
221 | # define RT_HASH_LOCK_SZ 512 | |
222 | # else | |
223 | # define RT_HASH_LOCK_SZ 256 | |
224 | # endif | |
22c047cc ED |
225 | #endif |
226 | ||
227 | static spinlock_t *rt_hash_locks; | |
228 | # define rt_hash_lock_addr(slot) &rt_hash_locks[(slot) & (RT_HASH_LOCK_SZ - 1)] | |
1ff1cc20 PE |
229 | |
230 | static __init void rt_hash_lock_init(void) | |
231 | { | |
232 | int i; | |
233 | ||
234 | rt_hash_locks = kmalloc(sizeof(spinlock_t) * RT_HASH_LOCK_SZ, | |
235 | GFP_KERNEL); | |
236 | if (!rt_hash_locks) | |
237 | panic("IP: failed to allocate rt_hash_locks\n"); | |
238 | ||
239 | for (i = 0; i < RT_HASH_LOCK_SZ; i++) | |
240 | spin_lock_init(&rt_hash_locks[i]); | |
241 | } | |
22c047cc ED |
242 | #else |
243 | # define rt_hash_lock_addr(slot) NULL | |
1ff1cc20 PE |
244 | |
245 | static inline void rt_hash_lock_init(void) | |
246 | { | |
247 | } | |
22c047cc | 248 | #endif |
1da177e4 | 249 | |
817bc4db SH |
250 | static struct rt_hash_bucket *rt_hash_table __read_mostly; |
251 | static unsigned rt_hash_mask __read_mostly; | |
252 | static unsigned int rt_hash_log __read_mostly; | |
1da177e4 | 253 | |
2f970d83 | 254 | static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat); |
27f39c73 | 255 | #define RT_CACHE_STAT_INC(field) __this_cpu_inc(rt_cache_stat.field) |
1da177e4 | 256 | |
b00180de | 257 | static inline unsigned int rt_hash(__be32 daddr, __be32 saddr, int idx, |
0eae88f3 | 258 | int genid) |
1da177e4 | 259 | { |
0eae88f3 | 260 | return jhash_3words((__force u32)daddr, (__force u32)saddr, |
b00180de | 261 | idx, genid) |
29e75252 | 262 | & rt_hash_mask; |
1da177e4 LT |
263 | } |
264 | ||
e84f84f2 DL |
265 | static inline int rt_genid(struct net *net) |
266 | { | |
267 | return atomic_read(&net->ipv4.rt_genid); | |
268 | } | |
269 | ||
1da177e4 LT |
270 | #ifdef CONFIG_PROC_FS |
271 | struct rt_cache_iter_state { | |
a75e936f | 272 | struct seq_net_private p; |
1da177e4 | 273 | int bucket; |
29e75252 | 274 | int genid; |
1da177e4 LT |
275 | }; |
276 | ||
1218854a | 277 | static struct rtable *rt_cache_get_first(struct seq_file *seq) |
1da177e4 | 278 | { |
1218854a | 279 | struct rt_cache_iter_state *st = seq->private; |
1da177e4 | 280 | struct rtable *r = NULL; |
1da177e4 LT |
281 | |
282 | for (st->bucket = rt_hash_mask; st->bucket >= 0; --st->bucket) { | |
a6272665 ED |
283 | if (!rt_hash_table[st->bucket].chain) |
284 | continue; | |
1da177e4 | 285 | rcu_read_lock_bh(); |
a898def2 | 286 | r = rcu_dereference_bh(rt_hash_table[st->bucket].chain); |
29e75252 | 287 | while (r) { |
d8d1f30b | 288 | if (dev_net(r->dst.dev) == seq_file_net(seq) && |
a75e936f | 289 | r->rt_genid == st->genid) |
29e75252 | 290 | return r; |
d8d1f30b | 291 | r = rcu_dereference_bh(r->dst.rt_next); |
29e75252 | 292 | } |
1da177e4 LT |
293 | rcu_read_unlock_bh(); |
294 | } | |
29e75252 | 295 | return r; |
1da177e4 LT |
296 | } |
297 | ||
1218854a | 298 | static struct rtable *__rt_cache_get_next(struct seq_file *seq, |
642d6318 | 299 | struct rtable *r) |
1da177e4 | 300 | { |
1218854a | 301 | struct rt_cache_iter_state *st = seq->private; |
a6272665 | 302 | |
d8d1f30b | 303 | r = r->dst.rt_next; |
1da177e4 LT |
304 | while (!r) { |
305 | rcu_read_unlock_bh(); | |
a6272665 ED |
306 | do { |
307 | if (--st->bucket < 0) | |
308 | return NULL; | |
309 | } while (!rt_hash_table[st->bucket].chain); | |
1da177e4 LT |
310 | rcu_read_lock_bh(); |
311 | r = rt_hash_table[st->bucket].chain; | |
312 | } | |
a898def2 | 313 | return rcu_dereference_bh(r); |
1da177e4 LT |
314 | } |
315 | ||
1218854a | 316 | static struct rtable *rt_cache_get_next(struct seq_file *seq, |
642d6318 DL |
317 | struct rtable *r) |
318 | { | |
1218854a YH |
319 | struct rt_cache_iter_state *st = seq->private; |
320 | while ((r = __rt_cache_get_next(seq, r)) != NULL) { | |
d8d1f30b | 321 | if (dev_net(r->dst.dev) != seq_file_net(seq)) |
a75e936f | 322 | continue; |
642d6318 DL |
323 | if (r->rt_genid == st->genid) |
324 | break; | |
325 | } | |
326 | return r; | |
327 | } | |
328 | ||
1218854a | 329 | static struct rtable *rt_cache_get_idx(struct seq_file *seq, loff_t pos) |
1da177e4 | 330 | { |
1218854a | 331 | struct rtable *r = rt_cache_get_first(seq); |
1da177e4 LT |
332 | |
333 | if (r) | |
1218854a | 334 | while (pos && (r = rt_cache_get_next(seq, r))) |
1da177e4 LT |
335 | --pos; |
336 | return pos ? NULL : r; | |
337 | } | |
338 | ||
339 | static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos) | |
340 | { | |
29e75252 | 341 | struct rt_cache_iter_state *st = seq->private; |
29e75252 | 342 | if (*pos) |
1218854a | 343 | return rt_cache_get_idx(seq, *pos - 1); |
e84f84f2 | 344 | st->genid = rt_genid(seq_file_net(seq)); |
29e75252 | 345 | return SEQ_START_TOKEN; |
1da177e4 LT |
346 | } |
347 | ||
348 | static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
349 | { | |
29e75252 | 350 | struct rtable *r; |
1da177e4 LT |
351 | |
352 | if (v == SEQ_START_TOKEN) | |
1218854a | 353 | r = rt_cache_get_first(seq); |
1da177e4 | 354 | else |
1218854a | 355 | r = rt_cache_get_next(seq, v); |
1da177e4 LT |
356 | ++*pos; |
357 | return r; | |
358 | } | |
359 | ||
360 | static void rt_cache_seq_stop(struct seq_file *seq, void *v) | |
361 | { | |
362 | if (v && v != SEQ_START_TOKEN) | |
363 | rcu_read_unlock_bh(); | |
364 | } | |
365 | ||
366 | static int rt_cache_seq_show(struct seq_file *seq, void *v) | |
367 | { | |
368 | if (v == SEQ_START_TOKEN) | |
369 | seq_printf(seq, "%-127s\n", | |
370 | "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t" | |
371 | "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t" | |
372 | "HHUptod\tSpecDst"); | |
373 | else { | |
374 | struct rtable *r = v; | |
5e659e4c | 375 | int len; |
1da177e4 | 376 | |
0eae88f3 ED |
377 | seq_printf(seq, "%s\t%08X\t%08X\t%8X\t%d\t%u\t%d\t" |
378 | "%08X\t%d\t%u\t%u\t%02X\t%d\t%1d\t%08X%n", | |
d8d1f30b | 379 | r->dst.dev ? r->dst.dev->name : "*", |
0eae88f3 ED |
380 | (__force u32)r->rt_dst, |
381 | (__force u32)r->rt_gateway, | |
d8d1f30b CG |
382 | r->rt_flags, atomic_read(&r->dst.__refcnt), |
383 | r->dst.__use, 0, (__force u32)r->rt_src, | |
384 | (dst_metric(&r->dst, RTAX_ADVMSS) ? | |
385 | (int)dst_metric(&r->dst, RTAX_ADVMSS) + 40 : 0), | |
386 | dst_metric(&r->dst, RTAX_WINDOW), | |
387 | (int)((dst_metric(&r->dst, RTAX_RTT) >> 3) + | |
388 | dst_metric(&r->dst, RTAX_RTTVAR)), | |
1da177e4 | 389 | r->fl.fl4_tos, |
d8d1f30b CG |
390 | r->dst.hh ? atomic_read(&r->dst.hh->hh_refcnt) : -1, |
391 | r->dst.hh ? (r->dst.hh->hh_output == | |
1da177e4 | 392 | dev_queue_xmit) : 0, |
5e659e4c PE |
393 | r->rt_spec_dst, &len); |
394 | ||
395 | seq_printf(seq, "%*s\n", 127 - len, ""); | |
e905a9ed YH |
396 | } |
397 | return 0; | |
1da177e4 LT |
398 | } |
399 | ||
f690808e | 400 | static const struct seq_operations rt_cache_seq_ops = { |
1da177e4 LT |
401 | .start = rt_cache_seq_start, |
402 | .next = rt_cache_seq_next, | |
403 | .stop = rt_cache_seq_stop, | |
404 | .show = rt_cache_seq_show, | |
405 | }; | |
406 | ||
407 | static int rt_cache_seq_open(struct inode *inode, struct file *file) | |
408 | { | |
a75e936f | 409 | return seq_open_net(inode, file, &rt_cache_seq_ops, |
cf7732e4 | 410 | sizeof(struct rt_cache_iter_state)); |
1da177e4 LT |
411 | } |
412 | ||
9a32144e | 413 | static const struct file_operations rt_cache_seq_fops = { |
1da177e4 LT |
414 | .owner = THIS_MODULE, |
415 | .open = rt_cache_seq_open, | |
416 | .read = seq_read, | |
417 | .llseek = seq_lseek, | |
a75e936f | 418 | .release = seq_release_net, |
1da177e4 LT |
419 | }; |
420 | ||
421 | ||
422 | static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos) | |
423 | { | |
424 | int cpu; | |
425 | ||
426 | if (*pos == 0) | |
427 | return SEQ_START_TOKEN; | |
428 | ||
0f23174a | 429 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
430 | if (!cpu_possible(cpu)) |
431 | continue; | |
432 | *pos = cpu+1; | |
2f970d83 | 433 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
434 | } |
435 | return NULL; | |
436 | } | |
437 | ||
438 | static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
439 | { | |
440 | int cpu; | |
441 | ||
0f23174a | 442 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
443 | if (!cpu_possible(cpu)) |
444 | continue; | |
445 | *pos = cpu+1; | |
2f970d83 | 446 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
447 | } |
448 | return NULL; | |
e905a9ed | 449 | |
1da177e4 LT |
450 | } |
451 | ||
452 | static void rt_cpu_seq_stop(struct seq_file *seq, void *v) | |
453 | { | |
454 | ||
455 | } | |
456 | ||
457 | static int rt_cpu_seq_show(struct seq_file *seq, void *v) | |
458 | { | |
459 | struct rt_cache_stat *st = v; | |
460 | ||
461 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 462 | 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 |
463 | return 0; |
464 | } | |
e905a9ed | 465 | |
1da177e4 LT |
466 | seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x " |
467 | " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n", | |
fc66f95c | 468 | dst_entries_get_slow(&ipv4_dst_ops), |
1da177e4 LT |
469 | st->in_hit, |
470 | st->in_slow_tot, | |
471 | st->in_slow_mc, | |
472 | st->in_no_route, | |
473 | st->in_brd, | |
474 | st->in_martian_dst, | |
475 | st->in_martian_src, | |
476 | ||
477 | st->out_hit, | |
478 | st->out_slow_tot, | |
e905a9ed | 479 | st->out_slow_mc, |
1da177e4 LT |
480 | |
481 | st->gc_total, | |
482 | st->gc_ignored, | |
483 | st->gc_goal_miss, | |
484 | st->gc_dst_overflow, | |
485 | st->in_hlist_search, | |
486 | st->out_hlist_search | |
487 | ); | |
488 | return 0; | |
489 | } | |
490 | ||
f690808e | 491 | static const struct seq_operations rt_cpu_seq_ops = { |
1da177e4 LT |
492 | .start = rt_cpu_seq_start, |
493 | .next = rt_cpu_seq_next, | |
494 | .stop = rt_cpu_seq_stop, | |
495 | .show = rt_cpu_seq_show, | |
496 | }; | |
497 | ||
498 | ||
499 | static int rt_cpu_seq_open(struct inode *inode, struct file *file) | |
500 | { | |
501 | return seq_open(file, &rt_cpu_seq_ops); | |
502 | } | |
503 | ||
9a32144e | 504 | static const struct file_operations rt_cpu_seq_fops = { |
1da177e4 LT |
505 | .owner = THIS_MODULE, |
506 | .open = rt_cpu_seq_open, | |
507 | .read = seq_read, | |
508 | .llseek = seq_lseek, | |
509 | .release = seq_release, | |
510 | }; | |
511 | ||
78c686e9 | 512 | #ifdef CONFIG_NET_CLS_ROUTE |
a661c419 | 513 | static int rt_acct_proc_show(struct seq_file *m, void *v) |
78c686e9 | 514 | { |
a661c419 AD |
515 | struct ip_rt_acct *dst, *src; |
516 | unsigned int i, j; | |
517 | ||
518 | dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL); | |
519 | if (!dst) | |
520 | return -ENOMEM; | |
521 | ||
522 | for_each_possible_cpu(i) { | |
523 | src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i); | |
524 | for (j = 0; j < 256; j++) { | |
525 | dst[j].o_bytes += src[j].o_bytes; | |
526 | dst[j].o_packets += src[j].o_packets; | |
527 | dst[j].i_bytes += src[j].i_bytes; | |
528 | dst[j].i_packets += src[j].i_packets; | |
529 | } | |
78c686e9 PE |
530 | } |
531 | ||
a661c419 AD |
532 | seq_write(m, dst, 256 * sizeof(struct ip_rt_acct)); |
533 | kfree(dst); | |
534 | return 0; | |
535 | } | |
78c686e9 | 536 | |
a661c419 AD |
537 | static int rt_acct_proc_open(struct inode *inode, struct file *file) |
538 | { | |
539 | return single_open(file, rt_acct_proc_show, NULL); | |
78c686e9 | 540 | } |
a661c419 AD |
541 | |
542 | static const struct file_operations rt_acct_proc_fops = { | |
543 | .owner = THIS_MODULE, | |
544 | .open = rt_acct_proc_open, | |
545 | .read = seq_read, | |
546 | .llseek = seq_lseek, | |
547 | .release = single_release, | |
548 | }; | |
78c686e9 | 549 | #endif |
107f1634 | 550 | |
73b38711 | 551 | static int __net_init ip_rt_do_proc_init(struct net *net) |
107f1634 PE |
552 | { |
553 | struct proc_dir_entry *pde; | |
554 | ||
555 | pde = proc_net_fops_create(net, "rt_cache", S_IRUGO, | |
556 | &rt_cache_seq_fops); | |
557 | if (!pde) | |
558 | goto err1; | |
559 | ||
77020720 WC |
560 | pde = proc_create("rt_cache", S_IRUGO, |
561 | net->proc_net_stat, &rt_cpu_seq_fops); | |
107f1634 PE |
562 | if (!pde) |
563 | goto err2; | |
564 | ||
107f1634 | 565 | #ifdef CONFIG_NET_CLS_ROUTE |
a661c419 | 566 | pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops); |
107f1634 PE |
567 | if (!pde) |
568 | goto err3; | |
569 | #endif | |
570 | return 0; | |
571 | ||
572 | #ifdef CONFIG_NET_CLS_ROUTE | |
573 | err3: | |
574 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
575 | #endif | |
576 | err2: | |
577 | remove_proc_entry("rt_cache", net->proc_net); | |
578 | err1: | |
579 | return -ENOMEM; | |
580 | } | |
73b38711 DL |
581 | |
582 | static void __net_exit ip_rt_do_proc_exit(struct net *net) | |
583 | { | |
584 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
585 | remove_proc_entry("rt_cache", net->proc_net); | |
0a931acf | 586 | #ifdef CONFIG_NET_CLS_ROUTE |
73b38711 | 587 | remove_proc_entry("rt_acct", net->proc_net); |
0a931acf | 588 | #endif |
73b38711 DL |
589 | } |
590 | ||
591 | static struct pernet_operations ip_rt_proc_ops __net_initdata = { | |
592 | .init = ip_rt_do_proc_init, | |
593 | .exit = ip_rt_do_proc_exit, | |
594 | }; | |
595 | ||
596 | static int __init ip_rt_proc_init(void) | |
597 | { | |
598 | return register_pernet_subsys(&ip_rt_proc_ops); | |
599 | } | |
600 | ||
107f1634 | 601 | #else |
73b38711 | 602 | static inline int ip_rt_proc_init(void) |
107f1634 PE |
603 | { |
604 | return 0; | |
605 | } | |
1da177e4 | 606 | #endif /* CONFIG_PROC_FS */ |
e905a9ed | 607 | |
5969f71d | 608 | static inline void rt_free(struct rtable *rt) |
1da177e4 | 609 | { |
d8d1f30b | 610 | call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free); |
1da177e4 LT |
611 | } |
612 | ||
5969f71d | 613 | static inline void rt_drop(struct rtable *rt) |
1da177e4 | 614 | { |
1da177e4 | 615 | ip_rt_put(rt); |
d8d1f30b | 616 | call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free); |
1da177e4 LT |
617 | } |
618 | ||
5969f71d | 619 | static inline int rt_fast_clean(struct rtable *rth) |
1da177e4 LT |
620 | { |
621 | /* Kill broadcast/multicast entries very aggresively, if they | |
622 | collide in hash table with more useful entries */ | |
623 | return (rth->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && | |
d8d1f30b | 624 | rth->fl.iif && rth->dst.rt_next; |
1da177e4 LT |
625 | } |
626 | ||
5969f71d | 627 | static inline int rt_valuable(struct rtable *rth) |
1da177e4 LT |
628 | { |
629 | return (rth->rt_flags & (RTCF_REDIRECTED | RTCF_NOTIFY)) || | |
d8d1f30b | 630 | rth->dst.expires; |
1da177e4 LT |
631 | } |
632 | ||
633 | static int rt_may_expire(struct rtable *rth, unsigned long tmo1, unsigned long tmo2) | |
634 | { | |
635 | unsigned long age; | |
636 | int ret = 0; | |
637 | ||
d8d1f30b | 638 | if (atomic_read(&rth->dst.__refcnt)) |
1da177e4 LT |
639 | goto out; |
640 | ||
641 | ret = 1; | |
d8d1f30b CG |
642 | if (rth->dst.expires && |
643 | time_after_eq(jiffies, rth->dst.expires)) | |
1da177e4 LT |
644 | goto out; |
645 | ||
d8d1f30b | 646 | age = jiffies - rth->dst.lastuse; |
1da177e4 LT |
647 | ret = 0; |
648 | if ((age <= tmo1 && !rt_fast_clean(rth)) || | |
649 | (age <= tmo2 && rt_valuable(rth))) | |
650 | goto out; | |
651 | ret = 1; | |
652 | out: return ret; | |
653 | } | |
654 | ||
655 | /* Bits of score are: | |
656 | * 31: very valuable | |
657 | * 30: not quite useless | |
658 | * 29..0: usage counter | |
659 | */ | |
660 | static inline u32 rt_score(struct rtable *rt) | |
661 | { | |
d8d1f30b | 662 | u32 score = jiffies - rt->dst.lastuse; |
1da177e4 LT |
663 | |
664 | score = ~score & ~(3<<30); | |
665 | ||
666 | if (rt_valuable(rt)) | |
667 | score |= (1<<31); | |
668 | ||
669 | if (!rt->fl.iif || | |
670 | !(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL))) | |
671 | score |= (1<<30); | |
672 | ||
673 | return score; | |
674 | } | |
675 | ||
1080d709 NH |
676 | static inline bool rt_caching(const struct net *net) |
677 | { | |
678 | return net->ipv4.current_rt_cache_rebuild_count <= | |
679 | net->ipv4.sysctl_rt_cache_rebuild_count; | |
680 | } | |
681 | ||
682 | static inline bool compare_hash_inputs(const struct flowi *fl1, | |
683 | const struct flowi *fl2) | |
684 | { | |
0eae88f3 ED |
685 | return ((((__force u32)fl1->nl_u.ip4_u.daddr ^ (__force u32)fl2->nl_u.ip4_u.daddr) | |
686 | ((__force u32)fl1->nl_u.ip4_u.saddr ^ (__force u32)fl2->nl_u.ip4_u.saddr) | | |
1080d709 NH |
687 | (fl1->iif ^ fl2->iif)) == 0); |
688 | } | |
689 | ||
1da177e4 LT |
690 | static inline int compare_keys(struct flowi *fl1, struct flowi *fl2) |
691 | { | |
0eae88f3 ED |
692 | return (((__force u32)fl1->nl_u.ip4_u.daddr ^ (__force u32)fl2->nl_u.ip4_u.daddr) | |
693 | ((__force u32)fl1->nl_u.ip4_u.saddr ^ (__force u32)fl2->nl_u.ip4_u.saddr) | | |
47dcf0cb | 694 | (fl1->mark ^ fl2->mark) | |
0eae88f3 | 695 | (*(u16 *)&fl1->nl_u.ip4_u.tos ^ *(u16 *)&fl2->nl_u.ip4_u.tos) | |
8238b218 DM |
696 | (fl1->oif ^ fl2->oif) | |
697 | (fl1->iif ^ fl2->iif)) == 0; | |
1da177e4 LT |
698 | } |
699 | ||
b5921910 DL |
700 | static inline int compare_netns(struct rtable *rt1, struct rtable *rt2) |
701 | { | |
d8d1f30b | 702 | return net_eq(dev_net(rt1->dst.dev), dev_net(rt2->dst.dev)); |
b5921910 DL |
703 | } |
704 | ||
e84f84f2 DL |
705 | static inline int rt_is_expired(struct rtable *rth) |
706 | { | |
d8d1f30b | 707 | return rth->rt_genid != rt_genid(dev_net(rth->dst.dev)); |
e84f84f2 DL |
708 | } |
709 | ||
beb659bd ED |
710 | /* |
711 | * Perform a full scan of hash table and free all entries. | |
712 | * Can be called by a softirq or a process. | |
713 | * In the later case, we want to be reschedule if necessary | |
714 | */ | |
715 | static void rt_do_flush(int process_context) | |
716 | { | |
717 | unsigned int i; | |
718 | struct rtable *rth, *next; | |
32cb5b4e | 719 | struct rtable * tail; |
beb659bd ED |
720 | |
721 | for (i = 0; i <= rt_hash_mask; i++) { | |
722 | if (process_context && need_resched()) | |
723 | cond_resched(); | |
724 | rth = rt_hash_table[i].chain; | |
725 | if (!rth) | |
726 | continue; | |
727 | ||
728 | spin_lock_bh(rt_hash_lock_addr(i)); | |
32cb5b4e DL |
729 | #ifdef CONFIG_NET_NS |
730 | { | |
731 | struct rtable ** prev, * p; | |
732 | ||
733 | rth = rt_hash_table[i].chain; | |
734 | ||
735 | /* defer releasing the head of the list after spin_unlock */ | |
d8d1f30b | 736 | for (tail = rth; tail; tail = tail->dst.rt_next) |
32cb5b4e DL |
737 | if (!rt_is_expired(tail)) |
738 | break; | |
739 | if (rth != tail) | |
740 | rt_hash_table[i].chain = tail; | |
741 | ||
742 | /* call rt_free on entries after the tail requiring flush */ | |
743 | prev = &rt_hash_table[i].chain; | |
744 | for (p = *prev; p; p = next) { | |
d8d1f30b | 745 | next = p->dst.rt_next; |
32cb5b4e | 746 | if (!rt_is_expired(p)) { |
d8d1f30b | 747 | prev = &p->dst.rt_next; |
32cb5b4e DL |
748 | } else { |
749 | *prev = next; | |
750 | rt_free(p); | |
751 | } | |
752 | } | |
753 | } | |
754 | #else | |
beb659bd ED |
755 | rth = rt_hash_table[i].chain; |
756 | rt_hash_table[i].chain = NULL; | |
32cb5b4e DL |
757 | tail = NULL; |
758 | #endif | |
beb659bd ED |
759 | spin_unlock_bh(rt_hash_lock_addr(i)); |
760 | ||
32cb5b4e | 761 | for (; rth != tail; rth = next) { |
d8d1f30b | 762 | next = rth->dst.rt_next; |
beb659bd ED |
763 | rt_free(rth); |
764 | } | |
765 | } | |
766 | } | |
767 | ||
1080d709 NH |
768 | /* |
769 | * While freeing expired entries, we compute average chain length | |
770 | * and standard deviation, using fixed-point arithmetic. | |
771 | * This to have an estimation of rt_chain_length_max | |
772 | * rt_chain_length_max = max(elasticity, AVG + 4*SD) | |
773 | * We use 3 bits for frational part, and 29 (or 61) for magnitude. | |
774 | */ | |
775 | ||
776 | #define FRACT_BITS 3 | |
777 | #define ONE (1UL << FRACT_BITS) | |
778 | ||
98376387 ED |
779 | /* |
780 | * Given a hash chain and an item in this hash chain, | |
781 | * find if a previous entry has the same hash_inputs | |
782 | * (but differs on tos, mark or oif) | |
783 | * Returns 0 if an alias is found. | |
784 | * Returns ONE if rth has no alias before itself. | |
785 | */ | |
786 | static int has_noalias(const struct rtable *head, const struct rtable *rth) | |
787 | { | |
788 | const struct rtable *aux = head; | |
789 | ||
790 | while (aux != rth) { | |
791 | if (compare_hash_inputs(&aux->fl, &rth->fl)) | |
792 | return 0; | |
d8d1f30b | 793 | aux = aux->dst.rt_next; |
98376387 ED |
794 | } |
795 | return ONE; | |
796 | } | |
797 | ||
beb659bd | 798 | static void rt_check_expire(void) |
1da177e4 | 799 | { |
bb1d23b0 ED |
800 | static unsigned int rover; |
801 | unsigned int i = rover, goal; | |
98376387 | 802 | struct rtable *rth, **rthp; |
cf8da764 | 803 | unsigned long samples = 0; |
1080d709 | 804 | unsigned long sum = 0, sum2 = 0; |
125bb8f5 | 805 | unsigned long delta; |
bb1d23b0 ED |
806 | u64 mult; |
807 | ||
125bb8f5 ED |
808 | delta = jiffies - expires_ljiffies; |
809 | expires_ljiffies = jiffies; | |
810 | mult = ((u64)delta) << rt_hash_log; | |
bb1d23b0 ED |
811 | if (ip_rt_gc_timeout > 1) |
812 | do_div(mult, ip_rt_gc_timeout); | |
813 | goal = (unsigned int)mult; | |
39c90ece ED |
814 | if (goal > rt_hash_mask) |
815 | goal = rt_hash_mask + 1; | |
bb1d23b0 | 816 | for (; goal > 0; goal--) { |
1da177e4 | 817 | unsigned long tmo = ip_rt_gc_timeout; |
cf8da764 | 818 | unsigned long length; |
1da177e4 LT |
819 | |
820 | i = (i + 1) & rt_hash_mask; | |
821 | rthp = &rt_hash_table[i].chain; | |
822 | ||
d90bf5a9 ED |
823 | if (need_resched()) |
824 | cond_resched(); | |
825 | ||
1080d709 NH |
826 | samples++; |
827 | ||
cfcabdcc | 828 | if (*rthp == NULL) |
bb1d23b0 | 829 | continue; |
cf8da764 | 830 | length = 0; |
39c90ece | 831 | spin_lock_bh(rt_hash_lock_addr(i)); |
1da177e4 | 832 | while ((rth = *rthp) != NULL) { |
d8d1f30b | 833 | prefetch(rth->dst.rt_next); |
e84f84f2 | 834 | if (rt_is_expired(rth)) { |
d8d1f30b | 835 | *rthp = rth->dst.rt_next; |
29e75252 ED |
836 | rt_free(rth); |
837 | continue; | |
838 | } | |
d8d1f30b | 839 | if (rth->dst.expires) { |
1da177e4 | 840 | /* Entry is expired even if it is in use */ |
d8d1f30b | 841 | if (time_before_eq(jiffies, rth->dst.expires)) { |
1ddbcb00 | 842 | nofree: |
1da177e4 | 843 | tmo >>= 1; |
d8d1f30b | 844 | rthp = &rth->dst.rt_next; |
1080d709 | 845 | /* |
1ddbcb00 | 846 | * We only count entries on |
1080d709 NH |
847 | * a chain with equal hash inputs once |
848 | * so that entries for different QOS | |
849 | * levels, and other non-hash input | |
850 | * attributes don't unfairly skew | |
851 | * the length computation | |
852 | */ | |
98376387 | 853 | length += has_noalias(rt_hash_table[i].chain, rth); |
1da177e4 LT |
854 | continue; |
855 | } | |
1ddbcb00 ED |
856 | } else if (!rt_may_expire(rth, tmo, ip_rt_gc_timeout)) |
857 | goto nofree; | |
1da177e4 LT |
858 | |
859 | /* Cleanup aged off entries. */ | |
d8d1f30b | 860 | *rthp = rth->dst.rt_next; |
e905a9ed | 861 | rt_free(rth); |
1da177e4 | 862 | } |
39c90ece | 863 | spin_unlock_bh(rt_hash_lock_addr(i)); |
1080d709 NH |
864 | sum += length; |
865 | sum2 += length*length; | |
866 | } | |
867 | if (samples) { | |
868 | unsigned long avg = sum / samples; | |
869 | unsigned long sd = int_sqrt(sum2 / samples - avg*avg); | |
870 | rt_chain_length_max = max_t(unsigned long, | |
871 | ip_rt_gc_elasticity, | |
872 | (avg + 4*sd) >> FRACT_BITS); | |
1da177e4 LT |
873 | } |
874 | rover = i; | |
beb659bd ED |
875 | } |
876 | ||
877 | /* | |
878 | * rt_worker_func() is run in process context. | |
29e75252 | 879 | * we call rt_check_expire() to scan part of the hash table |
beb659bd ED |
880 | */ |
881 | static void rt_worker_func(struct work_struct *work) | |
882 | { | |
29e75252 | 883 | rt_check_expire(); |
39c90ece | 884 | schedule_delayed_work(&expires_work, ip_rt_gc_interval); |
1da177e4 LT |
885 | } |
886 | ||
29e75252 ED |
887 | /* |
888 | * Pertubation of rt_genid by a small quantity [1..256] | |
889 | * Using 8 bits of shuffling ensure we can call rt_cache_invalidate() | |
890 | * many times (2^24) without giving recent rt_genid. | |
891 | * Jenkins hash is strong enough that litle changes of rt_genid are OK. | |
1da177e4 | 892 | */ |
86c657f6 | 893 | static void rt_cache_invalidate(struct net *net) |
1da177e4 | 894 | { |
29e75252 | 895 | unsigned char shuffle; |
1da177e4 | 896 | |
29e75252 | 897 | get_random_bytes(&shuffle, sizeof(shuffle)); |
e84f84f2 | 898 | atomic_add(shuffle + 1U, &net->ipv4.rt_genid); |
1da177e4 LT |
899 | } |
900 | ||
29e75252 ED |
901 | /* |
902 | * delay < 0 : invalidate cache (fast : entries will be deleted later) | |
903 | * delay >= 0 : invalidate & flush cache (can be long) | |
904 | */ | |
76e6ebfb | 905 | void rt_cache_flush(struct net *net, int delay) |
1da177e4 | 906 | { |
86c657f6 | 907 | rt_cache_invalidate(net); |
29e75252 ED |
908 | if (delay >= 0) |
909 | rt_do_flush(!in_softirq()); | |
1da177e4 LT |
910 | } |
911 | ||
a5ee1551 EB |
912 | /* Flush previous cache invalidated entries from the cache */ |
913 | void rt_cache_flush_batch(void) | |
914 | { | |
915 | rt_do_flush(!in_softirq()); | |
916 | } | |
917 | ||
1080d709 NH |
918 | static void rt_emergency_hash_rebuild(struct net *net) |
919 | { | |
3ee94372 | 920 | if (net_ratelimit()) |
1080d709 | 921 | printk(KERN_WARNING "Route hash chain too long!\n"); |
3ee94372 | 922 | rt_cache_invalidate(net); |
1080d709 NH |
923 | } |
924 | ||
1da177e4 LT |
925 | /* |
926 | Short description of GC goals. | |
927 | ||
928 | We want to build algorithm, which will keep routing cache | |
929 | at some equilibrium point, when number of aged off entries | |
930 | is kept approximately equal to newly generated ones. | |
931 | ||
932 | Current expiration strength is variable "expire". | |
933 | We try to adjust it dynamically, so that if networking | |
934 | is idle expires is large enough to keep enough of warm entries, | |
935 | and when load increases it reduces to limit cache size. | |
936 | */ | |
937 | ||
569d3645 | 938 | static int rt_garbage_collect(struct dst_ops *ops) |
1da177e4 LT |
939 | { |
940 | static unsigned long expire = RT_GC_TIMEOUT; | |
941 | static unsigned long last_gc; | |
942 | static int rover; | |
943 | static int equilibrium; | |
944 | struct rtable *rth, **rthp; | |
945 | unsigned long now = jiffies; | |
946 | int goal; | |
fc66f95c | 947 | int entries = dst_entries_get_fast(&ipv4_dst_ops); |
1da177e4 LT |
948 | |
949 | /* | |
950 | * Garbage collection is pretty expensive, | |
951 | * do not make it too frequently. | |
952 | */ | |
953 | ||
954 | RT_CACHE_STAT_INC(gc_total); | |
955 | ||
956 | if (now - last_gc < ip_rt_gc_min_interval && | |
fc66f95c | 957 | entries < ip_rt_max_size) { |
1da177e4 LT |
958 | RT_CACHE_STAT_INC(gc_ignored); |
959 | goto out; | |
960 | } | |
961 | ||
fc66f95c | 962 | entries = dst_entries_get_slow(&ipv4_dst_ops); |
1da177e4 | 963 | /* Calculate number of entries, which we want to expire now. */ |
fc66f95c | 964 | goal = entries - (ip_rt_gc_elasticity << rt_hash_log); |
1da177e4 LT |
965 | if (goal <= 0) { |
966 | if (equilibrium < ipv4_dst_ops.gc_thresh) | |
967 | equilibrium = ipv4_dst_ops.gc_thresh; | |
fc66f95c | 968 | goal = entries - equilibrium; |
1da177e4 | 969 | if (goal > 0) { |
b790cedd | 970 | equilibrium += min_t(unsigned int, goal >> 1, rt_hash_mask + 1); |
fc66f95c | 971 | goal = entries - equilibrium; |
1da177e4 LT |
972 | } |
973 | } else { | |
974 | /* We are in dangerous area. Try to reduce cache really | |
975 | * aggressively. | |
976 | */ | |
b790cedd | 977 | goal = max_t(unsigned int, goal >> 1, rt_hash_mask + 1); |
fc66f95c | 978 | equilibrium = entries - goal; |
1da177e4 LT |
979 | } |
980 | ||
981 | if (now - last_gc >= ip_rt_gc_min_interval) | |
982 | last_gc = now; | |
983 | ||
984 | if (goal <= 0) { | |
985 | equilibrium += goal; | |
986 | goto work_done; | |
987 | } | |
988 | ||
989 | do { | |
990 | int i, k; | |
991 | ||
992 | for (i = rt_hash_mask, k = rover; i >= 0; i--) { | |
993 | unsigned long tmo = expire; | |
994 | ||
995 | k = (k + 1) & rt_hash_mask; | |
996 | rthp = &rt_hash_table[k].chain; | |
22c047cc | 997 | spin_lock_bh(rt_hash_lock_addr(k)); |
1da177e4 | 998 | while ((rth = *rthp) != NULL) { |
e84f84f2 | 999 | if (!rt_is_expired(rth) && |
29e75252 | 1000 | !rt_may_expire(rth, tmo, expire)) { |
1da177e4 | 1001 | tmo >>= 1; |
d8d1f30b | 1002 | rthp = &rth->dst.rt_next; |
1da177e4 LT |
1003 | continue; |
1004 | } | |
d8d1f30b | 1005 | *rthp = rth->dst.rt_next; |
1da177e4 LT |
1006 | rt_free(rth); |
1007 | goal--; | |
1da177e4 | 1008 | } |
22c047cc | 1009 | spin_unlock_bh(rt_hash_lock_addr(k)); |
1da177e4 LT |
1010 | if (goal <= 0) |
1011 | break; | |
1012 | } | |
1013 | rover = k; | |
1014 | ||
1015 | if (goal <= 0) | |
1016 | goto work_done; | |
1017 | ||
1018 | /* Goal is not achieved. We stop process if: | |
1019 | ||
1020 | - if expire reduced to zero. Otherwise, expire is halfed. | |
1021 | - if table is not full. | |
1022 | - if we are called from interrupt. | |
1023 | - jiffies check is just fallback/debug loop breaker. | |
1024 | We will not spin here for long time in any case. | |
1025 | */ | |
1026 | ||
1027 | RT_CACHE_STAT_INC(gc_goal_miss); | |
1028 | ||
1029 | if (expire == 0) | |
1030 | break; | |
1031 | ||
1032 | expire >>= 1; | |
1033 | #if RT_CACHE_DEBUG >= 2 | |
1034 | printk(KERN_DEBUG "expire>> %u %d %d %d\n", expire, | |
fc66f95c | 1035 | dst_entries_get_fast(&ipv4_dst_ops), goal, i); |
1da177e4 LT |
1036 | #endif |
1037 | ||
fc66f95c | 1038 | if (dst_entries_get_fast(&ipv4_dst_ops) < ip_rt_max_size) |
1da177e4 LT |
1039 | goto out; |
1040 | } while (!in_softirq() && time_before_eq(jiffies, now)); | |
1041 | ||
fc66f95c ED |
1042 | if (dst_entries_get_fast(&ipv4_dst_ops) < ip_rt_max_size) |
1043 | goto out; | |
1044 | if (dst_entries_get_slow(&ipv4_dst_ops) < ip_rt_max_size) | |
1da177e4 LT |
1045 | goto out; |
1046 | if (net_ratelimit()) | |
1047 | printk(KERN_WARNING "dst cache overflow\n"); | |
1048 | RT_CACHE_STAT_INC(gc_dst_overflow); | |
1049 | return 1; | |
1050 | ||
1051 | work_done: | |
1052 | expire += ip_rt_gc_min_interval; | |
1053 | if (expire > ip_rt_gc_timeout || | |
fc66f95c ED |
1054 | dst_entries_get_fast(&ipv4_dst_ops) < ipv4_dst_ops.gc_thresh || |
1055 | dst_entries_get_slow(&ipv4_dst_ops) < ipv4_dst_ops.gc_thresh) | |
1da177e4 LT |
1056 | expire = ip_rt_gc_timeout; |
1057 | #if RT_CACHE_DEBUG >= 2 | |
1058 | printk(KERN_DEBUG "expire++ %u %d %d %d\n", expire, | |
fc66f95c | 1059 | dst_entries_get_fast(&ipv4_dst_ops), goal, rover); |
1da177e4 LT |
1060 | #endif |
1061 | out: return 0; | |
1062 | } | |
1063 | ||
98376387 ED |
1064 | /* |
1065 | * Returns number of entries in a hash chain that have different hash_inputs | |
1066 | */ | |
1067 | static int slow_chain_length(const struct rtable *head) | |
1068 | { | |
1069 | int length = 0; | |
1070 | const struct rtable *rth = head; | |
1071 | ||
1072 | while (rth) { | |
1073 | length += has_noalias(head, rth); | |
d8d1f30b | 1074 | rth = rth->dst.rt_next; |
98376387 ED |
1075 | } |
1076 | return length >> FRACT_BITS; | |
1077 | } | |
1078 | ||
511c3f92 | 1079 | static int rt_intern_hash(unsigned hash, struct rtable *rt, |
6a2bad70 | 1080 | struct rtable **rp, struct sk_buff *skb, int ifindex) |
1da177e4 LT |
1081 | { |
1082 | struct rtable *rth, **rthp; | |
1083 | unsigned long now; | |
1084 | struct rtable *cand, **candp; | |
1085 | u32 min_score; | |
1086 | int chain_length; | |
1087 | int attempts = !in_softirq(); | |
1088 | ||
1089 | restart: | |
1090 | chain_length = 0; | |
1091 | min_score = ~(u32)0; | |
1092 | cand = NULL; | |
1093 | candp = NULL; | |
1094 | now = jiffies; | |
1095 | ||
d8d1f30b | 1096 | if (!rt_caching(dev_net(rt->dst.dev))) { |
73e42897 NH |
1097 | /* |
1098 | * If we're not caching, just tell the caller we | |
1099 | * were successful and don't touch the route. The | |
1100 | * caller hold the sole reference to the cache entry, and | |
1101 | * it will be released when the caller is done with it. | |
1102 | * If we drop it here, the callers have no way to resolve routes | |
1103 | * when we're not caching. Instead, just point *rp at rt, so | |
1104 | * the caller gets a single use out of the route | |
b6280b47 NH |
1105 | * Note that we do rt_free on this new route entry, so that |
1106 | * once its refcount hits zero, we are still able to reap it | |
1107 | * (Thanks Alexey) | |
1108 | * Note also the rt_free uses call_rcu. We don't actually | |
1109 | * need rcu protection here, this is just our path to get | |
1110 | * on the route gc list. | |
73e42897 | 1111 | */ |
b6280b47 | 1112 | |
c7d4426a | 1113 | rt->dst.flags |= DST_NOCACHE; |
b6280b47 | 1114 | if (rt->rt_type == RTN_UNICAST || rt->fl.iif == 0) { |
d8d1f30b | 1115 | int err = arp_bind_neighbour(&rt->dst); |
b6280b47 NH |
1116 | if (err) { |
1117 | if (net_ratelimit()) | |
1118 | printk(KERN_WARNING | |
1119 | "Neighbour table failure & not caching routes.\n"); | |
1120 | rt_drop(rt); | |
1121 | return err; | |
1122 | } | |
1123 | } | |
1124 | ||
1125 | rt_free(rt); | |
1126 | goto skip_hashing; | |
1080d709 NH |
1127 | } |
1128 | ||
1da177e4 LT |
1129 | rthp = &rt_hash_table[hash].chain; |
1130 | ||
22c047cc | 1131 | spin_lock_bh(rt_hash_lock_addr(hash)); |
1da177e4 | 1132 | while ((rth = *rthp) != NULL) { |
e84f84f2 | 1133 | if (rt_is_expired(rth)) { |
d8d1f30b | 1134 | *rthp = rth->dst.rt_next; |
29e75252 ED |
1135 | rt_free(rth); |
1136 | continue; | |
1137 | } | |
b5921910 | 1138 | if (compare_keys(&rth->fl, &rt->fl) && compare_netns(rth, rt)) { |
1da177e4 | 1139 | /* Put it first */ |
d8d1f30b | 1140 | *rthp = rth->dst.rt_next; |
1da177e4 LT |
1141 | /* |
1142 | * Since lookup is lockfree, the deletion | |
1143 | * must be visible to another weakly ordered CPU before | |
1144 | * the insertion at the start of the hash chain. | |
1145 | */ | |
d8d1f30b | 1146 | rcu_assign_pointer(rth->dst.rt_next, |
1da177e4 LT |
1147 | rt_hash_table[hash].chain); |
1148 | /* | |
1149 | * Since lookup is lockfree, the update writes | |
1150 | * must be ordered for consistency on SMP. | |
1151 | */ | |
1152 | rcu_assign_pointer(rt_hash_table[hash].chain, rth); | |
1153 | ||
d8d1f30b | 1154 | dst_use(&rth->dst, now); |
22c047cc | 1155 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1156 | |
1157 | rt_drop(rt); | |
511c3f92 ED |
1158 | if (rp) |
1159 | *rp = rth; | |
1160 | else | |
d8d1f30b | 1161 | skb_dst_set(skb, &rth->dst); |
1da177e4 LT |
1162 | return 0; |
1163 | } | |
1164 | ||
d8d1f30b | 1165 | if (!atomic_read(&rth->dst.__refcnt)) { |
1da177e4 LT |
1166 | u32 score = rt_score(rth); |
1167 | ||
1168 | if (score <= min_score) { | |
1169 | cand = rth; | |
1170 | candp = rthp; | |
1171 | min_score = score; | |
1172 | } | |
1173 | } | |
1174 | ||
1175 | chain_length++; | |
1176 | ||
d8d1f30b | 1177 | rthp = &rth->dst.rt_next; |
1da177e4 LT |
1178 | } |
1179 | ||
1180 | if (cand) { | |
1181 | /* ip_rt_gc_elasticity used to be average length of chain | |
1182 | * length, when exceeded gc becomes really aggressive. | |
1183 | * | |
1184 | * The second limit is less certain. At the moment it allows | |
1185 | * only 2 entries per bucket. We will see. | |
1186 | */ | |
1187 | if (chain_length > ip_rt_gc_elasticity) { | |
d8d1f30b | 1188 | *candp = cand->dst.rt_next; |
1da177e4 LT |
1189 | rt_free(cand); |
1190 | } | |
1080d709 | 1191 | } else { |
98376387 ED |
1192 | if (chain_length > rt_chain_length_max && |
1193 | slow_chain_length(rt_hash_table[hash].chain) > rt_chain_length_max) { | |
d8d1f30b | 1194 | struct net *net = dev_net(rt->dst.dev); |
1080d709 | 1195 | int num = ++net->ipv4.current_rt_cache_rebuild_count; |
b35ecb5d | 1196 | if (!rt_caching(net)) { |
1080d709 | 1197 | printk(KERN_WARNING "%s: %d rebuilds is over limit, route caching disabled\n", |
d8d1f30b | 1198 | rt->dst.dev->name, num); |
1080d709 | 1199 | } |
b35ecb5d | 1200 | rt_emergency_hash_rebuild(net); |
6a2bad70 PE |
1201 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1202 | ||
1203 | hash = rt_hash(rt->fl.fl4_dst, rt->fl.fl4_src, | |
1204 | ifindex, rt_genid(net)); | |
1205 | goto restart; | |
1080d709 | 1206 | } |
1da177e4 LT |
1207 | } |
1208 | ||
1209 | /* Try to bind route to arp only if it is output | |
1210 | route or unicast forwarding path. | |
1211 | */ | |
1212 | if (rt->rt_type == RTN_UNICAST || rt->fl.iif == 0) { | |
d8d1f30b | 1213 | int err = arp_bind_neighbour(&rt->dst); |
1da177e4 | 1214 | if (err) { |
22c047cc | 1215 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1216 | |
1217 | if (err != -ENOBUFS) { | |
1218 | rt_drop(rt); | |
1219 | return err; | |
1220 | } | |
1221 | ||
1222 | /* Neighbour tables are full and nothing | |
1223 | can be released. Try to shrink route cache, | |
1224 | it is most likely it holds some neighbour records. | |
1225 | */ | |
1226 | if (attempts-- > 0) { | |
1227 | int saved_elasticity = ip_rt_gc_elasticity; | |
1228 | int saved_int = ip_rt_gc_min_interval; | |
1229 | ip_rt_gc_elasticity = 1; | |
1230 | ip_rt_gc_min_interval = 0; | |
569d3645 | 1231 | rt_garbage_collect(&ipv4_dst_ops); |
1da177e4 LT |
1232 | ip_rt_gc_min_interval = saved_int; |
1233 | ip_rt_gc_elasticity = saved_elasticity; | |
1234 | goto restart; | |
1235 | } | |
1236 | ||
1237 | if (net_ratelimit()) | |
7e1b33e5 | 1238 | printk(KERN_WARNING "ipv4: Neighbour table overflow.\n"); |
1da177e4 LT |
1239 | rt_drop(rt); |
1240 | return -ENOBUFS; | |
1241 | } | |
1242 | } | |
1243 | ||
d8d1f30b | 1244 | rt->dst.rt_next = rt_hash_table[hash].chain; |
1080d709 | 1245 | |
1da177e4 | 1246 | #if RT_CACHE_DEBUG >= 2 |
d8d1f30b | 1247 | if (rt->dst.rt_next) { |
1da177e4 | 1248 | struct rtable *trt; |
b6280b47 NH |
1249 | printk(KERN_DEBUG "rt_cache @%02x: %pI4", |
1250 | hash, &rt->rt_dst); | |
d8d1f30b | 1251 | for (trt = rt->dst.rt_next; trt; trt = trt->dst.rt_next) |
673d57e7 | 1252 | printk(" . %pI4", &trt->rt_dst); |
1da177e4 LT |
1253 | printk("\n"); |
1254 | } | |
1255 | #endif | |
00269b54 ED |
1256 | /* |
1257 | * Since lookup is lockfree, we must make sure | |
1258 | * previous writes to rt are comitted to memory | |
1259 | * before making rt visible to other CPUS. | |
1260 | */ | |
1ddbcb00 | 1261 | rcu_assign_pointer(rt_hash_table[hash].chain, rt); |
1080d709 | 1262 | |
22c047cc | 1263 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
73e42897 | 1264 | |
b6280b47 | 1265 | skip_hashing: |
511c3f92 ED |
1266 | if (rp) |
1267 | *rp = rt; | |
1268 | else | |
d8d1f30b | 1269 | skb_dst_set(skb, &rt->dst); |
1da177e4 LT |
1270 | return 0; |
1271 | } | |
1272 | ||
1273 | void rt_bind_peer(struct rtable *rt, int create) | |
1274 | { | |
1da177e4 LT |
1275 | struct inet_peer *peer; |
1276 | ||
1277 | peer = inet_getpeer(rt->rt_dst, create); | |
1278 | ||
49e8ab03 | 1279 | if (peer && cmpxchg(&rt->peer, NULL, peer) != NULL) |
1da177e4 LT |
1280 | inet_putpeer(peer); |
1281 | } | |
1282 | ||
1283 | /* | |
1284 | * Peer allocation may fail only in serious out-of-memory conditions. However | |
1285 | * we still can generate some output. | |
1286 | * Random ID selection looks a bit dangerous because we have no chances to | |
1287 | * select ID being unique in a reasonable period of time. | |
1288 | * But broken packet identifier may be better than no packet at all. | |
1289 | */ | |
1290 | static void ip_select_fb_ident(struct iphdr *iph) | |
1291 | { | |
1292 | static DEFINE_SPINLOCK(ip_fb_id_lock); | |
1293 | static u32 ip_fallback_id; | |
1294 | u32 salt; | |
1295 | ||
1296 | spin_lock_bh(&ip_fb_id_lock); | |
e448515c | 1297 | salt = secure_ip_id((__force __be32)ip_fallback_id ^ iph->daddr); |
1da177e4 LT |
1298 | iph->id = htons(salt & 0xFFFF); |
1299 | ip_fallback_id = salt; | |
1300 | spin_unlock_bh(&ip_fb_id_lock); | |
1301 | } | |
1302 | ||
1303 | void __ip_select_ident(struct iphdr *iph, struct dst_entry *dst, int more) | |
1304 | { | |
1305 | struct rtable *rt = (struct rtable *) dst; | |
1306 | ||
1307 | if (rt) { | |
1308 | if (rt->peer == NULL) | |
1309 | rt_bind_peer(rt, 1); | |
1310 | ||
1311 | /* If peer is attached to destination, it is never detached, | |
1312 | so that we need not to grab a lock to dereference it. | |
1313 | */ | |
1314 | if (rt->peer) { | |
1315 | iph->id = htons(inet_getid(rt->peer, more)); | |
1316 | return; | |
1317 | } | |
1318 | } else | |
e905a9ed | 1319 | printk(KERN_DEBUG "rt_bind_peer(0) @%p\n", |
9c2b3328 | 1320 | __builtin_return_address(0)); |
1da177e4 LT |
1321 | |
1322 | ip_select_fb_ident(iph); | |
1323 | } | |
4bc2f18b | 1324 | EXPORT_SYMBOL(__ip_select_ident); |
1da177e4 LT |
1325 | |
1326 | static void rt_del(unsigned hash, struct rtable *rt) | |
1327 | { | |
29e75252 | 1328 | struct rtable **rthp, *aux; |
1da177e4 | 1329 | |
29e75252 | 1330 | rthp = &rt_hash_table[hash].chain; |
22c047cc | 1331 | spin_lock_bh(rt_hash_lock_addr(hash)); |
1da177e4 | 1332 | ip_rt_put(rt); |
29e75252 | 1333 | while ((aux = *rthp) != NULL) { |
e84f84f2 | 1334 | if (aux == rt || rt_is_expired(aux)) { |
d8d1f30b | 1335 | *rthp = aux->dst.rt_next; |
29e75252 ED |
1336 | rt_free(aux); |
1337 | continue; | |
1da177e4 | 1338 | } |
d8d1f30b | 1339 | rthp = &aux->dst.rt_next; |
29e75252 | 1340 | } |
22c047cc | 1341 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1342 | } |
1343 | ||
ed7865a4 | 1344 | /* called in rcu_read_lock() section */ |
f7655229 AV |
1345 | void ip_rt_redirect(__be32 old_gw, __be32 daddr, __be32 new_gw, |
1346 | __be32 saddr, struct net_device *dev) | |
1da177e4 LT |
1347 | { |
1348 | int i, k; | |
ed7865a4 | 1349 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1350 | struct rtable *rth, **rthp; |
f7655229 | 1351 | __be32 skeys[2] = { saddr, 0 }; |
1da177e4 | 1352 | int ikeys[2] = { dev->ifindex, 0 }; |
8d71740c | 1353 | struct netevent_redirect netevent; |
317805b8 | 1354 | struct net *net; |
1da177e4 | 1355 | |
1da177e4 LT |
1356 | if (!in_dev) |
1357 | return; | |
1358 | ||
c346dca1 | 1359 | net = dev_net(dev); |
9d4fb27d JP |
1360 | if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) || |
1361 | ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) || | |
1362 | ipv4_is_zeronet(new_gw)) | |
1da177e4 LT |
1363 | goto reject_redirect; |
1364 | ||
1080d709 NH |
1365 | if (!rt_caching(net)) |
1366 | goto reject_redirect; | |
1367 | ||
1da177e4 LT |
1368 | if (!IN_DEV_SHARED_MEDIA(in_dev)) { |
1369 | if (!inet_addr_onlink(in_dev, new_gw, old_gw)) | |
1370 | goto reject_redirect; | |
1371 | if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev)) | |
1372 | goto reject_redirect; | |
1373 | } else { | |
317805b8 | 1374 | if (inet_addr_type(net, new_gw) != RTN_UNICAST) |
1da177e4 LT |
1375 | goto reject_redirect; |
1376 | } | |
1377 | ||
1378 | for (i = 0; i < 2; i++) { | |
1379 | for (k = 0; k < 2; k++) { | |
b00180de | 1380 | unsigned hash = rt_hash(daddr, skeys[i], ikeys[k], |
e84f84f2 | 1381 | rt_genid(net)); |
1da177e4 LT |
1382 | |
1383 | rthp=&rt_hash_table[hash].chain; | |
1384 | ||
1da177e4 LT |
1385 | while ((rth = rcu_dereference(*rthp)) != NULL) { |
1386 | struct rtable *rt; | |
1387 | ||
1388 | if (rth->fl.fl4_dst != daddr || | |
1389 | rth->fl.fl4_src != skeys[i] || | |
1da177e4 | 1390 | rth->fl.oif != ikeys[k] || |
29e75252 | 1391 | rth->fl.iif != 0 || |
e84f84f2 | 1392 | rt_is_expired(rth) || |
d8d1f30b CG |
1393 | !net_eq(dev_net(rth->dst.dev), net)) { |
1394 | rthp = &rth->dst.rt_next; | |
1da177e4 LT |
1395 | continue; |
1396 | } | |
1397 | ||
1398 | if (rth->rt_dst != daddr || | |
1399 | rth->rt_src != saddr || | |
d8d1f30b | 1400 | rth->dst.error || |
1da177e4 | 1401 | rth->rt_gateway != old_gw || |
d8d1f30b | 1402 | rth->dst.dev != dev) |
1da177e4 LT |
1403 | break; |
1404 | ||
d8d1f30b | 1405 | dst_hold(&rth->dst); |
1da177e4 LT |
1406 | |
1407 | rt = dst_alloc(&ipv4_dst_ops); | |
1408 | if (rt == NULL) { | |
1409 | ip_rt_put(rth); | |
1da177e4 LT |
1410 | return; |
1411 | } | |
1412 | ||
1413 | /* Copy all the information. */ | |
1414 | *rt = *rth; | |
d8d1f30b CG |
1415 | rt->dst.__use = 1; |
1416 | atomic_set(&rt->dst.__refcnt, 1); | |
1417 | rt->dst.child = NULL; | |
1418 | if (rt->dst.dev) | |
1419 | dev_hold(rt->dst.dev); | |
1da177e4 LT |
1420 | if (rt->idev) |
1421 | in_dev_hold(rt->idev); | |
d8d1f30b CG |
1422 | rt->dst.obsolete = -1; |
1423 | rt->dst.lastuse = jiffies; | |
1424 | rt->dst.path = &rt->dst; | |
1425 | rt->dst.neighbour = NULL; | |
1426 | rt->dst.hh = NULL; | |
def8b4fa | 1427 | #ifdef CONFIG_XFRM |
d8d1f30b | 1428 | rt->dst.xfrm = NULL; |
def8b4fa | 1429 | #endif |
e84f84f2 | 1430 | rt->rt_genid = rt_genid(net); |
1da177e4 LT |
1431 | rt->rt_flags |= RTCF_REDIRECTED; |
1432 | ||
1433 | /* Gateway is different ... */ | |
1434 | rt->rt_gateway = new_gw; | |
1435 | ||
1436 | /* Redirect received -> path was valid */ | |
d8d1f30b | 1437 | dst_confirm(&rth->dst); |
1da177e4 LT |
1438 | |
1439 | if (rt->peer) | |
1440 | atomic_inc(&rt->peer->refcnt); | |
1441 | ||
d8d1f30b CG |
1442 | if (arp_bind_neighbour(&rt->dst) || |
1443 | !(rt->dst.neighbour->nud_state & | |
1da177e4 | 1444 | NUD_VALID)) { |
d8d1f30b CG |
1445 | if (rt->dst.neighbour) |
1446 | neigh_event_send(rt->dst.neighbour, NULL); | |
1da177e4 LT |
1447 | ip_rt_put(rth); |
1448 | rt_drop(rt); | |
1449 | goto do_next; | |
1450 | } | |
e905a9ed | 1451 | |
d8d1f30b CG |
1452 | netevent.old = &rth->dst; |
1453 | netevent.new = &rt->dst; | |
e905a9ed YH |
1454 | call_netevent_notifiers(NETEVENT_REDIRECT, |
1455 | &netevent); | |
1da177e4 LT |
1456 | |
1457 | rt_del(hash, rth); | |
6a2bad70 | 1458 | if (!rt_intern_hash(hash, rt, &rt, NULL, rt->fl.oif)) |
1da177e4 LT |
1459 | ip_rt_put(rt); |
1460 | goto do_next; | |
1461 | } | |
1da177e4 LT |
1462 | do_next: |
1463 | ; | |
1464 | } | |
1465 | } | |
1da177e4 LT |
1466 | return; |
1467 | ||
1468 | reject_redirect: | |
1469 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1470 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) | |
673d57e7 HH |
1471 | printk(KERN_INFO "Redirect from %pI4 on %s about %pI4 ignored.\n" |
1472 | " Advised path = %pI4 -> %pI4\n", | |
1473 | &old_gw, dev->name, &new_gw, | |
1474 | &saddr, &daddr); | |
1da177e4 | 1475 | #endif |
ed7865a4 | 1476 | ; |
1da177e4 LT |
1477 | } |
1478 | ||
1479 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst) | |
1480 | { | |
ee6b9673 | 1481 | struct rtable *rt = (struct rtable *)dst; |
1da177e4 LT |
1482 | struct dst_entry *ret = dst; |
1483 | ||
1484 | if (rt) { | |
d11a4dc1 | 1485 | if (dst->obsolete > 0) { |
1da177e4 LT |
1486 | ip_rt_put(rt); |
1487 | ret = NULL; | |
1488 | } else if ((rt->rt_flags & RTCF_REDIRECTED) || | |
d8d1f30b CG |
1489 | (rt->dst.expires && |
1490 | time_after_eq(jiffies, rt->dst.expires))) { | |
8c7bc840 | 1491 | unsigned hash = rt_hash(rt->fl.fl4_dst, rt->fl.fl4_src, |
b00180de | 1492 | rt->fl.oif, |
e84f84f2 | 1493 | rt_genid(dev_net(dst->dev))); |
1da177e4 | 1494 | #if RT_CACHE_DEBUG >= 1 |
673d57e7 HH |
1495 | printk(KERN_DEBUG "ipv4_negative_advice: redirect to %pI4/%02x dropped\n", |
1496 | &rt->rt_dst, rt->fl.fl4_tos); | |
1da177e4 LT |
1497 | #endif |
1498 | rt_del(hash, rt); | |
1499 | ret = NULL; | |
1500 | } | |
1501 | } | |
1502 | return ret; | |
1503 | } | |
1504 | ||
1505 | /* | |
1506 | * Algorithm: | |
1507 | * 1. The first ip_rt_redirect_number redirects are sent | |
1508 | * with exponential backoff, then we stop sending them at all, | |
1509 | * assuming that the host ignores our redirects. | |
1510 | * 2. If we did not see packets requiring redirects | |
1511 | * during ip_rt_redirect_silence, we assume that the host | |
1512 | * forgot redirected route and start to send redirects again. | |
1513 | * | |
1514 | * This algorithm is much cheaper and more intelligent than dumb load limiting | |
1515 | * in icmp.c. | |
1516 | * | |
1517 | * NOTE. Do not forget to inhibit load limiting for redirects (redundant) | |
1518 | * and "frag. need" (breaks PMTU discovery) in icmp.c. | |
1519 | */ | |
1520 | ||
1521 | void ip_rt_send_redirect(struct sk_buff *skb) | |
1522 | { | |
511c3f92 | 1523 | struct rtable *rt = skb_rtable(skb); |
30038fc6 ED |
1524 | struct in_device *in_dev; |
1525 | int log_martians; | |
1da177e4 | 1526 | |
30038fc6 | 1527 | rcu_read_lock(); |
d8d1f30b | 1528 | in_dev = __in_dev_get_rcu(rt->dst.dev); |
30038fc6 ED |
1529 | if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) { |
1530 | rcu_read_unlock(); | |
1da177e4 | 1531 | return; |
30038fc6 ED |
1532 | } |
1533 | log_martians = IN_DEV_LOG_MARTIANS(in_dev); | |
1534 | rcu_read_unlock(); | |
1da177e4 LT |
1535 | |
1536 | /* No redirected packets during ip_rt_redirect_silence; | |
1537 | * reset the algorithm. | |
1538 | */ | |
d8d1f30b CG |
1539 | if (time_after(jiffies, rt->dst.rate_last + ip_rt_redirect_silence)) |
1540 | rt->dst.rate_tokens = 0; | |
1da177e4 LT |
1541 | |
1542 | /* Too many ignored redirects; do not send anything | |
d8d1f30b | 1543 | * set dst.rate_last to the last seen redirected packet. |
1da177e4 | 1544 | */ |
d8d1f30b CG |
1545 | if (rt->dst.rate_tokens >= ip_rt_redirect_number) { |
1546 | rt->dst.rate_last = jiffies; | |
30038fc6 | 1547 | return; |
1da177e4 LT |
1548 | } |
1549 | ||
1550 | /* Check for load limit; set rate_last to the latest sent | |
1551 | * redirect. | |
1552 | */ | |
d8d1f30b | 1553 | if (rt->dst.rate_tokens == 0 || |
14fb8a76 | 1554 | time_after(jiffies, |
d8d1f30b CG |
1555 | (rt->dst.rate_last + |
1556 | (ip_rt_redirect_load << rt->dst.rate_tokens)))) { | |
1da177e4 | 1557 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway); |
d8d1f30b CG |
1558 | rt->dst.rate_last = jiffies; |
1559 | ++rt->dst.rate_tokens; | |
1da177e4 | 1560 | #ifdef CONFIG_IP_ROUTE_VERBOSE |
30038fc6 | 1561 | if (log_martians && |
d8d1f30b | 1562 | rt->dst.rate_tokens == ip_rt_redirect_number && |
1da177e4 | 1563 | net_ratelimit()) |
673d57e7 HH |
1564 | printk(KERN_WARNING "host %pI4/if%d ignores redirects for %pI4 to %pI4.\n", |
1565 | &rt->rt_src, rt->rt_iif, | |
1566 | &rt->rt_dst, &rt->rt_gateway); | |
1da177e4 LT |
1567 | #endif |
1568 | } | |
1da177e4 LT |
1569 | } |
1570 | ||
1571 | static int ip_error(struct sk_buff *skb) | |
1572 | { | |
511c3f92 | 1573 | struct rtable *rt = skb_rtable(skb); |
1da177e4 LT |
1574 | unsigned long now; |
1575 | int code; | |
1576 | ||
d8d1f30b | 1577 | switch (rt->dst.error) { |
1da177e4 LT |
1578 | case EINVAL: |
1579 | default: | |
1580 | goto out; | |
1581 | case EHOSTUNREACH: | |
1582 | code = ICMP_HOST_UNREACH; | |
1583 | break; | |
1584 | case ENETUNREACH: | |
1585 | code = ICMP_NET_UNREACH; | |
d8d1f30b | 1586 | IP_INC_STATS_BH(dev_net(rt->dst.dev), |
7c73a6fa | 1587 | IPSTATS_MIB_INNOROUTES); |
1da177e4 LT |
1588 | break; |
1589 | case EACCES: | |
1590 | code = ICMP_PKT_FILTERED; | |
1591 | break; | |
1592 | } | |
1593 | ||
1594 | now = jiffies; | |
d8d1f30b CG |
1595 | rt->dst.rate_tokens += now - rt->dst.rate_last; |
1596 | if (rt->dst.rate_tokens > ip_rt_error_burst) | |
1597 | rt->dst.rate_tokens = ip_rt_error_burst; | |
1598 | rt->dst.rate_last = now; | |
1599 | if (rt->dst.rate_tokens >= ip_rt_error_cost) { | |
1600 | rt->dst.rate_tokens -= ip_rt_error_cost; | |
1da177e4 LT |
1601 | icmp_send(skb, ICMP_DEST_UNREACH, code, 0); |
1602 | } | |
1603 | ||
1604 | out: kfree_skb(skb); | |
1605 | return 0; | |
e905a9ed | 1606 | } |
1da177e4 LT |
1607 | |
1608 | /* | |
1609 | * The last two values are not from the RFC but | |
1610 | * are needed for AMPRnet AX.25 paths. | |
1611 | */ | |
1612 | ||
9b5b5cff | 1613 | static const unsigned short mtu_plateau[] = |
1da177e4 LT |
1614 | {32000, 17914, 8166, 4352, 2002, 1492, 576, 296, 216, 128 }; |
1615 | ||
5969f71d | 1616 | static inline unsigned short guess_mtu(unsigned short old_mtu) |
1da177e4 LT |
1617 | { |
1618 | int i; | |
e905a9ed | 1619 | |
1da177e4 LT |
1620 | for (i = 0; i < ARRAY_SIZE(mtu_plateau); i++) |
1621 | if (old_mtu > mtu_plateau[i]) | |
1622 | return mtu_plateau[i]; | |
1623 | return 68; | |
1624 | } | |
1625 | ||
b5921910 | 1626 | unsigned short ip_rt_frag_needed(struct net *net, struct iphdr *iph, |
0010e465 TT |
1627 | unsigned short new_mtu, |
1628 | struct net_device *dev) | |
1da177e4 | 1629 | { |
0010e465 | 1630 | int i, k; |
1da177e4 LT |
1631 | unsigned short old_mtu = ntohs(iph->tot_len); |
1632 | struct rtable *rth; | |
0010e465 | 1633 | int ikeys[2] = { dev->ifindex, 0 }; |
e448515c AV |
1634 | __be32 skeys[2] = { iph->saddr, 0, }; |
1635 | __be32 daddr = iph->daddr; | |
1da177e4 LT |
1636 | unsigned short est_mtu = 0; |
1637 | ||
0010e465 TT |
1638 | for (k = 0; k < 2; k++) { |
1639 | for (i = 0; i < 2; i++) { | |
b00180de | 1640 | unsigned hash = rt_hash(daddr, skeys[i], ikeys[k], |
e84f84f2 | 1641 | rt_genid(net)); |
0010e465 TT |
1642 | |
1643 | rcu_read_lock(); | |
1644 | for (rth = rcu_dereference(rt_hash_table[hash].chain); rth; | |
d8d1f30b | 1645 | rth = rcu_dereference(rth->dst.rt_next)) { |
1da177e4 LT |
1646 | unsigned short mtu = new_mtu; |
1647 | ||
0010e465 TT |
1648 | if (rth->fl.fl4_dst != daddr || |
1649 | rth->fl.fl4_src != skeys[i] || | |
1650 | rth->rt_dst != daddr || | |
1651 | rth->rt_src != iph->saddr || | |
1652 | rth->fl.oif != ikeys[k] || | |
1653 | rth->fl.iif != 0 || | |
d8d1f30b CG |
1654 | dst_metric_locked(&rth->dst, RTAX_MTU) || |
1655 | !net_eq(dev_net(rth->dst.dev), net) || | |
6c3b8fc6 | 1656 | rt_is_expired(rth)) |
0010e465 TT |
1657 | continue; |
1658 | ||
1da177e4 LT |
1659 | if (new_mtu < 68 || new_mtu >= old_mtu) { |
1660 | ||
1661 | /* BSD 4.2 compatibility hack :-( */ | |
1662 | if (mtu == 0 && | |
d8d1f30b | 1663 | old_mtu >= dst_mtu(&rth->dst) && |
1da177e4 LT |
1664 | old_mtu >= 68 + (iph->ihl << 2)) |
1665 | old_mtu -= iph->ihl << 2; | |
1666 | ||
1667 | mtu = guess_mtu(old_mtu); | |
1668 | } | |
d8d1f30b CG |
1669 | if (mtu <= dst_mtu(&rth->dst)) { |
1670 | if (mtu < dst_mtu(&rth->dst)) { | |
1671 | dst_confirm(&rth->dst); | |
1da177e4 LT |
1672 | if (mtu < ip_rt_min_pmtu) { |
1673 | mtu = ip_rt_min_pmtu; | |
d8d1f30b | 1674 | rth->dst.metrics[RTAX_LOCK-1] |= |
1da177e4 LT |
1675 | (1 << RTAX_MTU); |
1676 | } | |
d8d1f30b CG |
1677 | rth->dst.metrics[RTAX_MTU-1] = mtu; |
1678 | dst_set_expires(&rth->dst, | |
1da177e4 LT |
1679 | ip_rt_mtu_expires); |
1680 | } | |
1681 | est_mtu = mtu; | |
1682 | } | |
1683 | } | |
0010e465 | 1684 | rcu_read_unlock(); |
1da177e4 | 1685 | } |
1da177e4 LT |
1686 | } |
1687 | return est_mtu ? : new_mtu; | |
1688 | } | |
1689 | ||
1690 | static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu) | |
1691 | { | |
6d273f8d | 1692 | if (dst_mtu(dst) > mtu && mtu >= 68 && |
1da177e4 LT |
1693 | !(dst_metric_locked(dst, RTAX_MTU))) { |
1694 | if (mtu < ip_rt_min_pmtu) { | |
1695 | mtu = ip_rt_min_pmtu; | |
1696 | dst->metrics[RTAX_LOCK-1] |= (1 << RTAX_MTU); | |
1697 | } | |
1698 | dst->metrics[RTAX_MTU-1] = mtu; | |
1699 | dst_set_expires(dst, ip_rt_mtu_expires); | |
8d71740c | 1700 | call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst); |
1da177e4 LT |
1701 | } |
1702 | } | |
1703 | ||
1704 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie) | |
1705 | { | |
d11a4dc1 TT |
1706 | if (rt_is_expired((struct rtable *)dst)) |
1707 | return NULL; | |
1708 | return dst; | |
1da177e4 LT |
1709 | } |
1710 | ||
1711 | static void ipv4_dst_destroy(struct dst_entry *dst) | |
1712 | { | |
1713 | struct rtable *rt = (struct rtable *) dst; | |
1714 | struct inet_peer *peer = rt->peer; | |
1715 | struct in_device *idev = rt->idev; | |
1716 | ||
1717 | if (peer) { | |
1718 | rt->peer = NULL; | |
1719 | inet_putpeer(peer); | |
1720 | } | |
1721 | ||
1722 | if (idev) { | |
1723 | rt->idev = NULL; | |
1724 | in_dev_put(idev); | |
1725 | } | |
1726 | } | |
1727 | ||
1728 | static void ipv4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, | |
1729 | int how) | |
1730 | { | |
1731 | struct rtable *rt = (struct rtable *) dst; | |
1732 | struct in_device *idev = rt->idev; | |
c346dca1 | 1733 | if (dev != dev_net(dev)->loopback_dev && idev && idev->dev == dev) { |
5a3e55d6 | 1734 | struct in_device *loopback_idev = |
c346dca1 | 1735 | in_dev_get(dev_net(dev)->loopback_dev); |
1da177e4 LT |
1736 | if (loopback_idev) { |
1737 | rt->idev = loopback_idev; | |
1738 | in_dev_put(idev); | |
1739 | } | |
1740 | } | |
1741 | } | |
1742 | ||
1743 | static void ipv4_link_failure(struct sk_buff *skb) | |
1744 | { | |
1745 | struct rtable *rt; | |
1746 | ||
1747 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0); | |
1748 | ||
511c3f92 | 1749 | rt = skb_rtable(skb); |
1da177e4 | 1750 | if (rt) |
d8d1f30b | 1751 | dst_set_expires(&rt->dst, 0); |
1da177e4 LT |
1752 | } |
1753 | ||
1754 | static int ip_rt_bug(struct sk_buff *skb) | |
1755 | { | |
673d57e7 HH |
1756 | printk(KERN_DEBUG "ip_rt_bug: %pI4 -> %pI4, %s\n", |
1757 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
1da177e4 LT |
1758 | skb->dev ? skb->dev->name : "?"); |
1759 | kfree_skb(skb); | |
1760 | return 0; | |
1761 | } | |
1762 | ||
1763 | /* | |
1764 | We do not cache source address of outgoing interface, | |
1765 | because it is used only by IP RR, TS and SRR options, | |
1766 | so that it out of fast path. | |
1767 | ||
1768 | BTW remember: "addr" is allowed to be not aligned | |
1769 | in IP options! | |
1770 | */ | |
1771 | ||
1772 | void ip_rt_get_source(u8 *addr, struct rtable *rt) | |
1773 | { | |
a61ced5d | 1774 | __be32 src; |
1da177e4 LT |
1775 | struct fib_result res; |
1776 | ||
1777 | if (rt->fl.iif == 0) | |
1778 | src = rt->rt_src; | |
ebc0ffae ED |
1779 | else { |
1780 | rcu_read_lock(); | |
1781 | if (fib_lookup(dev_net(rt->dst.dev), &rt->fl, &res) == 0) | |
1782 | src = FIB_RES_PREFSRC(res); | |
1783 | else | |
1784 | src = inet_select_addr(rt->dst.dev, rt->rt_gateway, | |
1da177e4 | 1785 | RT_SCOPE_UNIVERSE); |
ebc0ffae ED |
1786 | rcu_read_unlock(); |
1787 | } | |
1da177e4 LT |
1788 | memcpy(addr, &src, 4); |
1789 | } | |
1790 | ||
1791 | #ifdef CONFIG_NET_CLS_ROUTE | |
1792 | static void set_class_tag(struct rtable *rt, u32 tag) | |
1793 | { | |
d8d1f30b CG |
1794 | if (!(rt->dst.tclassid & 0xFFFF)) |
1795 | rt->dst.tclassid |= tag & 0xFFFF; | |
1796 | if (!(rt->dst.tclassid & 0xFFFF0000)) | |
1797 | rt->dst.tclassid |= tag & 0xFFFF0000; | |
1da177e4 LT |
1798 | } |
1799 | #endif | |
1800 | ||
1801 | static void rt_set_nexthop(struct rtable *rt, struct fib_result *res, u32 itag) | |
1802 | { | |
1803 | struct fib_info *fi = res->fi; | |
1804 | ||
1805 | if (fi) { | |
1806 | if (FIB_RES_GW(*res) && | |
1807 | FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) | |
1808 | rt->rt_gateway = FIB_RES_GW(*res); | |
d8d1f30b CG |
1809 | memcpy(rt->dst.metrics, fi->fib_metrics, |
1810 | sizeof(rt->dst.metrics)); | |
1da177e4 | 1811 | if (fi->fib_mtu == 0) { |
d8d1f30b CG |
1812 | rt->dst.metrics[RTAX_MTU-1] = rt->dst.dev->mtu; |
1813 | if (dst_metric_locked(&rt->dst, RTAX_MTU) && | |
1da177e4 | 1814 | rt->rt_gateway != rt->rt_dst && |
d8d1f30b CG |
1815 | rt->dst.dev->mtu > 576) |
1816 | rt->dst.metrics[RTAX_MTU-1] = 576; | |
1da177e4 LT |
1817 | } |
1818 | #ifdef CONFIG_NET_CLS_ROUTE | |
d8d1f30b | 1819 | rt->dst.tclassid = FIB_RES_NH(*res).nh_tclassid; |
1da177e4 LT |
1820 | #endif |
1821 | } else | |
d8d1f30b CG |
1822 | rt->dst.metrics[RTAX_MTU-1]= rt->dst.dev->mtu; |
1823 | ||
1824 | if (dst_metric(&rt->dst, RTAX_HOPLIMIT) == 0) | |
1825 | rt->dst.metrics[RTAX_HOPLIMIT-1] = sysctl_ip_default_ttl; | |
1826 | if (dst_mtu(&rt->dst) > IP_MAX_MTU) | |
1827 | rt->dst.metrics[RTAX_MTU-1] = IP_MAX_MTU; | |
1828 | if (dst_metric(&rt->dst, RTAX_ADVMSS) == 0) | |
1829 | rt->dst.metrics[RTAX_ADVMSS-1] = max_t(unsigned int, rt->dst.dev->mtu - 40, | |
1da177e4 | 1830 | ip_rt_min_advmss); |
d8d1f30b CG |
1831 | if (dst_metric(&rt->dst, RTAX_ADVMSS) > 65535 - 40) |
1832 | rt->dst.metrics[RTAX_ADVMSS-1] = 65535 - 40; | |
1da177e4 LT |
1833 | |
1834 | #ifdef CONFIG_NET_CLS_ROUTE | |
1835 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
1836 | set_class_tag(rt, fib_rules_tclass(res)); | |
1837 | #endif | |
1838 | set_class_tag(rt, itag); | |
1839 | #endif | |
e905a9ed | 1840 | rt->rt_type = res->type; |
1da177e4 LT |
1841 | } |
1842 | ||
96d36220 | 1843 | /* called in rcu_read_lock() section */ |
9e12bb22 | 1844 | static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
1845 | u8 tos, struct net_device *dev, int our) |
1846 | { | |
96d36220 | 1847 | unsigned int hash; |
1da177e4 | 1848 | struct rtable *rth; |
a61ced5d | 1849 | __be32 spec_dst; |
96d36220 | 1850 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1851 | u32 itag = 0; |
b5f7e755 | 1852 | int err; |
1da177e4 LT |
1853 | |
1854 | /* Primary sanity checks. */ | |
1855 | ||
1856 | if (in_dev == NULL) | |
1857 | return -EINVAL; | |
1858 | ||
1e637c74 | 1859 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
f97c1e0c | 1860 | ipv4_is_loopback(saddr) || skb->protocol != htons(ETH_P_IP)) |
1da177e4 LT |
1861 | goto e_inval; |
1862 | ||
f97c1e0c JP |
1863 | if (ipv4_is_zeronet(saddr)) { |
1864 | if (!ipv4_is_local_multicast(daddr)) | |
1da177e4 LT |
1865 | goto e_inval; |
1866 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); | |
b5f7e755 ED |
1867 | } else { |
1868 | err = fib_validate_source(saddr, 0, tos, 0, dev, &spec_dst, | |
1869 | &itag, 0); | |
1870 | if (err < 0) | |
1871 | goto e_err; | |
1872 | } | |
1da177e4 LT |
1873 | rth = dst_alloc(&ipv4_dst_ops); |
1874 | if (!rth) | |
1875 | goto e_nobufs; | |
1876 | ||
d8d1f30b CG |
1877 | rth->dst.output = ip_rt_bug; |
1878 | rth->dst.obsolete = -1; | |
1da177e4 | 1879 | |
d8d1f30b CG |
1880 | atomic_set(&rth->dst.__refcnt, 1); |
1881 | rth->dst.flags= DST_HOST; | |
42f811b8 | 1882 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
d8d1f30b | 1883 | rth->dst.flags |= DST_NOPOLICY; |
1da177e4 LT |
1884 | rth->fl.fl4_dst = daddr; |
1885 | rth->rt_dst = daddr; | |
1886 | rth->fl.fl4_tos = tos; | |
47dcf0cb | 1887 | rth->fl.mark = skb->mark; |
1da177e4 LT |
1888 | rth->fl.fl4_src = saddr; |
1889 | rth->rt_src = saddr; | |
1890 | #ifdef CONFIG_NET_CLS_ROUTE | |
d8d1f30b | 1891 | rth->dst.tclassid = itag; |
1da177e4 LT |
1892 | #endif |
1893 | rth->rt_iif = | |
1894 | rth->fl.iif = dev->ifindex; | |
d8d1f30b CG |
1895 | rth->dst.dev = init_net.loopback_dev; |
1896 | dev_hold(rth->dst.dev); | |
1897 | rth->idev = in_dev_get(rth->dst.dev); | |
1da177e4 LT |
1898 | rth->fl.oif = 0; |
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))); |
6a2bad70 | 1916 | return rt_intern_hash(hash, rth, NULL, skb, dev->ifindex); |
1da177e4 LT |
1917 | |
1918 | e_nobufs: | |
1da177e4 | 1919 | return -ENOBUFS; |
1da177e4 | 1920 | e_inval: |
96d36220 | 1921 | return -EINVAL; |
b5f7e755 | 1922 | e_err: |
b5f7e755 | 1923 | return err; |
1da177e4 LT |
1924 | } |
1925 | ||
1926 | ||
1927 | static void ip_handle_martian_source(struct net_device *dev, | |
1928 | struct in_device *in_dev, | |
1929 | struct sk_buff *skb, | |
9e12bb22 AV |
1930 | __be32 daddr, |
1931 | __be32 saddr) | |
1da177e4 LT |
1932 | { |
1933 | RT_CACHE_STAT_INC(in_martian_src); | |
1934 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1935 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) { | |
1936 | /* | |
1937 | * RFC1812 recommendation, if source is martian, | |
1938 | * the only hint is MAC header. | |
1939 | */ | |
673d57e7 HH |
1940 | printk(KERN_WARNING "martian source %pI4 from %pI4, on dev %s\n", |
1941 | &daddr, &saddr, dev->name); | |
98e399f8 | 1942 | if (dev->hard_header_len && skb_mac_header_was_set(skb)) { |
1da177e4 | 1943 | int i; |
98e399f8 | 1944 | const unsigned char *p = skb_mac_header(skb); |
1da177e4 LT |
1945 | printk(KERN_WARNING "ll header: "); |
1946 | for (i = 0; i < dev->hard_header_len; i++, p++) { | |
1947 | printk("%02x", *p); | |
1948 | if (i < (dev->hard_header_len - 1)) | |
1949 | printk(":"); | |
1950 | } | |
1951 | printk("\n"); | |
1952 | } | |
1953 | } | |
1954 | #endif | |
1955 | } | |
1956 | ||
47360228 | 1957 | /* called in rcu_read_lock() section */ |
5969f71d SH |
1958 | static int __mkroute_input(struct sk_buff *skb, |
1959 | struct fib_result *res, | |
1960 | struct in_device *in_dev, | |
1961 | __be32 daddr, __be32 saddr, u32 tos, | |
1962 | struct rtable **result) | |
1da177e4 | 1963 | { |
1da177e4 LT |
1964 | struct rtable *rth; |
1965 | int err; | |
1966 | struct in_device *out_dev; | |
47360228 | 1967 | unsigned int flags = 0; |
d9c9df8c AV |
1968 | __be32 spec_dst; |
1969 | u32 itag; | |
1da177e4 LT |
1970 | |
1971 | /* get a working reference to the output device */ | |
47360228 | 1972 | out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res)); |
1da177e4 LT |
1973 | if (out_dev == NULL) { |
1974 | if (net_ratelimit()) | |
1975 | printk(KERN_CRIT "Bug in ip_route_input" \ | |
1976 | "_slow(). Please, report\n"); | |
1977 | return -EINVAL; | |
1978 | } | |
1979 | ||
1980 | ||
e905a9ed | 1981 | err = fib_validate_source(saddr, daddr, tos, FIB_RES_OIF(*res), |
b0c110ca | 1982 | in_dev->dev, &spec_dst, &itag, skb->mark); |
1da177e4 | 1983 | if (err < 0) { |
e905a9ed | 1984 | ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr, |
1da177e4 | 1985 | saddr); |
e905a9ed | 1986 | |
1da177e4 LT |
1987 | goto cleanup; |
1988 | } | |
1989 | ||
1990 | if (err) | |
1991 | flags |= RTCF_DIRECTSRC; | |
1992 | ||
51b77cae | 1993 | if (out_dev == in_dev && err && |
1da177e4 LT |
1994 | (IN_DEV_SHARED_MEDIA(out_dev) || |
1995 | inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) | |
1996 | flags |= RTCF_DOREDIRECT; | |
1997 | ||
1998 | if (skb->protocol != htons(ETH_P_IP)) { | |
1999 | /* Not IP (i.e. ARP). Do not create route, if it is | |
2000 | * invalid for proxy arp. DNAT routes are always valid. | |
65324144 JDB |
2001 | * |
2002 | * Proxy arp feature have been extended to allow, ARP | |
2003 | * replies back to the same interface, to support | |
2004 | * Private VLAN switch technologies. See arp.c. | |
1da177e4 | 2005 | */ |
65324144 JDB |
2006 | if (out_dev == in_dev && |
2007 | IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) { | |
1da177e4 LT |
2008 | err = -EINVAL; |
2009 | goto cleanup; | |
2010 | } | |
2011 | } | |
2012 | ||
2013 | ||
2014 | rth = dst_alloc(&ipv4_dst_ops); | |
2015 | if (!rth) { | |
2016 | err = -ENOBUFS; | |
2017 | goto cleanup; | |
2018 | } | |
2019 | ||
d8d1f30b CG |
2020 | atomic_set(&rth->dst.__refcnt, 1); |
2021 | rth->dst.flags= DST_HOST; | |
42f811b8 | 2022 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
d8d1f30b | 2023 | rth->dst.flags |= DST_NOPOLICY; |
42f811b8 | 2024 | if (IN_DEV_CONF_GET(out_dev, NOXFRM)) |
d8d1f30b | 2025 | rth->dst.flags |= DST_NOXFRM; |
1da177e4 LT |
2026 | rth->fl.fl4_dst = daddr; |
2027 | rth->rt_dst = daddr; | |
2028 | rth->fl.fl4_tos = tos; | |
47dcf0cb | 2029 | rth->fl.mark = skb->mark; |
1da177e4 LT |
2030 | rth->fl.fl4_src = saddr; |
2031 | rth->rt_src = saddr; | |
2032 | rth->rt_gateway = daddr; | |
2033 | rth->rt_iif = | |
2034 | rth->fl.iif = in_dev->dev->ifindex; | |
d8d1f30b CG |
2035 | rth->dst.dev = (out_dev)->dev; |
2036 | dev_hold(rth->dst.dev); | |
2037 | rth->idev = in_dev_get(rth->dst.dev); | |
1da177e4 LT |
2038 | rth->fl.oif = 0; |
2039 | rth->rt_spec_dst= spec_dst; | |
2040 | ||
d8d1f30b CG |
2041 | rth->dst.obsolete = -1; |
2042 | rth->dst.input = ip_forward; | |
2043 | rth->dst.output = ip_output; | |
2044 | rth->rt_genid = rt_genid(dev_net(rth->dst.dev)); | |
1da177e4 LT |
2045 | |
2046 | rt_set_nexthop(rth, res, itag); | |
2047 | ||
2048 | rth->rt_flags = flags; | |
2049 | ||
2050 | *result = rth; | |
2051 | err = 0; | |
2052 | cleanup: | |
1da177e4 | 2053 | return err; |
e905a9ed | 2054 | } |
1da177e4 | 2055 | |
5969f71d SH |
2056 | static int ip_mkroute_input(struct sk_buff *skb, |
2057 | struct fib_result *res, | |
2058 | const struct flowi *fl, | |
2059 | struct in_device *in_dev, | |
2060 | __be32 daddr, __be32 saddr, u32 tos) | |
1da177e4 | 2061 | { |
7abaa27c | 2062 | struct rtable* rth = NULL; |
1da177e4 LT |
2063 | int err; |
2064 | unsigned hash; | |
2065 | ||
2066 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
2067 | if (res->fi && res->fi->fib_nhs > 1 && fl->oif == 0) | |
2068 | fib_select_multipath(fl, res); | |
2069 | #endif | |
2070 | ||
2071 | /* create a routing cache entry */ | |
2072 | err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth); | |
2073 | if (err) | |
2074 | return err; | |
1da177e4 LT |
2075 | |
2076 | /* put it into the cache */ | |
e84f84f2 | 2077 | hash = rt_hash(daddr, saddr, fl->iif, |
d8d1f30b | 2078 | rt_genid(dev_net(rth->dst.dev))); |
6a2bad70 | 2079 | return rt_intern_hash(hash, rth, NULL, skb, fl->iif); |
1da177e4 LT |
2080 | } |
2081 | ||
1da177e4 LT |
2082 | /* |
2083 | * NOTE. We drop all the packets that has local source | |
2084 | * addresses, because every properly looped back packet | |
2085 | * must have correct destination already attached by output routine. | |
2086 | * | |
2087 | * Such approach solves two big problems: | |
2088 | * 1. Not simplex devices are handled properly. | |
2089 | * 2. IP spoofing attempts are filtered with 100% of guarantee. | |
ebc0ffae | 2090 | * called with rcu_read_lock() |
1da177e4 LT |
2091 | */ |
2092 | ||
9e12bb22 | 2093 | static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
2094 | u8 tos, struct net_device *dev) |
2095 | { | |
2096 | struct fib_result res; | |
96d36220 | 2097 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 LT |
2098 | struct flowi fl = { .nl_u = { .ip4_u = |
2099 | { .daddr = daddr, | |
2100 | .saddr = saddr, | |
2101 | .tos = tos, | |
2102 | .scope = RT_SCOPE_UNIVERSE, | |
1da177e4 | 2103 | } }, |
47dcf0cb | 2104 | .mark = skb->mark, |
1da177e4 LT |
2105 | .iif = dev->ifindex }; |
2106 | unsigned flags = 0; | |
2107 | u32 itag = 0; | |
2108 | struct rtable * rth; | |
2109 | unsigned hash; | |
9e12bb22 | 2110 | __be32 spec_dst; |
1da177e4 | 2111 | int err = -EINVAL; |
c346dca1 | 2112 | struct net * net = dev_net(dev); |
1da177e4 LT |
2113 | |
2114 | /* IP on this device is disabled. */ | |
2115 | ||
2116 | if (!in_dev) | |
2117 | goto out; | |
2118 | ||
2119 | /* Check for the most weird martians, which can be not detected | |
2120 | by fib_lookup. | |
2121 | */ | |
2122 | ||
1e637c74 | 2123 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
f97c1e0c | 2124 | ipv4_is_loopback(saddr)) |
1da177e4 LT |
2125 | goto martian_source; |
2126 | ||
e448515c | 2127 | if (daddr == htonl(0xFFFFFFFF) || (saddr == 0 && daddr == 0)) |
1da177e4 LT |
2128 | goto brd_input; |
2129 | ||
2130 | /* Accept zero addresses only to limited broadcast; | |
2131 | * I even do not know to fix it or not. Waiting for complains :-) | |
2132 | */ | |
f97c1e0c | 2133 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
2134 | goto martian_source; |
2135 | ||
1e637c74 | 2136 | if (ipv4_is_lbcast(daddr) || ipv4_is_zeronet(daddr) || |
f97c1e0c | 2137 | ipv4_is_loopback(daddr)) |
1da177e4 LT |
2138 | goto martian_destination; |
2139 | ||
2140 | /* | |
2141 | * Now we are ready to route packet. | |
2142 | */ | |
ebc0ffae ED |
2143 | err = fib_lookup(net, &fl, &res); |
2144 | if (err != 0) { | |
1da177e4 | 2145 | if (!IN_DEV_FORWARD(in_dev)) |
2c2910a4 | 2146 | goto e_hostunreach; |
1da177e4 LT |
2147 | goto no_route; |
2148 | } | |
1da177e4 LT |
2149 | |
2150 | RT_CACHE_STAT_INC(in_slow_tot); | |
2151 | ||
2152 | if (res.type == RTN_BROADCAST) | |
2153 | goto brd_input; | |
2154 | ||
2155 | if (res.type == RTN_LOCAL) { | |
b5f7e755 | 2156 | err = fib_validate_source(saddr, daddr, tos, |
ebc0ffae ED |
2157 | net->loopback_dev->ifindex, |
2158 | dev, &spec_dst, &itag, skb->mark); | |
b5f7e755 ED |
2159 | if (err < 0) |
2160 | goto martian_source_keep_err; | |
2161 | if (err) | |
1da177e4 LT |
2162 | flags |= RTCF_DIRECTSRC; |
2163 | spec_dst = daddr; | |
2164 | goto local_input; | |
2165 | } | |
2166 | ||
2167 | if (!IN_DEV_FORWARD(in_dev)) | |
2c2910a4 | 2168 | goto e_hostunreach; |
1da177e4 LT |
2169 | if (res.type != RTN_UNICAST) |
2170 | goto martian_destination; | |
2171 | ||
2172 | err = ip_mkroute_input(skb, &res, &fl, in_dev, daddr, saddr, tos); | |
1da177e4 LT |
2173 | out: return err; |
2174 | ||
2175 | brd_input: | |
2176 | if (skb->protocol != htons(ETH_P_IP)) | |
2177 | goto e_inval; | |
2178 | ||
f97c1e0c | 2179 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
2180 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
2181 | else { | |
2182 | err = fib_validate_source(saddr, 0, tos, 0, dev, &spec_dst, | |
b0c110ca | 2183 | &itag, skb->mark); |
1da177e4 | 2184 | if (err < 0) |
b5f7e755 | 2185 | goto martian_source_keep_err; |
1da177e4 LT |
2186 | if (err) |
2187 | flags |= RTCF_DIRECTSRC; | |
2188 | } | |
2189 | flags |= RTCF_BROADCAST; | |
2190 | res.type = RTN_BROADCAST; | |
2191 | RT_CACHE_STAT_INC(in_brd); | |
2192 | ||
2193 | local_input: | |
2194 | rth = dst_alloc(&ipv4_dst_ops); | |
2195 | if (!rth) | |
2196 | goto e_nobufs; | |
2197 | ||
d8d1f30b CG |
2198 | rth->dst.output= ip_rt_bug; |
2199 | rth->dst.obsolete = -1; | |
e84f84f2 | 2200 | rth->rt_genid = rt_genid(net); |
1da177e4 | 2201 | |
d8d1f30b CG |
2202 | atomic_set(&rth->dst.__refcnt, 1); |
2203 | rth->dst.flags= DST_HOST; | |
42f811b8 | 2204 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
d8d1f30b | 2205 | rth->dst.flags |= DST_NOPOLICY; |
1da177e4 LT |
2206 | rth->fl.fl4_dst = daddr; |
2207 | rth->rt_dst = daddr; | |
2208 | rth->fl.fl4_tos = tos; | |
47dcf0cb | 2209 | rth->fl.mark = skb->mark; |
1da177e4 LT |
2210 | rth->fl.fl4_src = saddr; |
2211 | rth->rt_src = saddr; | |
2212 | #ifdef CONFIG_NET_CLS_ROUTE | |
d8d1f30b | 2213 | rth->dst.tclassid = itag; |
1da177e4 LT |
2214 | #endif |
2215 | rth->rt_iif = | |
2216 | rth->fl.iif = dev->ifindex; | |
d8d1f30b CG |
2217 | rth->dst.dev = net->loopback_dev; |
2218 | dev_hold(rth->dst.dev); | |
2219 | rth->idev = in_dev_get(rth->dst.dev); | |
1da177e4 LT |
2220 | rth->rt_gateway = daddr; |
2221 | rth->rt_spec_dst= spec_dst; | |
d8d1f30b | 2222 | rth->dst.input= ip_local_deliver; |
1da177e4 LT |
2223 | rth->rt_flags = flags|RTCF_LOCAL; |
2224 | if (res.type == RTN_UNREACHABLE) { | |
d8d1f30b CG |
2225 | rth->dst.input= ip_error; |
2226 | rth->dst.error= -err; | |
1da177e4 LT |
2227 | rth->rt_flags &= ~RTCF_LOCAL; |
2228 | } | |
2229 | rth->rt_type = res.type; | |
e84f84f2 | 2230 | hash = rt_hash(daddr, saddr, fl.iif, rt_genid(net)); |
6a2bad70 | 2231 | err = rt_intern_hash(hash, rth, NULL, skb, fl.iif); |
ebc0ffae | 2232 | goto out; |
1da177e4 LT |
2233 | |
2234 | no_route: | |
2235 | RT_CACHE_STAT_INC(in_no_route); | |
2236 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE); | |
2237 | res.type = RTN_UNREACHABLE; | |
7f53878d MC |
2238 | if (err == -ESRCH) |
2239 | err = -ENETUNREACH; | |
1da177e4 LT |
2240 | goto local_input; |
2241 | ||
2242 | /* | |
2243 | * Do not cache martian addresses: they should be logged (RFC1812) | |
2244 | */ | |
2245 | martian_destination: | |
2246 | RT_CACHE_STAT_INC(in_martian_dst); | |
2247 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
2248 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) | |
673d57e7 HH |
2249 | printk(KERN_WARNING "martian destination %pI4 from %pI4, dev %s\n", |
2250 | &daddr, &saddr, dev->name); | |
1da177e4 | 2251 | #endif |
2c2910a4 DE |
2252 | |
2253 | e_hostunreach: | |
e905a9ed | 2254 | err = -EHOSTUNREACH; |
ebc0ffae | 2255 | goto out; |
2c2910a4 | 2256 | |
1da177e4 LT |
2257 | e_inval: |
2258 | err = -EINVAL; | |
ebc0ffae | 2259 | goto out; |
1da177e4 LT |
2260 | |
2261 | e_nobufs: | |
2262 | err = -ENOBUFS; | |
ebc0ffae | 2263 | goto out; |
1da177e4 LT |
2264 | |
2265 | martian_source: | |
b5f7e755 ED |
2266 | err = -EINVAL; |
2267 | martian_source_keep_err: | |
1da177e4 | 2268 | ip_handle_martian_source(dev, in_dev, skb, daddr, saddr); |
ebc0ffae | 2269 | goto out; |
1da177e4 LT |
2270 | } |
2271 | ||
407eadd9 ED |
2272 | int ip_route_input_common(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
2273 | u8 tos, struct net_device *dev, bool noref) | |
1da177e4 LT |
2274 | { |
2275 | struct rtable * rth; | |
2276 | unsigned hash; | |
2277 | int iif = dev->ifindex; | |
b5921910 | 2278 | struct net *net; |
96d36220 | 2279 | int res; |
1da177e4 | 2280 | |
c346dca1 | 2281 | net = dev_net(dev); |
1080d709 | 2282 | |
96d36220 ED |
2283 | rcu_read_lock(); |
2284 | ||
1080d709 NH |
2285 | if (!rt_caching(net)) |
2286 | goto skip_cache; | |
2287 | ||
1da177e4 | 2288 | tos &= IPTOS_RT_MASK; |
e84f84f2 | 2289 | hash = rt_hash(daddr, saddr, iif, rt_genid(net)); |
1da177e4 | 2290 | |
1da177e4 | 2291 | for (rth = rcu_dereference(rt_hash_table[hash].chain); rth; |
d8d1f30b | 2292 | rth = rcu_dereference(rth->dst.rt_next)) { |
0eae88f3 ED |
2293 | if ((((__force u32)rth->fl.fl4_dst ^ (__force u32)daddr) | |
2294 | ((__force u32)rth->fl.fl4_src ^ (__force u32)saddr) | | |
c0b8c32b SH |
2295 | (rth->fl.iif ^ iif) | |
2296 | rth->fl.oif | | |
2297 | (rth->fl.fl4_tos ^ tos)) == 0 && | |
47dcf0cb | 2298 | rth->fl.mark == skb->mark && |
d8d1f30b | 2299 | net_eq(dev_net(rth->dst.dev), net) && |
e84f84f2 | 2300 | !rt_is_expired(rth)) { |
407eadd9 | 2301 | if (noref) { |
d8d1f30b CG |
2302 | dst_use_noref(&rth->dst, jiffies); |
2303 | skb_dst_set_noref(skb, &rth->dst); | |
407eadd9 | 2304 | } else { |
d8d1f30b CG |
2305 | dst_use(&rth->dst, jiffies); |
2306 | skb_dst_set(skb, &rth->dst); | |
407eadd9 | 2307 | } |
1da177e4 LT |
2308 | RT_CACHE_STAT_INC(in_hit); |
2309 | rcu_read_unlock(); | |
1da177e4 LT |
2310 | return 0; |
2311 | } | |
2312 | RT_CACHE_STAT_INC(in_hlist_search); | |
2313 | } | |
1da177e4 | 2314 | |
1080d709 | 2315 | skip_cache: |
1da177e4 LT |
2316 | /* Multicast recognition logic is moved from route cache to here. |
2317 | The problem was that too many Ethernet cards have broken/missing | |
2318 | hardware multicast filters :-( As result the host on multicasting | |
2319 | network acquires a lot of useless route cache entries, sort of | |
2320 | SDR messages from all the world. Now we try to get rid of them. | |
2321 | Really, provided software IP multicast filter is organized | |
2322 | reasonably (at least, hashed), it does not result in a slowdown | |
2323 | comparing with route cache reject entries. | |
2324 | Note, that multicast routers are not affected, because | |
2325 | route cache entry is created eventually. | |
2326 | */ | |
f97c1e0c | 2327 | if (ipv4_is_multicast(daddr)) { |
96d36220 | 2328 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 2329 | |
96d36220 | 2330 | if (in_dev) { |
1da177e4 | 2331 | int our = ip_check_mc(in_dev, daddr, saddr, |
96d36220 | 2332 | ip_hdr(skb)->protocol); |
1da177e4 LT |
2333 | if (our |
2334 | #ifdef CONFIG_IP_MROUTE | |
9d4fb27d JP |
2335 | || |
2336 | (!ipv4_is_local_multicast(daddr) && | |
2337 | IN_DEV_MFORWARD(in_dev)) | |
1da177e4 | 2338 | #endif |
9d4fb27d | 2339 | ) { |
96d36220 ED |
2340 | int res = ip_route_input_mc(skb, daddr, saddr, |
2341 | tos, dev, our); | |
1da177e4 | 2342 | rcu_read_unlock(); |
96d36220 | 2343 | return res; |
1da177e4 LT |
2344 | } |
2345 | } | |
2346 | rcu_read_unlock(); | |
2347 | return -EINVAL; | |
2348 | } | |
96d36220 ED |
2349 | res = ip_route_input_slow(skb, daddr, saddr, tos, dev); |
2350 | rcu_read_unlock(); | |
2351 | return res; | |
1da177e4 | 2352 | } |
407eadd9 | 2353 | EXPORT_SYMBOL(ip_route_input_common); |
1da177e4 | 2354 | |
ebc0ffae | 2355 | /* called with rcu_read_lock() */ |
5969f71d SH |
2356 | static int __mkroute_output(struct rtable **result, |
2357 | struct fib_result *res, | |
2358 | const struct flowi *fl, | |
2359 | const struct flowi *oldflp, | |
2360 | struct net_device *dev_out, | |
2361 | unsigned flags) | |
1da177e4 LT |
2362 | { |
2363 | struct rtable *rth; | |
2364 | struct in_device *in_dev; | |
2365 | u32 tos = RT_FL_TOS(oldflp); | |
1da177e4 | 2366 | |
dd28d1a0 | 2367 | if (ipv4_is_loopback(fl->fl4_src) && !(dev_out->flags & IFF_LOOPBACK)) |
1da177e4 LT |
2368 | return -EINVAL; |
2369 | ||
e448515c | 2370 | if (fl->fl4_dst == htonl(0xFFFFFFFF)) |
1da177e4 | 2371 | res->type = RTN_BROADCAST; |
f97c1e0c | 2372 | else if (ipv4_is_multicast(fl->fl4_dst)) |
1da177e4 | 2373 | res->type = RTN_MULTICAST; |
1e637c74 | 2374 | else if (ipv4_is_lbcast(fl->fl4_dst) || ipv4_is_zeronet(fl->fl4_dst)) |
1da177e4 LT |
2375 | return -EINVAL; |
2376 | ||
2377 | if (dev_out->flags & IFF_LOOPBACK) | |
2378 | flags |= RTCF_LOCAL; | |
2379 | ||
dd28d1a0 | 2380 | in_dev = __in_dev_get_rcu(dev_out); |
ebc0ffae | 2381 | if (!in_dev) |
1da177e4 | 2382 | return -EINVAL; |
ebc0ffae | 2383 | |
1da177e4 LT |
2384 | if (res->type == RTN_BROADCAST) { |
2385 | flags |= RTCF_BROADCAST | RTCF_LOCAL; | |
ebc0ffae | 2386 | res->fi = NULL; |
1da177e4 | 2387 | } else if (res->type == RTN_MULTICAST) { |
dd28d1a0 | 2388 | flags |= RTCF_MULTICAST | RTCF_LOCAL; |
e905a9ed | 2389 | if (!ip_check_mc(in_dev, oldflp->fl4_dst, oldflp->fl4_src, |
1da177e4 LT |
2390 | oldflp->proto)) |
2391 | flags &= ~RTCF_LOCAL; | |
2392 | /* If multicast route do not exist use | |
dd28d1a0 ED |
2393 | * default one, but do not gateway in this case. |
2394 | * Yes, it is hack. | |
1da177e4 | 2395 | */ |
ebc0ffae | 2396 | if (res->fi && res->prefixlen < 4) |
1da177e4 | 2397 | res->fi = NULL; |
1da177e4 LT |
2398 | } |
2399 | ||
2400 | ||
2401 | rth = dst_alloc(&ipv4_dst_ops); | |
8391d07b | 2402 | if (!rth) |
dd28d1a0 | 2403 | return -ENOBUFS; |
8391d07b | 2404 | |
dd28d1a0 | 2405 | in_dev_hold(in_dev); |
dd28d1a0 | 2406 | rth->idev = in_dev; |
1da177e4 | 2407 | |
d8d1f30b CG |
2408 | atomic_set(&rth->dst.__refcnt, 1); |
2409 | rth->dst.flags= DST_HOST; | |
42f811b8 | 2410 | if (IN_DEV_CONF_GET(in_dev, NOXFRM)) |
d8d1f30b | 2411 | rth->dst.flags |= DST_NOXFRM; |
42f811b8 | 2412 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
d8d1f30b | 2413 | rth->dst.flags |= DST_NOPOLICY; |
1da177e4 LT |
2414 | |
2415 | rth->fl.fl4_dst = oldflp->fl4_dst; | |
2416 | rth->fl.fl4_tos = tos; | |
2417 | rth->fl.fl4_src = oldflp->fl4_src; | |
2418 | rth->fl.oif = oldflp->oif; | |
47dcf0cb | 2419 | rth->fl.mark = oldflp->mark; |
1da177e4 LT |
2420 | rth->rt_dst = fl->fl4_dst; |
2421 | rth->rt_src = fl->fl4_src; | |
2422 | rth->rt_iif = oldflp->oif ? : dev_out->ifindex; | |
e905a9ed | 2423 | /* get references to the devices that are to be hold by the routing |
1da177e4 | 2424 | cache entry */ |
d8d1f30b | 2425 | rth->dst.dev = dev_out; |
1da177e4 | 2426 | dev_hold(dev_out); |
1da177e4 LT |
2427 | rth->rt_gateway = fl->fl4_dst; |
2428 | rth->rt_spec_dst= fl->fl4_src; | |
2429 | ||
d8d1f30b CG |
2430 | rth->dst.output=ip_output; |
2431 | rth->dst.obsolete = -1; | |
e84f84f2 | 2432 | rth->rt_genid = rt_genid(dev_net(dev_out)); |
1da177e4 LT |
2433 | |
2434 | RT_CACHE_STAT_INC(out_slow_tot); | |
2435 | ||
2436 | if (flags & RTCF_LOCAL) { | |
d8d1f30b | 2437 | rth->dst.input = ip_local_deliver; |
1da177e4 LT |
2438 | rth->rt_spec_dst = fl->fl4_dst; |
2439 | } | |
2440 | if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { | |
2441 | rth->rt_spec_dst = fl->fl4_src; | |
e905a9ed | 2442 | if (flags & RTCF_LOCAL && |
1da177e4 | 2443 | !(dev_out->flags & IFF_LOOPBACK)) { |
d8d1f30b | 2444 | rth->dst.output = ip_mc_output; |
1da177e4 LT |
2445 | RT_CACHE_STAT_INC(out_slow_mc); |
2446 | } | |
2447 | #ifdef CONFIG_IP_MROUTE | |
2448 | if (res->type == RTN_MULTICAST) { | |
2449 | if (IN_DEV_MFORWARD(in_dev) && | |
f97c1e0c | 2450 | !ipv4_is_local_multicast(oldflp->fl4_dst)) { |
d8d1f30b CG |
2451 | rth->dst.input = ip_mr_input; |
2452 | rth->dst.output = ip_mc_output; | |
1da177e4 LT |
2453 | } |
2454 | } | |
2455 | #endif | |
2456 | } | |
2457 | ||
2458 | rt_set_nexthop(rth, res, 0); | |
2459 | ||
2460 | rth->rt_flags = flags; | |
1da177e4 | 2461 | *result = rth; |
dd28d1a0 | 2462 | return 0; |
1da177e4 LT |
2463 | } |
2464 | ||
ebc0ffae | 2465 | /* called with rcu_read_lock() */ |
5969f71d SH |
2466 | static int ip_mkroute_output(struct rtable **rp, |
2467 | struct fib_result *res, | |
2468 | const struct flowi *fl, | |
2469 | const struct flowi *oldflp, | |
2470 | struct net_device *dev_out, | |
2471 | unsigned flags) | |
1da177e4 | 2472 | { |
7abaa27c | 2473 | struct rtable *rth = NULL; |
1da177e4 LT |
2474 | int err = __mkroute_output(&rth, res, fl, oldflp, dev_out, flags); |
2475 | unsigned hash; | |
2476 | if (err == 0) { | |
b00180de | 2477 | hash = rt_hash(oldflp->fl4_dst, oldflp->fl4_src, oldflp->oif, |
e84f84f2 | 2478 | rt_genid(dev_net(dev_out))); |
6a2bad70 | 2479 | err = rt_intern_hash(hash, rth, rp, NULL, oldflp->oif); |
1da177e4 | 2480 | } |
e905a9ed | 2481 | |
1da177e4 LT |
2482 | return err; |
2483 | } | |
2484 | ||
1da177e4 LT |
2485 | /* |
2486 | * Major route resolver routine. | |
0197aa38 | 2487 | * called with rcu_read_lock(); |
1da177e4 LT |
2488 | */ |
2489 | ||
b40afd0e DL |
2490 | static int ip_route_output_slow(struct net *net, struct rtable **rp, |
2491 | const struct flowi *oldflp) | |
1da177e4 LT |
2492 | { |
2493 | u32 tos = RT_FL_TOS(oldflp); | |
2494 | struct flowi fl = { .nl_u = { .ip4_u = | |
2495 | { .daddr = oldflp->fl4_dst, | |
2496 | .saddr = oldflp->fl4_src, | |
2497 | .tos = tos & IPTOS_RT_MASK, | |
2498 | .scope = ((tos & RTO_ONLINK) ? | |
2499 | RT_SCOPE_LINK : | |
2500 | RT_SCOPE_UNIVERSE), | |
1da177e4 | 2501 | } }, |
47dcf0cb | 2502 | .mark = oldflp->mark, |
b40afd0e | 2503 | .iif = net->loopback_dev->ifindex, |
1da177e4 LT |
2504 | .oif = oldflp->oif }; |
2505 | struct fib_result res; | |
0197aa38 | 2506 | unsigned int flags = 0; |
1da177e4 | 2507 | struct net_device *dev_out = NULL; |
1da177e4 LT |
2508 | int err; |
2509 | ||
2510 | ||
2511 | res.fi = NULL; | |
2512 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
2513 | res.r = NULL; | |
2514 | #endif | |
2515 | ||
2516 | if (oldflp->fl4_src) { | |
2517 | err = -EINVAL; | |
f97c1e0c | 2518 | if (ipv4_is_multicast(oldflp->fl4_src) || |
1e637c74 | 2519 | ipv4_is_lbcast(oldflp->fl4_src) || |
f97c1e0c | 2520 | ipv4_is_zeronet(oldflp->fl4_src)) |
1da177e4 LT |
2521 | goto out; |
2522 | ||
1da177e4 LT |
2523 | /* I removed check for oif == dev_out->oif here. |
2524 | It was wrong for two reasons: | |
1ab35276 DL |
2525 | 1. ip_dev_find(net, saddr) can return wrong iface, if saddr |
2526 | is assigned to multiple interfaces. | |
1da177e4 LT |
2527 | 2. Moreover, we are allowed to send packets with saddr |
2528 | of another iface. --ANK | |
2529 | */ | |
2530 | ||
9d4fb27d JP |
2531 | if (oldflp->oif == 0 && |
2532 | (ipv4_is_multicast(oldflp->fl4_dst) || | |
2533 | oldflp->fl4_dst == htonl(0xFFFFFFFF))) { | |
a210d01a | 2534 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
0197aa38 | 2535 | dev_out = __ip_dev_find(net, oldflp->fl4_src, false); |
a210d01a JA |
2536 | if (dev_out == NULL) |
2537 | goto out; | |
2538 | ||
1da177e4 LT |
2539 | /* Special hack: user can direct multicasts |
2540 | and limited broadcast via necessary interface | |
2541 | without fiddling with IP_MULTICAST_IF or IP_PKTINFO. | |
2542 | This hack is not just for fun, it allows | |
2543 | vic,vat and friends to work. | |
2544 | They bind socket to loopback, set ttl to zero | |
2545 | and expect that it will work. | |
2546 | From the viewpoint of routing cache they are broken, | |
2547 | because we are not allowed to build multicast path | |
2548 | with loopback source addr (look, routing cache | |
2549 | cannot know, that ttl is zero, so that packet | |
2550 | will not leave this host and route is valid). | |
2551 | Luckily, this hack is good workaround. | |
2552 | */ | |
2553 | ||
2554 | fl.oif = dev_out->ifindex; | |
2555 | goto make_route; | |
2556 | } | |
a210d01a JA |
2557 | |
2558 | if (!(oldflp->flags & FLOWI_FLAG_ANYSRC)) { | |
2559 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ | |
0197aa38 | 2560 | if (!__ip_dev_find(net, oldflp->fl4_src, false)) |
a210d01a | 2561 | goto out; |
a210d01a | 2562 | } |
1da177e4 LT |
2563 | } |
2564 | ||
2565 | ||
2566 | if (oldflp->oif) { | |
0197aa38 | 2567 | dev_out = dev_get_by_index_rcu(net, oldflp->oif); |
1da177e4 LT |
2568 | err = -ENODEV; |
2569 | if (dev_out == NULL) | |
2570 | goto out; | |
e5ed6399 HX |
2571 | |
2572 | /* RACE: Check return value of inet_select_addr instead. */ | |
0197aa38 | 2573 | if (rcu_dereference(dev_out->ip_ptr) == NULL) |
1da177e4 | 2574 | goto out; /* Wrong error code */ |
1da177e4 | 2575 | |
f97c1e0c JP |
2576 | if (ipv4_is_local_multicast(oldflp->fl4_dst) || |
2577 | oldflp->fl4_dst == htonl(0xFFFFFFFF)) { | |
1da177e4 LT |
2578 | if (!fl.fl4_src) |
2579 | fl.fl4_src = inet_select_addr(dev_out, 0, | |
2580 | RT_SCOPE_LINK); | |
2581 | goto make_route; | |
2582 | } | |
2583 | if (!fl.fl4_src) { | |
f97c1e0c | 2584 | if (ipv4_is_multicast(oldflp->fl4_dst)) |
1da177e4 LT |
2585 | fl.fl4_src = inet_select_addr(dev_out, 0, |
2586 | fl.fl4_scope); | |
2587 | else if (!oldflp->fl4_dst) | |
2588 | fl.fl4_src = inet_select_addr(dev_out, 0, | |
2589 | RT_SCOPE_HOST); | |
2590 | } | |
2591 | } | |
2592 | ||
2593 | if (!fl.fl4_dst) { | |
2594 | fl.fl4_dst = fl.fl4_src; | |
2595 | if (!fl.fl4_dst) | |
2596 | fl.fl4_dst = fl.fl4_src = htonl(INADDR_LOOPBACK); | |
b40afd0e | 2597 | dev_out = net->loopback_dev; |
b40afd0e | 2598 | fl.oif = net->loopback_dev->ifindex; |
1da177e4 LT |
2599 | res.type = RTN_LOCAL; |
2600 | flags |= RTCF_LOCAL; | |
2601 | goto make_route; | |
2602 | } | |
2603 | ||
b40afd0e | 2604 | if (fib_lookup(net, &fl, &res)) { |
1da177e4 LT |
2605 | res.fi = NULL; |
2606 | if (oldflp->oif) { | |
2607 | /* Apparently, routing tables are wrong. Assume, | |
2608 | that the destination is on link. | |
2609 | ||
2610 | WHY? DW. | |
2611 | Because we are allowed to send to iface | |
2612 | even if it has NO routes and NO assigned | |
2613 | addresses. When oif is specified, routing | |
2614 | tables are looked up with only one purpose: | |
2615 | to catch if destination is gatewayed, rather than | |
2616 | direct. Moreover, if MSG_DONTROUTE is set, | |
2617 | we send packet, ignoring both routing tables | |
2618 | and ifaddr state. --ANK | |
2619 | ||
2620 | ||
2621 | We could make it even if oif is unknown, | |
2622 | likely IPv6, but we do not. | |
2623 | */ | |
2624 | ||
2625 | if (fl.fl4_src == 0) | |
2626 | fl.fl4_src = inet_select_addr(dev_out, 0, | |
2627 | RT_SCOPE_LINK); | |
2628 | res.type = RTN_UNICAST; | |
2629 | goto make_route; | |
2630 | } | |
1da177e4 LT |
2631 | err = -ENETUNREACH; |
2632 | goto out; | |
2633 | } | |
1da177e4 LT |
2634 | |
2635 | if (res.type == RTN_LOCAL) { | |
2636 | if (!fl.fl4_src) | |
2637 | fl.fl4_src = fl.fl4_dst; | |
b40afd0e | 2638 | dev_out = net->loopback_dev; |
1da177e4 | 2639 | fl.oif = dev_out->ifindex; |
1da177e4 LT |
2640 | res.fi = NULL; |
2641 | flags |= RTCF_LOCAL; | |
2642 | goto make_route; | |
2643 | } | |
2644 | ||
2645 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
2646 | if (res.fi->fib_nhs > 1 && fl.oif == 0) | |
2647 | fib_select_multipath(&fl, &res); | |
2648 | else | |
2649 | #endif | |
2650 | if (!res.prefixlen && res.type == RTN_UNICAST && !fl.oif) | |
b40afd0e | 2651 | fib_select_default(net, &fl, &res); |
1da177e4 LT |
2652 | |
2653 | if (!fl.fl4_src) | |
2654 | fl.fl4_src = FIB_RES_PREFSRC(res); | |
2655 | ||
1da177e4 | 2656 | dev_out = FIB_RES_DEV(res); |
1da177e4 LT |
2657 | fl.oif = dev_out->ifindex; |
2658 | ||
2659 | ||
2660 | make_route: | |
2661 | err = ip_mkroute_output(rp, &res, &fl, oldflp, dev_out, flags); | |
2662 | ||
1da177e4 LT |
2663 | out: return err; |
2664 | } | |
2665 | ||
611c183e DL |
2666 | int __ip_route_output_key(struct net *net, struct rtable **rp, |
2667 | const struct flowi *flp) | |
1da177e4 | 2668 | { |
0197aa38 ED |
2669 | unsigned int hash; |
2670 | int res; | |
1da177e4 LT |
2671 | struct rtable *rth; |
2672 | ||
1080d709 NH |
2673 | if (!rt_caching(net)) |
2674 | goto slow_output; | |
2675 | ||
e84f84f2 | 2676 | hash = rt_hash(flp->fl4_dst, flp->fl4_src, flp->oif, rt_genid(net)); |
1da177e4 LT |
2677 | |
2678 | rcu_read_lock_bh(); | |
a898def2 | 2679 | for (rth = rcu_dereference_bh(rt_hash_table[hash].chain); rth; |
d8d1f30b | 2680 | rth = rcu_dereference_bh(rth->dst.rt_next)) { |
1da177e4 LT |
2681 | if (rth->fl.fl4_dst == flp->fl4_dst && |
2682 | rth->fl.fl4_src == flp->fl4_src && | |
2683 | rth->fl.iif == 0 && | |
2684 | rth->fl.oif == flp->oif && | |
47dcf0cb | 2685 | rth->fl.mark == flp->mark && |
1da177e4 | 2686 | !((rth->fl.fl4_tos ^ flp->fl4_tos) & |
b5921910 | 2687 | (IPTOS_RT_MASK | RTO_ONLINK)) && |
d8d1f30b | 2688 | net_eq(dev_net(rth->dst.dev), net) && |
e84f84f2 | 2689 | !rt_is_expired(rth)) { |
d8d1f30b | 2690 | dst_use(&rth->dst, jiffies); |
1da177e4 LT |
2691 | RT_CACHE_STAT_INC(out_hit); |
2692 | rcu_read_unlock_bh(); | |
2693 | *rp = rth; | |
2694 | return 0; | |
2695 | } | |
2696 | RT_CACHE_STAT_INC(out_hlist_search); | |
2697 | } | |
2698 | rcu_read_unlock_bh(); | |
2699 | ||
1080d709 | 2700 | slow_output: |
0197aa38 ED |
2701 | rcu_read_lock(); |
2702 | res = ip_route_output_slow(net, rp, flp); | |
2703 | rcu_read_unlock(); | |
2704 | return res; | |
1da177e4 | 2705 | } |
d8c97a94 ACM |
2706 | EXPORT_SYMBOL_GPL(__ip_route_output_key); |
2707 | ||
ae2688d5 JW |
2708 | static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie) |
2709 | { | |
2710 | return NULL; | |
2711 | } | |
2712 | ||
14e50e57 DM |
2713 | static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu) |
2714 | { | |
2715 | } | |
2716 | ||
2717 | static struct dst_ops ipv4_dst_blackhole_ops = { | |
2718 | .family = AF_INET, | |
09640e63 | 2719 | .protocol = cpu_to_be16(ETH_P_IP), |
14e50e57 | 2720 | .destroy = ipv4_dst_destroy, |
ae2688d5 | 2721 | .check = ipv4_blackhole_dst_check, |
14e50e57 | 2722 | .update_pmtu = ipv4_rt_blackhole_update_pmtu, |
14e50e57 DM |
2723 | }; |
2724 | ||
2725 | ||
e84f84f2 | 2726 | static int ipv4_dst_blackhole(struct net *net, struct rtable **rp, struct flowi *flp) |
14e50e57 DM |
2727 | { |
2728 | struct rtable *ort = *rp; | |
2729 | struct rtable *rt = (struct rtable *) | |
2730 | dst_alloc(&ipv4_dst_blackhole_ops); | |
2731 | ||
2732 | if (rt) { | |
d8d1f30b | 2733 | struct dst_entry *new = &rt->dst; |
14e50e57 DM |
2734 | |
2735 | atomic_set(&new->__refcnt, 1); | |
2736 | new->__use = 1; | |
352e512c HX |
2737 | new->input = dst_discard; |
2738 | new->output = dst_discard; | |
d8d1f30b | 2739 | memcpy(new->metrics, ort->dst.metrics, RTAX_MAX*sizeof(u32)); |
14e50e57 | 2740 | |
d8d1f30b | 2741 | new->dev = ort->dst.dev; |
14e50e57 DM |
2742 | if (new->dev) |
2743 | dev_hold(new->dev); | |
2744 | ||
2745 | rt->fl = ort->fl; | |
2746 | ||
2747 | rt->idev = ort->idev; | |
2748 | if (rt->idev) | |
2749 | in_dev_hold(rt->idev); | |
e84f84f2 | 2750 | rt->rt_genid = rt_genid(net); |
14e50e57 DM |
2751 | rt->rt_flags = ort->rt_flags; |
2752 | rt->rt_type = ort->rt_type; | |
2753 | rt->rt_dst = ort->rt_dst; | |
2754 | rt->rt_src = ort->rt_src; | |
2755 | rt->rt_iif = ort->rt_iif; | |
2756 | rt->rt_gateway = ort->rt_gateway; | |
2757 | rt->rt_spec_dst = ort->rt_spec_dst; | |
2758 | rt->peer = ort->peer; | |
2759 | if (rt->peer) | |
2760 | atomic_inc(&rt->peer->refcnt); | |
2761 | ||
2762 | dst_free(new); | |
2763 | } | |
2764 | ||
d8d1f30b | 2765 | dst_release(&(*rp)->dst); |
14e50e57 | 2766 | *rp = rt; |
a02cec21 | 2767 | return rt ? 0 : -ENOMEM; |
14e50e57 DM |
2768 | } |
2769 | ||
f1b050bf DL |
2770 | int ip_route_output_flow(struct net *net, struct rtable **rp, struct flowi *flp, |
2771 | struct sock *sk, int flags) | |
1da177e4 LT |
2772 | { |
2773 | int err; | |
2774 | ||
f1b050bf | 2775 | if ((err = __ip_route_output_key(net, rp, flp)) != 0) |
1da177e4 LT |
2776 | return err; |
2777 | ||
2778 | if (flp->proto) { | |
2779 | if (!flp->fl4_src) | |
2780 | flp->fl4_src = (*rp)->rt_src; | |
2781 | if (!flp->fl4_dst) | |
2782 | flp->fl4_dst = (*rp)->rt_dst; | |
52479b62 | 2783 | err = __xfrm_lookup(net, (struct dst_entry **)rp, flp, sk, |
bb72845e | 2784 | flags ? XFRM_LOOKUP_WAIT : 0); |
14e50e57 | 2785 | if (err == -EREMOTE) |
e84f84f2 | 2786 | err = ipv4_dst_blackhole(net, rp, flp); |
14e50e57 DM |
2787 | |
2788 | return err; | |
1da177e4 LT |
2789 | } |
2790 | ||
2791 | return 0; | |
2792 | } | |
d8c97a94 ACM |
2793 | EXPORT_SYMBOL_GPL(ip_route_output_flow); |
2794 | ||
f206351a | 2795 | int ip_route_output_key(struct net *net, struct rtable **rp, struct flowi *flp) |
1da177e4 | 2796 | { |
f206351a | 2797 | return ip_route_output_flow(net, rp, flp, NULL, 0); |
1da177e4 | 2798 | } |
4bc2f18b | 2799 | EXPORT_SYMBOL(ip_route_output_key); |
1da177e4 | 2800 | |
4feb88e5 BT |
2801 | static int rt_fill_info(struct net *net, |
2802 | struct sk_buff *skb, u32 pid, u32 seq, int event, | |
b6544c0b | 2803 | int nowait, unsigned int flags) |
1da177e4 | 2804 | { |
511c3f92 | 2805 | struct rtable *rt = skb_rtable(skb); |
1da177e4 | 2806 | struct rtmsg *r; |
be403ea1 | 2807 | struct nlmsghdr *nlh; |
e3703b3d TG |
2808 | long expires; |
2809 | u32 id = 0, ts = 0, tsage = 0, error; | |
be403ea1 TG |
2810 | |
2811 | nlh = nlmsg_put(skb, pid, seq, event, sizeof(*r), flags); | |
2812 | if (nlh == NULL) | |
26932566 | 2813 | return -EMSGSIZE; |
be403ea1 TG |
2814 | |
2815 | r = nlmsg_data(nlh); | |
1da177e4 LT |
2816 | r->rtm_family = AF_INET; |
2817 | r->rtm_dst_len = 32; | |
2818 | r->rtm_src_len = 0; | |
2819 | r->rtm_tos = rt->fl.fl4_tos; | |
2820 | r->rtm_table = RT_TABLE_MAIN; | |
be403ea1 | 2821 | NLA_PUT_U32(skb, RTA_TABLE, RT_TABLE_MAIN); |
1da177e4 LT |
2822 | r->rtm_type = rt->rt_type; |
2823 | r->rtm_scope = RT_SCOPE_UNIVERSE; | |
2824 | r->rtm_protocol = RTPROT_UNSPEC; | |
2825 | r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED; | |
2826 | if (rt->rt_flags & RTCF_NOTIFY) | |
2827 | r->rtm_flags |= RTM_F_NOTIFY; | |
be403ea1 | 2828 | |
17fb2c64 | 2829 | NLA_PUT_BE32(skb, RTA_DST, rt->rt_dst); |
be403ea1 | 2830 | |
1da177e4 LT |
2831 | if (rt->fl.fl4_src) { |
2832 | r->rtm_src_len = 32; | |
17fb2c64 | 2833 | NLA_PUT_BE32(skb, RTA_SRC, rt->fl.fl4_src); |
1da177e4 | 2834 | } |
d8d1f30b CG |
2835 | if (rt->dst.dev) |
2836 | NLA_PUT_U32(skb, RTA_OIF, rt->dst.dev->ifindex); | |
1da177e4 | 2837 | #ifdef CONFIG_NET_CLS_ROUTE |
d8d1f30b CG |
2838 | if (rt->dst.tclassid) |
2839 | NLA_PUT_U32(skb, RTA_FLOW, rt->dst.tclassid); | |
1da177e4 LT |
2840 | #endif |
2841 | if (rt->fl.iif) | |
17fb2c64 | 2842 | NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_spec_dst); |
1da177e4 | 2843 | else if (rt->rt_src != rt->fl.fl4_src) |
17fb2c64 | 2844 | NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_src); |
be403ea1 | 2845 | |
1da177e4 | 2846 | if (rt->rt_dst != rt->rt_gateway) |
17fb2c64 | 2847 | NLA_PUT_BE32(skb, RTA_GATEWAY, rt->rt_gateway); |
be403ea1 | 2848 | |
d8d1f30b | 2849 | if (rtnetlink_put_metrics(skb, rt->dst.metrics) < 0) |
be403ea1 TG |
2850 | goto nla_put_failure; |
2851 | ||
963bfeee ED |
2852 | if (rt->fl.mark) |
2853 | NLA_PUT_BE32(skb, RTA_MARK, rt->fl.mark); | |
2854 | ||
d8d1f30b CG |
2855 | error = rt->dst.error; |
2856 | expires = rt->dst.expires ? rt->dst.expires - jiffies : 0; | |
1da177e4 | 2857 | if (rt->peer) { |
317fe0e6 | 2858 | inet_peer_refcheck(rt->peer); |
2c1409a0 | 2859 | id = atomic_read(&rt->peer->ip_id_count) & 0xffff; |
1da177e4 | 2860 | if (rt->peer->tcp_ts_stamp) { |
e3703b3d | 2861 | ts = rt->peer->tcp_ts; |
9d729f72 | 2862 | tsage = get_seconds() - rt->peer->tcp_ts_stamp; |
1da177e4 LT |
2863 | } |
2864 | } | |
be403ea1 | 2865 | |
1da177e4 LT |
2866 | if (rt->fl.iif) { |
2867 | #ifdef CONFIG_IP_MROUTE | |
e448515c | 2868 | __be32 dst = rt->rt_dst; |
1da177e4 | 2869 | |
f97c1e0c | 2870 | if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) && |
4feb88e5 BT |
2871 | IPV4_DEVCONF_ALL(net, MC_FORWARDING)) { |
2872 | int err = ipmr_get_route(net, skb, r, nowait); | |
1da177e4 LT |
2873 | if (err <= 0) { |
2874 | if (!nowait) { | |
2875 | if (err == 0) | |
2876 | return 0; | |
be403ea1 | 2877 | goto nla_put_failure; |
1da177e4 LT |
2878 | } else { |
2879 | if (err == -EMSGSIZE) | |
be403ea1 | 2880 | goto nla_put_failure; |
e3703b3d | 2881 | error = err; |
1da177e4 LT |
2882 | } |
2883 | } | |
2884 | } else | |
2885 | #endif | |
be403ea1 | 2886 | NLA_PUT_U32(skb, RTA_IIF, rt->fl.iif); |
1da177e4 LT |
2887 | } |
2888 | ||
d8d1f30b | 2889 | if (rtnl_put_cacheinfo(skb, &rt->dst, id, ts, tsage, |
e3703b3d TG |
2890 | expires, error) < 0) |
2891 | goto nla_put_failure; | |
be403ea1 TG |
2892 | |
2893 | return nlmsg_end(skb, nlh); | |
1da177e4 | 2894 | |
be403ea1 | 2895 | nla_put_failure: |
26932566 PM |
2896 | nlmsg_cancel(skb, nlh); |
2897 | return -EMSGSIZE; | |
1da177e4 LT |
2898 | } |
2899 | ||
63f3444f | 2900 | static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg) |
1da177e4 | 2901 | { |
3b1e0a65 | 2902 | struct net *net = sock_net(in_skb->sk); |
d889ce3b TG |
2903 | struct rtmsg *rtm; |
2904 | struct nlattr *tb[RTA_MAX+1]; | |
1da177e4 | 2905 | struct rtable *rt = NULL; |
9e12bb22 AV |
2906 | __be32 dst = 0; |
2907 | __be32 src = 0; | |
2908 | u32 iif; | |
d889ce3b | 2909 | int err; |
963bfeee | 2910 | int mark; |
1da177e4 LT |
2911 | struct sk_buff *skb; |
2912 | ||
d889ce3b TG |
2913 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy); |
2914 | if (err < 0) | |
2915 | goto errout; | |
2916 | ||
2917 | rtm = nlmsg_data(nlh); | |
2918 | ||
1da177e4 | 2919 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
d889ce3b TG |
2920 | if (skb == NULL) { |
2921 | err = -ENOBUFS; | |
2922 | goto errout; | |
2923 | } | |
1da177e4 LT |
2924 | |
2925 | /* Reserve room for dummy headers, this skb can pass | |
2926 | through good chunk of routing engine. | |
2927 | */ | |
459a98ed | 2928 | skb_reset_mac_header(skb); |
c1d2bbe1 | 2929 | skb_reset_network_header(skb); |
d2c962b8 SH |
2930 | |
2931 | /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */ | |
eddc9ec5 | 2932 | ip_hdr(skb)->protocol = IPPROTO_ICMP; |
1da177e4 LT |
2933 | skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr)); |
2934 | ||
17fb2c64 AV |
2935 | src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0; |
2936 | dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0; | |
d889ce3b | 2937 | iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0; |
963bfeee | 2938 | mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0; |
1da177e4 LT |
2939 | |
2940 | if (iif) { | |
d889ce3b TG |
2941 | struct net_device *dev; |
2942 | ||
1937504d | 2943 | dev = __dev_get_by_index(net, iif); |
d889ce3b TG |
2944 | if (dev == NULL) { |
2945 | err = -ENODEV; | |
2946 | goto errout_free; | |
2947 | } | |
2948 | ||
1da177e4 LT |
2949 | skb->protocol = htons(ETH_P_IP); |
2950 | skb->dev = dev; | |
963bfeee | 2951 | skb->mark = mark; |
1da177e4 LT |
2952 | local_bh_disable(); |
2953 | err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev); | |
2954 | local_bh_enable(); | |
d889ce3b | 2955 | |
511c3f92 | 2956 | rt = skb_rtable(skb); |
d8d1f30b CG |
2957 | if (err == 0 && rt->dst.error) |
2958 | err = -rt->dst.error; | |
1da177e4 | 2959 | } else { |
d889ce3b TG |
2960 | struct flowi fl = { |
2961 | .nl_u = { | |
2962 | .ip4_u = { | |
2963 | .daddr = dst, | |
2964 | .saddr = src, | |
2965 | .tos = rtm->rtm_tos, | |
2966 | }, | |
2967 | }, | |
2968 | .oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0, | |
963bfeee | 2969 | .mark = mark, |
d889ce3b | 2970 | }; |
1937504d | 2971 | err = ip_route_output_key(net, &rt, &fl); |
1da177e4 | 2972 | } |
d889ce3b | 2973 | |
1da177e4 | 2974 | if (err) |
d889ce3b | 2975 | goto errout_free; |
1da177e4 | 2976 | |
d8d1f30b | 2977 | skb_dst_set(skb, &rt->dst); |
1da177e4 LT |
2978 | if (rtm->rtm_flags & RTM_F_NOTIFY) |
2979 | rt->rt_flags |= RTCF_NOTIFY; | |
2980 | ||
4feb88e5 | 2981 | err = rt_fill_info(net, skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, |
1937504d | 2982 | RTM_NEWROUTE, 0, 0); |
d889ce3b TG |
2983 | if (err <= 0) |
2984 | goto errout_free; | |
1da177e4 | 2985 | |
1937504d | 2986 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid); |
d889ce3b | 2987 | errout: |
2942e900 | 2988 | return err; |
1da177e4 | 2989 | |
d889ce3b | 2990 | errout_free: |
1da177e4 | 2991 | kfree_skb(skb); |
d889ce3b | 2992 | goto errout; |
1da177e4 LT |
2993 | } |
2994 | ||
2995 | int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb) | |
2996 | { | |
2997 | struct rtable *rt; | |
2998 | int h, s_h; | |
2999 | int idx, s_idx; | |
1937504d DL |
3000 | struct net *net; |
3001 | ||
3b1e0a65 | 3002 | net = sock_net(skb->sk); |
1da177e4 LT |
3003 | |
3004 | s_h = cb->args[0]; | |
d8c92830 ED |
3005 | if (s_h < 0) |
3006 | s_h = 0; | |
1da177e4 | 3007 | s_idx = idx = cb->args[1]; |
a6272665 ED |
3008 | for (h = s_h; h <= rt_hash_mask; h++, s_idx = 0) { |
3009 | if (!rt_hash_table[h].chain) | |
3010 | continue; | |
1da177e4 | 3011 | rcu_read_lock_bh(); |
a898def2 | 3012 | for (rt = rcu_dereference_bh(rt_hash_table[h].chain), idx = 0; rt; |
d8d1f30b CG |
3013 | rt = rcu_dereference_bh(rt->dst.rt_next), idx++) { |
3014 | if (!net_eq(dev_net(rt->dst.dev), net) || idx < s_idx) | |
1da177e4 | 3015 | continue; |
e84f84f2 | 3016 | if (rt_is_expired(rt)) |
29e75252 | 3017 | continue; |
d8d1f30b | 3018 | skb_dst_set_noref(skb, &rt->dst); |
4feb88e5 | 3019 | if (rt_fill_info(net, skb, NETLINK_CB(cb->skb).pid, |
e905a9ed | 3020 | cb->nlh->nlmsg_seq, RTM_NEWROUTE, |
b6544c0b | 3021 | 1, NLM_F_MULTI) <= 0) { |
adf30907 | 3022 | skb_dst_drop(skb); |
1da177e4 LT |
3023 | rcu_read_unlock_bh(); |
3024 | goto done; | |
3025 | } | |
adf30907 | 3026 | skb_dst_drop(skb); |
1da177e4 LT |
3027 | } |
3028 | rcu_read_unlock_bh(); | |
3029 | } | |
3030 | ||
3031 | done: | |
3032 | cb->args[0] = h; | |
3033 | cb->args[1] = idx; | |
3034 | return skb->len; | |
3035 | } | |
3036 | ||
3037 | void ip_rt_multicast_event(struct in_device *in_dev) | |
3038 | { | |
76e6ebfb | 3039 | rt_cache_flush(dev_net(in_dev->dev), 0); |
1da177e4 LT |
3040 | } |
3041 | ||
3042 | #ifdef CONFIG_SYSCTL | |
81c684d1 | 3043 | static int ipv4_sysctl_rtcache_flush(ctl_table *__ctl, int write, |
8d65af78 | 3044 | void __user *buffer, |
1da177e4 LT |
3045 | size_t *lenp, loff_t *ppos) |
3046 | { | |
3047 | if (write) { | |
639e104f | 3048 | int flush_delay; |
81c684d1 | 3049 | ctl_table ctl; |
39a23e75 | 3050 | struct net *net; |
639e104f | 3051 | |
81c684d1 DL |
3052 | memcpy(&ctl, __ctl, sizeof(ctl)); |
3053 | ctl.data = &flush_delay; | |
8d65af78 | 3054 | proc_dointvec(&ctl, write, buffer, lenp, ppos); |
639e104f | 3055 | |
81c684d1 | 3056 | net = (struct net *)__ctl->extra1; |
39a23e75 | 3057 | rt_cache_flush(net, flush_delay); |
1da177e4 | 3058 | return 0; |
e905a9ed | 3059 | } |
1da177e4 LT |
3060 | |
3061 | return -EINVAL; | |
3062 | } | |
3063 | ||
eeb61f71 | 3064 | static ctl_table ipv4_route_table[] = { |
1da177e4 | 3065 | { |
1da177e4 LT |
3066 | .procname = "gc_thresh", |
3067 | .data = &ipv4_dst_ops.gc_thresh, | |
3068 | .maxlen = sizeof(int), | |
3069 | .mode = 0644, | |
6d9f239a | 3070 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3071 | }, |
3072 | { | |
1da177e4 LT |
3073 | .procname = "max_size", |
3074 | .data = &ip_rt_max_size, | |
3075 | .maxlen = sizeof(int), | |
3076 | .mode = 0644, | |
6d9f239a | 3077 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3078 | }, |
3079 | { | |
3080 | /* Deprecated. Use gc_min_interval_ms */ | |
e905a9ed | 3081 | |
1da177e4 LT |
3082 | .procname = "gc_min_interval", |
3083 | .data = &ip_rt_gc_min_interval, | |
3084 | .maxlen = sizeof(int), | |
3085 | .mode = 0644, | |
6d9f239a | 3086 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3087 | }, |
3088 | { | |
1da177e4 LT |
3089 | .procname = "gc_min_interval_ms", |
3090 | .data = &ip_rt_gc_min_interval, | |
3091 | .maxlen = sizeof(int), | |
3092 | .mode = 0644, | |
6d9f239a | 3093 | .proc_handler = proc_dointvec_ms_jiffies, |
1da177e4 LT |
3094 | }, |
3095 | { | |
1da177e4 LT |
3096 | .procname = "gc_timeout", |
3097 | .data = &ip_rt_gc_timeout, | |
3098 | .maxlen = sizeof(int), | |
3099 | .mode = 0644, | |
6d9f239a | 3100 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3101 | }, |
3102 | { | |
1da177e4 LT |
3103 | .procname = "gc_interval", |
3104 | .data = &ip_rt_gc_interval, | |
3105 | .maxlen = sizeof(int), | |
3106 | .mode = 0644, | |
6d9f239a | 3107 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3108 | }, |
3109 | { | |
1da177e4 LT |
3110 | .procname = "redirect_load", |
3111 | .data = &ip_rt_redirect_load, | |
3112 | .maxlen = sizeof(int), | |
3113 | .mode = 0644, | |
6d9f239a | 3114 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3115 | }, |
3116 | { | |
1da177e4 LT |
3117 | .procname = "redirect_number", |
3118 | .data = &ip_rt_redirect_number, | |
3119 | .maxlen = sizeof(int), | |
3120 | .mode = 0644, | |
6d9f239a | 3121 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3122 | }, |
3123 | { | |
1da177e4 LT |
3124 | .procname = "redirect_silence", |
3125 | .data = &ip_rt_redirect_silence, | |
3126 | .maxlen = sizeof(int), | |
3127 | .mode = 0644, | |
6d9f239a | 3128 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3129 | }, |
3130 | { | |
1da177e4 LT |
3131 | .procname = "error_cost", |
3132 | .data = &ip_rt_error_cost, | |
3133 | .maxlen = sizeof(int), | |
3134 | .mode = 0644, | |
6d9f239a | 3135 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3136 | }, |
3137 | { | |
1da177e4 LT |
3138 | .procname = "error_burst", |
3139 | .data = &ip_rt_error_burst, | |
3140 | .maxlen = sizeof(int), | |
3141 | .mode = 0644, | |
6d9f239a | 3142 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3143 | }, |
3144 | { | |
1da177e4 LT |
3145 | .procname = "gc_elasticity", |
3146 | .data = &ip_rt_gc_elasticity, | |
3147 | .maxlen = sizeof(int), | |
3148 | .mode = 0644, | |
6d9f239a | 3149 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3150 | }, |
3151 | { | |
1da177e4 LT |
3152 | .procname = "mtu_expires", |
3153 | .data = &ip_rt_mtu_expires, | |
3154 | .maxlen = sizeof(int), | |
3155 | .mode = 0644, | |
6d9f239a | 3156 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3157 | }, |
3158 | { | |
1da177e4 LT |
3159 | .procname = "min_pmtu", |
3160 | .data = &ip_rt_min_pmtu, | |
3161 | .maxlen = sizeof(int), | |
3162 | .mode = 0644, | |
6d9f239a | 3163 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3164 | }, |
3165 | { | |
1da177e4 LT |
3166 | .procname = "min_adv_mss", |
3167 | .data = &ip_rt_min_advmss, | |
3168 | .maxlen = sizeof(int), | |
3169 | .mode = 0644, | |
6d9f239a | 3170 | .proc_handler = proc_dointvec, |
1da177e4 | 3171 | }, |
f8572d8f | 3172 | { } |
1da177e4 | 3173 | }; |
39a23e75 | 3174 | |
2f4520d3 AV |
3175 | static struct ctl_table empty[1]; |
3176 | ||
3177 | static struct ctl_table ipv4_skeleton[] = | |
3178 | { | |
f8572d8f | 3179 | { .procname = "route", |
d994af0d | 3180 | .mode = 0555, .child = ipv4_route_table}, |
f8572d8f | 3181 | { .procname = "neigh", |
d994af0d | 3182 | .mode = 0555, .child = empty}, |
2f4520d3 AV |
3183 | { } |
3184 | }; | |
3185 | ||
3186 | static __net_initdata struct ctl_path ipv4_path[] = { | |
f8572d8f EB |
3187 | { .procname = "net", }, |
3188 | { .procname = "ipv4", }, | |
39a23e75 DL |
3189 | { }, |
3190 | }; | |
3191 | ||
39a23e75 DL |
3192 | static struct ctl_table ipv4_route_flush_table[] = { |
3193 | { | |
39a23e75 DL |
3194 | .procname = "flush", |
3195 | .maxlen = sizeof(int), | |
3196 | .mode = 0200, | |
6d9f239a | 3197 | .proc_handler = ipv4_sysctl_rtcache_flush, |
39a23e75 | 3198 | }, |
f8572d8f | 3199 | { }, |
39a23e75 DL |
3200 | }; |
3201 | ||
2f4520d3 | 3202 | static __net_initdata struct ctl_path ipv4_route_path[] = { |
f8572d8f EB |
3203 | { .procname = "net", }, |
3204 | { .procname = "ipv4", }, | |
3205 | { .procname = "route", }, | |
2f4520d3 AV |
3206 | { }, |
3207 | }; | |
3208 | ||
39a23e75 DL |
3209 | static __net_init int sysctl_route_net_init(struct net *net) |
3210 | { | |
3211 | struct ctl_table *tbl; | |
3212 | ||
3213 | tbl = ipv4_route_flush_table; | |
09ad9bc7 | 3214 | if (!net_eq(net, &init_net)) { |
39a23e75 DL |
3215 | tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL); |
3216 | if (tbl == NULL) | |
3217 | goto err_dup; | |
3218 | } | |
3219 | tbl[0].extra1 = net; | |
3220 | ||
3221 | net->ipv4.route_hdr = | |
3222 | register_net_sysctl_table(net, ipv4_route_path, tbl); | |
3223 | if (net->ipv4.route_hdr == NULL) | |
3224 | goto err_reg; | |
3225 | return 0; | |
3226 | ||
3227 | err_reg: | |
3228 | if (tbl != ipv4_route_flush_table) | |
3229 | kfree(tbl); | |
3230 | err_dup: | |
3231 | return -ENOMEM; | |
3232 | } | |
3233 | ||
3234 | static __net_exit void sysctl_route_net_exit(struct net *net) | |
3235 | { | |
3236 | struct ctl_table *tbl; | |
3237 | ||
3238 | tbl = net->ipv4.route_hdr->ctl_table_arg; | |
3239 | unregister_net_sysctl_table(net->ipv4.route_hdr); | |
3240 | BUG_ON(tbl == ipv4_route_flush_table); | |
3241 | kfree(tbl); | |
3242 | } | |
3243 | ||
3244 | static __net_initdata struct pernet_operations sysctl_route_ops = { | |
3245 | .init = sysctl_route_net_init, | |
3246 | .exit = sysctl_route_net_exit, | |
3247 | }; | |
1da177e4 LT |
3248 | #endif |
3249 | ||
3ee94372 | 3250 | static __net_init int rt_genid_init(struct net *net) |
9f5e97e5 | 3251 | { |
3ee94372 NH |
3252 | get_random_bytes(&net->ipv4.rt_genid, |
3253 | sizeof(net->ipv4.rt_genid)); | |
9f5e97e5 DL |
3254 | return 0; |
3255 | } | |
3256 | ||
3ee94372 NH |
3257 | static __net_initdata struct pernet_operations rt_genid_ops = { |
3258 | .init = rt_genid_init, | |
9f5e97e5 DL |
3259 | }; |
3260 | ||
3261 | ||
1da177e4 | 3262 | #ifdef CONFIG_NET_CLS_ROUTE |
7d720c3e | 3263 | struct ip_rt_acct __percpu *ip_rt_acct __read_mostly; |
1da177e4 LT |
3264 | #endif /* CONFIG_NET_CLS_ROUTE */ |
3265 | ||
3266 | static __initdata unsigned long rhash_entries; | |
3267 | static int __init set_rhash_entries(char *str) | |
3268 | { | |
3269 | if (!str) | |
3270 | return 0; | |
3271 | rhash_entries = simple_strtoul(str, &str, 0); | |
3272 | return 1; | |
3273 | } | |
3274 | __setup("rhash_entries=", set_rhash_entries); | |
3275 | ||
3276 | int __init ip_rt_init(void) | |
3277 | { | |
424c4b70 | 3278 | int rc = 0; |
1da177e4 | 3279 | |
1da177e4 | 3280 | #ifdef CONFIG_NET_CLS_ROUTE |
0dcec8c2 | 3281 | ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct)); |
1da177e4 LT |
3282 | if (!ip_rt_acct) |
3283 | panic("IP: failed to allocate ip_rt_acct\n"); | |
1da177e4 LT |
3284 | #endif |
3285 | ||
e5d679f3 AD |
3286 | ipv4_dst_ops.kmem_cachep = |
3287 | kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0, | |
20c2df83 | 3288 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 3289 | |
14e50e57 DM |
3290 | ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep; |
3291 | ||
fc66f95c ED |
3292 | if (dst_entries_init(&ipv4_dst_ops) < 0) |
3293 | panic("IP: failed to allocate ipv4_dst_ops counter\n"); | |
3294 | ||
3295 | if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0) | |
3296 | panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n"); | |
3297 | ||
424c4b70 ED |
3298 | rt_hash_table = (struct rt_hash_bucket *) |
3299 | alloc_large_system_hash("IP route cache", | |
3300 | sizeof(struct rt_hash_bucket), | |
3301 | rhash_entries, | |
4481374c | 3302 | (totalram_pages >= 128 * 1024) ? |
18955cfc | 3303 | 15 : 17, |
8d1502de | 3304 | 0, |
424c4b70 ED |
3305 | &rt_hash_log, |
3306 | &rt_hash_mask, | |
c9503e0f | 3307 | rhash_entries ? 0 : 512 * 1024); |
22c047cc ED |
3308 | memset(rt_hash_table, 0, (rt_hash_mask + 1) * sizeof(struct rt_hash_bucket)); |
3309 | rt_hash_lock_init(); | |
1da177e4 LT |
3310 | |
3311 | ipv4_dst_ops.gc_thresh = (rt_hash_mask + 1); | |
3312 | ip_rt_max_size = (rt_hash_mask + 1) * 16; | |
3313 | ||
1da177e4 LT |
3314 | devinet_init(); |
3315 | ip_fib_init(); | |
3316 | ||
1da177e4 LT |
3317 | /* All the timers, started at system startup tend |
3318 | to synchronize. Perturb it a bit. | |
3319 | */ | |
125bb8f5 ED |
3320 | INIT_DELAYED_WORK_DEFERRABLE(&expires_work, rt_worker_func); |
3321 | expires_ljiffies = jiffies; | |
39c90ece ED |
3322 | schedule_delayed_work(&expires_work, |
3323 | net_random() % ip_rt_gc_interval + ip_rt_gc_interval); | |
1da177e4 | 3324 | |
73b38711 | 3325 | if (ip_rt_proc_init()) |
107f1634 | 3326 | printk(KERN_ERR "Unable to create route proc files\n"); |
1da177e4 LT |
3327 | #ifdef CONFIG_XFRM |
3328 | xfrm_init(); | |
a33bc5c1 | 3329 | xfrm4_init(ip_rt_max_size); |
1da177e4 | 3330 | #endif |
63f3444f TG |
3331 | rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL); |
3332 | ||
39a23e75 DL |
3333 | #ifdef CONFIG_SYSCTL |
3334 | register_pernet_subsys(&sysctl_route_ops); | |
3335 | #endif | |
3ee94372 | 3336 | register_pernet_subsys(&rt_genid_ops); |
1da177e4 LT |
3337 | return rc; |
3338 | } | |
3339 | ||
a1bc6eb4 | 3340 | #ifdef CONFIG_SYSCTL |
eeb61f71 AV |
3341 | /* |
3342 | * We really need to sanitize the damn ipv4 init order, then all | |
3343 | * this nonsense will go away. | |
3344 | */ | |
3345 | void __init ip_static_sysctl_init(void) | |
3346 | { | |
2f4520d3 | 3347 | register_sysctl_paths(ipv4_path, ipv4_skeleton); |
eeb61f71 | 3348 | } |
a1bc6eb4 | 3349 | #endif |