]>
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> | |
352e512c | 93 | #include <net/dst.h> |
457c4cbc | 94 | #include <net/net_namespace.h> |
1da177e4 LT |
95 | #include <net/protocol.h> |
96 | #include <net/ip.h> | |
97 | #include <net/route.h> | |
98 | #include <net/inetpeer.h> | |
99 | #include <net/sock.h> | |
100 | #include <net/ip_fib.h> | |
101 | #include <net/arp.h> | |
102 | #include <net/tcp.h> | |
103 | #include <net/icmp.h> | |
104 | #include <net/xfrm.h> | |
8d71740c | 105 | #include <net/netevent.h> |
63f3444f | 106 | #include <net/rtnetlink.h> |
1da177e4 LT |
107 | #ifdef CONFIG_SYSCTL |
108 | #include <linux/sysctl.h> | |
109 | #endif | |
110 | ||
111 | #define RT_FL_TOS(oldflp) \ | |
112 | ((u32)(oldflp->fl4_tos & (IPTOS_RT_MASK | RTO_ONLINK))) | |
113 | ||
114 | #define IP_MAX_MTU 0xFFF0 | |
115 | ||
116 | #define RT_GC_TIMEOUT (300*HZ) | |
117 | ||
1da177e4 | 118 | static int ip_rt_max_size; |
817bc4db SH |
119 | static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT; |
120 | static int ip_rt_gc_interval __read_mostly = 60 * HZ; | |
121 | static int ip_rt_gc_min_interval __read_mostly = HZ / 2; | |
122 | static int ip_rt_redirect_number __read_mostly = 9; | |
123 | static int ip_rt_redirect_load __read_mostly = HZ / 50; | |
124 | static int ip_rt_redirect_silence __read_mostly = ((HZ / 50) << (9 + 1)); | |
125 | static int ip_rt_error_cost __read_mostly = HZ; | |
126 | static int ip_rt_error_burst __read_mostly = 5 * HZ; | |
127 | static int ip_rt_gc_elasticity __read_mostly = 8; | |
128 | static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ; | |
129 | static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20; | |
130 | static int ip_rt_min_advmss __read_mostly = 256; | |
131 | static int ip_rt_secret_interval __read_mostly = 10 * 60 * HZ; | |
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, |
e2422970 | 162 | .entries = ATOMIC_INIT(0), |
1da177e4 LT |
163 | }; |
164 | ||
165 | #define ECN_OR_COST(class) TC_PRIO_##class | |
166 | ||
4839c52b | 167 | const __u8 ip_tos2prio[16] = { |
1da177e4 LT |
168 | TC_PRIO_BESTEFFORT, |
169 | ECN_OR_COST(FILLER), | |
170 | TC_PRIO_BESTEFFORT, | |
171 | ECN_OR_COST(BESTEFFORT), | |
172 | TC_PRIO_BULK, | |
173 | ECN_OR_COST(BULK), | |
174 | TC_PRIO_BULK, | |
175 | ECN_OR_COST(BULK), | |
176 | TC_PRIO_INTERACTIVE, | |
177 | ECN_OR_COST(INTERACTIVE), | |
178 | TC_PRIO_INTERACTIVE, | |
179 | ECN_OR_COST(INTERACTIVE), | |
180 | TC_PRIO_INTERACTIVE_BULK, | |
181 | ECN_OR_COST(INTERACTIVE_BULK), | |
182 | TC_PRIO_INTERACTIVE_BULK, | |
183 | ECN_OR_COST(INTERACTIVE_BULK) | |
184 | }; | |
185 | ||
186 | ||
187 | /* | |
188 | * Route cache. | |
189 | */ | |
190 | ||
191 | /* The locking scheme is rather straight forward: | |
192 | * | |
193 | * 1) Read-Copy Update protects the buckets of the central route hash. | |
194 | * 2) Only writers remove entries, and they hold the lock | |
195 | * as they look at rtable reference counts. | |
196 | * 3) Only readers acquire references to rtable entries, | |
197 | * they do so with atomic increments and with the | |
198 | * lock held. | |
199 | */ | |
200 | ||
201 | struct rt_hash_bucket { | |
202 | struct rtable *chain; | |
22c047cc | 203 | }; |
1080d709 | 204 | |
8a25d5de IM |
205 | #if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK) || \ |
206 | defined(CONFIG_PROVE_LOCKING) | |
22c047cc ED |
207 | /* |
208 | * Instead of using one spinlock for each rt_hash_bucket, we use a table of spinlocks | |
209 | * The size of this table is a power of two and depends on the number of CPUS. | |
62051200 | 210 | * (on lockdep we have a quite big spinlock_t, so keep the size down there) |
22c047cc | 211 | */ |
62051200 IM |
212 | #ifdef CONFIG_LOCKDEP |
213 | # define RT_HASH_LOCK_SZ 256 | |
22c047cc | 214 | #else |
62051200 IM |
215 | # if NR_CPUS >= 32 |
216 | # define RT_HASH_LOCK_SZ 4096 | |
217 | # elif NR_CPUS >= 16 | |
218 | # define RT_HASH_LOCK_SZ 2048 | |
219 | # elif NR_CPUS >= 8 | |
220 | # define RT_HASH_LOCK_SZ 1024 | |
221 | # elif NR_CPUS >= 4 | |
222 | # define RT_HASH_LOCK_SZ 512 | |
223 | # else | |
224 | # define RT_HASH_LOCK_SZ 256 | |
225 | # endif | |
22c047cc ED |
226 | #endif |
227 | ||
228 | static spinlock_t *rt_hash_locks; | |
229 | # define rt_hash_lock_addr(slot) &rt_hash_locks[(slot) & (RT_HASH_LOCK_SZ - 1)] | |
1ff1cc20 PE |
230 | |
231 | static __init void rt_hash_lock_init(void) | |
232 | { | |
233 | int i; | |
234 | ||
235 | rt_hash_locks = kmalloc(sizeof(spinlock_t) * RT_HASH_LOCK_SZ, | |
236 | GFP_KERNEL); | |
237 | if (!rt_hash_locks) | |
238 | panic("IP: failed to allocate rt_hash_locks\n"); | |
239 | ||
240 | for (i = 0; i < RT_HASH_LOCK_SZ; i++) | |
241 | spin_lock_init(&rt_hash_locks[i]); | |
242 | } | |
22c047cc ED |
243 | #else |
244 | # define rt_hash_lock_addr(slot) NULL | |
1ff1cc20 PE |
245 | |
246 | static inline void rt_hash_lock_init(void) | |
247 | { | |
248 | } | |
22c047cc | 249 | #endif |
1da177e4 | 250 | |
817bc4db SH |
251 | static struct rt_hash_bucket *rt_hash_table __read_mostly; |
252 | static unsigned rt_hash_mask __read_mostly; | |
253 | static unsigned int rt_hash_log __read_mostly; | |
1da177e4 | 254 | |
2f970d83 | 255 | static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat); |
dbd2915c | 256 | #define RT_CACHE_STAT_INC(field) \ |
bfe5d834 | 257 | (__raw_get_cpu_var(rt_cache_stat).field++) |
1da177e4 | 258 | |
b00180de DL |
259 | static inline unsigned int rt_hash(__be32 daddr, __be32 saddr, int idx, |
260 | int genid) | |
1da177e4 | 261 | { |
1294fc4a SH |
262 | return jhash_3words((__force u32)(__be32)(daddr), |
263 | (__force u32)(__be32)(saddr), | |
b00180de | 264 | idx, genid) |
29e75252 | 265 | & rt_hash_mask; |
1da177e4 LT |
266 | } |
267 | ||
e84f84f2 DL |
268 | static inline int rt_genid(struct net *net) |
269 | { | |
270 | return atomic_read(&net->ipv4.rt_genid); | |
271 | } | |
272 | ||
1da177e4 LT |
273 | #ifdef CONFIG_PROC_FS |
274 | struct rt_cache_iter_state { | |
a75e936f | 275 | struct seq_net_private p; |
1da177e4 | 276 | int bucket; |
29e75252 | 277 | int genid; |
1da177e4 LT |
278 | }; |
279 | ||
1218854a | 280 | static struct rtable *rt_cache_get_first(struct seq_file *seq) |
1da177e4 | 281 | { |
1218854a | 282 | struct rt_cache_iter_state *st = seq->private; |
1da177e4 | 283 | struct rtable *r = NULL; |
1da177e4 LT |
284 | |
285 | for (st->bucket = rt_hash_mask; st->bucket >= 0; --st->bucket) { | |
a6272665 ED |
286 | if (!rt_hash_table[st->bucket].chain) |
287 | continue; | |
1da177e4 | 288 | rcu_read_lock_bh(); |
a898def2 | 289 | r = rcu_dereference_bh(rt_hash_table[st->bucket].chain); |
29e75252 | 290 | while (r) { |
1218854a | 291 | if (dev_net(r->u.dst.dev) == seq_file_net(seq) && |
a75e936f | 292 | r->rt_genid == st->genid) |
29e75252 | 293 | return r; |
a898def2 | 294 | r = rcu_dereference_bh(r->u.dst.rt_next); |
29e75252 | 295 | } |
1da177e4 LT |
296 | rcu_read_unlock_bh(); |
297 | } | |
29e75252 | 298 | return r; |
1da177e4 LT |
299 | } |
300 | ||
1218854a | 301 | static struct rtable *__rt_cache_get_next(struct seq_file *seq, |
642d6318 | 302 | struct rtable *r) |
1da177e4 | 303 | { |
1218854a | 304 | struct rt_cache_iter_state *st = seq->private; |
a6272665 | 305 | |
093c2ca4 | 306 | r = r->u.dst.rt_next; |
1da177e4 LT |
307 | while (!r) { |
308 | rcu_read_unlock_bh(); | |
a6272665 ED |
309 | do { |
310 | if (--st->bucket < 0) | |
311 | return NULL; | |
312 | } while (!rt_hash_table[st->bucket].chain); | |
1da177e4 LT |
313 | rcu_read_lock_bh(); |
314 | r = rt_hash_table[st->bucket].chain; | |
315 | } | |
a898def2 | 316 | return rcu_dereference_bh(r); |
1da177e4 LT |
317 | } |
318 | ||
1218854a | 319 | static struct rtable *rt_cache_get_next(struct seq_file *seq, |
642d6318 DL |
320 | struct rtable *r) |
321 | { | |
1218854a YH |
322 | struct rt_cache_iter_state *st = seq->private; |
323 | while ((r = __rt_cache_get_next(seq, r)) != NULL) { | |
324 | if (dev_net(r->u.dst.dev) != seq_file_net(seq)) | |
a75e936f | 325 | continue; |
642d6318 DL |
326 | if (r->rt_genid == st->genid) |
327 | break; | |
328 | } | |
329 | return r; | |
330 | } | |
331 | ||
1218854a | 332 | static struct rtable *rt_cache_get_idx(struct seq_file *seq, loff_t pos) |
1da177e4 | 333 | { |
1218854a | 334 | struct rtable *r = rt_cache_get_first(seq); |
1da177e4 LT |
335 | |
336 | if (r) | |
1218854a | 337 | while (pos && (r = rt_cache_get_next(seq, r))) |
1da177e4 LT |
338 | --pos; |
339 | return pos ? NULL : r; | |
340 | } | |
341 | ||
342 | static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos) | |
343 | { | |
29e75252 | 344 | struct rt_cache_iter_state *st = seq->private; |
29e75252 | 345 | if (*pos) |
1218854a | 346 | return rt_cache_get_idx(seq, *pos - 1); |
e84f84f2 | 347 | st->genid = rt_genid(seq_file_net(seq)); |
29e75252 | 348 | return SEQ_START_TOKEN; |
1da177e4 LT |
349 | } |
350 | ||
351 | static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
352 | { | |
29e75252 | 353 | struct rtable *r; |
1da177e4 LT |
354 | |
355 | if (v == SEQ_START_TOKEN) | |
1218854a | 356 | r = rt_cache_get_first(seq); |
1da177e4 | 357 | else |
1218854a | 358 | r = rt_cache_get_next(seq, v); |
1da177e4 LT |
359 | ++*pos; |
360 | return r; | |
361 | } | |
362 | ||
363 | static void rt_cache_seq_stop(struct seq_file *seq, void *v) | |
364 | { | |
365 | if (v && v != SEQ_START_TOKEN) | |
366 | rcu_read_unlock_bh(); | |
367 | } | |
368 | ||
369 | static int rt_cache_seq_show(struct seq_file *seq, void *v) | |
370 | { | |
371 | if (v == SEQ_START_TOKEN) | |
372 | seq_printf(seq, "%-127s\n", | |
373 | "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t" | |
374 | "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t" | |
375 | "HHUptod\tSpecDst"); | |
376 | else { | |
377 | struct rtable *r = v; | |
5e659e4c | 378 | int len; |
1da177e4 | 379 | |
5e659e4c PE |
380 | seq_printf(seq, "%s\t%08lX\t%08lX\t%8X\t%d\t%u\t%d\t" |
381 | "%08lX\t%d\t%u\t%u\t%02X\t%d\t%1d\t%08X%n", | |
1da177e4 LT |
382 | r->u.dst.dev ? r->u.dst.dev->name : "*", |
383 | (unsigned long)r->rt_dst, (unsigned long)r->rt_gateway, | |
384 | r->rt_flags, atomic_read(&r->u.dst.__refcnt), | |
385 | r->u.dst.__use, 0, (unsigned long)r->rt_src, | |
386 | (dst_metric(&r->u.dst, RTAX_ADVMSS) ? | |
387 | (int)dst_metric(&r->u.dst, RTAX_ADVMSS) + 40 : 0), | |
388 | dst_metric(&r->u.dst, RTAX_WINDOW), | |
389 | (int)((dst_metric(&r->u.dst, RTAX_RTT) >> 3) + | |
390 | dst_metric(&r->u.dst, RTAX_RTTVAR)), | |
391 | r->fl.fl4_tos, | |
392 | r->u.dst.hh ? atomic_read(&r->u.dst.hh->hh_refcnt) : -1, | |
393 | r->u.dst.hh ? (r->u.dst.hh->hh_output == | |
394 | dev_queue_xmit) : 0, | |
5e659e4c PE |
395 | r->rt_spec_dst, &len); |
396 | ||
397 | seq_printf(seq, "%*s\n", 127 - len, ""); | |
e905a9ed YH |
398 | } |
399 | return 0; | |
1da177e4 LT |
400 | } |
401 | ||
f690808e | 402 | static const struct seq_operations rt_cache_seq_ops = { |
1da177e4 LT |
403 | .start = rt_cache_seq_start, |
404 | .next = rt_cache_seq_next, | |
405 | .stop = rt_cache_seq_stop, | |
406 | .show = rt_cache_seq_show, | |
407 | }; | |
408 | ||
409 | static int rt_cache_seq_open(struct inode *inode, struct file *file) | |
410 | { | |
a75e936f | 411 | return seq_open_net(inode, file, &rt_cache_seq_ops, |
cf7732e4 | 412 | sizeof(struct rt_cache_iter_state)); |
1da177e4 LT |
413 | } |
414 | ||
9a32144e | 415 | static const struct file_operations rt_cache_seq_fops = { |
1da177e4 LT |
416 | .owner = THIS_MODULE, |
417 | .open = rt_cache_seq_open, | |
418 | .read = seq_read, | |
419 | .llseek = seq_lseek, | |
a75e936f | 420 | .release = seq_release_net, |
1da177e4 LT |
421 | }; |
422 | ||
423 | ||
424 | static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos) | |
425 | { | |
426 | int cpu; | |
427 | ||
428 | if (*pos == 0) | |
429 | return SEQ_START_TOKEN; | |
430 | ||
0f23174a | 431 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
432 | if (!cpu_possible(cpu)) |
433 | continue; | |
434 | *pos = cpu+1; | |
2f970d83 | 435 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
436 | } |
437 | return NULL; | |
438 | } | |
439 | ||
440 | static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
441 | { | |
442 | int cpu; | |
443 | ||
0f23174a | 444 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
445 | if (!cpu_possible(cpu)) |
446 | continue; | |
447 | *pos = cpu+1; | |
2f970d83 | 448 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
449 | } |
450 | return NULL; | |
e905a9ed | 451 | |
1da177e4 LT |
452 | } |
453 | ||
454 | static void rt_cpu_seq_stop(struct seq_file *seq, void *v) | |
455 | { | |
456 | ||
457 | } | |
458 | ||
459 | static int rt_cpu_seq_show(struct seq_file *seq, void *v) | |
460 | { | |
461 | struct rt_cache_stat *st = v; | |
462 | ||
463 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 464 | 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 |
465 | return 0; |
466 | } | |
e905a9ed | 467 | |
1da177e4 LT |
468 | seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x " |
469 | " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n", | |
470 | atomic_read(&ipv4_dst_ops.entries), | |
471 | st->in_hit, | |
472 | st->in_slow_tot, | |
473 | st->in_slow_mc, | |
474 | st->in_no_route, | |
475 | st->in_brd, | |
476 | st->in_martian_dst, | |
477 | st->in_martian_src, | |
478 | ||
479 | st->out_hit, | |
480 | st->out_slow_tot, | |
e905a9ed | 481 | st->out_slow_mc, |
1da177e4 LT |
482 | |
483 | st->gc_total, | |
484 | st->gc_ignored, | |
485 | st->gc_goal_miss, | |
486 | st->gc_dst_overflow, | |
487 | st->in_hlist_search, | |
488 | st->out_hlist_search | |
489 | ); | |
490 | return 0; | |
491 | } | |
492 | ||
f690808e | 493 | static const struct seq_operations rt_cpu_seq_ops = { |
1da177e4 LT |
494 | .start = rt_cpu_seq_start, |
495 | .next = rt_cpu_seq_next, | |
496 | .stop = rt_cpu_seq_stop, | |
497 | .show = rt_cpu_seq_show, | |
498 | }; | |
499 | ||
500 | ||
501 | static int rt_cpu_seq_open(struct inode *inode, struct file *file) | |
502 | { | |
503 | return seq_open(file, &rt_cpu_seq_ops); | |
504 | } | |
505 | ||
9a32144e | 506 | static const struct file_operations rt_cpu_seq_fops = { |
1da177e4 LT |
507 | .owner = THIS_MODULE, |
508 | .open = rt_cpu_seq_open, | |
509 | .read = seq_read, | |
510 | .llseek = seq_lseek, | |
511 | .release = seq_release, | |
512 | }; | |
513 | ||
78c686e9 | 514 | #ifdef CONFIG_NET_CLS_ROUTE |
a661c419 | 515 | static int rt_acct_proc_show(struct seq_file *m, void *v) |
78c686e9 | 516 | { |
a661c419 AD |
517 | struct ip_rt_acct *dst, *src; |
518 | unsigned int i, j; | |
519 | ||
520 | dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL); | |
521 | if (!dst) | |
522 | return -ENOMEM; | |
523 | ||
524 | for_each_possible_cpu(i) { | |
525 | src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i); | |
526 | for (j = 0; j < 256; j++) { | |
527 | dst[j].o_bytes += src[j].o_bytes; | |
528 | dst[j].o_packets += src[j].o_packets; | |
529 | dst[j].i_bytes += src[j].i_bytes; | |
530 | dst[j].i_packets += src[j].i_packets; | |
531 | } | |
78c686e9 PE |
532 | } |
533 | ||
a661c419 AD |
534 | seq_write(m, dst, 256 * sizeof(struct ip_rt_acct)); |
535 | kfree(dst); | |
536 | return 0; | |
537 | } | |
78c686e9 | 538 | |
a661c419 AD |
539 | static int rt_acct_proc_open(struct inode *inode, struct file *file) |
540 | { | |
541 | return single_open(file, rt_acct_proc_show, NULL); | |
78c686e9 | 542 | } |
a661c419 AD |
543 | |
544 | static const struct file_operations rt_acct_proc_fops = { | |
545 | .owner = THIS_MODULE, | |
546 | .open = rt_acct_proc_open, | |
547 | .read = seq_read, | |
548 | .llseek = seq_lseek, | |
549 | .release = single_release, | |
550 | }; | |
78c686e9 | 551 | #endif |
107f1634 | 552 | |
73b38711 | 553 | static int __net_init ip_rt_do_proc_init(struct net *net) |
107f1634 PE |
554 | { |
555 | struct proc_dir_entry *pde; | |
556 | ||
557 | pde = proc_net_fops_create(net, "rt_cache", S_IRUGO, | |
558 | &rt_cache_seq_fops); | |
559 | if (!pde) | |
560 | goto err1; | |
561 | ||
77020720 WC |
562 | pde = proc_create("rt_cache", S_IRUGO, |
563 | net->proc_net_stat, &rt_cpu_seq_fops); | |
107f1634 PE |
564 | if (!pde) |
565 | goto err2; | |
566 | ||
107f1634 | 567 | #ifdef CONFIG_NET_CLS_ROUTE |
a661c419 | 568 | pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops); |
107f1634 PE |
569 | if (!pde) |
570 | goto err3; | |
571 | #endif | |
572 | return 0; | |
573 | ||
574 | #ifdef CONFIG_NET_CLS_ROUTE | |
575 | err3: | |
576 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
577 | #endif | |
578 | err2: | |
579 | remove_proc_entry("rt_cache", net->proc_net); | |
580 | err1: | |
581 | return -ENOMEM; | |
582 | } | |
73b38711 DL |
583 | |
584 | static void __net_exit ip_rt_do_proc_exit(struct net *net) | |
585 | { | |
586 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
587 | remove_proc_entry("rt_cache", net->proc_net); | |
0a931acf | 588 | #ifdef CONFIG_NET_CLS_ROUTE |
73b38711 | 589 | remove_proc_entry("rt_acct", net->proc_net); |
0a931acf | 590 | #endif |
73b38711 DL |
591 | } |
592 | ||
593 | static struct pernet_operations ip_rt_proc_ops __net_initdata = { | |
594 | .init = ip_rt_do_proc_init, | |
595 | .exit = ip_rt_do_proc_exit, | |
596 | }; | |
597 | ||
598 | static int __init ip_rt_proc_init(void) | |
599 | { | |
600 | return register_pernet_subsys(&ip_rt_proc_ops); | |
601 | } | |
602 | ||
107f1634 | 603 | #else |
73b38711 | 604 | static inline int ip_rt_proc_init(void) |
107f1634 PE |
605 | { |
606 | return 0; | |
607 | } | |
1da177e4 | 608 | #endif /* CONFIG_PROC_FS */ |
e905a9ed | 609 | |
5969f71d | 610 | static inline void rt_free(struct rtable *rt) |
1da177e4 | 611 | { |
1da177e4 LT |
612 | call_rcu_bh(&rt->u.dst.rcu_head, dst_rcu_free); |
613 | } | |
614 | ||
5969f71d | 615 | static inline void rt_drop(struct rtable *rt) |
1da177e4 | 616 | { |
1da177e4 LT |
617 | ip_rt_put(rt); |
618 | call_rcu_bh(&rt->u.dst.rcu_head, dst_rcu_free); | |
619 | } | |
620 | ||
5969f71d | 621 | static inline int rt_fast_clean(struct rtable *rth) |
1da177e4 LT |
622 | { |
623 | /* Kill broadcast/multicast entries very aggresively, if they | |
624 | collide in hash table with more useful entries */ | |
625 | return (rth->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && | |
093c2ca4 | 626 | rth->fl.iif && rth->u.dst.rt_next; |
1da177e4 LT |
627 | } |
628 | ||
5969f71d | 629 | static inline int rt_valuable(struct rtable *rth) |
1da177e4 LT |
630 | { |
631 | return (rth->rt_flags & (RTCF_REDIRECTED | RTCF_NOTIFY)) || | |
632 | rth->u.dst.expires; | |
633 | } | |
634 | ||
635 | static int rt_may_expire(struct rtable *rth, unsigned long tmo1, unsigned long tmo2) | |
636 | { | |
637 | unsigned long age; | |
638 | int ret = 0; | |
639 | ||
640 | if (atomic_read(&rth->u.dst.__refcnt)) | |
641 | goto out; | |
642 | ||
643 | ret = 1; | |
644 | if (rth->u.dst.expires && | |
645 | time_after_eq(jiffies, rth->u.dst.expires)) | |
646 | goto out; | |
647 | ||
648 | age = jiffies - rth->u.dst.lastuse; | |
649 | ret = 0; | |
650 | if ((age <= tmo1 && !rt_fast_clean(rth)) || | |
651 | (age <= tmo2 && rt_valuable(rth))) | |
652 | goto out; | |
653 | ret = 1; | |
654 | out: return ret; | |
655 | } | |
656 | ||
657 | /* Bits of score are: | |
658 | * 31: very valuable | |
659 | * 30: not quite useless | |
660 | * 29..0: usage counter | |
661 | */ | |
662 | static inline u32 rt_score(struct rtable *rt) | |
663 | { | |
664 | u32 score = jiffies - rt->u.dst.lastuse; | |
665 | ||
666 | score = ~score & ~(3<<30); | |
667 | ||
668 | if (rt_valuable(rt)) | |
669 | score |= (1<<31); | |
670 | ||
671 | if (!rt->fl.iif || | |
672 | !(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL))) | |
673 | score |= (1<<30); | |
674 | ||
675 | return score; | |
676 | } | |
677 | ||
1080d709 NH |
678 | static inline bool rt_caching(const struct net *net) |
679 | { | |
680 | return net->ipv4.current_rt_cache_rebuild_count <= | |
681 | net->ipv4.sysctl_rt_cache_rebuild_count; | |
682 | } | |
683 | ||
684 | static inline bool compare_hash_inputs(const struct flowi *fl1, | |
685 | const struct flowi *fl2) | |
686 | { | |
687 | return (__force u32)(((fl1->nl_u.ip4_u.daddr ^ fl2->nl_u.ip4_u.daddr) | | |
688 | (fl1->nl_u.ip4_u.saddr ^ fl2->nl_u.ip4_u.saddr) | | |
689 | (fl1->iif ^ fl2->iif)) == 0); | |
690 | } | |
691 | ||
1da177e4 LT |
692 | static inline int compare_keys(struct flowi *fl1, struct flowi *fl2) |
693 | { | |
714e85be AV |
694 | return ((__force u32)((fl1->nl_u.ip4_u.daddr ^ fl2->nl_u.ip4_u.daddr) | |
695 | (fl1->nl_u.ip4_u.saddr ^ fl2->nl_u.ip4_u.saddr)) | | |
47dcf0cb | 696 | (fl1->mark ^ fl2->mark) | |
8238b218 DM |
697 | (*(u16 *)&fl1->nl_u.ip4_u.tos ^ |
698 | *(u16 *)&fl2->nl_u.ip4_u.tos) | | |
699 | (fl1->oif ^ fl2->oif) | | |
700 | (fl1->iif ^ fl2->iif)) == 0; | |
1da177e4 LT |
701 | } |
702 | ||
b5921910 DL |
703 | static inline int compare_netns(struct rtable *rt1, struct rtable *rt2) |
704 | { | |
09ad9bc7 | 705 | return net_eq(dev_net(rt1->u.dst.dev), dev_net(rt2->u.dst.dev)); |
b5921910 DL |
706 | } |
707 | ||
e84f84f2 DL |
708 | static inline int rt_is_expired(struct rtable *rth) |
709 | { | |
710 | return rth->rt_genid != rt_genid(dev_net(rth->u.dst.dev)); | |
711 | } | |
712 | ||
beb659bd ED |
713 | /* |
714 | * Perform a full scan of hash table and free all entries. | |
715 | * Can be called by a softirq or a process. | |
716 | * In the later case, we want to be reschedule if necessary | |
717 | */ | |
718 | static void rt_do_flush(int process_context) | |
719 | { | |
720 | unsigned int i; | |
721 | struct rtable *rth, *next; | |
32cb5b4e | 722 | struct rtable * tail; |
beb659bd ED |
723 | |
724 | for (i = 0; i <= rt_hash_mask; i++) { | |
725 | if (process_context && need_resched()) | |
726 | cond_resched(); | |
727 | rth = rt_hash_table[i].chain; | |
728 | if (!rth) | |
729 | continue; | |
730 | ||
731 | spin_lock_bh(rt_hash_lock_addr(i)); | |
32cb5b4e DL |
732 | #ifdef CONFIG_NET_NS |
733 | { | |
734 | struct rtable ** prev, * p; | |
735 | ||
736 | rth = rt_hash_table[i].chain; | |
737 | ||
738 | /* defer releasing the head of the list after spin_unlock */ | |
739 | for (tail = rth; tail; tail = tail->u.dst.rt_next) | |
740 | if (!rt_is_expired(tail)) | |
741 | break; | |
742 | if (rth != tail) | |
743 | rt_hash_table[i].chain = tail; | |
744 | ||
745 | /* call rt_free on entries after the tail requiring flush */ | |
746 | prev = &rt_hash_table[i].chain; | |
747 | for (p = *prev; p; p = next) { | |
748 | next = p->u.dst.rt_next; | |
749 | if (!rt_is_expired(p)) { | |
750 | prev = &p->u.dst.rt_next; | |
751 | } else { | |
752 | *prev = next; | |
753 | rt_free(p); | |
754 | } | |
755 | } | |
756 | } | |
757 | #else | |
beb659bd ED |
758 | rth = rt_hash_table[i].chain; |
759 | rt_hash_table[i].chain = NULL; | |
32cb5b4e DL |
760 | tail = NULL; |
761 | #endif | |
beb659bd ED |
762 | spin_unlock_bh(rt_hash_lock_addr(i)); |
763 | ||
32cb5b4e | 764 | for (; rth != tail; rth = next) { |
beb659bd ED |
765 | next = rth->u.dst.rt_next; |
766 | rt_free(rth); | |
767 | } | |
768 | } | |
769 | } | |
770 | ||
1080d709 NH |
771 | /* |
772 | * While freeing expired entries, we compute average chain length | |
773 | * and standard deviation, using fixed-point arithmetic. | |
774 | * This to have an estimation of rt_chain_length_max | |
775 | * rt_chain_length_max = max(elasticity, AVG + 4*SD) | |
776 | * We use 3 bits for frational part, and 29 (or 61) for magnitude. | |
777 | */ | |
778 | ||
779 | #define FRACT_BITS 3 | |
780 | #define ONE (1UL << FRACT_BITS) | |
781 | ||
98376387 ED |
782 | /* |
783 | * Given a hash chain and an item in this hash chain, | |
784 | * find if a previous entry has the same hash_inputs | |
785 | * (but differs on tos, mark or oif) | |
786 | * Returns 0 if an alias is found. | |
787 | * Returns ONE if rth has no alias before itself. | |
788 | */ | |
789 | static int has_noalias(const struct rtable *head, const struct rtable *rth) | |
790 | { | |
791 | const struct rtable *aux = head; | |
792 | ||
793 | while (aux != rth) { | |
794 | if (compare_hash_inputs(&aux->fl, &rth->fl)) | |
795 | return 0; | |
796 | aux = aux->u.dst.rt_next; | |
797 | } | |
798 | return ONE; | |
799 | } | |
800 | ||
beb659bd | 801 | static void rt_check_expire(void) |
1da177e4 | 802 | { |
bb1d23b0 ED |
803 | static unsigned int rover; |
804 | unsigned int i = rover, goal; | |
98376387 | 805 | struct rtable *rth, **rthp; |
cf8da764 | 806 | unsigned long samples = 0; |
1080d709 | 807 | unsigned long sum = 0, sum2 = 0; |
125bb8f5 | 808 | unsigned long delta; |
bb1d23b0 ED |
809 | u64 mult; |
810 | ||
125bb8f5 ED |
811 | delta = jiffies - expires_ljiffies; |
812 | expires_ljiffies = jiffies; | |
813 | mult = ((u64)delta) << rt_hash_log; | |
bb1d23b0 ED |
814 | if (ip_rt_gc_timeout > 1) |
815 | do_div(mult, ip_rt_gc_timeout); | |
816 | goal = (unsigned int)mult; | |
39c90ece ED |
817 | if (goal > rt_hash_mask) |
818 | goal = rt_hash_mask + 1; | |
bb1d23b0 | 819 | for (; goal > 0; goal--) { |
1da177e4 | 820 | unsigned long tmo = ip_rt_gc_timeout; |
cf8da764 | 821 | unsigned long length; |
1da177e4 LT |
822 | |
823 | i = (i + 1) & rt_hash_mask; | |
824 | rthp = &rt_hash_table[i].chain; | |
825 | ||
d90bf5a9 ED |
826 | if (need_resched()) |
827 | cond_resched(); | |
828 | ||
1080d709 NH |
829 | samples++; |
830 | ||
cfcabdcc | 831 | if (*rthp == NULL) |
bb1d23b0 | 832 | continue; |
cf8da764 | 833 | length = 0; |
39c90ece | 834 | spin_lock_bh(rt_hash_lock_addr(i)); |
1da177e4 | 835 | while ((rth = *rthp) != NULL) { |
1ddbcb00 | 836 | prefetch(rth->u.dst.rt_next); |
e84f84f2 | 837 | if (rt_is_expired(rth)) { |
29e75252 ED |
838 | *rthp = rth->u.dst.rt_next; |
839 | rt_free(rth); | |
840 | continue; | |
841 | } | |
1da177e4 LT |
842 | if (rth->u.dst.expires) { |
843 | /* Entry is expired even if it is in use */ | |
39c90ece | 844 | if (time_before_eq(jiffies, rth->u.dst.expires)) { |
1ddbcb00 | 845 | nofree: |
1da177e4 | 846 | tmo >>= 1; |
093c2ca4 | 847 | rthp = &rth->u.dst.rt_next; |
1080d709 | 848 | /* |
1ddbcb00 | 849 | * We only count entries on |
1080d709 NH |
850 | * a chain with equal hash inputs once |
851 | * so that entries for different QOS | |
852 | * levels, and other non-hash input | |
853 | * attributes don't unfairly skew | |
854 | * the length computation | |
855 | */ | |
98376387 | 856 | length += has_noalias(rt_hash_table[i].chain, rth); |
1da177e4 LT |
857 | continue; |
858 | } | |
1ddbcb00 ED |
859 | } else if (!rt_may_expire(rth, tmo, ip_rt_gc_timeout)) |
860 | goto nofree; | |
1da177e4 LT |
861 | |
862 | /* Cleanup aged off entries. */ | |
093c2ca4 | 863 | *rthp = rth->u.dst.rt_next; |
e905a9ed | 864 | rt_free(rth); |
1da177e4 | 865 | } |
39c90ece | 866 | spin_unlock_bh(rt_hash_lock_addr(i)); |
1080d709 NH |
867 | sum += length; |
868 | sum2 += length*length; | |
869 | } | |
870 | if (samples) { | |
871 | unsigned long avg = sum / samples; | |
872 | unsigned long sd = int_sqrt(sum2 / samples - avg*avg); | |
873 | rt_chain_length_max = max_t(unsigned long, | |
874 | ip_rt_gc_elasticity, | |
875 | (avg + 4*sd) >> FRACT_BITS); | |
1da177e4 LT |
876 | } |
877 | rover = i; | |
beb659bd ED |
878 | } |
879 | ||
880 | /* | |
881 | * rt_worker_func() is run in process context. | |
29e75252 | 882 | * we call rt_check_expire() to scan part of the hash table |
beb659bd ED |
883 | */ |
884 | static void rt_worker_func(struct work_struct *work) | |
885 | { | |
29e75252 | 886 | rt_check_expire(); |
39c90ece | 887 | schedule_delayed_work(&expires_work, ip_rt_gc_interval); |
1da177e4 LT |
888 | } |
889 | ||
29e75252 ED |
890 | /* |
891 | * Pertubation of rt_genid by a small quantity [1..256] | |
892 | * Using 8 bits of shuffling ensure we can call rt_cache_invalidate() | |
893 | * many times (2^24) without giving recent rt_genid. | |
894 | * Jenkins hash is strong enough that litle changes of rt_genid are OK. | |
1da177e4 | 895 | */ |
86c657f6 | 896 | static void rt_cache_invalidate(struct net *net) |
1da177e4 | 897 | { |
29e75252 | 898 | unsigned char shuffle; |
1da177e4 | 899 | |
29e75252 | 900 | get_random_bytes(&shuffle, sizeof(shuffle)); |
e84f84f2 | 901 | atomic_add(shuffle + 1U, &net->ipv4.rt_genid); |
1da177e4 LT |
902 | } |
903 | ||
29e75252 ED |
904 | /* |
905 | * delay < 0 : invalidate cache (fast : entries will be deleted later) | |
906 | * delay >= 0 : invalidate & flush cache (can be long) | |
907 | */ | |
76e6ebfb | 908 | void rt_cache_flush(struct net *net, int delay) |
1da177e4 | 909 | { |
86c657f6 | 910 | rt_cache_invalidate(net); |
29e75252 ED |
911 | if (delay >= 0) |
912 | rt_do_flush(!in_softirq()); | |
1da177e4 LT |
913 | } |
914 | ||
a5ee1551 EB |
915 | /* Flush previous cache invalidated entries from the cache */ |
916 | void rt_cache_flush_batch(void) | |
917 | { | |
918 | rt_do_flush(!in_softirq()); | |
919 | } | |
920 | ||
beb659bd | 921 | /* |
29e75252 | 922 | * We change rt_genid and let gc do the cleanup |
beb659bd | 923 | */ |
9f5e97e5 | 924 | static void rt_secret_rebuild(unsigned long __net) |
1da177e4 | 925 | { |
9f5e97e5 | 926 | struct net *net = (struct net *)__net; |
86c657f6 | 927 | rt_cache_invalidate(net); |
9f5e97e5 | 928 | mod_timer(&net->ipv4.rt_secret_timer, jiffies + ip_rt_secret_interval); |
1da177e4 LT |
929 | } |
930 | ||
1080d709 NH |
931 | static void rt_secret_rebuild_oneshot(struct net *net) |
932 | { | |
933 | del_timer_sync(&net->ipv4.rt_secret_timer); | |
934 | rt_cache_invalidate(net); | |
858a18a6 VG |
935 | if (ip_rt_secret_interval) |
936 | mod_timer(&net->ipv4.rt_secret_timer, jiffies + ip_rt_secret_interval); | |
1080d709 NH |
937 | } |
938 | ||
939 | static void rt_emergency_hash_rebuild(struct net *net) | |
940 | { | |
941 | if (net_ratelimit()) { | |
942 | printk(KERN_WARNING "Route hash chain too long!\n"); | |
943 | printk(KERN_WARNING "Adjust your secret_interval!\n"); | |
944 | } | |
945 | ||
946 | rt_secret_rebuild_oneshot(net); | |
947 | } | |
948 | ||
1da177e4 LT |
949 | /* |
950 | Short description of GC goals. | |
951 | ||
952 | We want to build algorithm, which will keep routing cache | |
953 | at some equilibrium point, when number of aged off entries | |
954 | is kept approximately equal to newly generated ones. | |
955 | ||
956 | Current expiration strength is variable "expire". | |
957 | We try to adjust it dynamically, so that if networking | |
958 | is idle expires is large enough to keep enough of warm entries, | |
959 | and when load increases it reduces to limit cache size. | |
960 | */ | |
961 | ||
569d3645 | 962 | static int rt_garbage_collect(struct dst_ops *ops) |
1da177e4 LT |
963 | { |
964 | static unsigned long expire = RT_GC_TIMEOUT; | |
965 | static unsigned long last_gc; | |
966 | static int rover; | |
967 | static int equilibrium; | |
968 | struct rtable *rth, **rthp; | |
969 | unsigned long now = jiffies; | |
970 | int goal; | |
971 | ||
972 | /* | |
973 | * Garbage collection is pretty expensive, | |
974 | * do not make it too frequently. | |
975 | */ | |
976 | ||
977 | RT_CACHE_STAT_INC(gc_total); | |
978 | ||
979 | if (now - last_gc < ip_rt_gc_min_interval && | |
980 | atomic_read(&ipv4_dst_ops.entries) < ip_rt_max_size) { | |
981 | RT_CACHE_STAT_INC(gc_ignored); | |
982 | goto out; | |
983 | } | |
984 | ||
985 | /* Calculate number of entries, which we want to expire now. */ | |
986 | goal = atomic_read(&ipv4_dst_ops.entries) - | |
987 | (ip_rt_gc_elasticity << rt_hash_log); | |
988 | if (goal <= 0) { | |
989 | if (equilibrium < ipv4_dst_ops.gc_thresh) | |
990 | equilibrium = ipv4_dst_ops.gc_thresh; | |
991 | goal = atomic_read(&ipv4_dst_ops.entries) - equilibrium; | |
992 | if (goal > 0) { | |
b790cedd | 993 | equilibrium += min_t(unsigned int, goal >> 1, rt_hash_mask + 1); |
1da177e4 LT |
994 | goal = atomic_read(&ipv4_dst_ops.entries) - equilibrium; |
995 | } | |
996 | } else { | |
997 | /* We are in dangerous area. Try to reduce cache really | |
998 | * aggressively. | |
999 | */ | |
b790cedd | 1000 | goal = max_t(unsigned int, goal >> 1, rt_hash_mask + 1); |
1da177e4 LT |
1001 | equilibrium = atomic_read(&ipv4_dst_ops.entries) - goal; |
1002 | } | |
1003 | ||
1004 | if (now - last_gc >= ip_rt_gc_min_interval) | |
1005 | last_gc = now; | |
1006 | ||
1007 | if (goal <= 0) { | |
1008 | equilibrium += goal; | |
1009 | goto work_done; | |
1010 | } | |
1011 | ||
1012 | do { | |
1013 | int i, k; | |
1014 | ||
1015 | for (i = rt_hash_mask, k = rover; i >= 0; i--) { | |
1016 | unsigned long tmo = expire; | |
1017 | ||
1018 | k = (k + 1) & rt_hash_mask; | |
1019 | rthp = &rt_hash_table[k].chain; | |
22c047cc | 1020 | spin_lock_bh(rt_hash_lock_addr(k)); |
1da177e4 | 1021 | while ((rth = *rthp) != NULL) { |
e84f84f2 | 1022 | if (!rt_is_expired(rth) && |
29e75252 | 1023 | !rt_may_expire(rth, tmo, expire)) { |
1da177e4 | 1024 | tmo >>= 1; |
093c2ca4 | 1025 | rthp = &rth->u.dst.rt_next; |
1da177e4 LT |
1026 | continue; |
1027 | } | |
093c2ca4 | 1028 | *rthp = rth->u.dst.rt_next; |
1da177e4 LT |
1029 | rt_free(rth); |
1030 | goal--; | |
1da177e4 | 1031 | } |
22c047cc | 1032 | spin_unlock_bh(rt_hash_lock_addr(k)); |
1da177e4 LT |
1033 | if (goal <= 0) |
1034 | break; | |
1035 | } | |
1036 | rover = k; | |
1037 | ||
1038 | if (goal <= 0) | |
1039 | goto work_done; | |
1040 | ||
1041 | /* Goal is not achieved. We stop process if: | |
1042 | ||
1043 | - if expire reduced to zero. Otherwise, expire is halfed. | |
1044 | - if table is not full. | |
1045 | - if we are called from interrupt. | |
1046 | - jiffies check is just fallback/debug loop breaker. | |
1047 | We will not spin here for long time in any case. | |
1048 | */ | |
1049 | ||
1050 | RT_CACHE_STAT_INC(gc_goal_miss); | |
1051 | ||
1052 | if (expire == 0) | |
1053 | break; | |
1054 | ||
1055 | expire >>= 1; | |
1056 | #if RT_CACHE_DEBUG >= 2 | |
1057 | printk(KERN_DEBUG "expire>> %u %d %d %d\n", expire, | |
1058 | atomic_read(&ipv4_dst_ops.entries), goal, i); | |
1059 | #endif | |
1060 | ||
1061 | if (atomic_read(&ipv4_dst_ops.entries) < ip_rt_max_size) | |
1062 | goto out; | |
1063 | } while (!in_softirq() && time_before_eq(jiffies, now)); | |
1064 | ||
1065 | if (atomic_read(&ipv4_dst_ops.entries) < ip_rt_max_size) | |
1066 | goto out; | |
1067 | if (net_ratelimit()) | |
1068 | printk(KERN_WARNING "dst cache overflow\n"); | |
1069 | RT_CACHE_STAT_INC(gc_dst_overflow); | |
1070 | return 1; | |
1071 | ||
1072 | work_done: | |
1073 | expire += ip_rt_gc_min_interval; | |
1074 | if (expire > ip_rt_gc_timeout || | |
1075 | atomic_read(&ipv4_dst_ops.entries) < ipv4_dst_ops.gc_thresh) | |
1076 | expire = ip_rt_gc_timeout; | |
1077 | #if RT_CACHE_DEBUG >= 2 | |
1078 | printk(KERN_DEBUG "expire++ %u %d %d %d\n", expire, | |
1079 | atomic_read(&ipv4_dst_ops.entries), goal, rover); | |
1080 | #endif | |
1081 | out: return 0; | |
1082 | } | |
1083 | ||
98376387 ED |
1084 | /* |
1085 | * Returns number of entries in a hash chain that have different hash_inputs | |
1086 | */ | |
1087 | static int slow_chain_length(const struct rtable *head) | |
1088 | { | |
1089 | int length = 0; | |
1090 | const struct rtable *rth = head; | |
1091 | ||
1092 | while (rth) { | |
1093 | length += has_noalias(head, rth); | |
1094 | rth = rth->u.dst.rt_next; | |
1095 | } | |
1096 | return length >> FRACT_BITS; | |
1097 | } | |
1098 | ||
511c3f92 ED |
1099 | static int rt_intern_hash(unsigned hash, struct rtable *rt, |
1100 | struct rtable **rp, struct sk_buff *skb) | |
1da177e4 LT |
1101 | { |
1102 | struct rtable *rth, **rthp; | |
1103 | unsigned long now; | |
1104 | struct rtable *cand, **candp; | |
1105 | u32 min_score; | |
1106 | int chain_length; | |
1107 | int attempts = !in_softirq(); | |
1108 | ||
1109 | restart: | |
1110 | chain_length = 0; | |
1111 | min_score = ~(u32)0; | |
1112 | cand = NULL; | |
1113 | candp = NULL; | |
1114 | now = jiffies; | |
1115 | ||
1080d709 | 1116 | if (!rt_caching(dev_net(rt->u.dst.dev))) { |
73e42897 NH |
1117 | /* |
1118 | * If we're not caching, just tell the caller we | |
1119 | * were successful and don't touch the route. The | |
1120 | * caller hold the sole reference to the cache entry, and | |
1121 | * it will be released when the caller is done with it. | |
1122 | * If we drop it here, the callers have no way to resolve routes | |
1123 | * when we're not caching. Instead, just point *rp at rt, so | |
1124 | * the caller gets a single use out of the route | |
b6280b47 NH |
1125 | * Note that we do rt_free on this new route entry, so that |
1126 | * once its refcount hits zero, we are still able to reap it | |
1127 | * (Thanks Alexey) | |
1128 | * Note also the rt_free uses call_rcu. We don't actually | |
1129 | * need rcu protection here, this is just our path to get | |
1130 | * on the route gc list. | |
73e42897 | 1131 | */ |
b6280b47 NH |
1132 | |
1133 | if (rt->rt_type == RTN_UNICAST || rt->fl.iif == 0) { | |
1134 | int err = arp_bind_neighbour(&rt->u.dst); | |
1135 | if (err) { | |
1136 | if (net_ratelimit()) | |
1137 | printk(KERN_WARNING | |
1138 | "Neighbour table failure & not caching routes.\n"); | |
1139 | rt_drop(rt); | |
1140 | return err; | |
1141 | } | |
1142 | } | |
1143 | ||
1144 | rt_free(rt); | |
1145 | goto skip_hashing; | |
1080d709 NH |
1146 | } |
1147 | ||
1da177e4 LT |
1148 | rthp = &rt_hash_table[hash].chain; |
1149 | ||
22c047cc | 1150 | spin_lock_bh(rt_hash_lock_addr(hash)); |
1da177e4 | 1151 | while ((rth = *rthp) != NULL) { |
e84f84f2 | 1152 | if (rt_is_expired(rth)) { |
29e75252 ED |
1153 | *rthp = rth->u.dst.rt_next; |
1154 | rt_free(rth); | |
1155 | continue; | |
1156 | } | |
b5921910 | 1157 | if (compare_keys(&rth->fl, &rt->fl) && compare_netns(rth, rt)) { |
1da177e4 | 1158 | /* Put it first */ |
093c2ca4 | 1159 | *rthp = rth->u.dst.rt_next; |
1da177e4 LT |
1160 | /* |
1161 | * Since lookup is lockfree, the deletion | |
1162 | * must be visible to another weakly ordered CPU before | |
1163 | * the insertion at the start of the hash chain. | |
1164 | */ | |
093c2ca4 | 1165 | rcu_assign_pointer(rth->u.dst.rt_next, |
1da177e4 LT |
1166 | rt_hash_table[hash].chain); |
1167 | /* | |
1168 | * Since lookup is lockfree, the update writes | |
1169 | * must be ordered for consistency on SMP. | |
1170 | */ | |
1171 | rcu_assign_pointer(rt_hash_table[hash].chain, rth); | |
1172 | ||
03f49f34 | 1173 | dst_use(&rth->u.dst, now); |
22c047cc | 1174 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1175 | |
1176 | rt_drop(rt); | |
511c3f92 ED |
1177 | if (rp) |
1178 | *rp = rth; | |
1179 | else | |
adf30907 | 1180 | skb_dst_set(skb, &rth->u.dst); |
1da177e4 LT |
1181 | return 0; |
1182 | } | |
1183 | ||
1184 | if (!atomic_read(&rth->u.dst.__refcnt)) { | |
1185 | u32 score = rt_score(rth); | |
1186 | ||
1187 | if (score <= min_score) { | |
1188 | cand = rth; | |
1189 | candp = rthp; | |
1190 | min_score = score; | |
1191 | } | |
1192 | } | |
1193 | ||
1194 | chain_length++; | |
1195 | ||
093c2ca4 | 1196 | rthp = &rth->u.dst.rt_next; |
1da177e4 LT |
1197 | } |
1198 | ||
1199 | if (cand) { | |
1200 | /* ip_rt_gc_elasticity used to be average length of chain | |
1201 | * length, when exceeded gc becomes really aggressive. | |
1202 | * | |
1203 | * The second limit is less certain. At the moment it allows | |
1204 | * only 2 entries per bucket. We will see. | |
1205 | */ | |
1206 | if (chain_length > ip_rt_gc_elasticity) { | |
093c2ca4 | 1207 | *candp = cand->u.dst.rt_next; |
1da177e4 LT |
1208 | rt_free(cand); |
1209 | } | |
1080d709 | 1210 | } else { |
98376387 ED |
1211 | if (chain_length > rt_chain_length_max && |
1212 | slow_chain_length(rt_hash_table[hash].chain) > rt_chain_length_max) { | |
1080d709 NH |
1213 | struct net *net = dev_net(rt->u.dst.dev); |
1214 | int num = ++net->ipv4.current_rt_cache_rebuild_count; | |
1215 | if (!rt_caching(dev_net(rt->u.dst.dev))) { | |
1216 | printk(KERN_WARNING "%s: %d rebuilds is over limit, route caching disabled\n", | |
1217 | rt->u.dst.dev->name, num); | |
1218 | } | |
1219 | rt_emergency_hash_rebuild(dev_net(rt->u.dst.dev)); | |
1220 | } | |
1da177e4 LT |
1221 | } |
1222 | ||
1223 | /* Try to bind route to arp only if it is output | |
1224 | route or unicast forwarding path. | |
1225 | */ | |
1226 | if (rt->rt_type == RTN_UNICAST || rt->fl.iif == 0) { | |
1227 | int err = arp_bind_neighbour(&rt->u.dst); | |
1228 | if (err) { | |
22c047cc | 1229 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1230 | |
1231 | if (err != -ENOBUFS) { | |
1232 | rt_drop(rt); | |
1233 | return err; | |
1234 | } | |
1235 | ||
1236 | /* Neighbour tables are full and nothing | |
1237 | can be released. Try to shrink route cache, | |
1238 | it is most likely it holds some neighbour records. | |
1239 | */ | |
1240 | if (attempts-- > 0) { | |
1241 | int saved_elasticity = ip_rt_gc_elasticity; | |
1242 | int saved_int = ip_rt_gc_min_interval; | |
1243 | ip_rt_gc_elasticity = 1; | |
1244 | ip_rt_gc_min_interval = 0; | |
569d3645 | 1245 | rt_garbage_collect(&ipv4_dst_ops); |
1da177e4 LT |
1246 | ip_rt_gc_min_interval = saved_int; |
1247 | ip_rt_gc_elasticity = saved_elasticity; | |
1248 | goto restart; | |
1249 | } | |
1250 | ||
1251 | if (net_ratelimit()) | |
1252 | printk(KERN_WARNING "Neighbour table overflow.\n"); | |
1253 | rt_drop(rt); | |
1254 | return -ENOBUFS; | |
1255 | } | |
1256 | } | |
1257 | ||
1ddbcb00 | 1258 | rt->u.dst.rt_next = rt_hash_table[hash].chain; |
1080d709 | 1259 | |
1da177e4 | 1260 | #if RT_CACHE_DEBUG >= 2 |
093c2ca4 | 1261 | if (rt->u.dst.rt_next) { |
1da177e4 | 1262 | struct rtable *trt; |
b6280b47 NH |
1263 | printk(KERN_DEBUG "rt_cache @%02x: %pI4", |
1264 | hash, &rt->rt_dst); | |
093c2ca4 | 1265 | for (trt = rt->u.dst.rt_next; trt; trt = trt->u.dst.rt_next) |
673d57e7 | 1266 | printk(" . %pI4", &trt->rt_dst); |
1da177e4 LT |
1267 | printk("\n"); |
1268 | } | |
1269 | #endif | |
00269b54 ED |
1270 | /* |
1271 | * Since lookup is lockfree, we must make sure | |
1272 | * previous writes to rt are comitted to memory | |
1273 | * before making rt visible to other CPUS. | |
1274 | */ | |
1ddbcb00 | 1275 | rcu_assign_pointer(rt_hash_table[hash].chain, rt); |
1080d709 | 1276 | |
22c047cc | 1277 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
73e42897 | 1278 | |
b6280b47 | 1279 | skip_hashing: |
511c3f92 ED |
1280 | if (rp) |
1281 | *rp = rt; | |
1282 | else | |
adf30907 | 1283 | skb_dst_set(skb, &rt->u.dst); |
1da177e4 LT |
1284 | return 0; |
1285 | } | |
1286 | ||
1287 | void rt_bind_peer(struct rtable *rt, int create) | |
1288 | { | |
1289 | static DEFINE_SPINLOCK(rt_peer_lock); | |
1290 | struct inet_peer *peer; | |
1291 | ||
1292 | peer = inet_getpeer(rt->rt_dst, create); | |
1293 | ||
1294 | spin_lock_bh(&rt_peer_lock); | |
1295 | if (rt->peer == NULL) { | |
1296 | rt->peer = peer; | |
1297 | peer = NULL; | |
1298 | } | |
1299 | spin_unlock_bh(&rt_peer_lock); | |
1300 | if (peer) | |
1301 | inet_putpeer(peer); | |
1302 | } | |
1303 | ||
1304 | /* | |
1305 | * Peer allocation may fail only in serious out-of-memory conditions. However | |
1306 | * we still can generate some output. | |
1307 | * Random ID selection looks a bit dangerous because we have no chances to | |
1308 | * select ID being unique in a reasonable period of time. | |
1309 | * But broken packet identifier may be better than no packet at all. | |
1310 | */ | |
1311 | static void ip_select_fb_ident(struct iphdr *iph) | |
1312 | { | |
1313 | static DEFINE_SPINLOCK(ip_fb_id_lock); | |
1314 | static u32 ip_fallback_id; | |
1315 | u32 salt; | |
1316 | ||
1317 | spin_lock_bh(&ip_fb_id_lock); | |
e448515c | 1318 | salt = secure_ip_id((__force __be32)ip_fallback_id ^ iph->daddr); |
1da177e4 LT |
1319 | iph->id = htons(salt & 0xFFFF); |
1320 | ip_fallback_id = salt; | |
1321 | spin_unlock_bh(&ip_fb_id_lock); | |
1322 | } | |
1323 | ||
1324 | void __ip_select_ident(struct iphdr *iph, struct dst_entry *dst, int more) | |
1325 | { | |
1326 | struct rtable *rt = (struct rtable *) dst; | |
1327 | ||
1328 | if (rt) { | |
1329 | if (rt->peer == NULL) | |
1330 | rt_bind_peer(rt, 1); | |
1331 | ||
1332 | /* If peer is attached to destination, it is never detached, | |
1333 | so that we need not to grab a lock to dereference it. | |
1334 | */ | |
1335 | if (rt->peer) { | |
1336 | iph->id = htons(inet_getid(rt->peer, more)); | |
1337 | return; | |
1338 | } | |
1339 | } else | |
e905a9ed | 1340 | printk(KERN_DEBUG "rt_bind_peer(0) @%p\n", |
9c2b3328 | 1341 | __builtin_return_address(0)); |
1da177e4 LT |
1342 | |
1343 | ip_select_fb_ident(iph); | |
1344 | } | |
1345 | ||
1346 | static void rt_del(unsigned hash, struct rtable *rt) | |
1347 | { | |
29e75252 | 1348 | struct rtable **rthp, *aux; |
1da177e4 | 1349 | |
29e75252 | 1350 | rthp = &rt_hash_table[hash].chain; |
22c047cc | 1351 | spin_lock_bh(rt_hash_lock_addr(hash)); |
1da177e4 | 1352 | ip_rt_put(rt); |
29e75252 | 1353 | while ((aux = *rthp) != NULL) { |
e84f84f2 | 1354 | if (aux == rt || rt_is_expired(aux)) { |
29e75252 ED |
1355 | *rthp = aux->u.dst.rt_next; |
1356 | rt_free(aux); | |
1357 | continue; | |
1da177e4 | 1358 | } |
29e75252 ED |
1359 | rthp = &aux->u.dst.rt_next; |
1360 | } | |
22c047cc | 1361 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1362 | } |
1363 | ||
f7655229 AV |
1364 | void ip_rt_redirect(__be32 old_gw, __be32 daddr, __be32 new_gw, |
1365 | __be32 saddr, struct net_device *dev) | |
1da177e4 LT |
1366 | { |
1367 | int i, k; | |
1368 | struct in_device *in_dev = in_dev_get(dev); | |
1369 | struct rtable *rth, **rthp; | |
f7655229 | 1370 | __be32 skeys[2] = { saddr, 0 }; |
1da177e4 | 1371 | int ikeys[2] = { dev->ifindex, 0 }; |
8d71740c | 1372 | struct netevent_redirect netevent; |
317805b8 | 1373 | struct net *net; |
1da177e4 | 1374 | |
1da177e4 LT |
1375 | if (!in_dev) |
1376 | return; | |
1377 | ||
c346dca1 | 1378 | net = dev_net(dev); |
9d4fb27d JP |
1379 | if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) || |
1380 | ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) || | |
1381 | ipv4_is_zeronet(new_gw)) | |
1da177e4 LT |
1382 | goto reject_redirect; |
1383 | ||
1080d709 NH |
1384 | if (!rt_caching(net)) |
1385 | goto reject_redirect; | |
1386 | ||
1da177e4 LT |
1387 | if (!IN_DEV_SHARED_MEDIA(in_dev)) { |
1388 | if (!inet_addr_onlink(in_dev, new_gw, old_gw)) | |
1389 | goto reject_redirect; | |
1390 | if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev)) | |
1391 | goto reject_redirect; | |
1392 | } else { | |
317805b8 | 1393 | if (inet_addr_type(net, new_gw) != RTN_UNICAST) |
1da177e4 LT |
1394 | goto reject_redirect; |
1395 | } | |
1396 | ||
1397 | for (i = 0; i < 2; i++) { | |
1398 | for (k = 0; k < 2; k++) { | |
b00180de | 1399 | unsigned hash = rt_hash(daddr, skeys[i], ikeys[k], |
e84f84f2 | 1400 | rt_genid(net)); |
1da177e4 LT |
1401 | |
1402 | rthp=&rt_hash_table[hash].chain; | |
1403 | ||
1404 | rcu_read_lock(); | |
1405 | while ((rth = rcu_dereference(*rthp)) != NULL) { | |
1406 | struct rtable *rt; | |
1407 | ||
1408 | if (rth->fl.fl4_dst != daddr || | |
1409 | rth->fl.fl4_src != skeys[i] || | |
1da177e4 | 1410 | rth->fl.oif != ikeys[k] || |
29e75252 | 1411 | rth->fl.iif != 0 || |
e84f84f2 | 1412 | rt_is_expired(rth) || |
878628fb | 1413 | !net_eq(dev_net(rth->u.dst.dev), net)) { |
093c2ca4 | 1414 | rthp = &rth->u.dst.rt_next; |
1da177e4 LT |
1415 | continue; |
1416 | } | |
1417 | ||
1418 | if (rth->rt_dst != daddr || | |
1419 | rth->rt_src != saddr || | |
1420 | rth->u.dst.error || | |
1421 | rth->rt_gateway != old_gw || | |
1422 | rth->u.dst.dev != dev) | |
1423 | break; | |
1424 | ||
1425 | dst_hold(&rth->u.dst); | |
1426 | rcu_read_unlock(); | |
1427 | ||
1428 | rt = dst_alloc(&ipv4_dst_ops); | |
1429 | if (rt == NULL) { | |
1430 | ip_rt_put(rth); | |
1431 | in_dev_put(in_dev); | |
1432 | return; | |
1433 | } | |
1434 | ||
1435 | /* Copy all the information. */ | |
1436 | *rt = *rth; | |
1da177e4 LT |
1437 | rt->u.dst.__use = 1; |
1438 | atomic_set(&rt->u.dst.__refcnt, 1); | |
1439 | rt->u.dst.child = NULL; | |
1440 | if (rt->u.dst.dev) | |
1441 | dev_hold(rt->u.dst.dev); | |
1442 | if (rt->idev) | |
1443 | in_dev_hold(rt->idev); | |
d11a4dc1 | 1444 | rt->u.dst.obsolete = -1; |
1da177e4 LT |
1445 | rt->u.dst.lastuse = jiffies; |
1446 | rt->u.dst.path = &rt->u.dst; | |
1447 | rt->u.dst.neighbour = NULL; | |
1448 | rt->u.dst.hh = NULL; | |
def8b4fa | 1449 | #ifdef CONFIG_XFRM |
1da177e4 | 1450 | rt->u.dst.xfrm = NULL; |
def8b4fa | 1451 | #endif |
e84f84f2 | 1452 | rt->rt_genid = rt_genid(net); |
1da177e4 LT |
1453 | rt->rt_flags |= RTCF_REDIRECTED; |
1454 | ||
1455 | /* Gateway is different ... */ | |
1456 | rt->rt_gateway = new_gw; | |
1457 | ||
1458 | /* Redirect received -> path was valid */ | |
1459 | dst_confirm(&rth->u.dst); | |
1460 | ||
1461 | if (rt->peer) | |
1462 | atomic_inc(&rt->peer->refcnt); | |
1463 | ||
1464 | if (arp_bind_neighbour(&rt->u.dst) || | |
1465 | !(rt->u.dst.neighbour->nud_state & | |
1466 | NUD_VALID)) { | |
1467 | if (rt->u.dst.neighbour) | |
1468 | neigh_event_send(rt->u.dst.neighbour, NULL); | |
1469 | ip_rt_put(rth); | |
1470 | rt_drop(rt); | |
1471 | goto do_next; | |
1472 | } | |
e905a9ed | 1473 | |
8d71740c TT |
1474 | netevent.old = &rth->u.dst; |
1475 | netevent.new = &rt->u.dst; | |
e905a9ed YH |
1476 | call_netevent_notifiers(NETEVENT_REDIRECT, |
1477 | &netevent); | |
1da177e4 LT |
1478 | |
1479 | rt_del(hash, rth); | |
511c3f92 | 1480 | if (!rt_intern_hash(hash, rt, &rt, NULL)) |
1da177e4 LT |
1481 | ip_rt_put(rt); |
1482 | goto do_next; | |
1483 | } | |
1484 | rcu_read_unlock(); | |
1485 | do_next: | |
1486 | ; | |
1487 | } | |
1488 | } | |
1489 | in_dev_put(in_dev); | |
1490 | return; | |
1491 | ||
1492 | reject_redirect: | |
1493 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1494 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) | |
673d57e7 HH |
1495 | printk(KERN_INFO "Redirect from %pI4 on %s about %pI4 ignored.\n" |
1496 | " Advised path = %pI4 -> %pI4\n", | |
1497 | &old_gw, dev->name, &new_gw, | |
1498 | &saddr, &daddr); | |
1da177e4 LT |
1499 | #endif |
1500 | in_dev_put(in_dev); | |
1501 | } | |
1502 | ||
1503 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst) | |
1504 | { | |
ee6b9673 | 1505 | struct rtable *rt = (struct rtable *)dst; |
1da177e4 LT |
1506 | struct dst_entry *ret = dst; |
1507 | ||
1508 | if (rt) { | |
d11a4dc1 | 1509 | if (dst->obsolete > 0) { |
1da177e4 LT |
1510 | ip_rt_put(rt); |
1511 | ret = NULL; | |
1512 | } else if ((rt->rt_flags & RTCF_REDIRECTED) || | |
1513 | rt->u.dst.expires) { | |
8c7bc840 | 1514 | unsigned hash = rt_hash(rt->fl.fl4_dst, rt->fl.fl4_src, |
b00180de | 1515 | rt->fl.oif, |
e84f84f2 | 1516 | rt_genid(dev_net(dst->dev))); |
1da177e4 | 1517 | #if RT_CACHE_DEBUG >= 1 |
673d57e7 HH |
1518 | printk(KERN_DEBUG "ipv4_negative_advice: redirect to %pI4/%02x dropped\n", |
1519 | &rt->rt_dst, rt->fl.fl4_tos); | |
1da177e4 LT |
1520 | #endif |
1521 | rt_del(hash, rt); | |
1522 | ret = NULL; | |
1523 | } | |
1524 | } | |
1525 | return ret; | |
1526 | } | |
1527 | ||
1528 | /* | |
1529 | * Algorithm: | |
1530 | * 1. The first ip_rt_redirect_number redirects are sent | |
1531 | * with exponential backoff, then we stop sending them at all, | |
1532 | * assuming that the host ignores our redirects. | |
1533 | * 2. If we did not see packets requiring redirects | |
1534 | * during ip_rt_redirect_silence, we assume that the host | |
1535 | * forgot redirected route and start to send redirects again. | |
1536 | * | |
1537 | * This algorithm is much cheaper and more intelligent than dumb load limiting | |
1538 | * in icmp.c. | |
1539 | * | |
1540 | * NOTE. Do not forget to inhibit load limiting for redirects (redundant) | |
1541 | * and "frag. need" (breaks PMTU discovery) in icmp.c. | |
1542 | */ | |
1543 | ||
1544 | void ip_rt_send_redirect(struct sk_buff *skb) | |
1545 | { | |
511c3f92 | 1546 | struct rtable *rt = skb_rtable(skb); |
30038fc6 ED |
1547 | struct in_device *in_dev; |
1548 | int log_martians; | |
1da177e4 | 1549 | |
30038fc6 ED |
1550 | rcu_read_lock(); |
1551 | in_dev = __in_dev_get_rcu(rt->u.dst.dev); | |
1552 | if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) { | |
1553 | rcu_read_unlock(); | |
1da177e4 | 1554 | return; |
30038fc6 ED |
1555 | } |
1556 | log_martians = IN_DEV_LOG_MARTIANS(in_dev); | |
1557 | rcu_read_unlock(); | |
1da177e4 LT |
1558 | |
1559 | /* No redirected packets during ip_rt_redirect_silence; | |
1560 | * reset the algorithm. | |
1561 | */ | |
1562 | if (time_after(jiffies, rt->u.dst.rate_last + ip_rt_redirect_silence)) | |
1563 | rt->u.dst.rate_tokens = 0; | |
1564 | ||
1565 | /* Too many ignored redirects; do not send anything | |
1566 | * set u.dst.rate_last to the last seen redirected packet. | |
1567 | */ | |
1568 | if (rt->u.dst.rate_tokens >= ip_rt_redirect_number) { | |
1569 | rt->u.dst.rate_last = jiffies; | |
30038fc6 | 1570 | return; |
1da177e4 LT |
1571 | } |
1572 | ||
1573 | /* Check for load limit; set rate_last to the latest sent | |
1574 | * redirect. | |
1575 | */ | |
14fb8a76 LY |
1576 | if (rt->u.dst.rate_tokens == 0 || |
1577 | time_after(jiffies, | |
1da177e4 LT |
1578 | (rt->u.dst.rate_last + |
1579 | (ip_rt_redirect_load << rt->u.dst.rate_tokens)))) { | |
1580 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway); | |
1581 | rt->u.dst.rate_last = jiffies; | |
1582 | ++rt->u.dst.rate_tokens; | |
1583 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
30038fc6 | 1584 | if (log_martians && |
1da177e4 LT |
1585 | rt->u.dst.rate_tokens == ip_rt_redirect_number && |
1586 | net_ratelimit()) | |
673d57e7 HH |
1587 | printk(KERN_WARNING "host %pI4/if%d ignores redirects for %pI4 to %pI4.\n", |
1588 | &rt->rt_src, rt->rt_iif, | |
1589 | &rt->rt_dst, &rt->rt_gateway); | |
1da177e4 LT |
1590 | #endif |
1591 | } | |
1da177e4 LT |
1592 | } |
1593 | ||
1594 | static int ip_error(struct sk_buff *skb) | |
1595 | { | |
511c3f92 | 1596 | struct rtable *rt = skb_rtable(skb); |
1da177e4 LT |
1597 | unsigned long now; |
1598 | int code; | |
1599 | ||
1600 | switch (rt->u.dst.error) { | |
1601 | case EINVAL: | |
1602 | default: | |
1603 | goto out; | |
1604 | case EHOSTUNREACH: | |
1605 | code = ICMP_HOST_UNREACH; | |
1606 | break; | |
1607 | case ENETUNREACH: | |
1608 | code = ICMP_NET_UNREACH; | |
7c73a6fa PE |
1609 | IP_INC_STATS_BH(dev_net(rt->u.dst.dev), |
1610 | IPSTATS_MIB_INNOROUTES); | |
1da177e4 LT |
1611 | break; |
1612 | case EACCES: | |
1613 | code = ICMP_PKT_FILTERED; | |
1614 | break; | |
1615 | } | |
1616 | ||
1617 | now = jiffies; | |
1618 | rt->u.dst.rate_tokens += now - rt->u.dst.rate_last; | |
1619 | if (rt->u.dst.rate_tokens > ip_rt_error_burst) | |
1620 | rt->u.dst.rate_tokens = ip_rt_error_burst; | |
1621 | rt->u.dst.rate_last = now; | |
1622 | if (rt->u.dst.rate_tokens >= ip_rt_error_cost) { | |
1623 | rt->u.dst.rate_tokens -= ip_rt_error_cost; | |
1624 | icmp_send(skb, ICMP_DEST_UNREACH, code, 0); | |
1625 | } | |
1626 | ||
1627 | out: kfree_skb(skb); | |
1628 | return 0; | |
e905a9ed | 1629 | } |
1da177e4 LT |
1630 | |
1631 | /* | |
1632 | * The last two values are not from the RFC but | |
1633 | * are needed for AMPRnet AX.25 paths. | |
1634 | */ | |
1635 | ||
9b5b5cff | 1636 | static const unsigned short mtu_plateau[] = |
1da177e4 LT |
1637 | {32000, 17914, 8166, 4352, 2002, 1492, 576, 296, 216, 128 }; |
1638 | ||
5969f71d | 1639 | static inline unsigned short guess_mtu(unsigned short old_mtu) |
1da177e4 LT |
1640 | { |
1641 | int i; | |
e905a9ed | 1642 | |
1da177e4 LT |
1643 | for (i = 0; i < ARRAY_SIZE(mtu_plateau); i++) |
1644 | if (old_mtu > mtu_plateau[i]) | |
1645 | return mtu_plateau[i]; | |
1646 | return 68; | |
1647 | } | |
1648 | ||
b5921910 | 1649 | unsigned short ip_rt_frag_needed(struct net *net, struct iphdr *iph, |
0010e465 TT |
1650 | unsigned short new_mtu, |
1651 | struct net_device *dev) | |
1da177e4 | 1652 | { |
0010e465 | 1653 | int i, k; |
1da177e4 LT |
1654 | unsigned short old_mtu = ntohs(iph->tot_len); |
1655 | struct rtable *rth; | |
0010e465 | 1656 | int ikeys[2] = { dev->ifindex, 0 }; |
e448515c AV |
1657 | __be32 skeys[2] = { iph->saddr, 0, }; |
1658 | __be32 daddr = iph->daddr; | |
1da177e4 LT |
1659 | unsigned short est_mtu = 0; |
1660 | ||
0010e465 TT |
1661 | for (k = 0; k < 2; k++) { |
1662 | for (i = 0; i < 2; i++) { | |
b00180de | 1663 | unsigned hash = rt_hash(daddr, skeys[i], ikeys[k], |
e84f84f2 | 1664 | rt_genid(net)); |
0010e465 TT |
1665 | |
1666 | rcu_read_lock(); | |
1667 | for (rth = rcu_dereference(rt_hash_table[hash].chain); rth; | |
1668 | rth = rcu_dereference(rth->u.dst.rt_next)) { | |
1da177e4 LT |
1669 | unsigned short mtu = new_mtu; |
1670 | ||
0010e465 TT |
1671 | if (rth->fl.fl4_dst != daddr || |
1672 | rth->fl.fl4_src != skeys[i] || | |
1673 | rth->rt_dst != daddr || | |
1674 | rth->rt_src != iph->saddr || | |
1675 | rth->fl.oif != ikeys[k] || | |
1676 | rth->fl.iif != 0 || | |
1677 | dst_metric_locked(&rth->u.dst, RTAX_MTU) || | |
1678 | !net_eq(dev_net(rth->u.dst.dev), net) || | |
6c3b8fc6 | 1679 | rt_is_expired(rth)) |
0010e465 TT |
1680 | continue; |
1681 | ||
1da177e4 LT |
1682 | if (new_mtu < 68 || new_mtu >= old_mtu) { |
1683 | ||
1684 | /* BSD 4.2 compatibility hack :-( */ | |
1685 | if (mtu == 0 && | |
6d273f8d | 1686 | old_mtu >= dst_mtu(&rth->u.dst) && |
1da177e4 LT |
1687 | old_mtu >= 68 + (iph->ihl << 2)) |
1688 | old_mtu -= iph->ihl << 2; | |
1689 | ||
1690 | mtu = guess_mtu(old_mtu); | |
1691 | } | |
6d273f8d RR |
1692 | if (mtu <= dst_mtu(&rth->u.dst)) { |
1693 | if (mtu < dst_mtu(&rth->u.dst)) { | |
1da177e4 LT |
1694 | dst_confirm(&rth->u.dst); |
1695 | if (mtu < ip_rt_min_pmtu) { | |
1696 | mtu = ip_rt_min_pmtu; | |
1697 | rth->u.dst.metrics[RTAX_LOCK-1] |= | |
1698 | (1 << RTAX_MTU); | |
1699 | } | |
1700 | rth->u.dst.metrics[RTAX_MTU-1] = mtu; | |
1701 | dst_set_expires(&rth->u.dst, | |
1702 | ip_rt_mtu_expires); | |
1703 | } | |
1704 | est_mtu = mtu; | |
1705 | } | |
1706 | } | |
0010e465 | 1707 | rcu_read_unlock(); |
1da177e4 | 1708 | } |
1da177e4 LT |
1709 | } |
1710 | return est_mtu ? : new_mtu; | |
1711 | } | |
1712 | ||
1713 | static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu) | |
1714 | { | |
6d273f8d | 1715 | if (dst_mtu(dst) > mtu && mtu >= 68 && |
1da177e4 LT |
1716 | !(dst_metric_locked(dst, RTAX_MTU))) { |
1717 | if (mtu < ip_rt_min_pmtu) { | |
1718 | mtu = ip_rt_min_pmtu; | |
1719 | dst->metrics[RTAX_LOCK-1] |= (1 << RTAX_MTU); | |
1720 | } | |
1721 | dst->metrics[RTAX_MTU-1] = mtu; | |
1722 | dst_set_expires(dst, ip_rt_mtu_expires); | |
8d71740c | 1723 | call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst); |
1da177e4 LT |
1724 | } |
1725 | } | |
1726 | ||
1727 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie) | |
1728 | { | |
d11a4dc1 TT |
1729 | if (rt_is_expired((struct rtable *)dst)) |
1730 | return NULL; | |
1731 | return dst; | |
1da177e4 LT |
1732 | } |
1733 | ||
1734 | static void ipv4_dst_destroy(struct dst_entry *dst) | |
1735 | { | |
1736 | struct rtable *rt = (struct rtable *) dst; | |
1737 | struct inet_peer *peer = rt->peer; | |
1738 | struct in_device *idev = rt->idev; | |
1739 | ||
1740 | if (peer) { | |
1741 | rt->peer = NULL; | |
1742 | inet_putpeer(peer); | |
1743 | } | |
1744 | ||
1745 | if (idev) { | |
1746 | rt->idev = NULL; | |
1747 | in_dev_put(idev); | |
1748 | } | |
1749 | } | |
1750 | ||
1751 | static void ipv4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, | |
1752 | int how) | |
1753 | { | |
1754 | struct rtable *rt = (struct rtable *) dst; | |
1755 | struct in_device *idev = rt->idev; | |
c346dca1 | 1756 | if (dev != dev_net(dev)->loopback_dev && idev && idev->dev == dev) { |
5a3e55d6 | 1757 | struct in_device *loopback_idev = |
c346dca1 | 1758 | in_dev_get(dev_net(dev)->loopback_dev); |
1da177e4 LT |
1759 | if (loopback_idev) { |
1760 | rt->idev = loopback_idev; | |
1761 | in_dev_put(idev); | |
1762 | } | |
1763 | } | |
1764 | } | |
1765 | ||
1766 | static void ipv4_link_failure(struct sk_buff *skb) | |
1767 | { | |
1768 | struct rtable *rt; | |
1769 | ||
1770 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0); | |
1771 | ||
511c3f92 | 1772 | rt = skb_rtable(skb); |
1da177e4 LT |
1773 | if (rt) |
1774 | dst_set_expires(&rt->u.dst, 0); | |
1775 | } | |
1776 | ||
1777 | static int ip_rt_bug(struct sk_buff *skb) | |
1778 | { | |
673d57e7 HH |
1779 | printk(KERN_DEBUG "ip_rt_bug: %pI4 -> %pI4, %s\n", |
1780 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
1da177e4 LT |
1781 | skb->dev ? skb->dev->name : "?"); |
1782 | kfree_skb(skb); | |
1783 | return 0; | |
1784 | } | |
1785 | ||
1786 | /* | |
1787 | We do not cache source address of outgoing interface, | |
1788 | because it is used only by IP RR, TS and SRR options, | |
1789 | so that it out of fast path. | |
1790 | ||
1791 | BTW remember: "addr" is allowed to be not aligned | |
1792 | in IP options! | |
1793 | */ | |
1794 | ||
1795 | void ip_rt_get_source(u8 *addr, struct rtable *rt) | |
1796 | { | |
a61ced5d | 1797 | __be32 src; |
1da177e4 LT |
1798 | struct fib_result res; |
1799 | ||
1800 | if (rt->fl.iif == 0) | |
1801 | src = rt->rt_src; | |
c346dca1 | 1802 | else if (fib_lookup(dev_net(rt->u.dst.dev), &rt->fl, &res) == 0) { |
1da177e4 LT |
1803 | src = FIB_RES_PREFSRC(res); |
1804 | fib_res_put(&res); | |
1805 | } else | |
1806 | src = inet_select_addr(rt->u.dst.dev, rt->rt_gateway, | |
1807 | RT_SCOPE_UNIVERSE); | |
1808 | memcpy(addr, &src, 4); | |
1809 | } | |
1810 | ||
1811 | #ifdef CONFIG_NET_CLS_ROUTE | |
1812 | static void set_class_tag(struct rtable *rt, u32 tag) | |
1813 | { | |
1814 | if (!(rt->u.dst.tclassid & 0xFFFF)) | |
1815 | rt->u.dst.tclassid |= tag & 0xFFFF; | |
1816 | if (!(rt->u.dst.tclassid & 0xFFFF0000)) | |
1817 | rt->u.dst.tclassid |= tag & 0xFFFF0000; | |
1818 | } | |
1819 | #endif | |
1820 | ||
1821 | static void rt_set_nexthop(struct rtable *rt, struct fib_result *res, u32 itag) | |
1822 | { | |
1823 | struct fib_info *fi = res->fi; | |
1824 | ||
1825 | if (fi) { | |
1826 | if (FIB_RES_GW(*res) && | |
1827 | FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) | |
1828 | rt->rt_gateway = FIB_RES_GW(*res); | |
1829 | memcpy(rt->u.dst.metrics, fi->fib_metrics, | |
1830 | sizeof(rt->u.dst.metrics)); | |
1831 | if (fi->fib_mtu == 0) { | |
1832 | rt->u.dst.metrics[RTAX_MTU-1] = rt->u.dst.dev->mtu; | |
0bbeafd0 | 1833 | if (dst_metric_locked(&rt->u.dst, RTAX_MTU) && |
1da177e4 LT |
1834 | rt->rt_gateway != rt->rt_dst && |
1835 | rt->u.dst.dev->mtu > 576) | |
1836 | rt->u.dst.metrics[RTAX_MTU-1] = 576; | |
1837 | } | |
1838 | #ifdef CONFIG_NET_CLS_ROUTE | |
1839 | rt->u.dst.tclassid = FIB_RES_NH(*res).nh_tclassid; | |
1840 | #endif | |
1841 | } else | |
1842 | rt->u.dst.metrics[RTAX_MTU-1]= rt->u.dst.dev->mtu; | |
1843 | ||
5ffc02a1 | 1844 | if (dst_metric(&rt->u.dst, RTAX_HOPLIMIT) == 0) |
1da177e4 | 1845 | rt->u.dst.metrics[RTAX_HOPLIMIT-1] = sysctl_ip_default_ttl; |
6d273f8d | 1846 | if (dst_mtu(&rt->u.dst) > IP_MAX_MTU) |
1da177e4 | 1847 | rt->u.dst.metrics[RTAX_MTU-1] = IP_MAX_MTU; |
5ffc02a1 | 1848 | if (dst_metric(&rt->u.dst, RTAX_ADVMSS) == 0) |
1da177e4 LT |
1849 | rt->u.dst.metrics[RTAX_ADVMSS-1] = max_t(unsigned int, rt->u.dst.dev->mtu - 40, |
1850 | ip_rt_min_advmss); | |
5ffc02a1 | 1851 | if (dst_metric(&rt->u.dst, RTAX_ADVMSS) > 65535 - 40) |
1da177e4 LT |
1852 | rt->u.dst.metrics[RTAX_ADVMSS-1] = 65535 - 40; |
1853 | ||
1854 | #ifdef CONFIG_NET_CLS_ROUTE | |
1855 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
1856 | set_class_tag(rt, fib_rules_tclass(res)); | |
1857 | #endif | |
1858 | set_class_tag(rt, itag); | |
1859 | #endif | |
e905a9ed | 1860 | rt->rt_type = res->type; |
1da177e4 LT |
1861 | } |
1862 | ||
9e12bb22 | 1863 | static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
1864 | u8 tos, struct net_device *dev, int our) |
1865 | { | |
1866 | unsigned hash; | |
1867 | struct rtable *rth; | |
a61ced5d | 1868 | __be32 spec_dst; |
1da177e4 LT |
1869 | struct in_device *in_dev = in_dev_get(dev); |
1870 | u32 itag = 0; | |
1871 | ||
1872 | /* Primary sanity checks. */ | |
1873 | ||
1874 | if (in_dev == NULL) | |
1875 | return -EINVAL; | |
1876 | ||
1e637c74 | 1877 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
f97c1e0c | 1878 | ipv4_is_loopback(saddr) || skb->protocol != htons(ETH_P_IP)) |
1da177e4 LT |
1879 | goto e_inval; |
1880 | ||
f97c1e0c JP |
1881 | if (ipv4_is_zeronet(saddr)) { |
1882 | if (!ipv4_is_local_multicast(daddr)) | |
1da177e4 LT |
1883 | goto e_inval; |
1884 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); | |
1885 | } else if (fib_validate_source(saddr, 0, tos, 0, | |
b0c110ca | 1886 | dev, &spec_dst, &itag, 0) < 0) |
1da177e4 LT |
1887 | goto e_inval; |
1888 | ||
1889 | rth = dst_alloc(&ipv4_dst_ops); | |
1890 | if (!rth) | |
1891 | goto e_nobufs; | |
1892 | ||
d11a4dc1 TT |
1893 | rth->u.dst.output = ip_rt_bug; |
1894 | rth->u.dst.obsolete = -1; | |
1da177e4 LT |
1895 | |
1896 | atomic_set(&rth->u.dst.__refcnt, 1); | |
1897 | rth->u.dst.flags= DST_HOST; | |
42f811b8 | 1898 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
1da177e4 LT |
1899 | rth->u.dst.flags |= DST_NOPOLICY; |
1900 | rth->fl.fl4_dst = daddr; | |
1901 | rth->rt_dst = daddr; | |
1902 | rth->fl.fl4_tos = tos; | |
47dcf0cb | 1903 | rth->fl.mark = skb->mark; |
1da177e4 LT |
1904 | rth->fl.fl4_src = saddr; |
1905 | rth->rt_src = saddr; | |
1906 | #ifdef CONFIG_NET_CLS_ROUTE | |
1907 | rth->u.dst.tclassid = itag; | |
1908 | #endif | |
1909 | rth->rt_iif = | |
1910 | rth->fl.iif = dev->ifindex; | |
2774c7ab | 1911 | rth->u.dst.dev = init_net.loopback_dev; |
1da177e4 LT |
1912 | dev_hold(rth->u.dst.dev); |
1913 | rth->idev = in_dev_get(rth->u.dst.dev); | |
1914 | rth->fl.oif = 0; | |
1915 | rth->rt_gateway = daddr; | |
1916 | rth->rt_spec_dst= spec_dst; | |
e84f84f2 | 1917 | rth->rt_genid = rt_genid(dev_net(dev)); |
1da177e4 | 1918 | rth->rt_flags = RTCF_MULTICAST; |
29e75252 | 1919 | rth->rt_type = RTN_MULTICAST; |
1da177e4 LT |
1920 | if (our) { |
1921 | rth->u.dst.input= ip_local_deliver; | |
1922 | rth->rt_flags |= RTCF_LOCAL; | |
1923 | } | |
1924 | ||
1925 | #ifdef CONFIG_IP_MROUTE | |
f97c1e0c | 1926 | if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev)) |
1da177e4 LT |
1927 | rth->u.dst.input = ip_mr_input; |
1928 | #endif | |
1929 | RT_CACHE_STAT_INC(in_slow_mc); | |
1930 | ||
1931 | in_dev_put(in_dev); | |
e84f84f2 | 1932 | hash = rt_hash(daddr, saddr, dev->ifindex, rt_genid(dev_net(dev))); |
511c3f92 | 1933 | return rt_intern_hash(hash, rth, NULL, skb); |
1da177e4 LT |
1934 | |
1935 | e_nobufs: | |
1936 | in_dev_put(in_dev); | |
1937 | return -ENOBUFS; | |
1938 | ||
1939 | e_inval: | |
1940 | in_dev_put(in_dev); | |
1941 | return -EINVAL; | |
1942 | } | |
1943 | ||
1944 | ||
1945 | static void ip_handle_martian_source(struct net_device *dev, | |
1946 | struct in_device *in_dev, | |
1947 | struct sk_buff *skb, | |
9e12bb22 AV |
1948 | __be32 daddr, |
1949 | __be32 saddr) | |
1da177e4 LT |
1950 | { |
1951 | RT_CACHE_STAT_INC(in_martian_src); | |
1952 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1953 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) { | |
1954 | /* | |
1955 | * RFC1812 recommendation, if source is martian, | |
1956 | * the only hint is MAC header. | |
1957 | */ | |
673d57e7 HH |
1958 | printk(KERN_WARNING "martian source %pI4 from %pI4, on dev %s\n", |
1959 | &daddr, &saddr, dev->name); | |
98e399f8 | 1960 | if (dev->hard_header_len && skb_mac_header_was_set(skb)) { |
1da177e4 | 1961 | int i; |
98e399f8 | 1962 | const unsigned char *p = skb_mac_header(skb); |
1da177e4 LT |
1963 | printk(KERN_WARNING "ll header: "); |
1964 | for (i = 0; i < dev->hard_header_len; i++, p++) { | |
1965 | printk("%02x", *p); | |
1966 | if (i < (dev->hard_header_len - 1)) | |
1967 | printk(":"); | |
1968 | } | |
1969 | printk("\n"); | |
1970 | } | |
1971 | } | |
1972 | #endif | |
1973 | } | |
1974 | ||
5969f71d SH |
1975 | static int __mkroute_input(struct sk_buff *skb, |
1976 | struct fib_result *res, | |
1977 | struct in_device *in_dev, | |
1978 | __be32 daddr, __be32 saddr, u32 tos, | |
1979 | struct rtable **result) | |
1da177e4 LT |
1980 | { |
1981 | ||
1982 | struct rtable *rth; | |
1983 | int err; | |
1984 | struct in_device *out_dev; | |
1985 | unsigned flags = 0; | |
d9c9df8c AV |
1986 | __be32 spec_dst; |
1987 | u32 itag; | |
1da177e4 LT |
1988 | |
1989 | /* get a working reference to the output device */ | |
1990 | out_dev = in_dev_get(FIB_RES_DEV(*res)); | |
1991 | if (out_dev == NULL) { | |
1992 | if (net_ratelimit()) | |
1993 | printk(KERN_CRIT "Bug in ip_route_input" \ | |
1994 | "_slow(). Please, report\n"); | |
1995 | return -EINVAL; | |
1996 | } | |
1997 | ||
1998 | ||
e905a9ed | 1999 | err = fib_validate_source(saddr, daddr, tos, FIB_RES_OIF(*res), |
b0c110ca | 2000 | in_dev->dev, &spec_dst, &itag, skb->mark); |
1da177e4 | 2001 | if (err < 0) { |
e905a9ed | 2002 | ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr, |
1da177e4 | 2003 | saddr); |
e905a9ed | 2004 | |
1da177e4 LT |
2005 | err = -EINVAL; |
2006 | goto cleanup; | |
2007 | } | |
2008 | ||
2009 | if (err) | |
2010 | flags |= RTCF_DIRECTSRC; | |
2011 | ||
51b77cae | 2012 | if (out_dev == in_dev && err && |
1da177e4 LT |
2013 | (IN_DEV_SHARED_MEDIA(out_dev) || |
2014 | inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) | |
2015 | flags |= RTCF_DOREDIRECT; | |
2016 | ||
2017 | if (skb->protocol != htons(ETH_P_IP)) { | |
2018 | /* Not IP (i.e. ARP). Do not create route, if it is | |
2019 | * invalid for proxy arp. DNAT routes are always valid. | |
65324144 JDB |
2020 | * |
2021 | * Proxy arp feature have been extended to allow, ARP | |
2022 | * replies back to the same interface, to support | |
2023 | * Private VLAN switch technologies. See arp.c. | |
1da177e4 | 2024 | */ |
65324144 JDB |
2025 | if (out_dev == in_dev && |
2026 | IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) { | |
1da177e4 LT |
2027 | err = -EINVAL; |
2028 | goto cleanup; | |
2029 | } | |
2030 | } | |
2031 | ||
2032 | ||
2033 | rth = dst_alloc(&ipv4_dst_ops); | |
2034 | if (!rth) { | |
2035 | err = -ENOBUFS; | |
2036 | goto cleanup; | |
2037 | } | |
2038 | ||
ce723d8e | 2039 | atomic_set(&rth->u.dst.__refcnt, 1); |
1da177e4 | 2040 | rth->u.dst.flags= DST_HOST; |
42f811b8 | 2041 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
1da177e4 | 2042 | rth->u.dst.flags |= DST_NOPOLICY; |
42f811b8 | 2043 | if (IN_DEV_CONF_GET(out_dev, NOXFRM)) |
1da177e4 LT |
2044 | rth->u.dst.flags |= DST_NOXFRM; |
2045 | rth->fl.fl4_dst = daddr; | |
2046 | rth->rt_dst = daddr; | |
2047 | rth->fl.fl4_tos = tos; | |
47dcf0cb | 2048 | rth->fl.mark = skb->mark; |
1da177e4 LT |
2049 | rth->fl.fl4_src = saddr; |
2050 | rth->rt_src = saddr; | |
2051 | rth->rt_gateway = daddr; | |
2052 | rth->rt_iif = | |
2053 | rth->fl.iif = in_dev->dev->ifindex; | |
2054 | rth->u.dst.dev = (out_dev)->dev; | |
2055 | dev_hold(rth->u.dst.dev); | |
2056 | rth->idev = in_dev_get(rth->u.dst.dev); | |
2057 | rth->fl.oif = 0; | |
2058 | rth->rt_spec_dst= spec_dst; | |
2059 | ||
d11a4dc1 | 2060 | rth->u.dst.obsolete = -1; |
1da177e4 LT |
2061 | rth->u.dst.input = ip_forward; |
2062 | rth->u.dst.output = ip_output; | |
e84f84f2 | 2063 | rth->rt_genid = rt_genid(dev_net(rth->u.dst.dev)); |
1da177e4 LT |
2064 | |
2065 | rt_set_nexthop(rth, res, itag); | |
2066 | ||
2067 | rth->rt_flags = flags; | |
2068 | ||
2069 | *result = rth; | |
2070 | err = 0; | |
2071 | cleanup: | |
2072 | /* release the working reference to the output device */ | |
2073 | in_dev_put(out_dev); | |
2074 | return err; | |
e905a9ed | 2075 | } |
1da177e4 | 2076 | |
5969f71d SH |
2077 | static int ip_mkroute_input(struct sk_buff *skb, |
2078 | struct fib_result *res, | |
2079 | const struct flowi *fl, | |
2080 | struct in_device *in_dev, | |
2081 | __be32 daddr, __be32 saddr, u32 tos) | |
1da177e4 | 2082 | { |
7abaa27c | 2083 | struct rtable* rth = NULL; |
1da177e4 LT |
2084 | int err; |
2085 | unsigned hash; | |
2086 | ||
2087 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
2088 | if (res->fi && res->fi->fib_nhs > 1 && fl->oif == 0) | |
2089 | fib_select_multipath(fl, res); | |
2090 | #endif | |
2091 | ||
2092 | /* create a routing cache entry */ | |
2093 | err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth); | |
2094 | if (err) | |
2095 | return err; | |
1da177e4 LT |
2096 | |
2097 | /* put it into the cache */ | |
e84f84f2 DL |
2098 | hash = rt_hash(daddr, saddr, fl->iif, |
2099 | rt_genid(dev_net(rth->u.dst.dev))); | |
511c3f92 | 2100 | return rt_intern_hash(hash, rth, NULL, skb); |
1da177e4 LT |
2101 | } |
2102 | ||
1da177e4 LT |
2103 | /* |
2104 | * NOTE. We drop all the packets that has local source | |
2105 | * addresses, because every properly looped back packet | |
2106 | * must have correct destination already attached by output routine. | |
2107 | * | |
2108 | * Such approach solves two big problems: | |
2109 | * 1. Not simplex devices are handled properly. | |
2110 | * 2. IP spoofing attempts are filtered with 100% of guarantee. | |
2111 | */ | |
2112 | ||
9e12bb22 | 2113 | static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
2114 | u8 tos, struct net_device *dev) |
2115 | { | |
2116 | struct fib_result res; | |
2117 | struct in_device *in_dev = in_dev_get(dev); | |
2118 | struct flowi fl = { .nl_u = { .ip4_u = | |
2119 | { .daddr = daddr, | |
2120 | .saddr = saddr, | |
2121 | .tos = tos, | |
2122 | .scope = RT_SCOPE_UNIVERSE, | |
1da177e4 | 2123 | } }, |
47dcf0cb | 2124 | .mark = skb->mark, |
1da177e4 LT |
2125 | .iif = dev->ifindex }; |
2126 | unsigned flags = 0; | |
2127 | u32 itag = 0; | |
2128 | struct rtable * rth; | |
2129 | unsigned hash; | |
9e12bb22 | 2130 | __be32 spec_dst; |
1da177e4 LT |
2131 | int err = -EINVAL; |
2132 | int free_res = 0; | |
c346dca1 | 2133 | struct net * net = dev_net(dev); |
1da177e4 LT |
2134 | |
2135 | /* IP on this device is disabled. */ | |
2136 | ||
2137 | if (!in_dev) | |
2138 | goto out; | |
2139 | ||
2140 | /* Check for the most weird martians, which can be not detected | |
2141 | by fib_lookup. | |
2142 | */ | |
2143 | ||
1e637c74 | 2144 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
f97c1e0c | 2145 | ipv4_is_loopback(saddr)) |
1da177e4 LT |
2146 | goto martian_source; |
2147 | ||
e448515c | 2148 | if (daddr == htonl(0xFFFFFFFF) || (saddr == 0 && daddr == 0)) |
1da177e4 LT |
2149 | goto brd_input; |
2150 | ||
2151 | /* Accept zero addresses only to limited broadcast; | |
2152 | * I even do not know to fix it or not. Waiting for complains :-) | |
2153 | */ | |
f97c1e0c | 2154 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
2155 | goto martian_source; |
2156 | ||
1e637c74 | 2157 | if (ipv4_is_lbcast(daddr) || ipv4_is_zeronet(daddr) || |
f97c1e0c | 2158 | ipv4_is_loopback(daddr)) |
1da177e4 LT |
2159 | goto martian_destination; |
2160 | ||
2161 | /* | |
2162 | * Now we are ready to route packet. | |
2163 | */ | |
84a885f4 | 2164 | if ((err = fib_lookup(net, &fl, &res)) != 0) { |
1da177e4 | 2165 | if (!IN_DEV_FORWARD(in_dev)) |
2c2910a4 | 2166 | goto e_hostunreach; |
1da177e4 LT |
2167 | goto no_route; |
2168 | } | |
2169 | free_res = 1; | |
2170 | ||
2171 | RT_CACHE_STAT_INC(in_slow_tot); | |
2172 | ||
2173 | if (res.type == RTN_BROADCAST) | |
2174 | goto brd_input; | |
2175 | ||
2176 | if (res.type == RTN_LOCAL) { | |
2177 | int result; | |
2178 | result = fib_validate_source(saddr, daddr, tos, | |
84a885f4 | 2179 | net->loopback_dev->ifindex, |
b0c110ca | 2180 | dev, &spec_dst, &itag, skb->mark); |
1da177e4 LT |
2181 | if (result < 0) |
2182 | goto martian_source; | |
2183 | if (result) | |
2184 | flags |= RTCF_DIRECTSRC; | |
2185 | spec_dst = daddr; | |
2186 | goto local_input; | |
2187 | } | |
2188 | ||
2189 | if (!IN_DEV_FORWARD(in_dev)) | |
2c2910a4 | 2190 | goto e_hostunreach; |
1da177e4 LT |
2191 | if (res.type != RTN_UNICAST) |
2192 | goto martian_destination; | |
2193 | ||
2194 | err = ip_mkroute_input(skb, &res, &fl, in_dev, daddr, saddr, tos); | |
1da177e4 LT |
2195 | done: |
2196 | in_dev_put(in_dev); | |
2197 | if (free_res) | |
2198 | fib_res_put(&res); | |
2199 | out: return err; | |
2200 | ||
2201 | brd_input: | |
2202 | if (skb->protocol != htons(ETH_P_IP)) | |
2203 | goto e_inval; | |
2204 | ||
f97c1e0c | 2205 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
2206 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
2207 | else { | |
2208 | err = fib_validate_source(saddr, 0, tos, 0, dev, &spec_dst, | |
b0c110ca | 2209 | &itag, skb->mark); |
1da177e4 LT |
2210 | if (err < 0) |
2211 | goto martian_source; | |
2212 | if (err) | |
2213 | flags |= RTCF_DIRECTSRC; | |
2214 | } | |
2215 | flags |= RTCF_BROADCAST; | |
2216 | res.type = RTN_BROADCAST; | |
2217 | RT_CACHE_STAT_INC(in_brd); | |
2218 | ||
2219 | local_input: | |
2220 | rth = dst_alloc(&ipv4_dst_ops); | |
2221 | if (!rth) | |
2222 | goto e_nobufs; | |
2223 | ||
2224 | rth->u.dst.output= ip_rt_bug; | |
d11a4dc1 | 2225 | rth->u.dst.obsolete = -1; |
e84f84f2 | 2226 | rth->rt_genid = rt_genid(net); |
1da177e4 LT |
2227 | |
2228 | atomic_set(&rth->u.dst.__refcnt, 1); | |
2229 | rth->u.dst.flags= DST_HOST; | |
42f811b8 | 2230 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
1da177e4 LT |
2231 | rth->u.dst.flags |= DST_NOPOLICY; |
2232 | rth->fl.fl4_dst = daddr; | |
2233 | rth->rt_dst = daddr; | |
2234 | rth->fl.fl4_tos = tos; | |
47dcf0cb | 2235 | rth->fl.mark = skb->mark; |
1da177e4 LT |
2236 | rth->fl.fl4_src = saddr; |
2237 | rth->rt_src = saddr; | |
2238 | #ifdef CONFIG_NET_CLS_ROUTE | |
2239 | rth->u.dst.tclassid = itag; | |
2240 | #endif | |
2241 | rth->rt_iif = | |
2242 | rth->fl.iif = dev->ifindex; | |
84a885f4 | 2243 | rth->u.dst.dev = net->loopback_dev; |
1da177e4 LT |
2244 | dev_hold(rth->u.dst.dev); |
2245 | rth->idev = in_dev_get(rth->u.dst.dev); | |
2246 | rth->rt_gateway = daddr; | |
2247 | rth->rt_spec_dst= spec_dst; | |
2248 | rth->u.dst.input= ip_local_deliver; | |
2249 | rth->rt_flags = flags|RTCF_LOCAL; | |
2250 | if (res.type == RTN_UNREACHABLE) { | |
2251 | rth->u.dst.input= ip_error; | |
2252 | rth->u.dst.error= -err; | |
2253 | rth->rt_flags &= ~RTCF_LOCAL; | |
2254 | } | |
2255 | rth->rt_type = res.type; | |
e84f84f2 | 2256 | hash = rt_hash(daddr, saddr, fl.iif, rt_genid(net)); |
511c3f92 | 2257 | err = rt_intern_hash(hash, rth, NULL, skb); |
1da177e4 LT |
2258 | goto done; |
2259 | ||
2260 | no_route: | |
2261 | RT_CACHE_STAT_INC(in_no_route); | |
2262 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE); | |
2263 | res.type = RTN_UNREACHABLE; | |
7f53878d MC |
2264 | if (err == -ESRCH) |
2265 | err = -ENETUNREACH; | |
1da177e4 LT |
2266 | goto local_input; |
2267 | ||
2268 | /* | |
2269 | * Do not cache martian addresses: they should be logged (RFC1812) | |
2270 | */ | |
2271 | martian_destination: | |
2272 | RT_CACHE_STAT_INC(in_martian_dst); | |
2273 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
2274 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) | |
673d57e7 HH |
2275 | printk(KERN_WARNING "martian destination %pI4 from %pI4, dev %s\n", |
2276 | &daddr, &saddr, dev->name); | |
1da177e4 | 2277 | #endif |
2c2910a4 DE |
2278 | |
2279 | e_hostunreach: | |
e905a9ed YH |
2280 | err = -EHOSTUNREACH; |
2281 | goto done; | |
2c2910a4 | 2282 | |
1da177e4 LT |
2283 | e_inval: |
2284 | err = -EINVAL; | |
2285 | goto done; | |
2286 | ||
2287 | e_nobufs: | |
2288 | err = -ENOBUFS; | |
2289 | goto done; | |
2290 | ||
2291 | martian_source: | |
2292 | ip_handle_martian_source(dev, in_dev, skb, daddr, saddr); | |
2293 | goto e_inval; | |
2294 | } | |
2295 | ||
9e12bb22 | 2296 | int ip_route_input(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
2297 | u8 tos, struct net_device *dev) |
2298 | { | |
2299 | struct rtable * rth; | |
2300 | unsigned hash; | |
2301 | int iif = dev->ifindex; | |
b5921910 | 2302 | struct net *net; |
1da177e4 | 2303 | |
c346dca1 | 2304 | net = dev_net(dev); |
1080d709 NH |
2305 | |
2306 | if (!rt_caching(net)) | |
2307 | goto skip_cache; | |
2308 | ||
1da177e4 | 2309 | tos &= IPTOS_RT_MASK; |
e84f84f2 | 2310 | hash = rt_hash(daddr, saddr, iif, rt_genid(net)); |
1da177e4 LT |
2311 | |
2312 | rcu_read_lock(); | |
2313 | for (rth = rcu_dereference(rt_hash_table[hash].chain); rth; | |
093c2ca4 | 2314 | rth = rcu_dereference(rth->u.dst.rt_next)) { |
c0b8c32b SH |
2315 | if (((rth->fl.fl4_dst ^ daddr) | |
2316 | (rth->fl.fl4_src ^ saddr) | | |
2317 | (rth->fl.iif ^ iif) | | |
2318 | rth->fl.oif | | |
2319 | (rth->fl.fl4_tos ^ tos)) == 0 && | |
47dcf0cb | 2320 | rth->fl.mark == skb->mark && |
878628fb | 2321 | net_eq(dev_net(rth->u.dst.dev), net) && |
e84f84f2 | 2322 | !rt_is_expired(rth)) { |
03f49f34 | 2323 | dst_use(&rth->u.dst, jiffies); |
1da177e4 LT |
2324 | RT_CACHE_STAT_INC(in_hit); |
2325 | rcu_read_unlock(); | |
adf30907 | 2326 | skb_dst_set(skb, &rth->u.dst); |
1da177e4 LT |
2327 | return 0; |
2328 | } | |
2329 | RT_CACHE_STAT_INC(in_hlist_search); | |
2330 | } | |
2331 | rcu_read_unlock(); | |
2332 | ||
1080d709 | 2333 | skip_cache: |
1da177e4 LT |
2334 | /* Multicast recognition logic is moved from route cache to here. |
2335 | The problem was that too many Ethernet cards have broken/missing | |
2336 | hardware multicast filters :-( As result the host on multicasting | |
2337 | network acquires a lot of useless route cache entries, sort of | |
2338 | SDR messages from all the world. Now we try to get rid of them. | |
2339 | Really, provided software IP multicast filter is organized | |
2340 | reasonably (at least, hashed), it does not result in a slowdown | |
2341 | comparing with route cache reject entries. | |
2342 | Note, that multicast routers are not affected, because | |
2343 | route cache entry is created eventually. | |
2344 | */ | |
f97c1e0c | 2345 | if (ipv4_is_multicast(daddr)) { |
1da177e4 LT |
2346 | struct in_device *in_dev; |
2347 | ||
2348 | rcu_read_lock(); | |
e5ed6399 | 2349 | if ((in_dev = __in_dev_get_rcu(dev)) != NULL) { |
1da177e4 | 2350 | int our = ip_check_mc(in_dev, daddr, saddr, |
eddc9ec5 | 2351 | ip_hdr(skb)->protocol); |
1da177e4 LT |
2352 | if (our |
2353 | #ifdef CONFIG_IP_MROUTE | |
9d4fb27d JP |
2354 | || |
2355 | (!ipv4_is_local_multicast(daddr) && | |
2356 | IN_DEV_MFORWARD(in_dev)) | |
1da177e4 | 2357 | #endif |
9d4fb27d | 2358 | ) { |
1da177e4 LT |
2359 | rcu_read_unlock(); |
2360 | return ip_route_input_mc(skb, daddr, saddr, | |
2361 | tos, dev, our); | |
2362 | } | |
2363 | } | |
2364 | rcu_read_unlock(); | |
2365 | return -EINVAL; | |
2366 | } | |
2367 | return ip_route_input_slow(skb, daddr, saddr, tos, dev); | |
2368 | } | |
2369 | ||
5969f71d SH |
2370 | static int __mkroute_output(struct rtable **result, |
2371 | struct fib_result *res, | |
2372 | const struct flowi *fl, | |
2373 | const struct flowi *oldflp, | |
2374 | struct net_device *dev_out, | |
2375 | unsigned flags) | |
1da177e4 LT |
2376 | { |
2377 | struct rtable *rth; | |
2378 | struct in_device *in_dev; | |
2379 | u32 tos = RT_FL_TOS(oldflp); | |
2380 | int err = 0; | |
2381 | ||
f97c1e0c | 2382 | if (ipv4_is_loopback(fl->fl4_src) && !(dev_out->flags&IFF_LOOPBACK)) |
1da177e4 LT |
2383 | return -EINVAL; |
2384 | ||
e448515c | 2385 | if (fl->fl4_dst == htonl(0xFFFFFFFF)) |
1da177e4 | 2386 | res->type = RTN_BROADCAST; |
f97c1e0c | 2387 | else if (ipv4_is_multicast(fl->fl4_dst)) |
1da177e4 | 2388 | res->type = RTN_MULTICAST; |
1e637c74 | 2389 | else if (ipv4_is_lbcast(fl->fl4_dst) || ipv4_is_zeronet(fl->fl4_dst)) |
1da177e4 LT |
2390 | return -EINVAL; |
2391 | ||
2392 | if (dev_out->flags & IFF_LOOPBACK) | |
2393 | flags |= RTCF_LOCAL; | |
2394 | ||
2395 | /* get work reference to inet device */ | |
2396 | in_dev = in_dev_get(dev_out); | |
2397 | if (!in_dev) | |
2398 | return -EINVAL; | |
2399 | ||
2400 | if (res->type == RTN_BROADCAST) { | |
2401 | flags |= RTCF_BROADCAST | RTCF_LOCAL; | |
2402 | if (res->fi) { | |
2403 | fib_info_put(res->fi); | |
2404 | res->fi = NULL; | |
2405 | } | |
2406 | } else if (res->type == RTN_MULTICAST) { | |
2407 | flags |= RTCF_MULTICAST|RTCF_LOCAL; | |
e905a9ed | 2408 | if (!ip_check_mc(in_dev, oldflp->fl4_dst, oldflp->fl4_src, |
1da177e4 LT |
2409 | oldflp->proto)) |
2410 | flags &= ~RTCF_LOCAL; | |
2411 | /* If multicast route do not exist use | |
2412 | default one, but do not gateway in this case. | |
2413 | Yes, it is hack. | |
2414 | */ | |
2415 | if (res->fi && res->prefixlen < 4) { | |
2416 | fib_info_put(res->fi); | |
2417 | res->fi = NULL; | |
2418 | } | |
2419 | } | |
2420 | ||
2421 | ||
2422 | rth = dst_alloc(&ipv4_dst_ops); | |
2423 | if (!rth) { | |
2424 | err = -ENOBUFS; | |
2425 | goto cleanup; | |
e905a9ed | 2426 | } |
1da177e4 | 2427 | |
ce723d8e | 2428 | atomic_set(&rth->u.dst.__refcnt, 1); |
1da177e4 | 2429 | rth->u.dst.flags= DST_HOST; |
42f811b8 | 2430 | if (IN_DEV_CONF_GET(in_dev, NOXFRM)) |
1da177e4 | 2431 | rth->u.dst.flags |= DST_NOXFRM; |
42f811b8 | 2432 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
1da177e4 LT |
2433 | rth->u.dst.flags |= DST_NOPOLICY; |
2434 | ||
2435 | rth->fl.fl4_dst = oldflp->fl4_dst; | |
2436 | rth->fl.fl4_tos = tos; | |
2437 | rth->fl.fl4_src = oldflp->fl4_src; | |
2438 | rth->fl.oif = oldflp->oif; | |
47dcf0cb | 2439 | rth->fl.mark = oldflp->mark; |
1da177e4 LT |
2440 | rth->rt_dst = fl->fl4_dst; |
2441 | rth->rt_src = fl->fl4_src; | |
2442 | rth->rt_iif = oldflp->oif ? : dev_out->ifindex; | |
e905a9ed | 2443 | /* get references to the devices that are to be hold by the routing |
1da177e4 LT |
2444 | cache entry */ |
2445 | rth->u.dst.dev = dev_out; | |
2446 | dev_hold(dev_out); | |
2447 | rth->idev = in_dev_get(dev_out); | |
2448 | rth->rt_gateway = fl->fl4_dst; | |
2449 | rth->rt_spec_dst= fl->fl4_src; | |
2450 | ||
2451 | rth->u.dst.output=ip_output; | |
d11a4dc1 | 2452 | rth->u.dst.obsolete = -1; |
e84f84f2 | 2453 | rth->rt_genid = rt_genid(dev_net(dev_out)); |
1da177e4 LT |
2454 | |
2455 | RT_CACHE_STAT_INC(out_slow_tot); | |
2456 | ||
2457 | if (flags & RTCF_LOCAL) { | |
2458 | rth->u.dst.input = ip_local_deliver; | |
2459 | rth->rt_spec_dst = fl->fl4_dst; | |
2460 | } | |
2461 | if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { | |
2462 | rth->rt_spec_dst = fl->fl4_src; | |
e905a9ed | 2463 | if (flags & RTCF_LOCAL && |
1da177e4 LT |
2464 | !(dev_out->flags & IFF_LOOPBACK)) { |
2465 | rth->u.dst.output = ip_mc_output; | |
2466 | RT_CACHE_STAT_INC(out_slow_mc); | |
2467 | } | |
2468 | #ifdef CONFIG_IP_MROUTE | |
2469 | if (res->type == RTN_MULTICAST) { | |
2470 | if (IN_DEV_MFORWARD(in_dev) && | |
f97c1e0c | 2471 | !ipv4_is_local_multicast(oldflp->fl4_dst)) { |
1da177e4 LT |
2472 | rth->u.dst.input = ip_mr_input; |
2473 | rth->u.dst.output = ip_mc_output; | |
2474 | } | |
2475 | } | |
2476 | #endif | |
2477 | } | |
2478 | ||
2479 | rt_set_nexthop(rth, res, 0); | |
2480 | ||
2481 | rth->rt_flags = flags; | |
2482 | ||
2483 | *result = rth; | |
2484 | cleanup: | |
2485 | /* release work reference to inet device */ | |
2486 | in_dev_put(in_dev); | |
2487 | ||
2488 | return err; | |
2489 | } | |
2490 | ||
5969f71d SH |
2491 | static int ip_mkroute_output(struct rtable **rp, |
2492 | struct fib_result *res, | |
2493 | const struct flowi *fl, | |
2494 | const struct flowi *oldflp, | |
2495 | struct net_device *dev_out, | |
2496 | unsigned flags) | |
1da177e4 | 2497 | { |
7abaa27c | 2498 | struct rtable *rth = NULL; |
1da177e4 LT |
2499 | int err = __mkroute_output(&rth, res, fl, oldflp, dev_out, flags); |
2500 | unsigned hash; | |
2501 | if (err == 0) { | |
b00180de | 2502 | hash = rt_hash(oldflp->fl4_dst, oldflp->fl4_src, oldflp->oif, |
e84f84f2 | 2503 | rt_genid(dev_net(dev_out))); |
511c3f92 | 2504 | err = rt_intern_hash(hash, rth, rp, NULL); |
1da177e4 | 2505 | } |
e905a9ed | 2506 | |
1da177e4 LT |
2507 | return err; |
2508 | } | |
2509 | ||
1da177e4 LT |
2510 | /* |
2511 | * Major route resolver routine. | |
2512 | */ | |
2513 | ||
b40afd0e DL |
2514 | static int ip_route_output_slow(struct net *net, struct rtable **rp, |
2515 | const struct flowi *oldflp) | |
1da177e4 LT |
2516 | { |
2517 | u32 tos = RT_FL_TOS(oldflp); | |
2518 | struct flowi fl = { .nl_u = { .ip4_u = | |
2519 | { .daddr = oldflp->fl4_dst, | |
2520 | .saddr = oldflp->fl4_src, | |
2521 | .tos = tos & IPTOS_RT_MASK, | |
2522 | .scope = ((tos & RTO_ONLINK) ? | |
2523 | RT_SCOPE_LINK : | |
2524 | RT_SCOPE_UNIVERSE), | |
1da177e4 | 2525 | } }, |
47dcf0cb | 2526 | .mark = oldflp->mark, |
b40afd0e | 2527 | .iif = net->loopback_dev->ifindex, |
1da177e4 LT |
2528 | .oif = oldflp->oif }; |
2529 | struct fib_result res; | |
2530 | unsigned flags = 0; | |
2531 | struct net_device *dev_out = NULL; | |
2532 | int free_res = 0; | |
2533 | int err; | |
2534 | ||
2535 | ||
2536 | res.fi = NULL; | |
2537 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
2538 | res.r = NULL; | |
2539 | #endif | |
2540 | ||
2541 | if (oldflp->fl4_src) { | |
2542 | err = -EINVAL; | |
f97c1e0c | 2543 | if (ipv4_is_multicast(oldflp->fl4_src) || |
1e637c74 | 2544 | ipv4_is_lbcast(oldflp->fl4_src) || |
f97c1e0c | 2545 | ipv4_is_zeronet(oldflp->fl4_src)) |
1da177e4 LT |
2546 | goto out; |
2547 | ||
1da177e4 LT |
2548 | /* I removed check for oif == dev_out->oif here. |
2549 | It was wrong for two reasons: | |
1ab35276 DL |
2550 | 1. ip_dev_find(net, saddr) can return wrong iface, if saddr |
2551 | is assigned to multiple interfaces. | |
1da177e4 LT |
2552 | 2. Moreover, we are allowed to send packets with saddr |
2553 | of another iface. --ANK | |
2554 | */ | |
2555 | ||
9d4fb27d JP |
2556 | if (oldflp->oif == 0 && |
2557 | (ipv4_is_multicast(oldflp->fl4_dst) || | |
2558 | oldflp->fl4_dst == htonl(0xFFFFFFFF))) { | |
a210d01a JA |
2559 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
2560 | dev_out = ip_dev_find(net, oldflp->fl4_src); | |
2561 | if (dev_out == NULL) | |
2562 | goto out; | |
2563 | ||
1da177e4 LT |
2564 | /* Special hack: user can direct multicasts |
2565 | and limited broadcast via necessary interface | |
2566 | without fiddling with IP_MULTICAST_IF or IP_PKTINFO. | |
2567 | This hack is not just for fun, it allows | |
2568 | vic,vat and friends to work. | |
2569 | They bind socket to loopback, set ttl to zero | |
2570 | and expect that it will work. | |
2571 | From the viewpoint of routing cache they are broken, | |
2572 | because we are not allowed to build multicast path | |
2573 | with loopback source addr (look, routing cache | |
2574 | cannot know, that ttl is zero, so that packet | |
2575 | will not leave this host and route is valid). | |
2576 | Luckily, this hack is good workaround. | |
2577 | */ | |
2578 | ||
2579 | fl.oif = dev_out->ifindex; | |
2580 | goto make_route; | |
2581 | } | |
a210d01a JA |
2582 | |
2583 | if (!(oldflp->flags & FLOWI_FLAG_ANYSRC)) { | |
2584 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ | |
2585 | dev_out = ip_dev_find(net, oldflp->fl4_src); | |
2586 | if (dev_out == NULL) | |
2587 | goto out; | |
1da177e4 | 2588 | dev_put(dev_out); |
a210d01a JA |
2589 | dev_out = NULL; |
2590 | } | |
1da177e4 LT |
2591 | } |
2592 | ||
2593 | ||
2594 | if (oldflp->oif) { | |
b40afd0e | 2595 | dev_out = dev_get_by_index(net, oldflp->oif); |
1da177e4 LT |
2596 | err = -ENODEV; |
2597 | if (dev_out == NULL) | |
2598 | goto out; | |
e5ed6399 HX |
2599 | |
2600 | /* RACE: Check return value of inet_select_addr instead. */ | |
2601 | if (__in_dev_get_rtnl(dev_out) == NULL) { | |
1da177e4 LT |
2602 | dev_put(dev_out); |
2603 | goto out; /* Wrong error code */ | |
2604 | } | |
2605 | ||
f97c1e0c JP |
2606 | if (ipv4_is_local_multicast(oldflp->fl4_dst) || |
2607 | oldflp->fl4_dst == htonl(0xFFFFFFFF)) { | |
1da177e4 LT |
2608 | if (!fl.fl4_src) |
2609 | fl.fl4_src = inet_select_addr(dev_out, 0, | |
2610 | RT_SCOPE_LINK); | |
2611 | goto make_route; | |
2612 | } | |
2613 | if (!fl.fl4_src) { | |
f97c1e0c | 2614 | if (ipv4_is_multicast(oldflp->fl4_dst)) |
1da177e4 LT |
2615 | fl.fl4_src = inet_select_addr(dev_out, 0, |
2616 | fl.fl4_scope); | |
2617 | else if (!oldflp->fl4_dst) | |
2618 | fl.fl4_src = inet_select_addr(dev_out, 0, | |
2619 | RT_SCOPE_HOST); | |
2620 | } | |
2621 | } | |
2622 | ||
2623 | if (!fl.fl4_dst) { | |
2624 | fl.fl4_dst = fl.fl4_src; | |
2625 | if (!fl.fl4_dst) | |
2626 | fl.fl4_dst = fl.fl4_src = htonl(INADDR_LOOPBACK); | |
2627 | if (dev_out) | |
2628 | dev_put(dev_out); | |
b40afd0e | 2629 | dev_out = net->loopback_dev; |
1da177e4 | 2630 | dev_hold(dev_out); |
b40afd0e | 2631 | fl.oif = net->loopback_dev->ifindex; |
1da177e4 LT |
2632 | res.type = RTN_LOCAL; |
2633 | flags |= RTCF_LOCAL; | |
2634 | goto make_route; | |
2635 | } | |
2636 | ||
b40afd0e | 2637 | if (fib_lookup(net, &fl, &res)) { |
1da177e4 LT |
2638 | res.fi = NULL; |
2639 | if (oldflp->oif) { | |
2640 | /* Apparently, routing tables are wrong. Assume, | |
2641 | that the destination is on link. | |
2642 | ||
2643 | WHY? DW. | |
2644 | Because we are allowed to send to iface | |
2645 | even if it has NO routes and NO assigned | |
2646 | addresses. When oif is specified, routing | |
2647 | tables are looked up with only one purpose: | |
2648 | to catch if destination is gatewayed, rather than | |
2649 | direct. Moreover, if MSG_DONTROUTE is set, | |
2650 | we send packet, ignoring both routing tables | |
2651 | and ifaddr state. --ANK | |
2652 | ||
2653 | ||
2654 | We could make it even if oif is unknown, | |
2655 | likely IPv6, but we do not. | |
2656 | */ | |
2657 | ||
2658 | if (fl.fl4_src == 0) | |
2659 | fl.fl4_src = inet_select_addr(dev_out, 0, | |
2660 | RT_SCOPE_LINK); | |
2661 | res.type = RTN_UNICAST; | |
2662 | goto make_route; | |
2663 | } | |
2664 | if (dev_out) | |
2665 | dev_put(dev_out); | |
2666 | err = -ENETUNREACH; | |
2667 | goto out; | |
2668 | } | |
2669 | free_res = 1; | |
2670 | ||
2671 | if (res.type == RTN_LOCAL) { | |
2672 | if (!fl.fl4_src) | |
2673 | fl.fl4_src = fl.fl4_dst; | |
2674 | if (dev_out) | |
2675 | dev_put(dev_out); | |
b40afd0e | 2676 | dev_out = net->loopback_dev; |
1da177e4 LT |
2677 | dev_hold(dev_out); |
2678 | fl.oif = dev_out->ifindex; | |
2679 | if (res.fi) | |
2680 | fib_info_put(res.fi); | |
2681 | res.fi = NULL; | |
2682 | flags |= RTCF_LOCAL; | |
2683 | goto make_route; | |
2684 | } | |
2685 | ||
2686 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
2687 | if (res.fi->fib_nhs > 1 && fl.oif == 0) | |
2688 | fib_select_multipath(&fl, &res); | |
2689 | else | |
2690 | #endif | |
2691 | if (!res.prefixlen && res.type == RTN_UNICAST && !fl.oif) | |
b40afd0e | 2692 | fib_select_default(net, &fl, &res); |
1da177e4 LT |
2693 | |
2694 | if (!fl.fl4_src) | |
2695 | fl.fl4_src = FIB_RES_PREFSRC(res); | |
2696 | ||
2697 | if (dev_out) | |
2698 | dev_put(dev_out); | |
2699 | dev_out = FIB_RES_DEV(res); | |
2700 | dev_hold(dev_out); | |
2701 | fl.oif = dev_out->ifindex; | |
2702 | ||
2703 | ||
2704 | make_route: | |
2705 | err = ip_mkroute_output(rp, &res, &fl, oldflp, dev_out, flags); | |
2706 | ||
2707 | ||
2708 | if (free_res) | |
2709 | fib_res_put(&res); | |
2710 | if (dev_out) | |
2711 | dev_put(dev_out); | |
2712 | out: return err; | |
2713 | } | |
2714 | ||
611c183e DL |
2715 | int __ip_route_output_key(struct net *net, struct rtable **rp, |
2716 | const struct flowi *flp) | |
1da177e4 LT |
2717 | { |
2718 | unsigned hash; | |
2719 | struct rtable *rth; | |
2720 | ||
1080d709 NH |
2721 | if (!rt_caching(net)) |
2722 | goto slow_output; | |
2723 | ||
e84f84f2 | 2724 | hash = rt_hash(flp->fl4_dst, flp->fl4_src, flp->oif, rt_genid(net)); |
1da177e4 LT |
2725 | |
2726 | rcu_read_lock_bh(); | |
a898def2 PM |
2727 | for (rth = rcu_dereference_bh(rt_hash_table[hash].chain); rth; |
2728 | rth = rcu_dereference_bh(rth->u.dst.rt_next)) { | |
1da177e4 LT |
2729 | if (rth->fl.fl4_dst == flp->fl4_dst && |
2730 | rth->fl.fl4_src == flp->fl4_src && | |
2731 | rth->fl.iif == 0 && | |
2732 | rth->fl.oif == flp->oif && | |
47dcf0cb | 2733 | rth->fl.mark == flp->mark && |
1da177e4 | 2734 | !((rth->fl.fl4_tos ^ flp->fl4_tos) & |
b5921910 | 2735 | (IPTOS_RT_MASK | RTO_ONLINK)) && |
878628fb | 2736 | net_eq(dev_net(rth->u.dst.dev), net) && |
e84f84f2 | 2737 | !rt_is_expired(rth)) { |
03f49f34 | 2738 | dst_use(&rth->u.dst, jiffies); |
1da177e4 LT |
2739 | RT_CACHE_STAT_INC(out_hit); |
2740 | rcu_read_unlock_bh(); | |
2741 | *rp = rth; | |
2742 | return 0; | |
2743 | } | |
2744 | RT_CACHE_STAT_INC(out_hlist_search); | |
2745 | } | |
2746 | rcu_read_unlock_bh(); | |
2747 | ||
1080d709 | 2748 | slow_output: |
611c183e | 2749 | return ip_route_output_slow(net, rp, flp); |
1da177e4 LT |
2750 | } |
2751 | ||
d8c97a94 ACM |
2752 | EXPORT_SYMBOL_GPL(__ip_route_output_key); |
2753 | ||
14e50e57 DM |
2754 | static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu) |
2755 | { | |
2756 | } | |
2757 | ||
2758 | static struct dst_ops ipv4_dst_blackhole_ops = { | |
2759 | .family = AF_INET, | |
09640e63 | 2760 | .protocol = cpu_to_be16(ETH_P_IP), |
14e50e57 DM |
2761 | .destroy = ipv4_dst_destroy, |
2762 | .check = ipv4_dst_check, | |
2763 | .update_pmtu = ipv4_rt_blackhole_update_pmtu, | |
e2422970 | 2764 | .entries = ATOMIC_INIT(0), |
14e50e57 DM |
2765 | }; |
2766 | ||
2767 | ||
e84f84f2 | 2768 | static int ipv4_dst_blackhole(struct net *net, struct rtable **rp, struct flowi *flp) |
14e50e57 DM |
2769 | { |
2770 | struct rtable *ort = *rp; | |
2771 | struct rtable *rt = (struct rtable *) | |
2772 | dst_alloc(&ipv4_dst_blackhole_ops); | |
2773 | ||
2774 | if (rt) { | |
2775 | struct dst_entry *new = &rt->u.dst; | |
2776 | ||
2777 | atomic_set(&new->__refcnt, 1); | |
2778 | new->__use = 1; | |
352e512c HX |
2779 | new->input = dst_discard; |
2780 | new->output = dst_discard; | |
14e50e57 DM |
2781 | memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32)); |
2782 | ||
2783 | new->dev = ort->u.dst.dev; | |
2784 | if (new->dev) | |
2785 | dev_hold(new->dev); | |
2786 | ||
2787 | rt->fl = ort->fl; | |
2788 | ||
2789 | rt->idev = ort->idev; | |
2790 | if (rt->idev) | |
2791 | in_dev_hold(rt->idev); | |
e84f84f2 | 2792 | rt->rt_genid = rt_genid(net); |
14e50e57 DM |
2793 | rt->rt_flags = ort->rt_flags; |
2794 | rt->rt_type = ort->rt_type; | |
2795 | rt->rt_dst = ort->rt_dst; | |
2796 | rt->rt_src = ort->rt_src; | |
2797 | rt->rt_iif = ort->rt_iif; | |
2798 | rt->rt_gateway = ort->rt_gateway; | |
2799 | rt->rt_spec_dst = ort->rt_spec_dst; | |
2800 | rt->peer = ort->peer; | |
2801 | if (rt->peer) | |
2802 | atomic_inc(&rt->peer->refcnt); | |
2803 | ||
2804 | dst_free(new); | |
2805 | } | |
2806 | ||
2807 | dst_release(&(*rp)->u.dst); | |
2808 | *rp = rt; | |
2809 | return (rt ? 0 : -ENOMEM); | |
2810 | } | |
2811 | ||
f1b050bf DL |
2812 | int ip_route_output_flow(struct net *net, struct rtable **rp, struct flowi *flp, |
2813 | struct sock *sk, int flags) | |
1da177e4 LT |
2814 | { |
2815 | int err; | |
2816 | ||
f1b050bf | 2817 | if ((err = __ip_route_output_key(net, rp, flp)) != 0) |
1da177e4 LT |
2818 | return err; |
2819 | ||
2820 | if (flp->proto) { | |
2821 | if (!flp->fl4_src) | |
2822 | flp->fl4_src = (*rp)->rt_src; | |
2823 | if (!flp->fl4_dst) | |
2824 | flp->fl4_dst = (*rp)->rt_dst; | |
52479b62 | 2825 | err = __xfrm_lookup(net, (struct dst_entry **)rp, flp, sk, |
bb72845e | 2826 | flags ? XFRM_LOOKUP_WAIT : 0); |
14e50e57 | 2827 | if (err == -EREMOTE) |
e84f84f2 | 2828 | err = ipv4_dst_blackhole(net, rp, flp); |
14e50e57 DM |
2829 | |
2830 | return err; | |
1da177e4 LT |
2831 | } |
2832 | ||
2833 | return 0; | |
2834 | } | |
2835 | ||
d8c97a94 ACM |
2836 | EXPORT_SYMBOL_GPL(ip_route_output_flow); |
2837 | ||
f206351a | 2838 | int ip_route_output_key(struct net *net, struct rtable **rp, struct flowi *flp) |
1da177e4 | 2839 | { |
f206351a | 2840 | return ip_route_output_flow(net, rp, flp, NULL, 0); |
1da177e4 LT |
2841 | } |
2842 | ||
4feb88e5 BT |
2843 | static int rt_fill_info(struct net *net, |
2844 | struct sk_buff *skb, u32 pid, u32 seq, int event, | |
b6544c0b | 2845 | int nowait, unsigned int flags) |
1da177e4 | 2846 | { |
511c3f92 | 2847 | struct rtable *rt = skb_rtable(skb); |
1da177e4 | 2848 | struct rtmsg *r; |
be403ea1 | 2849 | struct nlmsghdr *nlh; |
e3703b3d TG |
2850 | long expires; |
2851 | u32 id = 0, ts = 0, tsage = 0, error; | |
be403ea1 TG |
2852 | |
2853 | nlh = nlmsg_put(skb, pid, seq, event, sizeof(*r), flags); | |
2854 | if (nlh == NULL) | |
26932566 | 2855 | return -EMSGSIZE; |
be403ea1 TG |
2856 | |
2857 | r = nlmsg_data(nlh); | |
1da177e4 LT |
2858 | r->rtm_family = AF_INET; |
2859 | r->rtm_dst_len = 32; | |
2860 | r->rtm_src_len = 0; | |
2861 | r->rtm_tos = rt->fl.fl4_tos; | |
2862 | r->rtm_table = RT_TABLE_MAIN; | |
be403ea1 | 2863 | NLA_PUT_U32(skb, RTA_TABLE, RT_TABLE_MAIN); |
1da177e4 LT |
2864 | r->rtm_type = rt->rt_type; |
2865 | r->rtm_scope = RT_SCOPE_UNIVERSE; | |
2866 | r->rtm_protocol = RTPROT_UNSPEC; | |
2867 | r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED; | |
2868 | if (rt->rt_flags & RTCF_NOTIFY) | |
2869 | r->rtm_flags |= RTM_F_NOTIFY; | |
be403ea1 | 2870 | |
17fb2c64 | 2871 | NLA_PUT_BE32(skb, RTA_DST, rt->rt_dst); |
be403ea1 | 2872 | |
1da177e4 LT |
2873 | if (rt->fl.fl4_src) { |
2874 | r->rtm_src_len = 32; | |
17fb2c64 | 2875 | NLA_PUT_BE32(skb, RTA_SRC, rt->fl.fl4_src); |
1da177e4 LT |
2876 | } |
2877 | if (rt->u.dst.dev) | |
be403ea1 | 2878 | NLA_PUT_U32(skb, RTA_OIF, rt->u.dst.dev->ifindex); |
1da177e4 LT |
2879 | #ifdef CONFIG_NET_CLS_ROUTE |
2880 | if (rt->u.dst.tclassid) | |
be403ea1 | 2881 | NLA_PUT_U32(skb, RTA_FLOW, rt->u.dst.tclassid); |
1da177e4 LT |
2882 | #endif |
2883 | if (rt->fl.iif) | |
17fb2c64 | 2884 | NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_spec_dst); |
1da177e4 | 2885 | else if (rt->rt_src != rt->fl.fl4_src) |
17fb2c64 | 2886 | NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_src); |
be403ea1 | 2887 | |
1da177e4 | 2888 | if (rt->rt_dst != rt->rt_gateway) |
17fb2c64 | 2889 | NLA_PUT_BE32(skb, RTA_GATEWAY, rt->rt_gateway); |
be403ea1 | 2890 | |
1da177e4 | 2891 | if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0) |
be403ea1 TG |
2892 | goto nla_put_failure; |
2893 | ||
e3703b3d TG |
2894 | error = rt->u.dst.error; |
2895 | expires = rt->u.dst.expires ? rt->u.dst.expires - jiffies : 0; | |
1da177e4 | 2896 | if (rt->peer) { |
2c1409a0 | 2897 | id = atomic_read(&rt->peer->ip_id_count) & 0xffff; |
1da177e4 | 2898 | if (rt->peer->tcp_ts_stamp) { |
e3703b3d | 2899 | ts = rt->peer->tcp_ts; |
9d729f72 | 2900 | tsage = get_seconds() - rt->peer->tcp_ts_stamp; |
1da177e4 LT |
2901 | } |
2902 | } | |
be403ea1 | 2903 | |
1da177e4 LT |
2904 | if (rt->fl.iif) { |
2905 | #ifdef CONFIG_IP_MROUTE | |
e448515c | 2906 | __be32 dst = rt->rt_dst; |
1da177e4 | 2907 | |
f97c1e0c | 2908 | if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) && |
4feb88e5 BT |
2909 | IPV4_DEVCONF_ALL(net, MC_FORWARDING)) { |
2910 | int err = ipmr_get_route(net, skb, r, nowait); | |
1da177e4 LT |
2911 | if (err <= 0) { |
2912 | if (!nowait) { | |
2913 | if (err == 0) | |
2914 | return 0; | |
be403ea1 | 2915 | goto nla_put_failure; |
1da177e4 LT |
2916 | } else { |
2917 | if (err == -EMSGSIZE) | |
be403ea1 | 2918 | goto nla_put_failure; |
e3703b3d | 2919 | error = err; |
1da177e4 LT |
2920 | } |
2921 | } | |
2922 | } else | |
2923 | #endif | |
be403ea1 | 2924 | NLA_PUT_U32(skb, RTA_IIF, rt->fl.iif); |
1da177e4 LT |
2925 | } |
2926 | ||
e3703b3d TG |
2927 | if (rtnl_put_cacheinfo(skb, &rt->u.dst, id, ts, tsage, |
2928 | expires, error) < 0) | |
2929 | goto nla_put_failure; | |
be403ea1 TG |
2930 | |
2931 | return nlmsg_end(skb, nlh); | |
1da177e4 | 2932 | |
be403ea1 | 2933 | nla_put_failure: |
26932566 PM |
2934 | nlmsg_cancel(skb, nlh); |
2935 | return -EMSGSIZE; | |
1da177e4 LT |
2936 | } |
2937 | ||
63f3444f | 2938 | static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg) |
1da177e4 | 2939 | { |
3b1e0a65 | 2940 | struct net *net = sock_net(in_skb->sk); |
d889ce3b TG |
2941 | struct rtmsg *rtm; |
2942 | struct nlattr *tb[RTA_MAX+1]; | |
1da177e4 | 2943 | struct rtable *rt = NULL; |
9e12bb22 AV |
2944 | __be32 dst = 0; |
2945 | __be32 src = 0; | |
2946 | u32 iif; | |
d889ce3b | 2947 | int err; |
1da177e4 LT |
2948 | struct sk_buff *skb; |
2949 | ||
d889ce3b TG |
2950 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy); |
2951 | if (err < 0) | |
2952 | goto errout; | |
2953 | ||
2954 | rtm = nlmsg_data(nlh); | |
2955 | ||
1da177e4 | 2956 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
d889ce3b TG |
2957 | if (skb == NULL) { |
2958 | err = -ENOBUFS; | |
2959 | goto errout; | |
2960 | } | |
1da177e4 LT |
2961 | |
2962 | /* Reserve room for dummy headers, this skb can pass | |
2963 | through good chunk of routing engine. | |
2964 | */ | |
459a98ed | 2965 | skb_reset_mac_header(skb); |
c1d2bbe1 | 2966 | skb_reset_network_header(skb); |
d2c962b8 SH |
2967 | |
2968 | /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */ | |
eddc9ec5 | 2969 | ip_hdr(skb)->protocol = IPPROTO_ICMP; |
1da177e4 LT |
2970 | skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr)); |
2971 | ||
17fb2c64 AV |
2972 | src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0; |
2973 | dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0; | |
d889ce3b | 2974 | iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0; |
1da177e4 LT |
2975 | |
2976 | if (iif) { | |
d889ce3b TG |
2977 | struct net_device *dev; |
2978 | ||
1937504d | 2979 | dev = __dev_get_by_index(net, iif); |
d889ce3b TG |
2980 | if (dev == NULL) { |
2981 | err = -ENODEV; | |
2982 | goto errout_free; | |
2983 | } | |
2984 | ||
1da177e4 LT |
2985 | skb->protocol = htons(ETH_P_IP); |
2986 | skb->dev = dev; | |
2987 | local_bh_disable(); | |
2988 | err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev); | |
2989 | local_bh_enable(); | |
d889ce3b | 2990 | |
511c3f92 | 2991 | rt = skb_rtable(skb); |
d889ce3b | 2992 | if (err == 0 && rt->u.dst.error) |
1da177e4 LT |
2993 | err = -rt->u.dst.error; |
2994 | } else { | |
d889ce3b TG |
2995 | struct flowi fl = { |
2996 | .nl_u = { | |
2997 | .ip4_u = { | |
2998 | .daddr = dst, | |
2999 | .saddr = src, | |
3000 | .tos = rtm->rtm_tos, | |
3001 | }, | |
3002 | }, | |
3003 | .oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0, | |
3004 | }; | |
1937504d | 3005 | err = ip_route_output_key(net, &rt, &fl); |
1da177e4 | 3006 | } |
d889ce3b | 3007 | |
1da177e4 | 3008 | if (err) |
d889ce3b | 3009 | goto errout_free; |
1da177e4 | 3010 | |
adf30907 | 3011 | skb_dst_set(skb, &rt->u.dst); |
1da177e4 LT |
3012 | if (rtm->rtm_flags & RTM_F_NOTIFY) |
3013 | rt->rt_flags |= RTCF_NOTIFY; | |
3014 | ||
4feb88e5 | 3015 | err = rt_fill_info(net, skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, |
1937504d | 3016 | RTM_NEWROUTE, 0, 0); |
d889ce3b TG |
3017 | if (err <= 0) |
3018 | goto errout_free; | |
1da177e4 | 3019 | |
1937504d | 3020 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid); |
d889ce3b | 3021 | errout: |
2942e900 | 3022 | return err; |
1da177e4 | 3023 | |
d889ce3b | 3024 | errout_free: |
1da177e4 | 3025 | kfree_skb(skb); |
d889ce3b | 3026 | goto errout; |
1da177e4 LT |
3027 | } |
3028 | ||
3029 | int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb) | |
3030 | { | |
3031 | struct rtable *rt; | |
3032 | int h, s_h; | |
3033 | int idx, s_idx; | |
1937504d DL |
3034 | struct net *net; |
3035 | ||
3b1e0a65 | 3036 | net = sock_net(skb->sk); |
1da177e4 LT |
3037 | |
3038 | s_h = cb->args[0]; | |
d8c92830 ED |
3039 | if (s_h < 0) |
3040 | s_h = 0; | |
1da177e4 | 3041 | s_idx = idx = cb->args[1]; |
a6272665 ED |
3042 | for (h = s_h; h <= rt_hash_mask; h++, s_idx = 0) { |
3043 | if (!rt_hash_table[h].chain) | |
3044 | continue; | |
1da177e4 | 3045 | rcu_read_lock_bh(); |
a898def2 PM |
3046 | for (rt = rcu_dereference_bh(rt_hash_table[h].chain), idx = 0; rt; |
3047 | rt = rcu_dereference_bh(rt->u.dst.rt_next), idx++) { | |
878628fb | 3048 | if (!net_eq(dev_net(rt->u.dst.dev), net) || idx < s_idx) |
1da177e4 | 3049 | continue; |
e84f84f2 | 3050 | if (rt_is_expired(rt)) |
29e75252 | 3051 | continue; |
adf30907 | 3052 | skb_dst_set(skb, dst_clone(&rt->u.dst)); |
4feb88e5 | 3053 | if (rt_fill_info(net, skb, NETLINK_CB(cb->skb).pid, |
e905a9ed | 3054 | cb->nlh->nlmsg_seq, RTM_NEWROUTE, |
b6544c0b | 3055 | 1, NLM_F_MULTI) <= 0) { |
adf30907 | 3056 | skb_dst_drop(skb); |
1da177e4 LT |
3057 | rcu_read_unlock_bh(); |
3058 | goto done; | |
3059 | } | |
adf30907 | 3060 | skb_dst_drop(skb); |
1da177e4 LT |
3061 | } |
3062 | rcu_read_unlock_bh(); | |
3063 | } | |
3064 | ||
3065 | done: | |
3066 | cb->args[0] = h; | |
3067 | cb->args[1] = idx; | |
3068 | return skb->len; | |
3069 | } | |
3070 | ||
3071 | void ip_rt_multicast_event(struct in_device *in_dev) | |
3072 | { | |
76e6ebfb | 3073 | rt_cache_flush(dev_net(in_dev->dev), 0); |
1da177e4 LT |
3074 | } |
3075 | ||
3076 | #ifdef CONFIG_SYSCTL | |
81c684d1 | 3077 | static int ipv4_sysctl_rtcache_flush(ctl_table *__ctl, int write, |
8d65af78 | 3078 | void __user *buffer, |
1da177e4 LT |
3079 | size_t *lenp, loff_t *ppos) |
3080 | { | |
3081 | if (write) { | |
639e104f | 3082 | int flush_delay; |
81c684d1 | 3083 | ctl_table ctl; |
39a23e75 | 3084 | struct net *net; |
639e104f | 3085 | |
81c684d1 DL |
3086 | memcpy(&ctl, __ctl, sizeof(ctl)); |
3087 | ctl.data = &flush_delay; | |
8d65af78 | 3088 | proc_dointvec(&ctl, write, buffer, lenp, ppos); |
639e104f | 3089 | |
81c684d1 | 3090 | net = (struct net *)__ctl->extra1; |
39a23e75 | 3091 | rt_cache_flush(net, flush_delay); |
1da177e4 | 3092 | return 0; |
e905a9ed | 3093 | } |
1da177e4 LT |
3094 | |
3095 | return -EINVAL; | |
3096 | } | |
3097 | ||
c6153b5b HX |
3098 | static void rt_secret_reschedule(int old) |
3099 | { | |
3100 | struct net *net; | |
3101 | int new = ip_rt_secret_interval; | |
3102 | int diff = new - old; | |
3103 | ||
3104 | if (!diff) | |
3105 | return; | |
3106 | ||
3107 | rtnl_lock(); | |
3108 | for_each_net(net) { | |
3109 | int deleted = del_timer_sync(&net->ipv4.rt_secret_timer); | |
858a18a6 | 3110 | long time; |
c6153b5b HX |
3111 | |
3112 | if (!new) | |
3113 | continue; | |
3114 | ||
3115 | if (deleted) { | |
858a18a6 | 3116 | time = net->ipv4.rt_secret_timer.expires - jiffies; |
c6153b5b HX |
3117 | |
3118 | if (time <= 0 || (time += diff) <= 0) | |
3119 | time = 0; | |
c6153b5b | 3120 | } else |
858a18a6 | 3121 | time = new; |
c6153b5b | 3122 | |
858a18a6 | 3123 | mod_timer(&net->ipv4.rt_secret_timer, jiffies + time); |
c6153b5b HX |
3124 | } |
3125 | rtnl_unlock(); | |
3126 | } | |
3127 | ||
3128 | static int ipv4_sysctl_rt_secret_interval(ctl_table *ctl, int write, | |
c6153b5b HX |
3129 | void __user *buffer, size_t *lenp, |
3130 | loff_t *ppos) | |
3131 | { | |
3132 | int old = ip_rt_secret_interval; | |
8d65af78 | 3133 | int ret = proc_dointvec_jiffies(ctl, write, buffer, lenp, ppos); |
c6153b5b HX |
3134 | |
3135 | rt_secret_reschedule(old); | |
3136 | ||
3137 | return ret; | |
3138 | } | |
3139 | ||
eeb61f71 | 3140 | static ctl_table ipv4_route_table[] = { |
1da177e4 | 3141 | { |
1da177e4 LT |
3142 | .procname = "gc_thresh", |
3143 | .data = &ipv4_dst_ops.gc_thresh, | |
3144 | .maxlen = sizeof(int), | |
3145 | .mode = 0644, | |
6d9f239a | 3146 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3147 | }, |
3148 | { | |
1da177e4 LT |
3149 | .procname = "max_size", |
3150 | .data = &ip_rt_max_size, | |
3151 | .maxlen = sizeof(int), | |
3152 | .mode = 0644, | |
6d9f239a | 3153 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3154 | }, |
3155 | { | |
3156 | /* Deprecated. Use gc_min_interval_ms */ | |
e905a9ed | 3157 | |
1da177e4 LT |
3158 | .procname = "gc_min_interval", |
3159 | .data = &ip_rt_gc_min_interval, | |
3160 | .maxlen = sizeof(int), | |
3161 | .mode = 0644, | |
6d9f239a | 3162 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3163 | }, |
3164 | { | |
1da177e4 LT |
3165 | .procname = "gc_min_interval_ms", |
3166 | .data = &ip_rt_gc_min_interval, | |
3167 | .maxlen = sizeof(int), | |
3168 | .mode = 0644, | |
6d9f239a | 3169 | .proc_handler = proc_dointvec_ms_jiffies, |
1da177e4 LT |
3170 | }, |
3171 | { | |
1da177e4 LT |
3172 | .procname = "gc_timeout", |
3173 | .data = &ip_rt_gc_timeout, | |
3174 | .maxlen = sizeof(int), | |
3175 | .mode = 0644, | |
6d9f239a | 3176 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3177 | }, |
3178 | { | |
1da177e4 LT |
3179 | .procname = "gc_interval", |
3180 | .data = &ip_rt_gc_interval, | |
3181 | .maxlen = sizeof(int), | |
3182 | .mode = 0644, | |
6d9f239a | 3183 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3184 | }, |
3185 | { | |
1da177e4 LT |
3186 | .procname = "redirect_load", |
3187 | .data = &ip_rt_redirect_load, | |
3188 | .maxlen = sizeof(int), | |
3189 | .mode = 0644, | |
6d9f239a | 3190 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3191 | }, |
3192 | { | |
1da177e4 LT |
3193 | .procname = "redirect_number", |
3194 | .data = &ip_rt_redirect_number, | |
3195 | .maxlen = sizeof(int), | |
3196 | .mode = 0644, | |
6d9f239a | 3197 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3198 | }, |
3199 | { | |
1da177e4 LT |
3200 | .procname = "redirect_silence", |
3201 | .data = &ip_rt_redirect_silence, | |
3202 | .maxlen = sizeof(int), | |
3203 | .mode = 0644, | |
6d9f239a | 3204 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3205 | }, |
3206 | { | |
1da177e4 LT |
3207 | .procname = "error_cost", |
3208 | .data = &ip_rt_error_cost, | |
3209 | .maxlen = sizeof(int), | |
3210 | .mode = 0644, | |
6d9f239a | 3211 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3212 | }, |
3213 | { | |
1da177e4 LT |
3214 | .procname = "error_burst", |
3215 | .data = &ip_rt_error_burst, | |
3216 | .maxlen = sizeof(int), | |
3217 | .mode = 0644, | |
6d9f239a | 3218 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3219 | }, |
3220 | { | |
1da177e4 LT |
3221 | .procname = "gc_elasticity", |
3222 | .data = &ip_rt_gc_elasticity, | |
3223 | .maxlen = sizeof(int), | |
3224 | .mode = 0644, | |
6d9f239a | 3225 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3226 | }, |
3227 | { | |
1da177e4 LT |
3228 | .procname = "mtu_expires", |
3229 | .data = &ip_rt_mtu_expires, | |
3230 | .maxlen = sizeof(int), | |
3231 | .mode = 0644, | |
6d9f239a | 3232 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3233 | }, |
3234 | { | |
1da177e4 LT |
3235 | .procname = "min_pmtu", |
3236 | .data = &ip_rt_min_pmtu, | |
3237 | .maxlen = sizeof(int), | |
3238 | .mode = 0644, | |
6d9f239a | 3239 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3240 | }, |
3241 | { | |
1da177e4 LT |
3242 | .procname = "min_adv_mss", |
3243 | .data = &ip_rt_min_advmss, | |
3244 | .maxlen = sizeof(int), | |
3245 | .mode = 0644, | |
6d9f239a | 3246 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3247 | }, |
3248 | { | |
1da177e4 LT |
3249 | .procname = "secret_interval", |
3250 | .data = &ip_rt_secret_interval, | |
3251 | .maxlen = sizeof(int), | |
3252 | .mode = 0644, | |
6d9f239a | 3253 | .proc_handler = ipv4_sysctl_rt_secret_interval, |
1da177e4 | 3254 | }, |
f8572d8f | 3255 | { } |
1da177e4 | 3256 | }; |
39a23e75 | 3257 | |
2f4520d3 AV |
3258 | static struct ctl_table empty[1]; |
3259 | ||
3260 | static struct ctl_table ipv4_skeleton[] = | |
3261 | { | |
f8572d8f | 3262 | { .procname = "route", |
d994af0d | 3263 | .mode = 0555, .child = ipv4_route_table}, |
f8572d8f | 3264 | { .procname = "neigh", |
d994af0d | 3265 | .mode = 0555, .child = empty}, |
2f4520d3 AV |
3266 | { } |
3267 | }; | |
3268 | ||
3269 | static __net_initdata struct ctl_path ipv4_path[] = { | |
f8572d8f EB |
3270 | { .procname = "net", }, |
3271 | { .procname = "ipv4", }, | |
39a23e75 DL |
3272 | { }, |
3273 | }; | |
3274 | ||
39a23e75 DL |
3275 | static struct ctl_table ipv4_route_flush_table[] = { |
3276 | { | |
39a23e75 DL |
3277 | .procname = "flush", |
3278 | .maxlen = sizeof(int), | |
3279 | .mode = 0200, | |
6d9f239a | 3280 | .proc_handler = ipv4_sysctl_rtcache_flush, |
39a23e75 | 3281 | }, |
f8572d8f | 3282 | { }, |
39a23e75 DL |
3283 | }; |
3284 | ||
2f4520d3 | 3285 | static __net_initdata struct ctl_path ipv4_route_path[] = { |
f8572d8f EB |
3286 | { .procname = "net", }, |
3287 | { .procname = "ipv4", }, | |
3288 | { .procname = "route", }, | |
2f4520d3 AV |
3289 | { }, |
3290 | }; | |
3291 | ||
39a23e75 DL |
3292 | static __net_init int sysctl_route_net_init(struct net *net) |
3293 | { | |
3294 | struct ctl_table *tbl; | |
3295 | ||
3296 | tbl = ipv4_route_flush_table; | |
09ad9bc7 | 3297 | if (!net_eq(net, &init_net)) { |
39a23e75 DL |
3298 | tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL); |
3299 | if (tbl == NULL) | |
3300 | goto err_dup; | |
3301 | } | |
3302 | tbl[0].extra1 = net; | |
3303 | ||
3304 | net->ipv4.route_hdr = | |
3305 | register_net_sysctl_table(net, ipv4_route_path, tbl); | |
3306 | if (net->ipv4.route_hdr == NULL) | |
3307 | goto err_reg; | |
3308 | return 0; | |
3309 | ||
3310 | err_reg: | |
3311 | if (tbl != ipv4_route_flush_table) | |
3312 | kfree(tbl); | |
3313 | err_dup: | |
3314 | return -ENOMEM; | |
3315 | } | |
3316 | ||
3317 | static __net_exit void sysctl_route_net_exit(struct net *net) | |
3318 | { | |
3319 | struct ctl_table *tbl; | |
3320 | ||
3321 | tbl = net->ipv4.route_hdr->ctl_table_arg; | |
3322 | unregister_net_sysctl_table(net->ipv4.route_hdr); | |
3323 | BUG_ON(tbl == ipv4_route_flush_table); | |
3324 | kfree(tbl); | |
3325 | } | |
3326 | ||
3327 | static __net_initdata struct pernet_operations sysctl_route_ops = { | |
3328 | .init = sysctl_route_net_init, | |
3329 | .exit = sysctl_route_net_exit, | |
3330 | }; | |
1da177e4 LT |
3331 | #endif |
3332 | ||
9f5e97e5 DL |
3333 | |
3334 | static __net_init int rt_secret_timer_init(struct net *net) | |
3335 | { | |
e84f84f2 DL |
3336 | atomic_set(&net->ipv4.rt_genid, |
3337 | (int) ((num_physpages ^ (num_physpages>>8)) ^ | |
3338 | (jiffies ^ (jiffies >> 7)))); | |
3339 | ||
9f5e97e5 DL |
3340 | net->ipv4.rt_secret_timer.function = rt_secret_rebuild; |
3341 | net->ipv4.rt_secret_timer.data = (unsigned long)net; | |
3342 | init_timer_deferrable(&net->ipv4.rt_secret_timer); | |
3343 | ||
c6153b5b HX |
3344 | if (ip_rt_secret_interval) { |
3345 | net->ipv4.rt_secret_timer.expires = | |
3346 | jiffies + net_random() % ip_rt_secret_interval + | |
3347 | ip_rt_secret_interval; | |
3348 | add_timer(&net->ipv4.rt_secret_timer); | |
3349 | } | |
9f5e97e5 DL |
3350 | return 0; |
3351 | } | |
3352 | ||
3353 | static __net_exit void rt_secret_timer_exit(struct net *net) | |
3354 | { | |
3355 | del_timer_sync(&net->ipv4.rt_secret_timer); | |
3356 | } | |
3357 | ||
3358 | static __net_initdata struct pernet_operations rt_secret_timer_ops = { | |
3359 | .init = rt_secret_timer_init, | |
3360 | .exit = rt_secret_timer_exit, | |
3361 | }; | |
3362 | ||
3363 | ||
1da177e4 | 3364 | #ifdef CONFIG_NET_CLS_ROUTE |
7d720c3e | 3365 | struct ip_rt_acct __percpu *ip_rt_acct __read_mostly; |
1da177e4 LT |
3366 | #endif /* CONFIG_NET_CLS_ROUTE */ |
3367 | ||
3368 | static __initdata unsigned long rhash_entries; | |
3369 | static int __init set_rhash_entries(char *str) | |
3370 | { | |
3371 | if (!str) | |
3372 | return 0; | |
3373 | rhash_entries = simple_strtoul(str, &str, 0); | |
3374 | return 1; | |
3375 | } | |
3376 | __setup("rhash_entries=", set_rhash_entries); | |
3377 | ||
3378 | int __init ip_rt_init(void) | |
3379 | { | |
424c4b70 | 3380 | int rc = 0; |
1da177e4 | 3381 | |
1da177e4 | 3382 | #ifdef CONFIG_NET_CLS_ROUTE |
0dcec8c2 | 3383 | ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct)); |
1da177e4 LT |
3384 | if (!ip_rt_acct) |
3385 | panic("IP: failed to allocate ip_rt_acct\n"); | |
1da177e4 LT |
3386 | #endif |
3387 | ||
e5d679f3 AD |
3388 | ipv4_dst_ops.kmem_cachep = |
3389 | kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0, | |
20c2df83 | 3390 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 3391 | |
14e50e57 DM |
3392 | ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep; |
3393 | ||
424c4b70 ED |
3394 | rt_hash_table = (struct rt_hash_bucket *) |
3395 | alloc_large_system_hash("IP route cache", | |
3396 | sizeof(struct rt_hash_bucket), | |
3397 | rhash_entries, | |
4481374c | 3398 | (totalram_pages >= 128 * 1024) ? |
18955cfc | 3399 | 15 : 17, |
8d1502de | 3400 | 0, |
424c4b70 ED |
3401 | &rt_hash_log, |
3402 | &rt_hash_mask, | |
c9503e0f | 3403 | rhash_entries ? 0 : 512 * 1024); |
22c047cc ED |
3404 | memset(rt_hash_table, 0, (rt_hash_mask + 1) * sizeof(struct rt_hash_bucket)); |
3405 | rt_hash_lock_init(); | |
1da177e4 LT |
3406 | |
3407 | ipv4_dst_ops.gc_thresh = (rt_hash_mask + 1); | |
3408 | ip_rt_max_size = (rt_hash_mask + 1) * 16; | |
3409 | ||
1da177e4 LT |
3410 | devinet_init(); |
3411 | ip_fib_init(); | |
3412 | ||
1da177e4 LT |
3413 | /* All the timers, started at system startup tend |
3414 | to synchronize. Perturb it a bit. | |
3415 | */ | |
125bb8f5 ED |
3416 | INIT_DELAYED_WORK_DEFERRABLE(&expires_work, rt_worker_func); |
3417 | expires_ljiffies = jiffies; | |
39c90ece ED |
3418 | schedule_delayed_work(&expires_work, |
3419 | net_random() % ip_rt_gc_interval + ip_rt_gc_interval); | |
1da177e4 | 3420 | |
9f5e97e5 DL |
3421 | if (register_pernet_subsys(&rt_secret_timer_ops)) |
3422 | printk(KERN_ERR "Unable to setup rt_secret_timer\n"); | |
1da177e4 | 3423 | |
73b38711 | 3424 | if (ip_rt_proc_init()) |
107f1634 | 3425 | printk(KERN_ERR "Unable to create route proc files\n"); |
1da177e4 LT |
3426 | #ifdef CONFIG_XFRM |
3427 | xfrm_init(); | |
a33bc5c1 | 3428 | xfrm4_init(ip_rt_max_size); |
1da177e4 | 3429 | #endif |
63f3444f TG |
3430 | rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL); |
3431 | ||
39a23e75 DL |
3432 | #ifdef CONFIG_SYSCTL |
3433 | register_pernet_subsys(&sysctl_route_ops); | |
3434 | #endif | |
1da177e4 LT |
3435 | return rc; |
3436 | } | |
3437 | ||
a1bc6eb4 | 3438 | #ifdef CONFIG_SYSCTL |
eeb61f71 AV |
3439 | /* |
3440 | * We really need to sanitize the damn ipv4 init order, then all | |
3441 | * this nonsense will go away. | |
3442 | */ | |
3443 | void __init ip_static_sysctl_init(void) | |
3444 | { | |
2f4520d3 | 3445 | register_sysctl_paths(ipv4_path, ipv4_skeleton); |
eeb61f71 | 3446 | } |
a1bc6eb4 | 3447 | #endif |
eeb61f71 | 3448 | |
1da177e4 LT |
3449 | EXPORT_SYMBOL(__ip_select_ident); |
3450 | EXPORT_SYMBOL(ip_route_input); | |
3451 | EXPORT_SYMBOL(ip_route_output_key); |