]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * ROUTE - implementation of the IP router. | |
7 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * Alan Cox, <gw4pts@gw4pts.ampr.org> | |
11 | * Linus Torvalds, <Linus.Torvalds@helsinki.fi> | |
12 | * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> | |
13 | * | |
14 | * Fixes: | |
15 | * Alan Cox : Verify area fixes. | |
16 | * Alan Cox : cli() protects routing changes | |
17 | * Rui Oliveira : ICMP routing table updates | |
18 | * (rco@di.uminho.pt) Routing table insertion and update | |
19 | * Linus Torvalds : Rewrote bits to be sensible | |
20 | * Alan Cox : Added BSD route gw semantics | |
e905a9ed | 21 | * Alan Cox : Super /proc >4K |
1da177e4 LT |
22 | * Alan Cox : MTU in route table |
23 | * Alan Cox : MSS actually. Also added the window | |
24 | * clamper. | |
25 | * Sam Lantinga : Fixed route matching in rt_del() | |
26 | * Alan Cox : Routing cache support. | |
27 | * Alan Cox : Removed compatibility cruft. | |
28 | * Alan Cox : RTF_REJECT support. | |
29 | * Alan Cox : TCP irtt support. | |
30 | * Jonathan Naylor : Added Metric support. | |
31 | * Miquel van Smoorenburg : BSD API fixes. | |
32 | * Miquel van Smoorenburg : Metrics. | |
33 | * Alan Cox : Use __u32 properly | |
34 | * Alan Cox : Aligned routing errors more closely with BSD | |
35 | * our system is still very different. | |
36 | * Alan Cox : Faster /proc handling | |
37 | * Alexey Kuznetsov : Massive rework to support tree based routing, | |
38 | * routing caches and better behaviour. | |
e905a9ed | 39 | * |
1da177e4 LT |
40 | * Olaf Erb : irtt wasn't being copied right. |
41 | * Bjorn Ekwall : Kerneld route support. | |
42 | * Alan Cox : Multicast fixed (I hope) | |
43 | * Pavel Krauz : Limited broadcast fixed | |
44 | * Mike McLagan : Routing by source | |
45 | * Alexey Kuznetsov : End of old history. Split to fib.c and | |
46 | * route.c and rewritten from scratch. | |
47 | * Andi Kleen : Load-limit warning messages. | |
48 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. | |
49 | * Vitaly E. Lavrov : Race condition in ip_route_input_slow. | |
50 | * Tobias Ringstrom : Uninitialized res.type in ip_route_output_slow. | |
51 | * Vladimir V. Ivanov : IP rule info (flowid) is really useful. | |
52 | * Marc Boucher : routing by fwmark | |
53 | * Robert Olsson : Added rt_cache statistics | |
54 | * Arnaldo C. Melo : Convert proc stuff to seq_file | |
bb1d23b0 | 55 | * Eric Dumazet : hashed spinlocks and rt_check_expire() fixes. |
cef2685e IS |
56 | * Ilia Sotnikov : Ignore TOS on PMTUD and Redirect |
57 | * Ilia Sotnikov : Removed TOS from hash calculations | |
1da177e4 LT |
58 | * |
59 | * This program is free software; you can redistribute it and/or | |
60 | * modify it under the terms of the GNU General Public License | |
61 | * as published by the Free Software Foundation; either version | |
62 | * 2 of the License, or (at your option) any later version. | |
63 | */ | |
64 | ||
1da177e4 LT |
65 | #include <linux/module.h> |
66 | #include <asm/uaccess.h> | |
67 | #include <asm/system.h> | |
68 | #include <linux/bitops.h> | |
69 | #include <linux/types.h> | |
70 | #include <linux/kernel.h> | |
1da177e4 | 71 | #include <linux/mm.h> |
424c4b70 | 72 | #include <linux/bootmem.h> |
1da177e4 LT |
73 | #include <linux/string.h> |
74 | #include <linux/socket.h> | |
75 | #include <linux/sockios.h> | |
76 | #include <linux/errno.h> | |
77 | #include <linux/in.h> | |
78 | #include <linux/inet.h> | |
79 | #include <linux/netdevice.h> | |
80 | #include <linux/proc_fs.h> | |
81 | #include <linux/init.h> | |
39c90ece | 82 | #include <linux/workqueue.h> |
1da177e4 | 83 | #include <linux/skbuff.h> |
1da177e4 LT |
84 | #include <linux/inetdevice.h> |
85 | #include <linux/igmp.h> | |
86 | #include <linux/pkt_sched.h> | |
87 | #include <linux/mroute.h> | |
88 | #include <linux/netfilter_ipv4.h> | |
89 | #include <linux/random.h> | |
90 | #include <linux/jhash.h> | |
91 | #include <linux/rcupdate.h> | |
92 | #include <linux/times.h> | |
5a0e3ad6 | 93 | #include <linux/slab.h> |
352e512c | 94 | #include <net/dst.h> |
457c4cbc | 95 | #include <net/net_namespace.h> |
1da177e4 LT |
96 | #include <net/protocol.h> |
97 | #include <net/ip.h> | |
98 | #include <net/route.h> | |
99 | #include <net/inetpeer.h> | |
100 | #include <net/sock.h> | |
101 | #include <net/ip_fib.h> | |
102 | #include <net/arp.h> | |
103 | #include <net/tcp.h> | |
104 | #include <net/icmp.h> | |
105 | #include <net/xfrm.h> | |
8d71740c | 106 | #include <net/netevent.h> |
63f3444f | 107 | #include <net/rtnetlink.h> |
1da177e4 LT |
108 | #ifdef CONFIG_SYSCTL |
109 | #include <linux/sysctl.h> | |
110 | #endif | |
111 | ||
112 | #define RT_FL_TOS(oldflp) \ | |
113 | ((u32)(oldflp->fl4_tos & (IPTOS_RT_MASK | RTO_ONLINK))) | |
114 | ||
115 | #define IP_MAX_MTU 0xFFF0 | |
116 | ||
117 | #define RT_GC_TIMEOUT (300*HZ) | |
118 | ||
1da177e4 | 119 | static int ip_rt_max_size; |
817bc4db SH |
120 | static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT; |
121 | static int ip_rt_gc_interval __read_mostly = 60 * HZ; | |
122 | static int ip_rt_gc_min_interval __read_mostly = HZ / 2; | |
123 | static int ip_rt_redirect_number __read_mostly = 9; | |
124 | static int ip_rt_redirect_load __read_mostly = HZ / 50; | |
125 | static int ip_rt_redirect_silence __read_mostly = ((HZ / 50) << (9 + 1)); | |
126 | static int ip_rt_error_cost __read_mostly = HZ; | |
127 | static int ip_rt_error_burst __read_mostly = 5 * HZ; | |
128 | static int ip_rt_gc_elasticity __read_mostly = 8; | |
129 | static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ; | |
130 | static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20; | |
131 | static int ip_rt_min_advmss __read_mostly = 256; | |
1080d709 | 132 | static int rt_chain_length_max __read_mostly = 20; |
1da177e4 | 133 | |
125bb8f5 ED |
134 | static struct delayed_work expires_work; |
135 | static unsigned long expires_ljiffies; | |
1da177e4 LT |
136 | |
137 | /* | |
138 | * Interface to generic destination cache. | |
139 | */ | |
140 | ||
141 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie); | |
142 | static void ipv4_dst_destroy(struct dst_entry *dst); | |
143 | static void ipv4_dst_ifdown(struct dst_entry *dst, | |
144 | struct net_device *dev, int how); | |
145 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst); | |
146 | static void ipv4_link_failure(struct sk_buff *skb); | |
147 | static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu); | |
569d3645 | 148 | static int rt_garbage_collect(struct dst_ops *ops); |
1da177e4 LT |
149 | |
150 | ||
151 | static struct dst_ops ipv4_dst_ops = { | |
152 | .family = AF_INET, | |
09640e63 | 153 | .protocol = cpu_to_be16(ETH_P_IP), |
1da177e4 LT |
154 | .gc = rt_garbage_collect, |
155 | .check = ipv4_dst_check, | |
156 | .destroy = ipv4_dst_destroy, | |
157 | .ifdown = ipv4_dst_ifdown, | |
158 | .negative_advice = ipv4_negative_advice, | |
159 | .link_failure = ipv4_link_failure, | |
160 | .update_pmtu = ip_rt_update_pmtu, | |
1ac06e03 | 161 | .local_out = __ip_local_out, |
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); |
27f39c73 | 256 | #define RT_CACHE_STAT_INC(field) __this_cpu_inc(rt_cache_stat.field) |
1da177e4 | 257 | |
b00180de | 258 | static inline unsigned int rt_hash(__be32 daddr, __be32 saddr, int idx, |
0eae88f3 | 259 | int genid) |
1da177e4 | 260 | { |
0eae88f3 | 261 | return jhash_3words((__force u32)daddr, (__force u32)saddr, |
b00180de | 262 | idx, genid) |
29e75252 | 263 | & rt_hash_mask; |
1da177e4 LT |
264 | } |
265 | ||
e84f84f2 DL |
266 | static inline int rt_genid(struct net *net) |
267 | { | |
268 | return atomic_read(&net->ipv4.rt_genid); | |
269 | } | |
270 | ||
1da177e4 LT |
271 | #ifdef CONFIG_PROC_FS |
272 | struct rt_cache_iter_state { | |
a75e936f | 273 | struct seq_net_private p; |
1da177e4 | 274 | int bucket; |
29e75252 | 275 | int genid; |
1da177e4 LT |
276 | }; |
277 | ||
1218854a | 278 | static struct rtable *rt_cache_get_first(struct seq_file *seq) |
1da177e4 | 279 | { |
1218854a | 280 | struct rt_cache_iter_state *st = seq->private; |
1da177e4 | 281 | struct rtable *r = NULL; |
1da177e4 LT |
282 | |
283 | for (st->bucket = rt_hash_mask; st->bucket >= 0; --st->bucket) { | |
a6272665 ED |
284 | if (!rt_hash_table[st->bucket].chain) |
285 | continue; | |
1da177e4 | 286 | rcu_read_lock_bh(); |
a898def2 | 287 | r = rcu_dereference_bh(rt_hash_table[st->bucket].chain); |
29e75252 | 288 | while (r) { |
d8d1f30b | 289 | if (dev_net(r->dst.dev) == seq_file_net(seq) && |
a75e936f | 290 | r->rt_genid == st->genid) |
29e75252 | 291 | return r; |
d8d1f30b | 292 | r = rcu_dereference_bh(r->dst.rt_next); |
29e75252 | 293 | } |
1da177e4 LT |
294 | rcu_read_unlock_bh(); |
295 | } | |
29e75252 | 296 | return r; |
1da177e4 LT |
297 | } |
298 | ||
1218854a | 299 | static struct rtable *__rt_cache_get_next(struct seq_file *seq, |
642d6318 | 300 | struct rtable *r) |
1da177e4 | 301 | { |
1218854a | 302 | struct rt_cache_iter_state *st = seq->private; |
a6272665 | 303 | |
d8d1f30b | 304 | r = r->dst.rt_next; |
1da177e4 LT |
305 | while (!r) { |
306 | rcu_read_unlock_bh(); | |
a6272665 ED |
307 | do { |
308 | if (--st->bucket < 0) | |
309 | return NULL; | |
310 | } while (!rt_hash_table[st->bucket].chain); | |
1da177e4 LT |
311 | rcu_read_lock_bh(); |
312 | r = rt_hash_table[st->bucket].chain; | |
313 | } | |
a898def2 | 314 | return rcu_dereference_bh(r); |
1da177e4 LT |
315 | } |
316 | ||
1218854a | 317 | static struct rtable *rt_cache_get_next(struct seq_file *seq, |
642d6318 DL |
318 | struct rtable *r) |
319 | { | |
1218854a YH |
320 | struct rt_cache_iter_state *st = seq->private; |
321 | while ((r = __rt_cache_get_next(seq, r)) != NULL) { | |
d8d1f30b | 322 | if (dev_net(r->dst.dev) != seq_file_net(seq)) |
a75e936f | 323 | continue; |
642d6318 DL |
324 | if (r->rt_genid == st->genid) |
325 | break; | |
326 | } | |
327 | return r; | |
328 | } | |
329 | ||
1218854a | 330 | static struct rtable *rt_cache_get_idx(struct seq_file *seq, loff_t pos) |
1da177e4 | 331 | { |
1218854a | 332 | struct rtable *r = rt_cache_get_first(seq); |
1da177e4 LT |
333 | |
334 | if (r) | |
1218854a | 335 | while (pos && (r = rt_cache_get_next(seq, r))) |
1da177e4 LT |
336 | --pos; |
337 | return pos ? NULL : r; | |
338 | } | |
339 | ||
340 | static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos) | |
341 | { | |
29e75252 | 342 | struct rt_cache_iter_state *st = seq->private; |
29e75252 | 343 | if (*pos) |
1218854a | 344 | return rt_cache_get_idx(seq, *pos - 1); |
e84f84f2 | 345 | st->genid = rt_genid(seq_file_net(seq)); |
29e75252 | 346 | return SEQ_START_TOKEN; |
1da177e4 LT |
347 | } |
348 | ||
349 | static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
350 | { | |
29e75252 | 351 | struct rtable *r; |
1da177e4 LT |
352 | |
353 | if (v == SEQ_START_TOKEN) | |
1218854a | 354 | r = rt_cache_get_first(seq); |
1da177e4 | 355 | else |
1218854a | 356 | r = rt_cache_get_next(seq, v); |
1da177e4 LT |
357 | ++*pos; |
358 | return r; | |
359 | } | |
360 | ||
361 | static void rt_cache_seq_stop(struct seq_file *seq, void *v) | |
362 | { | |
363 | if (v && v != SEQ_START_TOKEN) | |
364 | rcu_read_unlock_bh(); | |
365 | } | |
366 | ||
367 | static int rt_cache_seq_show(struct seq_file *seq, void *v) | |
368 | { | |
369 | if (v == SEQ_START_TOKEN) | |
370 | seq_printf(seq, "%-127s\n", | |
371 | "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t" | |
372 | "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t" | |
373 | "HHUptod\tSpecDst"); | |
374 | else { | |
375 | struct rtable *r = v; | |
5e659e4c | 376 | int len; |
1da177e4 | 377 | |
0eae88f3 ED |
378 | seq_printf(seq, "%s\t%08X\t%08X\t%8X\t%d\t%u\t%d\t" |
379 | "%08X\t%d\t%u\t%u\t%02X\t%d\t%1d\t%08X%n", | |
d8d1f30b | 380 | r->dst.dev ? r->dst.dev->name : "*", |
0eae88f3 ED |
381 | (__force u32)r->rt_dst, |
382 | (__force u32)r->rt_gateway, | |
d8d1f30b CG |
383 | r->rt_flags, atomic_read(&r->dst.__refcnt), |
384 | r->dst.__use, 0, (__force u32)r->rt_src, | |
385 | (dst_metric(&r->dst, RTAX_ADVMSS) ? | |
386 | (int)dst_metric(&r->dst, RTAX_ADVMSS) + 40 : 0), | |
387 | dst_metric(&r->dst, RTAX_WINDOW), | |
388 | (int)((dst_metric(&r->dst, RTAX_RTT) >> 3) + | |
389 | dst_metric(&r->dst, RTAX_RTTVAR)), | |
1da177e4 | 390 | r->fl.fl4_tos, |
d8d1f30b CG |
391 | r->dst.hh ? atomic_read(&r->dst.hh->hh_refcnt) : -1, |
392 | r->dst.hh ? (r->dst.hh->hh_output == | |
1da177e4 | 393 | dev_queue_xmit) : 0, |
5e659e4c PE |
394 | r->rt_spec_dst, &len); |
395 | ||
396 | seq_printf(seq, "%*s\n", 127 - len, ""); | |
e905a9ed YH |
397 | } |
398 | return 0; | |
1da177e4 LT |
399 | } |
400 | ||
f690808e | 401 | static const struct seq_operations rt_cache_seq_ops = { |
1da177e4 LT |
402 | .start = rt_cache_seq_start, |
403 | .next = rt_cache_seq_next, | |
404 | .stop = rt_cache_seq_stop, | |
405 | .show = rt_cache_seq_show, | |
406 | }; | |
407 | ||
408 | static int rt_cache_seq_open(struct inode *inode, struct file *file) | |
409 | { | |
a75e936f | 410 | return seq_open_net(inode, file, &rt_cache_seq_ops, |
cf7732e4 | 411 | sizeof(struct rt_cache_iter_state)); |
1da177e4 LT |
412 | } |
413 | ||
9a32144e | 414 | static const struct file_operations rt_cache_seq_fops = { |
1da177e4 LT |
415 | .owner = THIS_MODULE, |
416 | .open = rt_cache_seq_open, | |
417 | .read = seq_read, | |
418 | .llseek = seq_lseek, | |
a75e936f | 419 | .release = seq_release_net, |
1da177e4 LT |
420 | }; |
421 | ||
422 | ||
423 | static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos) | |
424 | { | |
425 | int cpu; | |
426 | ||
427 | if (*pos == 0) | |
428 | return SEQ_START_TOKEN; | |
429 | ||
0f23174a | 430 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
431 | if (!cpu_possible(cpu)) |
432 | continue; | |
433 | *pos = cpu+1; | |
2f970d83 | 434 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
435 | } |
436 | return NULL; | |
437 | } | |
438 | ||
439 | static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
440 | { | |
441 | int cpu; | |
442 | ||
0f23174a | 443 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
444 | if (!cpu_possible(cpu)) |
445 | continue; | |
446 | *pos = cpu+1; | |
2f970d83 | 447 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
448 | } |
449 | return NULL; | |
e905a9ed | 450 | |
1da177e4 LT |
451 | } |
452 | ||
453 | static void rt_cpu_seq_stop(struct seq_file *seq, void *v) | |
454 | { | |
455 | ||
456 | } | |
457 | ||
458 | static int rt_cpu_seq_show(struct seq_file *seq, void *v) | |
459 | { | |
460 | struct rt_cache_stat *st = v; | |
461 | ||
462 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 463 | 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 |
464 | return 0; |
465 | } | |
e905a9ed | 466 | |
1da177e4 LT |
467 | seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x " |
468 | " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n", | |
469 | atomic_read(&ipv4_dst_ops.entries), | |
470 | st->in_hit, | |
471 | st->in_slow_tot, | |
472 | st->in_slow_mc, | |
473 | st->in_no_route, | |
474 | st->in_brd, | |
475 | st->in_martian_dst, | |
476 | st->in_martian_src, | |
477 | ||
478 | st->out_hit, | |
479 | st->out_slow_tot, | |
e905a9ed | 480 | st->out_slow_mc, |
1da177e4 LT |
481 | |
482 | st->gc_total, | |
483 | st->gc_ignored, | |
484 | st->gc_goal_miss, | |
485 | st->gc_dst_overflow, | |
486 | st->in_hlist_search, | |
487 | st->out_hlist_search | |
488 | ); | |
489 | return 0; | |
490 | } | |
491 | ||
f690808e | 492 | static const struct seq_operations rt_cpu_seq_ops = { |
1da177e4 LT |
493 | .start = rt_cpu_seq_start, |
494 | .next = rt_cpu_seq_next, | |
495 | .stop = rt_cpu_seq_stop, | |
496 | .show = rt_cpu_seq_show, | |
497 | }; | |
498 | ||
499 | ||
500 | static int rt_cpu_seq_open(struct inode *inode, struct file *file) | |
501 | { | |
502 | return seq_open(file, &rt_cpu_seq_ops); | |
503 | } | |
504 | ||
9a32144e | 505 | static const struct file_operations rt_cpu_seq_fops = { |
1da177e4 LT |
506 | .owner = THIS_MODULE, |
507 | .open = rt_cpu_seq_open, | |
508 | .read = seq_read, | |
509 | .llseek = seq_lseek, | |
510 | .release = seq_release, | |
511 | }; | |
512 | ||
78c686e9 | 513 | #ifdef CONFIG_NET_CLS_ROUTE |
a661c419 | 514 | static int rt_acct_proc_show(struct seq_file *m, void *v) |
78c686e9 | 515 | { |
a661c419 AD |
516 | struct ip_rt_acct *dst, *src; |
517 | unsigned int i, j; | |
518 | ||
519 | dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL); | |
520 | if (!dst) | |
521 | return -ENOMEM; | |
522 | ||
523 | for_each_possible_cpu(i) { | |
524 | src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i); | |
525 | for (j = 0; j < 256; j++) { | |
526 | dst[j].o_bytes += src[j].o_bytes; | |
527 | dst[j].o_packets += src[j].o_packets; | |
528 | dst[j].i_bytes += src[j].i_bytes; | |
529 | dst[j].i_packets += src[j].i_packets; | |
530 | } | |
78c686e9 PE |
531 | } |
532 | ||
a661c419 AD |
533 | seq_write(m, dst, 256 * sizeof(struct ip_rt_acct)); |
534 | kfree(dst); | |
535 | return 0; | |
536 | } | |
78c686e9 | 537 | |
a661c419 AD |
538 | static int rt_acct_proc_open(struct inode *inode, struct file *file) |
539 | { | |
540 | return single_open(file, rt_acct_proc_show, NULL); | |
78c686e9 | 541 | } |
a661c419 AD |
542 | |
543 | static const struct file_operations rt_acct_proc_fops = { | |
544 | .owner = THIS_MODULE, | |
545 | .open = rt_acct_proc_open, | |
546 | .read = seq_read, | |
547 | .llseek = seq_lseek, | |
548 | .release = single_release, | |
549 | }; | |
78c686e9 | 550 | #endif |
107f1634 | 551 | |
73b38711 | 552 | static int __net_init ip_rt_do_proc_init(struct net *net) |
107f1634 PE |
553 | { |
554 | struct proc_dir_entry *pde; | |
555 | ||
556 | pde = proc_net_fops_create(net, "rt_cache", S_IRUGO, | |
557 | &rt_cache_seq_fops); | |
558 | if (!pde) | |
559 | goto err1; | |
560 | ||
77020720 WC |
561 | pde = proc_create("rt_cache", S_IRUGO, |
562 | net->proc_net_stat, &rt_cpu_seq_fops); | |
107f1634 PE |
563 | if (!pde) |
564 | goto err2; | |
565 | ||
107f1634 | 566 | #ifdef CONFIG_NET_CLS_ROUTE |
a661c419 | 567 | pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops); |
107f1634 PE |
568 | if (!pde) |
569 | goto err3; | |
570 | #endif | |
571 | return 0; | |
572 | ||
573 | #ifdef CONFIG_NET_CLS_ROUTE | |
574 | err3: | |
575 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
576 | #endif | |
577 | err2: | |
578 | remove_proc_entry("rt_cache", net->proc_net); | |
579 | err1: | |
580 | return -ENOMEM; | |
581 | } | |
73b38711 DL |
582 | |
583 | static void __net_exit ip_rt_do_proc_exit(struct net *net) | |
584 | { | |
585 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
586 | remove_proc_entry("rt_cache", net->proc_net); | |
0a931acf | 587 | #ifdef CONFIG_NET_CLS_ROUTE |
73b38711 | 588 | remove_proc_entry("rt_acct", net->proc_net); |
0a931acf | 589 | #endif |
73b38711 DL |
590 | } |
591 | ||
592 | static struct pernet_operations ip_rt_proc_ops __net_initdata = { | |
593 | .init = ip_rt_do_proc_init, | |
594 | .exit = ip_rt_do_proc_exit, | |
595 | }; | |
596 | ||
597 | static int __init ip_rt_proc_init(void) | |
598 | { | |
599 | return register_pernet_subsys(&ip_rt_proc_ops); | |
600 | } | |
601 | ||
107f1634 | 602 | #else |
73b38711 | 603 | static inline int ip_rt_proc_init(void) |
107f1634 PE |
604 | { |
605 | return 0; | |
606 | } | |
1da177e4 | 607 | #endif /* CONFIG_PROC_FS */ |
e905a9ed | 608 | |
5969f71d | 609 | static inline void rt_free(struct rtable *rt) |
1da177e4 | 610 | { |
d8d1f30b | 611 | call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free); |
1da177e4 LT |
612 | } |
613 | ||
5969f71d | 614 | static inline void rt_drop(struct rtable *rt) |
1da177e4 | 615 | { |
1da177e4 | 616 | ip_rt_put(rt); |
d8d1f30b | 617 | call_rcu_bh(&rt->dst.rcu_head, dst_rcu_free); |
1da177e4 LT |
618 | } |
619 | ||
5969f71d | 620 | static inline int rt_fast_clean(struct rtable *rth) |
1da177e4 LT |
621 | { |
622 | /* Kill broadcast/multicast entries very aggresively, if they | |
623 | collide in hash table with more useful entries */ | |
624 | return (rth->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) && | |
d8d1f30b | 625 | rth->fl.iif && rth->dst.rt_next; |
1da177e4 LT |
626 | } |
627 | ||
5969f71d | 628 | static inline int rt_valuable(struct rtable *rth) |
1da177e4 LT |
629 | { |
630 | return (rth->rt_flags & (RTCF_REDIRECTED | RTCF_NOTIFY)) || | |
d8d1f30b | 631 | rth->dst.expires; |
1da177e4 LT |
632 | } |
633 | ||
634 | static int rt_may_expire(struct rtable *rth, unsigned long tmo1, unsigned long tmo2) | |
635 | { | |
636 | unsigned long age; | |
637 | int ret = 0; | |
638 | ||
d8d1f30b | 639 | if (atomic_read(&rth->dst.__refcnt)) |
1da177e4 LT |
640 | goto out; |
641 | ||
642 | ret = 1; | |
d8d1f30b CG |
643 | if (rth->dst.expires && |
644 | time_after_eq(jiffies, rth->dst.expires)) | |
1da177e4 LT |
645 | goto out; |
646 | ||
d8d1f30b | 647 | age = jiffies - rth->dst.lastuse; |
1da177e4 LT |
648 | ret = 0; |
649 | if ((age <= tmo1 && !rt_fast_clean(rth)) || | |
650 | (age <= tmo2 && rt_valuable(rth))) | |
651 | goto out; | |
652 | ret = 1; | |
653 | out: return ret; | |
654 | } | |
655 | ||
656 | /* Bits of score are: | |
657 | * 31: very valuable | |
658 | * 30: not quite useless | |
659 | * 29..0: usage counter | |
660 | */ | |
661 | static inline u32 rt_score(struct rtable *rt) | |
662 | { | |
d8d1f30b | 663 | u32 score = jiffies - rt->dst.lastuse; |
1da177e4 LT |
664 | |
665 | score = ~score & ~(3<<30); | |
666 | ||
667 | if (rt_valuable(rt)) | |
668 | score |= (1<<31); | |
669 | ||
670 | if (!rt->fl.iif || | |
671 | !(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL))) | |
672 | score |= (1<<30); | |
673 | ||
674 | return score; | |
675 | } | |
676 | ||
1080d709 NH |
677 | static inline bool rt_caching(const struct net *net) |
678 | { | |
679 | return net->ipv4.current_rt_cache_rebuild_count <= | |
680 | net->ipv4.sysctl_rt_cache_rebuild_count; | |
681 | } | |
682 | ||
683 | static inline bool compare_hash_inputs(const struct flowi *fl1, | |
684 | const struct flowi *fl2) | |
685 | { | |
0eae88f3 ED |
686 | return ((((__force u32)fl1->nl_u.ip4_u.daddr ^ (__force u32)fl2->nl_u.ip4_u.daddr) | |
687 | ((__force u32)fl1->nl_u.ip4_u.saddr ^ (__force u32)fl2->nl_u.ip4_u.saddr) | | |
1080d709 NH |
688 | (fl1->iif ^ fl2->iif)) == 0); |
689 | } | |
690 | ||
1da177e4 LT |
691 | static inline int compare_keys(struct flowi *fl1, struct flowi *fl2) |
692 | { | |
0eae88f3 ED |
693 | return (((__force u32)fl1->nl_u.ip4_u.daddr ^ (__force u32)fl2->nl_u.ip4_u.daddr) | |
694 | ((__force u32)fl1->nl_u.ip4_u.saddr ^ (__force u32)fl2->nl_u.ip4_u.saddr) | | |
47dcf0cb | 695 | (fl1->mark ^ fl2->mark) | |
0eae88f3 | 696 | (*(u16 *)&fl1->nl_u.ip4_u.tos ^ *(u16 *)&fl2->nl_u.ip4_u.tos) | |
8238b218 DM |
697 | (fl1->oif ^ fl2->oif) | |
698 | (fl1->iif ^ fl2->iif)) == 0; | |
1da177e4 LT |
699 | } |
700 | ||
b5921910 DL |
701 | static inline int compare_netns(struct rtable *rt1, struct rtable *rt2) |
702 | { | |
d8d1f30b | 703 | return net_eq(dev_net(rt1->dst.dev), dev_net(rt2->dst.dev)); |
b5921910 DL |
704 | } |
705 | ||
e84f84f2 DL |
706 | static inline int rt_is_expired(struct rtable *rth) |
707 | { | |
d8d1f30b | 708 | return rth->rt_genid != rt_genid(dev_net(rth->dst.dev)); |
e84f84f2 DL |
709 | } |
710 | ||
beb659bd ED |
711 | /* |
712 | * Perform a full scan of hash table and free all entries. | |
713 | * Can be called by a softirq or a process. | |
714 | * In the later case, we want to be reschedule if necessary | |
715 | */ | |
716 | static void rt_do_flush(int process_context) | |
717 | { | |
718 | unsigned int i; | |
719 | struct rtable *rth, *next; | |
32cb5b4e | 720 | struct rtable * tail; |
beb659bd ED |
721 | |
722 | for (i = 0; i <= rt_hash_mask; i++) { | |
723 | if (process_context && need_resched()) | |
724 | cond_resched(); | |
725 | rth = rt_hash_table[i].chain; | |
726 | if (!rth) | |
727 | continue; | |
728 | ||
729 | spin_lock_bh(rt_hash_lock_addr(i)); | |
32cb5b4e DL |
730 | #ifdef CONFIG_NET_NS |
731 | { | |
732 | struct rtable ** prev, * p; | |
733 | ||
734 | rth = rt_hash_table[i].chain; | |
735 | ||
736 | /* defer releasing the head of the list after spin_unlock */ | |
d8d1f30b | 737 | for (tail = rth; tail; tail = tail->dst.rt_next) |
32cb5b4e DL |
738 | if (!rt_is_expired(tail)) |
739 | break; | |
740 | if (rth != tail) | |
741 | rt_hash_table[i].chain = tail; | |
742 | ||
743 | /* call rt_free on entries after the tail requiring flush */ | |
744 | prev = &rt_hash_table[i].chain; | |
745 | for (p = *prev; p; p = next) { | |
d8d1f30b | 746 | next = p->dst.rt_next; |
32cb5b4e | 747 | if (!rt_is_expired(p)) { |
d8d1f30b | 748 | prev = &p->dst.rt_next; |
32cb5b4e DL |
749 | } else { |
750 | *prev = next; | |
751 | rt_free(p); | |
752 | } | |
753 | } | |
754 | } | |
755 | #else | |
beb659bd ED |
756 | rth = rt_hash_table[i].chain; |
757 | rt_hash_table[i].chain = NULL; | |
32cb5b4e DL |
758 | tail = NULL; |
759 | #endif | |
beb659bd ED |
760 | spin_unlock_bh(rt_hash_lock_addr(i)); |
761 | ||
32cb5b4e | 762 | for (; rth != tail; rth = next) { |
d8d1f30b | 763 | next = rth->dst.rt_next; |
beb659bd ED |
764 | rt_free(rth); |
765 | } | |
766 | } | |
767 | } | |
768 | ||
1080d709 NH |
769 | /* |
770 | * While freeing expired entries, we compute average chain length | |
771 | * and standard deviation, using fixed-point arithmetic. | |
772 | * This to have an estimation of rt_chain_length_max | |
773 | * rt_chain_length_max = max(elasticity, AVG + 4*SD) | |
774 | * We use 3 bits for frational part, and 29 (or 61) for magnitude. | |
775 | */ | |
776 | ||
777 | #define FRACT_BITS 3 | |
778 | #define ONE (1UL << FRACT_BITS) | |
779 | ||
98376387 ED |
780 | /* |
781 | * Given a hash chain and an item in this hash chain, | |
782 | * find if a previous entry has the same hash_inputs | |
783 | * (but differs on tos, mark or oif) | |
784 | * Returns 0 if an alias is found. | |
785 | * Returns ONE if rth has no alias before itself. | |
786 | */ | |
787 | static int has_noalias(const struct rtable *head, const struct rtable *rth) | |
788 | { | |
789 | const struct rtable *aux = head; | |
790 | ||
791 | while (aux != rth) { | |
792 | if (compare_hash_inputs(&aux->fl, &rth->fl)) | |
793 | return 0; | |
d8d1f30b | 794 | aux = aux->dst.rt_next; |
98376387 ED |
795 | } |
796 | return ONE; | |
797 | } | |
798 | ||
beb659bd | 799 | static void rt_check_expire(void) |
1da177e4 | 800 | { |
bb1d23b0 ED |
801 | static unsigned int rover; |
802 | unsigned int i = rover, goal; | |
98376387 | 803 | struct rtable *rth, **rthp; |
cf8da764 | 804 | unsigned long samples = 0; |
1080d709 | 805 | unsigned long sum = 0, sum2 = 0; |
125bb8f5 | 806 | unsigned long delta; |
bb1d23b0 ED |
807 | u64 mult; |
808 | ||
125bb8f5 ED |
809 | delta = jiffies - expires_ljiffies; |
810 | expires_ljiffies = jiffies; | |
811 | mult = ((u64)delta) << rt_hash_log; | |
bb1d23b0 ED |
812 | if (ip_rt_gc_timeout > 1) |
813 | do_div(mult, ip_rt_gc_timeout); | |
814 | goal = (unsigned int)mult; | |
39c90ece ED |
815 | if (goal > rt_hash_mask) |
816 | goal = rt_hash_mask + 1; | |
bb1d23b0 | 817 | for (; goal > 0; goal--) { |
1da177e4 | 818 | unsigned long tmo = ip_rt_gc_timeout; |
cf8da764 | 819 | unsigned long length; |
1da177e4 LT |
820 | |
821 | i = (i + 1) & rt_hash_mask; | |
822 | rthp = &rt_hash_table[i].chain; | |
823 | ||
d90bf5a9 ED |
824 | if (need_resched()) |
825 | cond_resched(); | |
826 | ||
1080d709 NH |
827 | samples++; |
828 | ||
cfcabdcc | 829 | if (*rthp == NULL) |
bb1d23b0 | 830 | continue; |
cf8da764 | 831 | length = 0; |
39c90ece | 832 | spin_lock_bh(rt_hash_lock_addr(i)); |
1da177e4 | 833 | while ((rth = *rthp) != NULL) { |
d8d1f30b | 834 | prefetch(rth->dst.rt_next); |
e84f84f2 | 835 | if (rt_is_expired(rth)) { |
d8d1f30b | 836 | *rthp = rth->dst.rt_next; |
29e75252 ED |
837 | rt_free(rth); |
838 | continue; | |
839 | } | |
d8d1f30b | 840 | if (rth->dst.expires) { |
1da177e4 | 841 | /* Entry is expired even if it is in use */ |
d8d1f30b | 842 | if (time_before_eq(jiffies, rth->dst.expires)) { |
1ddbcb00 | 843 | nofree: |
1da177e4 | 844 | tmo >>= 1; |
d8d1f30b | 845 | rthp = &rth->dst.rt_next; |
1080d709 | 846 | /* |
1ddbcb00 | 847 | * We only count entries on |
1080d709 NH |
848 | * a chain with equal hash inputs once |
849 | * so that entries for different QOS | |
850 | * levels, and other non-hash input | |
851 | * attributes don't unfairly skew | |
852 | * the length computation | |
853 | */ | |
98376387 | 854 | length += has_noalias(rt_hash_table[i].chain, rth); |
1da177e4 LT |
855 | continue; |
856 | } | |
1ddbcb00 ED |
857 | } else if (!rt_may_expire(rth, tmo, ip_rt_gc_timeout)) |
858 | goto nofree; | |
1da177e4 LT |
859 | |
860 | /* Cleanup aged off entries. */ | |
d8d1f30b | 861 | *rthp = rth->dst.rt_next; |
e905a9ed | 862 | rt_free(rth); |
1da177e4 | 863 | } |
39c90ece | 864 | spin_unlock_bh(rt_hash_lock_addr(i)); |
1080d709 NH |
865 | sum += length; |
866 | sum2 += length*length; | |
867 | } | |
868 | if (samples) { | |
869 | unsigned long avg = sum / samples; | |
870 | unsigned long sd = int_sqrt(sum2 / samples - avg*avg); | |
871 | rt_chain_length_max = max_t(unsigned long, | |
872 | ip_rt_gc_elasticity, | |
873 | (avg + 4*sd) >> FRACT_BITS); | |
1da177e4 LT |
874 | } |
875 | rover = i; | |
beb659bd ED |
876 | } |
877 | ||
878 | /* | |
879 | * rt_worker_func() is run in process context. | |
29e75252 | 880 | * we call rt_check_expire() to scan part of the hash table |
beb659bd ED |
881 | */ |
882 | static void rt_worker_func(struct work_struct *work) | |
883 | { | |
29e75252 | 884 | rt_check_expire(); |
39c90ece | 885 | schedule_delayed_work(&expires_work, ip_rt_gc_interval); |
1da177e4 LT |
886 | } |
887 | ||
29e75252 ED |
888 | /* |
889 | * Pertubation of rt_genid by a small quantity [1..256] | |
890 | * Using 8 bits of shuffling ensure we can call rt_cache_invalidate() | |
891 | * many times (2^24) without giving recent rt_genid. | |
892 | * Jenkins hash is strong enough that litle changes of rt_genid are OK. | |
1da177e4 | 893 | */ |
86c657f6 | 894 | static void rt_cache_invalidate(struct net *net) |
1da177e4 | 895 | { |
29e75252 | 896 | unsigned char shuffle; |
1da177e4 | 897 | |
29e75252 | 898 | get_random_bytes(&shuffle, sizeof(shuffle)); |
e84f84f2 | 899 | atomic_add(shuffle + 1U, &net->ipv4.rt_genid); |
1da177e4 LT |
900 | } |
901 | ||
29e75252 ED |
902 | /* |
903 | * delay < 0 : invalidate cache (fast : entries will be deleted later) | |
904 | * delay >= 0 : invalidate & flush cache (can be long) | |
905 | */ | |
76e6ebfb | 906 | void rt_cache_flush(struct net *net, int delay) |
1da177e4 | 907 | { |
86c657f6 | 908 | rt_cache_invalidate(net); |
29e75252 ED |
909 | if (delay >= 0) |
910 | rt_do_flush(!in_softirq()); | |
1da177e4 LT |
911 | } |
912 | ||
a5ee1551 EB |
913 | /* Flush previous cache invalidated entries from the cache */ |
914 | void rt_cache_flush_batch(void) | |
915 | { | |
916 | rt_do_flush(!in_softirq()); | |
917 | } | |
918 | ||
1080d709 NH |
919 | static void rt_emergency_hash_rebuild(struct net *net) |
920 | { | |
3ee94372 | 921 | if (net_ratelimit()) |
1080d709 | 922 | printk(KERN_WARNING "Route hash chain too long!\n"); |
3ee94372 | 923 | rt_cache_invalidate(net); |
1080d709 NH |
924 | } |
925 | ||
1da177e4 LT |
926 | /* |
927 | Short description of GC goals. | |
928 | ||
929 | We want to build algorithm, which will keep routing cache | |
930 | at some equilibrium point, when number of aged off entries | |
931 | is kept approximately equal to newly generated ones. | |
932 | ||
933 | Current expiration strength is variable "expire". | |
934 | We try to adjust it dynamically, so that if networking | |
935 | is idle expires is large enough to keep enough of warm entries, | |
936 | and when load increases it reduces to limit cache size. | |
937 | */ | |
938 | ||
569d3645 | 939 | static int rt_garbage_collect(struct dst_ops *ops) |
1da177e4 LT |
940 | { |
941 | static unsigned long expire = RT_GC_TIMEOUT; | |
942 | static unsigned long last_gc; | |
943 | static int rover; | |
944 | static int equilibrium; | |
945 | struct rtable *rth, **rthp; | |
946 | unsigned long now = jiffies; | |
947 | int goal; | |
948 | ||
949 | /* | |
950 | * Garbage collection is pretty expensive, | |
951 | * do not make it too frequently. | |
952 | */ | |
953 | ||
954 | RT_CACHE_STAT_INC(gc_total); | |
955 | ||
956 | if (now - last_gc < ip_rt_gc_min_interval && | |
957 | atomic_read(&ipv4_dst_ops.entries) < ip_rt_max_size) { | |
958 | RT_CACHE_STAT_INC(gc_ignored); | |
959 | goto out; | |
960 | } | |
961 | ||
962 | /* Calculate number of entries, which we want to expire now. */ | |
963 | goal = atomic_read(&ipv4_dst_ops.entries) - | |
964 | (ip_rt_gc_elasticity << rt_hash_log); | |
965 | if (goal <= 0) { | |
966 | if (equilibrium < ipv4_dst_ops.gc_thresh) | |
967 | equilibrium = ipv4_dst_ops.gc_thresh; | |
968 | goal = atomic_read(&ipv4_dst_ops.entries) - equilibrium; | |
969 | if (goal > 0) { | |
b790cedd | 970 | equilibrium += min_t(unsigned int, goal >> 1, rt_hash_mask + 1); |
1da177e4 LT |
971 | goal = atomic_read(&ipv4_dst_ops.entries) - equilibrium; |
972 | } | |
973 | } else { | |
974 | /* We are in dangerous area. Try to reduce cache really | |
975 | * aggressively. | |
976 | */ | |
b790cedd | 977 | goal = max_t(unsigned int, goal >> 1, rt_hash_mask + 1); |
1da177e4 LT |
978 | equilibrium = atomic_read(&ipv4_dst_ops.entries) - goal; |
979 | } | |
980 | ||
981 | if (now - last_gc >= ip_rt_gc_min_interval) | |
982 | last_gc = now; | |
983 | ||
984 | if (goal <= 0) { | |
985 | equilibrium += goal; | |
986 | goto work_done; | |
987 | } | |
988 | ||
989 | do { | |
990 | int i, k; | |
991 | ||
992 | for (i = rt_hash_mask, k = rover; i >= 0; i--) { | |
993 | unsigned long tmo = expire; | |
994 | ||
995 | k = (k + 1) & rt_hash_mask; | |
996 | rthp = &rt_hash_table[k].chain; | |
22c047cc | 997 | spin_lock_bh(rt_hash_lock_addr(k)); |
1da177e4 | 998 | while ((rth = *rthp) != NULL) { |
e84f84f2 | 999 | if (!rt_is_expired(rth) && |
29e75252 | 1000 | !rt_may_expire(rth, tmo, expire)) { |
1da177e4 | 1001 | tmo >>= 1; |
d8d1f30b | 1002 | rthp = &rth->dst.rt_next; |
1da177e4 LT |
1003 | continue; |
1004 | } | |
d8d1f30b | 1005 | *rthp = rth->dst.rt_next; |
1da177e4 LT |
1006 | rt_free(rth); |
1007 | goal--; | |
1da177e4 | 1008 | } |
22c047cc | 1009 | spin_unlock_bh(rt_hash_lock_addr(k)); |
1da177e4 LT |
1010 | if (goal <= 0) |
1011 | break; | |
1012 | } | |
1013 | rover = k; | |
1014 | ||
1015 | if (goal <= 0) | |
1016 | goto work_done; | |
1017 | ||
1018 | /* Goal is not achieved. We stop process if: | |
1019 | ||
1020 | - if expire reduced to zero. Otherwise, expire is halfed. | |
1021 | - if table is not full. | |
1022 | - if we are called from interrupt. | |
1023 | - jiffies check is just fallback/debug loop breaker. | |
1024 | We will not spin here for long time in any case. | |
1025 | */ | |
1026 | ||
1027 | RT_CACHE_STAT_INC(gc_goal_miss); | |
1028 | ||
1029 | if (expire == 0) | |
1030 | break; | |
1031 | ||
1032 | expire >>= 1; | |
1033 | #if RT_CACHE_DEBUG >= 2 | |
1034 | printk(KERN_DEBUG "expire>> %u %d %d %d\n", expire, | |
1035 | atomic_read(&ipv4_dst_ops.entries), goal, i); | |
1036 | #endif | |
1037 | ||
1038 | if (atomic_read(&ipv4_dst_ops.entries) < ip_rt_max_size) | |
1039 | goto out; | |
1040 | } while (!in_softirq() && time_before_eq(jiffies, now)); | |
1041 | ||
1042 | if (atomic_read(&ipv4_dst_ops.entries) < ip_rt_max_size) | |
1043 | goto out; | |
1044 | if (net_ratelimit()) | |
1045 | printk(KERN_WARNING "dst cache overflow\n"); | |
1046 | RT_CACHE_STAT_INC(gc_dst_overflow); | |
1047 | return 1; | |
1048 | ||
1049 | work_done: | |
1050 | expire += ip_rt_gc_min_interval; | |
1051 | if (expire > ip_rt_gc_timeout || | |
1052 | atomic_read(&ipv4_dst_ops.entries) < ipv4_dst_ops.gc_thresh) | |
1053 | expire = ip_rt_gc_timeout; | |
1054 | #if RT_CACHE_DEBUG >= 2 | |
1055 | printk(KERN_DEBUG "expire++ %u %d %d %d\n", expire, | |
1056 | atomic_read(&ipv4_dst_ops.entries), goal, rover); | |
1057 | #endif | |
1058 | out: return 0; | |
1059 | } | |
1060 | ||
98376387 ED |
1061 | /* |
1062 | * Returns number of entries in a hash chain that have different hash_inputs | |
1063 | */ | |
1064 | static int slow_chain_length(const struct rtable *head) | |
1065 | { | |
1066 | int length = 0; | |
1067 | const struct rtable *rth = head; | |
1068 | ||
1069 | while (rth) { | |
1070 | length += has_noalias(head, rth); | |
d8d1f30b | 1071 | rth = rth->dst.rt_next; |
98376387 ED |
1072 | } |
1073 | return length >> FRACT_BITS; | |
1074 | } | |
1075 | ||
511c3f92 | 1076 | static int rt_intern_hash(unsigned hash, struct rtable *rt, |
6a2bad70 | 1077 | struct rtable **rp, struct sk_buff *skb, int ifindex) |
1da177e4 LT |
1078 | { |
1079 | struct rtable *rth, **rthp; | |
1080 | unsigned long now; | |
1081 | struct rtable *cand, **candp; | |
1082 | u32 min_score; | |
1083 | int chain_length; | |
1084 | int attempts = !in_softirq(); | |
1085 | ||
1086 | restart: | |
1087 | chain_length = 0; | |
1088 | min_score = ~(u32)0; | |
1089 | cand = NULL; | |
1090 | candp = NULL; | |
1091 | now = jiffies; | |
1092 | ||
d8d1f30b | 1093 | if (!rt_caching(dev_net(rt->dst.dev))) { |
73e42897 NH |
1094 | /* |
1095 | * If we're not caching, just tell the caller we | |
1096 | * were successful and don't touch the route. The | |
1097 | * caller hold the sole reference to the cache entry, and | |
1098 | * it will be released when the caller is done with it. | |
1099 | * If we drop it here, the callers have no way to resolve routes | |
1100 | * when we're not caching. Instead, just point *rp at rt, so | |
1101 | * the caller gets a single use out of the route | |
b6280b47 NH |
1102 | * Note that we do rt_free on this new route entry, so that |
1103 | * once its refcount hits zero, we are still able to reap it | |
1104 | * (Thanks Alexey) | |
1105 | * Note also the rt_free uses call_rcu. We don't actually | |
1106 | * need rcu protection here, this is just our path to get | |
1107 | * on the route gc list. | |
73e42897 | 1108 | */ |
b6280b47 | 1109 | |
c7d4426a | 1110 | rt->dst.flags |= DST_NOCACHE; |
b6280b47 | 1111 | if (rt->rt_type == RTN_UNICAST || rt->fl.iif == 0) { |
d8d1f30b | 1112 | int err = arp_bind_neighbour(&rt->dst); |
b6280b47 NH |
1113 | if (err) { |
1114 | if (net_ratelimit()) | |
1115 | printk(KERN_WARNING | |
1116 | "Neighbour table failure & not caching routes.\n"); | |
1117 | rt_drop(rt); | |
1118 | return err; | |
1119 | } | |
1120 | } | |
1121 | ||
1122 | rt_free(rt); | |
1123 | goto skip_hashing; | |
1080d709 NH |
1124 | } |
1125 | ||
1da177e4 LT |
1126 | rthp = &rt_hash_table[hash].chain; |
1127 | ||
22c047cc | 1128 | spin_lock_bh(rt_hash_lock_addr(hash)); |
1da177e4 | 1129 | while ((rth = *rthp) != NULL) { |
e84f84f2 | 1130 | if (rt_is_expired(rth)) { |
d8d1f30b | 1131 | *rthp = rth->dst.rt_next; |
29e75252 ED |
1132 | rt_free(rth); |
1133 | continue; | |
1134 | } | |
b5921910 | 1135 | if (compare_keys(&rth->fl, &rt->fl) && compare_netns(rth, rt)) { |
1da177e4 | 1136 | /* Put it first */ |
d8d1f30b | 1137 | *rthp = rth->dst.rt_next; |
1da177e4 LT |
1138 | /* |
1139 | * Since lookup is lockfree, the deletion | |
1140 | * must be visible to another weakly ordered CPU before | |
1141 | * the insertion at the start of the hash chain. | |
1142 | */ | |
d8d1f30b | 1143 | rcu_assign_pointer(rth->dst.rt_next, |
1da177e4 LT |
1144 | rt_hash_table[hash].chain); |
1145 | /* | |
1146 | * Since lookup is lockfree, the update writes | |
1147 | * must be ordered for consistency on SMP. | |
1148 | */ | |
1149 | rcu_assign_pointer(rt_hash_table[hash].chain, rth); | |
1150 | ||
d8d1f30b | 1151 | dst_use(&rth->dst, now); |
22c047cc | 1152 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1153 | |
1154 | rt_drop(rt); | |
511c3f92 ED |
1155 | if (rp) |
1156 | *rp = rth; | |
1157 | else | |
d8d1f30b | 1158 | skb_dst_set(skb, &rth->dst); |
1da177e4 LT |
1159 | return 0; |
1160 | } | |
1161 | ||
d8d1f30b | 1162 | if (!atomic_read(&rth->dst.__refcnt)) { |
1da177e4 LT |
1163 | u32 score = rt_score(rth); |
1164 | ||
1165 | if (score <= min_score) { | |
1166 | cand = rth; | |
1167 | candp = rthp; | |
1168 | min_score = score; | |
1169 | } | |
1170 | } | |
1171 | ||
1172 | chain_length++; | |
1173 | ||
d8d1f30b | 1174 | rthp = &rth->dst.rt_next; |
1da177e4 LT |
1175 | } |
1176 | ||
1177 | if (cand) { | |
1178 | /* ip_rt_gc_elasticity used to be average length of chain | |
1179 | * length, when exceeded gc becomes really aggressive. | |
1180 | * | |
1181 | * The second limit is less certain. At the moment it allows | |
1182 | * only 2 entries per bucket. We will see. | |
1183 | */ | |
1184 | if (chain_length > ip_rt_gc_elasticity) { | |
d8d1f30b | 1185 | *candp = cand->dst.rt_next; |
1da177e4 LT |
1186 | rt_free(cand); |
1187 | } | |
1080d709 | 1188 | } else { |
98376387 ED |
1189 | if (chain_length > rt_chain_length_max && |
1190 | slow_chain_length(rt_hash_table[hash].chain) > rt_chain_length_max) { | |
d8d1f30b | 1191 | struct net *net = dev_net(rt->dst.dev); |
1080d709 | 1192 | int num = ++net->ipv4.current_rt_cache_rebuild_count; |
b35ecb5d | 1193 | if (!rt_caching(net)) { |
1080d709 | 1194 | printk(KERN_WARNING "%s: %d rebuilds is over limit, route caching disabled\n", |
d8d1f30b | 1195 | rt->dst.dev->name, num); |
1080d709 | 1196 | } |
b35ecb5d | 1197 | rt_emergency_hash_rebuild(net); |
6a2bad70 PE |
1198 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1199 | ||
1200 | hash = rt_hash(rt->fl.fl4_dst, rt->fl.fl4_src, | |
1201 | ifindex, rt_genid(net)); | |
1202 | goto restart; | |
1080d709 | 1203 | } |
1da177e4 LT |
1204 | } |
1205 | ||
1206 | /* Try to bind route to arp only if it is output | |
1207 | route or unicast forwarding path. | |
1208 | */ | |
1209 | if (rt->rt_type == RTN_UNICAST || rt->fl.iif == 0) { | |
d8d1f30b | 1210 | int err = arp_bind_neighbour(&rt->dst); |
1da177e4 | 1211 | if (err) { |
22c047cc | 1212 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1213 | |
1214 | if (err != -ENOBUFS) { | |
1215 | rt_drop(rt); | |
1216 | return err; | |
1217 | } | |
1218 | ||
1219 | /* Neighbour tables are full and nothing | |
1220 | can be released. Try to shrink route cache, | |
1221 | it is most likely it holds some neighbour records. | |
1222 | */ | |
1223 | if (attempts-- > 0) { | |
1224 | int saved_elasticity = ip_rt_gc_elasticity; | |
1225 | int saved_int = ip_rt_gc_min_interval; | |
1226 | ip_rt_gc_elasticity = 1; | |
1227 | ip_rt_gc_min_interval = 0; | |
569d3645 | 1228 | rt_garbage_collect(&ipv4_dst_ops); |
1da177e4 LT |
1229 | ip_rt_gc_min_interval = saved_int; |
1230 | ip_rt_gc_elasticity = saved_elasticity; | |
1231 | goto restart; | |
1232 | } | |
1233 | ||
1234 | if (net_ratelimit()) | |
7e1b33e5 | 1235 | printk(KERN_WARNING "ipv4: Neighbour table overflow.\n"); |
1da177e4 LT |
1236 | rt_drop(rt); |
1237 | return -ENOBUFS; | |
1238 | } | |
1239 | } | |
1240 | ||
d8d1f30b | 1241 | rt->dst.rt_next = rt_hash_table[hash].chain; |
1080d709 | 1242 | |
1da177e4 | 1243 | #if RT_CACHE_DEBUG >= 2 |
d8d1f30b | 1244 | if (rt->dst.rt_next) { |
1da177e4 | 1245 | struct rtable *trt; |
b6280b47 NH |
1246 | printk(KERN_DEBUG "rt_cache @%02x: %pI4", |
1247 | hash, &rt->rt_dst); | |
d8d1f30b | 1248 | for (trt = rt->dst.rt_next; trt; trt = trt->dst.rt_next) |
673d57e7 | 1249 | printk(" . %pI4", &trt->rt_dst); |
1da177e4 LT |
1250 | printk("\n"); |
1251 | } | |
1252 | #endif | |
00269b54 ED |
1253 | /* |
1254 | * Since lookup is lockfree, we must make sure | |
1255 | * previous writes to rt are comitted to memory | |
1256 | * before making rt visible to other CPUS. | |
1257 | */ | |
1ddbcb00 | 1258 | rcu_assign_pointer(rt_hash_table[hash].chain, rt); |
1080d709 | 1259 | |
22c047cc | 1260 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
73e42897 | 1261 | |
b6280b47 | 1262 | skip_hashing: |
511c3f92 ED |
1263 | if (rp) |
1264 | *rp = rt; | |
1265 | else | |
d8d1f30b | 1266 | skb_dst_set(skb, &rt->dst); |
1da177e4 LT |
1267 | return 0; |
1268 | } | |
1269 | ||
1270 | void rt_bind_peer(struct rtable *rt, int create) | |
1271 | { | |
1da177e4 LT |
1272 | struct inet_peer *peer; |
1273 | ||
1274 | peer = inet_getpeer(rt->rt_dst, create); | |
1275 | ||
49e8ab03 | 1276 | if (peer && cmpxchg(&rt->peer, NULL, peer) != NULL) |
1da177e4 LT |
1277 | inet_putpeer(peer); |
1278 | } | |
1279 | ||
1280 | /* | |
1281 | * Peer allocation may fail only in serious out-of-memory conditions. However | |
1282 | * we still can generate some output. | |
1283 | * Random ID selection looks a bit dangerous because we have no chances to | |
1284 | * select ID being unique in a reasonable period of time. | |
1285 | * But broken packet identifier may be better than no packet at all. | |
1286 | */ | |
1287 | static void ip_select_fb_ident(struct iphdr *iph) | |
1288 | { | |
1289 | static DEFINE_SPINLOCK(ip_fb_id_lock); | |
1290 | static u32 ip_fallback_id; | |
1291 | u32 salt; | |
1292 | ||
1293 | spin_lock_bh(&ip_fb_id_lock); | |
e448515c | 1294 | salt = secure_ip_id((__force __be32)ip_fallback_id ^ iph->daddr); |
1da177e4 LT |
1295 | iph->id = htons(salt & 0xFFFF); |
1296 | ip_fallback_id = salt; | |
1297 | spin_unlock_bh(&ip_fb_id_lock); | |
1298 | } | |
1299 | ||
1300 | void __ip_select_ident(struct iphdr *iph, struct dst_entry *dst, int more) | |
1301 | { | |
1302 | struct rtable *rt = (struct rtable *) dst; | |
1303 | ||
1304 | if (rt) { | |
1305 | if (rt->peer == NULL) | |
1306 | rt_bind_peer(rt, 1); | |
1307 | ||
1308 | /* If peer is attached to destination, it is never detached, | |
1309 | so that we need not to grab a lock to dereference it. | |
1310 | */ | |
1311 | if (rt->peer) { | |
1312 | iph->id = htons(inet_getid(rt->peer, more)); | |
1313 | return; | |
1314 | } | |
1315 | } else | |
e905a9ed | 1316 | printk(KERN_DEBUG "rt_bind_peer(0) @%p\n", |
9c2b3328 | 1317 | __builtin_return_address(0)); |
1da177e4 LT |
1318 | |
1319 | ip_select_fb_ident(iph); | |
1320 | } | |
4bc2f18b | 1321 | EXPORT_SYMBOL(__ip_select_ident); |
1da177e4 LT |
1322 | |
1323 | static void rt_del(unsigned hash, struct rtable *rt) | |
1324 | { | |
29e75252 | 1325 | struct rtable **rthp, *aux; |
1da177e4 | 1326 | |
29e75252 | 1327 | rthp = &rt_hash_table[hash].chain; |
22c047cc | 1328 | spin_lock_bh(rt_hash_lock_addr(hash)); |
1da177e4 | 1329 | ip_rt_put(rt); |
29e75252 | 1330 | while ((aux = *rthp) != NULL) { |
e84f84f2 | 1331 | if (aux == rt || rt_is_expired(aux)) { |
d8d1f30b | 1332 | *rthp = aux->dst.rt_next; |
29e75252 ED |
1333 | rt_free(aux); |
1334 | continue; | |
1da177e4 | 1335 | } |
d8d1f30b | 1336 | rthp = &aux->dst.rt_next; |
29e75252 | 1337 | } |
22c047cc | 1338 | spin_unlock_bh(rt_hash_lock_addr(hash)); |
1da177e4 LT |
1339 | } |
1340 | ||
ed7865a4 | 1341 | /* called in rcu_read_lock() section */ |
f7655229 AV |
1342 | void ip_rt_redirect(__be32 old_gw, __be32 daddr, __be32 new_gw, |
1343 | __be32 saddr, struct net_device *dev) | |
1da177e4 LT |
1344 | { |
1345 | int i, k; | |
ed7865a4 | 1346 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1347 | struct rtable *rth, **rthp; |
f7655229 | 1348 | __be32 skeys[2] = { saddr, 0 }; |
1da177e4 | 1349 | int ikeys[2] = { dev->ifindex, 0 }; |
8d71740c | 1350 | struct netevent_redirect netevent; |
317805b8 | 1351 | struct net *net; |
1da177e4 | 1352 | |
1da177e4 LT |
1353 | if (!in_dev) |
1354 | return; | |
1355 | ||
c346dca1 | 1356 | net = dev_net(dev); |
9d4fb27d JP |
1357 | if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) || |
1358 | ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) || | |
1359 | ipv4_is_zeronet(new_gw)) | |
1da177e4 LT |
1360 | goto reject_redirect; |
1361 | ||
1080d709 NH |
1362 | if (!rt_caching(net)) |
1363 | goto reject_redirect; | |
1364 | ||
1da177e4 LT |
1365 | if (!IN_DEV_SHARED_MEDIA(in_dev)) { |
1366 | if (!inet_addr_onlink(in_dev, new_gw, old_gw)) | |
1367 | goto reject_redirect; | |
1368 | if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev)) | |
1369 | goto reject_redirect; | |
1370 | } else { | |
317805b8 | 1371 | if (inet_addr_type(net, new_gw) != RTN_UNICAST) |
1da177e4 LT |
1372 | goto reject_redirect; |
1373 | } | |
1374 | ||
1375 | for (i = 0; i < 2; i++) { | |
1376 | for (k = 0; k < 2; k++) { | |
b00180de | 1377 | unsigned hash = rt_hash(daddr, skeys[i], ikeys[k], |
e84f84f2 | 1378 | rt_genid(net)); |
1da177e4 LT |
1379 | |
1380 | rthp=&rt_hash_table[hash].chain; | |
1381 | ||
1da177e4 LT |
1382 | while ((rth = rcu_dereference(*rthp)) != NULL) { |
1383 | struct rtable *rt; | |
1384 | ||
1385 | if (rth->fl.fl4_dst != daddr || | |
1386 | rth->fl.fl4_src != skeys[i] || | |
1da177e4 | 1387 | rth->fl.oif != ikeys[k] || |
29e75252 | 1388 | rth->fl.iif != 0 || |
e84f84f2 | 1389 | rt_is_expired(rth) || |
d8d1f30b CG |
1390 | !net_eq(dev_net(rth->dst.dev), net)) { |
1391 | rthp = &rth->dst.rt_next; | |
1da177e4 LT |
1392 | continue; |
1393 | } | |
1394 | ||
1395 | if (rth->rt_dst != daddr || | |
1396 | rth->rt_src != saddr || | |
d8d1f30b | 1397 | rth->dst.error || |
1da177e4 | 1398 | rth->rt_gateway != old_gw || |
d8d1f30b | 1399 | rth->dst.dev != dev) |
1da177e4 LT |
1400 | break; |
1401 | ||
d8d1f30b | 1402 | dst_hold(&rth->dst); |
1da177e4 LT |
1403 | |
1404 | rt = dst_alloc(&ipv4_dst_ops); | |
1405 | if (rt == NULL) { | |
1406 | ip_rt_put(rth); | |
1da177e4 LT |
1407 | return; |
1408 | } | |
1409 | ||
1410 | /* Copy all the information. */ | |
1411 | *rt = *rth; | |
d8d1f30b CG |
1412 | rt->dst.__use = 1; |
1413 | atomic_set(&rt->dst.__refcnt, 1); | |
1414 | rt->dst.child = NULL; | |
1415 | if (rt->dst.dev) | |
1416 | dev_hold(rt->dst.dev); | |
1da177e4 LT |
1417 | if (rt->idev) |
1418 | in_dev_hold(rt->idev); | |
d8d1f30b CG |
1419 | rt->dst.obsolete = -1; |
1420 | rt->dst.lastuse = jiffies; | |
1421 | rt->dst.path = &rt->dst; | |
1422 | rt->dst.neighbour = NULL; | |
1423 | rt->dst.hh = NULL; | |
def8b4fa | 1424 | #ifdef CONFIG_XFRM |
d8d1f30b | 1425 | rt->dst.xfrm = NULL; |
def8b4fa | 1426 | #endif |
e84f84f2 | 1427 | rt->rt_genid = rt_genid(net); |
1da177e4 LT |
1428 | rt->rt_flags |= RTCF_REDIRECTED; |
1429 | ||
1430 | /* Gateway is different ... */ | |
1431 | rt->rt_gateway = new_gw; | |
1432 | ||
1433 | /* Redirect received -> path was valid */ | |
d8d1f30b | 1434 | dst_confirm(&rth->dst); |
1da177e4 LT |
1435 | |
1436 | if (rt->peer) | |
1437 | atomic_inc(&rt->peer->refcnt); | |
1438 | ||
d8d1f30b CG |
1439 | if (arp_bind_neighbour(&rt->dst) || |
1440 | !(rt->dst.neighbour->nud_state & | |
1da177e4 | 1441 | NUD_VALID)) { |
d8d1f30b CG |
1442 | if (rt->dst.neighbour) |
1443 | neigh_event_send(rt->dst.neighbour, NULL); | |
1da177e4 LT |
1444 | ip_rt_put(rth); |
1445 | rt_drop(rt); | |
1446 | goto do_next; | |
1447 | } | |
e905a9ed | 1448 | |
d8d1f30b CG |
1449 | netevent.old = &rth->dst; |
1450 | netevent.new = &rt->dst; | |
e905a9ed YH |
1451 | call_netevent_notifiers(NETEVENT_REDIRECT, |
1452 | &netevent); | |
1da177e4 LT |
1453 | |
1454 | rt_del(hash, rth); | |
6a2bad70 | 1455 | if (!rt_intern_hash(hash, rt, &rt, NULL, rt->fl.oif)) |
1da177e4 LT |
1456 | ip_rt_put(rt); |
1457 | goto do_next; | |
1458 | } | |
1da177e4 LT |
1459 | do_next: |
1460 | ; | |
1461 | } | |
1462 | } | |
1da177e4 LT |
1463 | return; |
1464 | ||
1465 | reject_redirect: | |
1466 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1467 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) | |
673d57e7 HH |
1468 | printk(KERN_INFO "Redirect from %pI4 on %s about %pI4 ignored.\n" |
1469 | " Advised path = %pI4 -> %pI4\n", | |
1470 | &old_gw, dev->name, &new_gw, | |
1471 | &saddr, &daddr); | |
1da177e4 | 1472 | #endif |
ed7865a4 | 1473 | ; |
1da177e4 LT |
1474 | } |
1475 | ||
1476 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst) | |
1477 | { | |
ee6b9673 | 1478 | struct rtable *rt = (struct rtable *)dst; |
1da177e4 LT |
1479 | struct dst_entry *ret = dst; |
1480 | ||
1481 | if (rt) { | |
d11a4dc1 | 1482 | if (dst->obsolete > 0) { |
1da177e4 LT |
1483 | ip_rt_put(rt); |
1484 | ret = NULL; | |
1485 | } else if ((rt->rt_flags & RTCF_REDIRECTED) || | |
d8d1f30b CG |
1486 | (rt->dst.expires && |
1487 | time_after_eq(jiffies, rt->dst.expires))) { | |
8c7bc840 | 1488 | unsigned hash = rt_hash(rt->fl.fl4_dst, rt->fl.fl4_src, |
b00180de | 1489 | rt->fl.oif, |
e84f84f2 | 1490 | rt_genid(dev_net(dst->dev))); |
1da177e4 | 1491 | #if RT_CACHE_DEBUG >= 1 |
673d57e7 HH |
1492 | printk(KERN_DEBUG "ipv4_negative_advice: redirect to %pI4/%02x dropped\n", |
1493 | &rt->rt_dst, rt->fl.fl4_tos); | |
1da177e4 LT |
1494 | #endif |
1495 | rt_del(hash, rt); | |
1496 | ret = NULL; | |
1497 | } | |
1498 | } | |
1499 | return ret; | |
1500 | } | |
1501 | ||
1502 | /* | |
1503 | * Algorithm: | |
1504 | * 1. The first ip_rt_redirect_number redirects are sent | |
1505 | * with exponential backoff, then we stop sending them at all, | |
1506 | * assuming that the host ignores our redirects. | |
1507 | * 2. If we did not see packets requiring redirects | |
1508 | * during ip_rt_redirect_silence, we assume that the host | |
1509 | * forgot redirected route and start to send redirects again. | |
1510 | * | |
1511 | * This algorithm is much cheaper and more intelligent than dumb load limiting | |
1512 | * in icmp.c. | |
1513 | * | |
1514 | * NOTE. Do not forget to inhibit load limiting for redirects (redundant) | |
1515 | * and "frag. need" (breaks PMTU discovery) in icmp.c. | |
1516 | */ | |
1517 | ||
1518 | void ip_rt_send_redirect(struct sk_buff *skb) | |
1519 | { | |
511c3f92 | 1520 | struct rtable *rt = skb_rtable(skb); |
30038fc6 ED |
1521 | struct in_device *in_dev; |
1522 | int log_martians; | |
1da177e4 | 1523 | |
30038fc6 | 1524 | rcu_read_lock(); |
d8d1f30b | 1525 | in_dev = __in_dev_get_rcu(rt->dst.dev); |
30038fc6 ED |
1526 | if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) { |
1527 | rcu_read_unlock(); | |
1da177e4 | 1528 | return; |
30038fc6 ED |
1529 | } |
1530 | log_martians = IN_DEV_LOG_MARTIANS(in_dev); | |
1531 | rcu_read_unlock(); | |
1da177e4 LT |
1532 | |
1533 | /* No redirected packets during ip_rt_redirect_silence; | |
1534 | * reset the algorithm. | |
1535 | */ | |
d8d1f30b CG |
1536 | if (time_after(jiffies, rt->dst.rate_last + ip_rt_redirect_silence)) |
1537 | rt->dst.rate_tokens = 0; | |
1da177e4 LT |
1538 | |
1539 | /* Too many ignored redirects; do not send anything | |
d8d1f30b | 1540 | * set dst.rate_last to the last seen redirected packet. |
1da177e4 | 1541 | */ |
d8d1f30b CG |
1542 | if (rt->dst.rate_tokens >= ip_rt_redirect_number) { |
1543 | rt->dst.rate_last = jiffies; | |
30038fc6 | 1544 | return; |
1da177e4 LT |
1545 | } |
1546 | ||
1547 | /* Check for load limit; set rate_last to the latest sent | |
1548 | * redirect. | |
1549 | */ | |
d8d1f30b | 1550 | if (rt->dst.rate_tokens == 0 || |
14fb8a76 | 1551 | time_after(jiffies, |
d8d1f30b CG |
1552 | (rt->dst.rate_last + |
1553 | (ip_rt_redirect_load << rt->dst.rate_tokens)))) { | |
1da177e4 | 1554 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, rt->rt_gateway); |
d8d1f30b CG |
1555 | rt->dst.rate_last = jiffies; |
1556 | ++rt->dst.rate_tokens; | |
1da177e4 | 1557 | #ifdef CONFIG_IP_ROUTE_VERBOSE |
30038fc6 | 1558 | if (log_martians && |
d8d1f30b | 1559 | rt->dst.rate_tokens == ip_rt_redirect_number && |
1da177e4 | 1560 | net_ratelimit()) |
673d57e7 HH |
1561 | printk(KERN_WARNING "host %pI4/if%d ignores redirects for %pI4 to %pI4.\n", |
1562 | &rt->rt_src, rt->rt_iif, | |
1563 | &rt->rt_dst, &rt->rt_gateway); | |
1da177e4 LT |
1564 | #endif |
1565 | } | |
1da177e4 LT |
1566 | } |
1567 | ||
1568 | static int ip_error(struct sk_buff *skb) | |
1569 | { | |
511c3f92 | 1570 | struct rtable *rt = skb_rtable(skb); |
1da177e4 LT |
1571 | unsigned long now; |
1572 | int code; | |
1573 | ||
d8d1f30b | 1574 | switch (rt->dst.error) { |
1da177e4 LT |
1575 | case EINVAL: |
1576 | default: | |
1577 | goto out; | |
1578 | case EHOSTUNREACH: | |
1579 | code = ICMP_HOST_UNREACH; | |
1580 | break; | |
1581 | case ENETUNREACH: | |
1582 | code = ICMP_NET_UNREACH; | |
d8d1f30b | 1583 | IP_INC_STATS_BH(dev_net(rt->dst.dev), |
7c73a6fa | 1584 | IPSTATS_MIB_INNOROUTES); |
1da177e4 LT |
1585 | break; |
1586 | case EACCES: | |
1587 | code = ICMP_PKT_FILTERED; | |
1588 | break; | |
1589 | } | |
1590 | ||
1591 | now = jiffies; | |
d8d1f30b CG |
1592 | rt->dst.rate_tokens += now - rt->dst.rate_last; |
1593 | if (rt->dst.rate_tokens > ip_rt_error_burst) | |
1594 | rt->dst.rate_tokens = ip_rt_error_burst; | |
1595 | rt->dst.rate_last = now; | |
1596 | if (rt->dst.rate_tokens >= ip_rt_error_cost) { | |
1597 | rt->dst.rate_tokens -= ip_rt_error_cost; | |
1da177e4 LT |
1598 | icmp_send(skb, ICMP_DEST_UNREACH, code, 0); |
1599 | } | |
1600 | ||
1601 | out: kfree_skb(skb); | |
1602 | return 0; | |
e905a9ed | 1603 | } |
1da177e4 LT |
1604 | |
1605 | /* | |
1606 | * The last two values are not from the RFC but | |
1607 | * are needed for AMPRnet AX.25 paths. | |
1608 | */ | |
1609 | ||
9b5b5cff | 1610 | static const unsigned short mtu_plateau[] = |
1da177e4 LT |
1611 | {32000, 17914, 8166, 4352, 2002, 1492, 576, 296, 216, 128 }; |
1612 | ||
5969f71d | 1613 | static inline unsigned short guess_mtu(unsigned short old_mtu) |
1da177e4 LT |
1614 | { |
1615 | int i; | |
e905a9ed | 1616 | |
1da177e4 LT |
1617 | for (i = 0; i < ARRAY_SIZE(mtu_plateau); i++) |
1618 | if (old_mtu > mtu_plateau[i]) | |
1619 | return mtu_plateau[i]; | |
1620 | return 68; | |
1621 | } | |
1622 | ||
b5921910 | 1623 | unsigned short ip_rt_frag_needed(struct net *net, struct iphdr *iph, |
0010e465 TT |
1624 | unsigned short new_mtu, |
1625 | struct net_device *dev) | |
1da177e4 | 1626 | { |
0010e465 | 1627 | int i, k; |
1da177e4 LT |
1628 | unsigned short old_mtu = ntohs(iph->tot_len); |
1629 | struct rtable *rth; | |
0010e465 | 1630 | int ikeys[2] = { dev->ifindex, 0 }; |
e448515c AV |
1631 | __be32 skeys[2] = { iph->saddr, 0, }; |
1632 | __be32 daddr = iph->daddr; | |
1da177e4 LT |
1633 | unsigned short est_mtu = 0; |
1634 | ||
0010e465 TT |
1635 | for (k = 0; k < 2; k++) { |
1636 | for (i = 0; i < 2; i++) { | |
b00180de | 1637 | unsigned hash = rt_hash(daddr, skeys[i], ikeys[k], |
e84f84f2 | 1638 | rt_genid(net)); |
0010e465 TT |
1639 | |
1640 | rcu_read_lock(); | |
1641 | for (rth = rcu_dereference(rt_hash_table[hash].chain); rth; | |
d8d1f30b | 1642 | rth = rcu_dereference(rth->dst.rt_next)) { |
1da177e4 LT |
1643 | unsigned short mtu = new_mtu; |
1644 | ||
0010e465 TT |
1645 | if (rth->fl.fl4_dst != daddr || |
1646 | rth->fl.fl4_src != skeys[i] || | |
1647 | rth->rt_dst != daddr || | |
1648 | rth->rt_src != iph->saddr || | |
1649 | rth->fl.oif != ikeys[k] || | |
1650 | rth->fl.iif != 0 || | |
d8d1f30b CG |
1651 | dst_metric_locked(&rth->dst, RTAX_MTU) || |
1652 | !net_eq(dev_net(rth->dst.dev), net) || | |
6c3b8fc6 | 1653 | rt_is_expired(rth)) |
0010e465 TT |
1654 | continue; |
1655 | ||
1da177e4 LT |
1656 | if (new_mtu < 68 || new_mtu >= old_mtu) { |
1657 | ||
1658 | /* BSD 4.2 compatibility hack :-( */ | |
1659 | if (mtu == 0 && | |
d8d1f30b | 1660 | old_mtu >= dst_mtu(&rth->dst) && |
1da177e4 LT |
1661 | old_mtu >= 68 + (iph->ihl << 2)) |
1662 | old_mtu -= iph->ihl << 2; | |
1663 | ||
1664 | mtu = guess_mtu(old_mtu); | |
1665 | } | |
d8d1f30b CG |
1666 | if (mtu <= dst_mtu(&rth->dst)) { |
1667 | if (mtu < dst_mtu(&rth->dst)) { | |
1668 | dst_confirm(&rth->dst); | |
1da177e4 LT |
1669 | if (mtu < ip_rt_min_pmtu) { |
1670 | mtu = ip_rt_min_pmtu; | |
d8d1f30b | 1671 | rth->dst.metrics[RTAX_LOCK-1] |= |
1da177e4 LT |
1672 | (1 << RTAX_MTU); |
1673 | } | |
d8d1f30b CG |
1674 | rth->dst.metrics[RTAX_MTU-1] = mtu; |
1675 | dst_set_expires(&rth->dst, | |
1da177e4 LT |
1676 | ip_rt_mtu_expires); |
1677 | } | |
1678 | est_mtu = mtu; | |
1679 | } | |
1680 | } | |
0010e465 | 1681 | rcu_read_unlock(); |
1da177e4 | 1682 | } |
1da177e4 LT |
1683 | } |
1684 | return est_mtu ? : new_mtu; | |
1685 | } | |
1686 | ||
1687 | static void ip_rt_update_pmtu(struct dst_entry *dst, u32 mtu) | |
1688 | { | |
6d273f8d | 1689 | if (dst_mtu(dst) > mtu && mtu >= 68 && |
1da177e4 LT |
1690 | !(dst_metric_locked(dst, RTAX_MTU))) { |
1691 | if (mtu < ip_rt_min_pmtu) { | |
1692 | mtu = ip_rt_min_pmtu; | |
1693 | dst->metrics[RTAX_LOCK-1] |= (1 << RTAX_MTU); | |
1694 | } | |
1695 | dst->metrics[RTAX_MTU-1] = mtu; | |
1696 | dst_set_expires(dst, ip_rt_mtu_expires); | |
8d71740c | 1697 | call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst); |
1da177e4 LT |
1698 | } |
1699 | } | |
1700 | ||
1701 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie) | |
1702 | { | |
d11a4dc1 TT |
1703 | if (rt_is_expired((struct rtable *)dst)) |
1704 | return NULL; | |
1705 | return dst; | |
1da177e4 LT |
1706 | } |
1707 | ||
1708 | static void ipv4_dst_destroy(struct dst_entry *dst) | |
1709 | { | |
1710 | struct rtable *rt = (struct rtable *) dst; | |
1711 | struct inet_peer *peer = rt->peer; | |
1712 | struct in_device *idev = rt->idev; | |
1713 | ||
1714 | if (peer) { | |
1715 | rt->peer = NULL; | |
1716 | inet_putpeer(peer); | |
1717 | } | |
1718 | ||
1719 | if (idev) { | |
1720 | rt->idev = NULL; | |
1721 | in_dev_put(idev); | |
1722 | } | |
1723 | } | |
1724 | ||
1725 | static void ipv4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, | |
1726 | int how) | |
1727 | { | |
1728 | struct rtable *rt = (struct rtable *) dst; | |
1729 | struct in_device *idev = rt->idev; | |
c346dca1 | 1730 | if (dev != dev_net(dev)->loopback_dev && idev && idev->dev == dev) { |
5a3e55d6 | 1731 | struct in_device *loopback_idev = |
c346dca1 | 1732 | in_dev_get(dev_net(dev)->loopback_dev); |
1da177e4 LT |
1733 | if (loopback_idev) { |
1734 | rt->idev = loopback_idev; | |
1735 | in_dev_put(idev); | |
1736 | } | |
1737 | } | |
1738 | } | |
1739 | ||
1740 | static void ipv4_link_failure(struct sk_buff *skb) | |
1741 | { | |
1742 | struct rtable *rt; | |
1743 | ||
1744 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0); | |
1745 | ||
511c3f92 | 1746 | rt = skb_rtable(skb); |
1da177e4 | 1747 | if (rt) |
d8d1f30b | 1748 | dst_set_expires(&rt->dst, 0); |
1da177e4 LT |
1749 | } |
1750 | ||
1751 | static int ip_rt_bug(struct sk_buff *skb) | |
1752 | { | |
673d57e7 HH |
1753 | printk(KERN_DEBUG "ip_rt_bug: %pI4 -> %pI4, %s\n", |
1754 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
1da177e4 LT |
1755 | skb->dev ? skb->dev->name : "?"); |
1756 | kfree_skb(skb); | |
1757 | return 0; | |
1758 | } | |
1759 | ||
1760 | /* | |
1761 | We do not cache source address of outgoing interface, | |
1762 | because it is used only by IP RR, TS and SRR options, | |
1763 | so that it out of fast path. | |
1764 | ||
1765 | BTW remember: "addr" is allowed to be not aligned | |
1766 | in IP options! | |
1767 | */ | |
1768 | ||
1769 | void ip_rt_get_source(u8 *addr, struct rtable *rt) | |
1770 | { | |
a61ced5d | 1771 | __be32 src; |
1da177e4 LT |
1772 | struct fib_result res; |
1773 | ||
1774 | if (rt->fl.iif == 0) | |
1775 | src = rt->rt_src; | |
ebc0ffae ED |
1776 | else { |
1777 | rcu_read_lock(); | |
1778 | if (fib_lookup(dev_net(rt->dst.dev), &rt->fl, &res) == 0) | |
1779 | src = FIB_RES_PREFSRC(res); | |
1780 | else | |
1781 | src = inet_select_addr(rt->dst.dev, rt->rt_gateway, | |
1da177e4 | 1782 | RT_SCOPE_UNIVERSE); |
ebc0ffae ED |
1783 | rcu_read_unlock(); |
1784 | } | |
1da177e4 LT |
1785 | memcpy(addr, &src, 4); |
1786 | } | |
1787 | ||
1788 | #ifdef CONFIG_NET_CLS_ROUTE | |
1789 | static void set_class_tag(struct rtable *rt, u32 tag) | |
1790 | { | |
d8d1f30b CG |
1791 | if (!(rt->dst.tclassid & 0xFFFF)) |
1792 | rt->dst.tclassid |= tag & 0xFFFF; | |
1793 | if (!(rt->dst.tclassid & 0xFFFF0000)) | |
1794 | rt->dst.tclassid |= tag & 0xFFFF0000; | |
1da177e4 LT |
1795 | } |
1796 | #endif | |
1797 | ||
1798 | static void rt_set_nexthop(struct rtable *rt, struct fib_result *res, u32 itag) | |
1799 | { | |
1800 | struct fib_info *fi = res->fi; | |
1801 | ||
1802 | if (fi) { | |
1803 | if (FIB_RES_GW(*res) && | |
1804 | FIB_RES_NH(*res).nh_scope == RT_SCOPE_LINK) | |
1805 | rt->rt_gateway = FIB_RES_GW(*res); | |
d8d1f30b CG |
1806 | memcpy(rt->dst.metrics, fi->fib_metrics, |
1807 | sizeof(rt->dst.metrics)); | |
1da177e4 | 1808 | if (fi->fib_mtu == 0) { |
d8d1f30b CG |
1809 | rt->dst.metrics[RTAX_MTU-1] = rt->dst.dev->mtu; |
1810 | if (dst_metric_locked(&rt->dst, RTAX_MTU) && | |
1da177e4 | 1811 | rt->rt_gateway != rt->rt_dst && |
d8d1f30b CG |
1812 | rt->dst.dev->mtu > 576) |
1813 | rt->dst.metrics[RTAX_MTU-1] = 576; | |
1da177e4 LT |
1814 | } |
1815 | #ifdef CONFIG_NET_CLS_ROUTE | |
d8d1f30b | 1816 | rt->dst.tclassid = FIB_RES_NH(*res).nh_tclassid; |
1da177e4 LT |
1817 | #endif |
1818 | } else | |
d8d1f30b CG |
1819 | rt->dst.metrics[RTAX_MTU-1]= rt->dst.dev->mtu; |
1820 | ||
1821 | if (dst_metric(&rt->dst, RTAX_HOPLIMIT) == 0) | |
1822 | rt->dst.metrics[RTAX_HOPLIMIT-1] = sysctl_ip_default_ttl; | |
1823 | if (dst_mtu(&rt->dst) > IP_MAX_MTU) | |
1824 | rt->dst.metrics[RTAX_MTU-1] = IP_MAX_MTU; | |
1825 | if (dst_metric(&rt->dst, RTAX_ADVMSS) == 0) | |
1826 | rt->dst.metrics[RTAX_ADVMSS-1] = max_t(unsigned int, rt->dst.dev->mtu - 40, | |
1da177e4 | 1827 | ip_rt_min_advmss); |
d8d1f30b CG |
1828 | if (dst_metric(&rt->dst, RTAX_ADVMSS) > 65535 - 40) |
1829 | rt->dst.metrics[RTAX_ADVMSS-1] = 65535 - 40; | |
1da177e4 LT |
1830 | |
1831 | #ifdef CONFIG_NET_CLS_ROUTE | |
1832 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
1833 | set_class_tag(rt, fib_rules_tclass(res)); | |
1834 | #endif | |
1835 | set_class_tag(rt, itag); | |
1836 | #endif | |
e905a9ed | 1837 | rt->rt_type = res->type; |
1da177e4 LT |
1838 | } |
1839 | ||
96d36220 | 1840 | /* called in rcu_read_lock() section */ |
9e12bb22 | 1841 | static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
1842 | u8 tos, struct net_device *dev, int our) |
1843 | { | |
96d36220 | 1844 | unsigned int hash; |
1da177e4 | 1845 | struct rtable *rth; |
a61ced5d | 1846 | __be32 spec_dst; |
96d36220 | 1847 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1848 | u32 itag = 0; |
b5f7e755 | 1849 | int err; |
1da177e4 LT |
1850 | |
1851 | /* Primary sanity checks. */ | |
1852 | ||
1853 | if (in_dev == NULL) | |
1854 | return -EINVAL; | |
1855 | ||
1e637c74 | 1856 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
f97c1e0c | 1857 | ipv4_is_loopback(saddr) || skb->protocol != htons(ETH_P_IP)) |
1da177e4 LT |
1858 | goto e_inval; |
1859 | ||
f97c1e0c JP |
1860 | if (ipv4_is_zeronet(saddr)) { |
1861 | if (!ipv4_is_local_multicast(daddr)) | |
1da177e4 LT |
1862 | goto e_inval; |
1863 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); | |
b5f7e755 ED |
1864 | } else { |
1865 | err = fib_validate_source(saddr, 0, tos, 0, dev, &spec_dst, | |
1866 | &itag, 0); | |
1867 | if (err < 0) | |
1868 | goto e_err; | |
1869 | } | |
1da177e4 LT |
1870 | rth = dst_alloc(&ipv4_dst_ops); |
1871 | if (!rth) | |
1872 | goto e_nobufs; | |
1873 | ||
d8d1f30b CG |
1874 | rth->dst.output = ip_rt_bug; |
1875 | rth->dst.obsolete = -1; | |
1da177e4 | 1876 | |
d8d1f30b CG |
1877 | atomic_set(&rth->dst.__refcnt, 1); |
1878 | rth->dst.flags= DST_HOST; | |
42f811b8 | 1879 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
d8d1f30b | 1880 | rth->dst.flags |= DST_NOPOLICY; |
1da177e4 LT |
1881 | rth->fl.fl4_dst = daddr; |
1882 | rth->rt_dst = daddr; | |
1883 | rth->fl.fl4_tos = tos; | |
47dcf0cb | 1884 | rth->fl.mark = skb->mark; |
1da177e4 LT |
1885 | rth->fl.fl4_src = saddr; |
1886 | rth->rt_src = saddr; | |
1887 | #ifdef CONFIG_NET_CLS_ROUTE | |
d8d1f30b | 1888 | rth->dst.tclassid = itag; |
1da177e4 LT |
1889 | #endif |
1890 | rth->rt_iif = | |
1891 | rth->fl.iif = dev->ifindex; | |
d8d1f30b CG |
1892 | rth->dst.dev = init_net.loopback_dev; |
1893 | dev_hold(rth->dst.dev); | |
1894 | rth->idev = in_dev_get(rth->dst.dev); | |
1da177e4 LT |
1895 | rth->fl.oif = 0; |
1896 | rth->rt_gateway = daddr; | |
1897 | rth->rt_spec_dst= spec_dst; | |
e84f84f2 | 1898 | rth->rt_genid = rt_genid(dev_net(dev)); |
1da177e4 | 1899 | rth->rt_flags = RTCF_MULTICAST; |
29e75252 | 1900 | rth->rt_type = RTN_MULTICAST; |
1da177e4 | 1901 | if (our) { |
d8d1f30b | 1902 | rth->dst.input= ip_local_deliver; |
1da177e4 LT |
1903 | rth->rt_flags |= RTCF_LOCAL; |
1904 | } | |
1905 | ||
1906 | #ifdef CONFIG_IP_MROUTE | |
f97c1e0c | 1907 | if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev)) |
d8d1f30b | 1908 | rth->dst.input = ip_mr_input; |
1da177e4 LT |
1909 | #endif |
1910 | RT_CACHE_STAT_INC(in_slow_mc); | |
1911 | ||
e84f84f2 | 1912 | hash = rt_hash(daddr, saddr, dev->ifindex, rt_genid(dev_net(dev))); |
6a2bad70 | 1913 | return rt_intern_hash(hash, rth, NULL, skb, dev->ifindex); |
1da177e4 LT |
1914 | |
1915 | e_nobufs: | |
1da177e4 | 1916 | return -ENOBUFS; |
1da177e4 | 1917 | e_inval: |
96d36220 | 1918 | return -EINVAL; |
b5f7e755 | 1919 | e_err: |
b5f7e755 | 1920 | return err; |
1da177e4 LT |
1921 | } |
1922 | ||
1923 | ||
1924 | static void ip_handle_martian_source(struct net_device *dev, | |
1925 | struct in_device *in_dev, | |
1926 | struct sk_buff *skb, | |
9e12bb22 AV |
1927 | __be32 daddr, |
1928 | __be32 saddr) | |
1da177e4 LT |
1929 | { |
1930 | RT_CACHE_STAT_INC(in_martian_src); | |
1931 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1932 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) { | |
1933 | /* | |
1934 | * RFC1812 recommendation, if source is martian, | |
1935 | * the only hint is MAC header. | |
1936 | */ | |
673d57e7 HH |
1937 | printk(KERN_WARNING "martian source %pI4 from %pI4, on dev %s\n", |
1938 | &daddr, &saddr, dev->name); | |
98e399f8 | 1939 | if (dev->hard_header_len && skb_mac_header_was_set(skb)) { |
1da177e4 | 1940 | int i; |
98e399f8 | 1941 | const unsigned char *p = skb_mac_header(skb); |
1da177e4 LT |
1942 | printk(KERN_WARNING "ll header: "); |
1943 | for (i = 0; i < dev->hard_header_len; i++, p++) { | |
1944 | printk("%02x", *p); | |
1945 | if (i < (dev->hard_header_len - 1)) | |
1946 | printk(":"); | |
1947 | } | |
1948 | printk("\n"); | |
1949 | } | |
1950 | } | |
1951 | #endif | |
1952 | } | |
1953 | ||
47360228 | 1954 | /* called in rcu_read_lock() section */ |
5969f71d SH |
1955 | static int __mkroute_input(struct sk_buff *skb, |
1956 | struct fib_result *res, | |
1957 | struct in_device *in_dev, | |
1958 | __be32 daddr, __be32 saddr, u32 tos, | |
1959 | struct rtable **result) | |
1da177e4 | 1960 | { |
1da177e4 LT |
1961 | struct rtable *rth; |
1962 | int err; | |
1963 | struct in_device *out_dev; | |
47360228 | 1964 | unsigned int flags = 0; |
d9c9df8c AV |
1965 | __be32 spec_dst; |
1966 | u32 itag; | |
1da177e4 LT |
1967 | |
1968 | /* get a working reference to the output device */ | |
47360228 | 1969 | out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res)); |
1da177e4 LT |
1970 | if (out_dev == NULL) { |
1971 | if (net_ratelimit()) | |
1972 | printk(KERN_CRIT "Bug in ip_route_input" \ | |
1973 | "_slow(). Please, report\n"); | |
1974 | return -EINVAL; | |
1975 | } | |
1976 | ||
1977 | ||
e905a9ed | 1978 | err = fib_validate_source(saddr, daddr, tos, FIB_RES_OIF(*res), |
b0c110ca | 1979 | in_dev->dev, &spec_dst, &itag, skb->mark); |
1da177e4 | 1980 | if (err < 0) { |
e905a9ed | 1981 | ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr, |
1da177e4 | 1982 | saddr); |
e905a9ed | 1983 | |
1da177e4 LT |
1984 | goto cleanup; |
1985 | } | |
1986 | ||
1987 | if (err) | |
1988 | flags |= RTCF_DIRECTSRC; | |
1989 | ||
51b77cae | 1990 | if (out_dev == in_dev && err && |
1da177e4 LT |
1991 | (IN_DEV_SHARED_MEDIA(out_dev) || |
1992 | inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) | |
1993 | flags |= RTCF_DOREDIRECT; | |
1994 | ||
1995 | if (skb->protocol != htons(ETH_P_IP)) { | |
1996 | /* Not IP (i.e. ARP). Do not create route, if it is | |
1997 | * invalid for proxy arp. DNAT routes are always valid. | |
65324144 JDB |
1998 | * |
1999 | * Proxy arp feature have been extended to allow, ARP | |
2000 | * replies back to the same interface, to support | |
2001 | * Private VLAN switch technologies. See arp.c. | |
1da177e4 | 2002 | */ |
65324144 JDB |
2003 | if (out_dev == in_dev && |
2004 | IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) { | |
1da177e4 LT |
2005 | err = -EINVAL; |
2006 | goto cleanup; | |
2007 | } | |
2008 | } | |
2009 | ||
2010 | ||
2011 | rth = dst_alloc(&ipv4_dst_ops); | |
2012 | if (!rth) { | |
2013 | err = -ENOBUFS; | |
2014 | goto cleanup; | |
2015 | } | |
2016 | ||
d8d1f30b CG |
2017 | atomic_set(&rth->dst.__refcnt, 1); |
2018 | rth->dst.flags= DST_HOST; | |
42f811b8 | 2019 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
d8d1f30b | 2020 | rth->dst.flags |= DST_NOPOLICY; |
42f811b8 | 2021 | if (IN_DEV_CONF_GET(out_dev, NOXFRM)) |
d8d1f30b | 2022 | rth->dst.flags |= DST_NOXFRM; |
1da177e4 LT |
2023 | rth->fl.fl4_dst = daddr; |
2024 | rth->rt_dst = daddr; | |
2025 | rth->fl.fl4_tos = tos; | |
47dcf0cb | 2026 | rth->fl.mark = skb->mark; |
1da177e4 LT |
2027 | rth->fl.fl4_src = saddr; |
2028 | rth->rt_src = saddr; | |
2029 | rth->rt_gateway = daddr; | |
2030 | rth->rt_iif = | |
2031 | rth->fl.iif = in_dev->dev->ifindex; | |
d8d1f30b CG |
2032 | rth->dst.dev = (out_dev)->dev; |
2033 | dev_hold(rth->dst.dev); | |
2034 | rth->idev = in_dev_get(rth->dst.dev); | |
1da177e4 LT |
2035 | rth->fl.oif = 0; |
2036 | rth->rt_spec_dst= spec_dst; | |
2037 | ||
d8d1f30b CG |
2038 | rth->dst.obsolete = -1; |
2039 | rth->dst.input = ip_forward; | |
2040 | rth->dst.output = ip_output; | |
2041 | rth->rt_genid = rt_genid(dev_net(rth->dst.dev)); | |
1da177e4 LT |
2042 | |
2043 | rt_set_nexthop(rth, res, itag); | |
2044 | ||
2045 | rth->rt_flags = flags; | |
2046 | ||
2047 | *result = rth; | |
2048 | err = 0; | |
2049 | cleanup: | |
1da177e4 | 2050 | return err; |
e905a9ed | 2051 | } |
1da177e4 | 2052 | |
5969f71d SH |
2053 | static int ip_mkroute_input(struct sk_buff *skb, |
2054 | struct fib_result *res, | |
2055 | const struct flowi *fl, | |
2056 | struct in_device *in_dev, | |
2057 | __be32 daddr, __be32 saddr, u32 tos) | |
1da177e4 | 2058 | { |
7abaa27c | 2059 | struct rtable* rth = NULL; |
1da177e4 LT |
2060 | int err; |
2061 | unsigned hash; | |
2062 | ||
2063 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
2064 | if (res->fi && res->fi->fib_nhs > 1 && fl->oif == 0) | |
2065 | fib_select_multipath(fl, res); | |
2066 | #endif | |
2067 | ||
2068 | /* create a routing cache entry */ | |
2069 | err = __mkroute_input(skb, res, in_dev, daddr, saddr, tos, &rth); | |
2070 | if (err) | |
2071 | return err; | |
1da177e4 LT |
2072 | |
2073 | /* put it into the cache */ | |
e84f84f2 | 2074 | hash = rt_hash(daddr, saddr, fl->iif, |
d8d1f30b | 2075 | rt_genid(dev_net(rth->dst.dev))); |
6a2bad70 | 2076 | return rt_intern_hash(hash, rth, NULL, skb, fl->iif); |
1da177e4 LT |
2077 | } |
2078 | ||
1da177e4 LT |
2079 | /* |
2080 | * NOTE. We drop all the packets that has local source | |
2081 | * addresses, because every properly looped back packet | |
2082 | * must have correct destination already attached by output routine. | |
2083 | * | |
2084 | * Such approach solves two big problems: | |
2085 | * 1. Not simplex devices are handled properly. | |
2086 | * 2. IP spoofing attempts are filtered with 100% of guarantee. | |
ebc0ffae | 2087 | * called with rcu_read_lock() |
1da177e4 LT |
2088 | */ |
2089 | ||
9e12bb22 | 2090 | static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
2091 | u8 tos, struct net_device *dev) |
2092 | { | |
2093 | struct fib_result res; | |
96d36220 | 2094 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 LT |
2095 | struct flowi fl = { .nl_u = { .ip4_u = |
2096 | { .daddr = daddr, | |
2097 | .saddr = saddr, | |
2098 | .tos = tos, | |
2099 | .scope = RT_SCOPE_UNIVERSE, | |
1da177e4 | 2100 | } }, |
47dcf0cb | 2101 | .mark = skb->mark, |
1da177e4 LT |
2102 | .iif = dev->ifindex }; |
2103 | unsigned flags = 0; | |
2104 | u32 itag = 0; | |
2105 | struct rtable * rth; | |
2106 | unsigned hash; | |
9e12bb22 | 2107 | __be32 spec_dst; |
1da177e4 | 2108 | int err = -EINVAL; |
c346dca1 | 2109 | struct net * net = dev_net(dev); |
1da177e4 LT |
2110 | |
2111 | /* IP on this device is disabled. */ | |
2112 | ||
2113 | if (!in_dev) | |
2114 | goto out; | |
2115 | ||
2116 | /* Check for the most weird martians, which can be not detected | |
2117 | by fib_lookup. | |
2118 | */ | |
2119 | ||
1e637c74 | 2120 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
f97c1e0c | 2121 | ipv4_is_loopback(saddr)) |
1da177e4 LT |
2122 | goto martian_source; |
2123 | ||
e448515c | 2124 | if (daddr == htonl(0xFFFFFFFF) || (saddr == 0 && daddr == 0)) |
1da177e4 LT |
2125 | goto brd_input; |
2126 | ||
2127 | /* Accept zero addresses only to limited broadcast; | |
2128 | * I even do not know to fix it or not. Waiting for complains :-) | |
2129 | */ | |
f97c1e0c | 2130 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
2131 | goto martian_source; |
2132 | ||
1e637c74 | 2133 | if (ipv4_is_lbcast(daddr) || ipv4_is_zeronet(daddr) || |
f97c1e0c | 2134 | ipv4_is_loopback(daddr)) |
1da177e4 LT |
2135 | goto martian_destination; |
2136 | ||
2137 | /* | |
2138 | * Now we are ready to route packet. | |
2139 | */ | |
ebc0ffae ED |
2140 | err = fib_lookup(net, &fl, &res); |
2141 | if (err != 0) { | |
1da177e4 | 2142 | if (!IN_DEV_FORWARD(in_dev)) |
2c2910a4 | 2143 | goto e_hostunreach; |
1da177e4 LT |
2144 | goto no_route; |
2145 | } | |
1da177e4 LT |
2146 | |
2147 | RT_CACHE_STAT_INC(in_slow_tot); | |
2148 | ||
2149 | if (res.type == RTN_BROADCAST) | |
2150 | goto brd_input; | |
2151 | ||
2152 | if (res.type == RTN_LOCAL) { | |
b5f7e755 | 2153 | err = fib_validate_source(saddr, daddr, tos, |
ebc0ffae ED |
2154 | net->loopback_dev->ifindex, |
2155 | dev, &spec_dst, &itag, skb->mark); | |
b5f7e755 ED |
2156 | if (err < 0) |
2157 | goto martian_source_keep_err; | |
2158 | if (err) | |
1da177e4 LT |
2159 | flags |= RTCF_DIRECTSRC; |
2160 | spec_dst = daddr; | |
2161 | goto local_input; | |
2162 | } | |
2163 | ||
2164 | if (!IN_DEV_FORWARD(in_dev)) | |
2c2910a4 | 2165 | goto e_hostunreach; |
1da177e4 LT |
2166 | if (res.type != RTN_UNICAST) |
2167 | goto martian_destination; | |
2168 | ||
2169 | err = ip_mkroute_input(skb, &res, &fl, in_dev, daddr, saddr, tos); | |
1da177e4 LT |
2170 | out: return err; |
2171 | ||
2172 | brd_input: | |
2173 | if (skb->protocol != htons(ETH_P_IP)) | |
2174 | goto e_inval; | |
2175 | ||
f97c1e0c | 2176 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
2177 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_LINK); |
2178 | else { | |
2179 | err = fib_validate_source(saddr, 0, tos, 0, dev, &spec_dst, | |
b0c110ca | 2180 | &itag, skb->mark); |
1da177e4 | 2181 | if (err < 0) |
b5f7e755 | 2182 | goto martian_source_keep_err; |
1da177e4 LT |
2183 | if (err) |
2184 | flags |= RTCF_DIRECTSRC; | |
2185 | } | |
2186 | flags |= RTCF_BROADCAST; | |
2187 | res.type = RTN_BROADCAST; | |
2188 | RT_CACHE_STAT_INC(in_brd); | |
2189 | ||
2190 | local_input: | |
2191 | rth = dst_alloc(&ipv4_dst_ops); | |
2192 | if (!rth) | |
2193 | goto e_nobufs; | |
2194 | ||
d8d1f30b CG |
2195 | rth->dst.output= ip_rt_bug; |
2196 | rth->dst.obsolete = -1; | |
e84f84f2 | 2197 | rth->rt_genid = rt_genid(net); |
1da177e4 | 2198 | |
d8d1f30b CG |
2199 | atomic_set(&rth->dst.__refcnt, 1); |
2200 | rth->dst.flags= DST_HOST; | |
42f811b8 | 2201 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
d8d1f30b | 2202 | rth->dst.flags |= DST_NOPOLICY; |
1da177e4 LT |
2203 | rth->fl.fl4_dst = daddr; |
2204 | rth->rt_dst = daddr; | |
2205 | rth->fl.fl4_tos = tos; | |
47dcf0cb | 2206 | rth->fl.mark = skb->mark; |
1da177e4 LT |
2207 | rth->fl.fl4_src = saddr; |
2208 | rth->rt_src = saddr; | |
2209 | #ifdef CONFIG_NET_CLS_ROUTE | |
d8d1f30b | 2210 | rth->dst.tclassid = itag; |
1da177e4 LT |
2211 | #endif |
2212 | rth->rt_iif = | |
2213 | rth->fl.iif = dev->ifindex; | |
d8d1f30b CG |
2214 | rth->dst.dev = net->loopback_dev; |
2215 | dev_hold(rth->dst.dev); | |
2216 | rth->idev = in_dev_get(rth->dst.dev); | |
1da177e4 LT |
2217 | rth->rt_gateway = daddr; |
2218 | rth->rt_spec_dst= spec_dst; | |
d8d1f30b | 2219 | rth->dst.input= ip_local_deliver; |
1da177e4 LT |
2220 | rth->rt_flags = flags|RTCF_LOCAL; |
2221 | if (res.type == RTN_UNREACHABLE) { | |
d8d1f30b CG |
2222 | rth->dst.input= ip_error; |
2223 | rth->dst.error= -err; | |
1da177e4 LT |
2224 | rth->rt_flags &= ~RTCF_LOCAL; |
2225 | } | |
2226 | rth->rt_type = res.type; | |
e84f84f2 | 2227 | hash = rt_hash(daddr, saddr, fl.iif, rt_genid(net)); |
6a2bad70 | 2228 | err = rt_intern_hash(hash, rth, NULL, skb, fl.iif); |
ebc0ffae | 2229 | goto out; |
1da177e4 LT |
2230 | |
2231 | no_route: | |
2232 | RT_CACHE_STAT_INC(in_no_route); | |
2233 | spec_dst = inet_select_addr(dev, 0, RT_SCOPE_UNIVERSE); | |
2234 | res.type = RTN_UNREACHABLE; | |
7f53878d MC |
2235 | if (err == -ESRCH) |
2236 | err = -ENETUNREACH; | |
1da177e4 LT |
2237 | goto local_input; |
2238 | ||
2239 | /* | |
2240 | * Do not cache martian addresses: they should be logged (RFC1812) | |
2241 | */ | |
2242 | martian_destination: | |
2243 | RT_CACHE_STAT_INC(in_martian_dst); | |
2244 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
2245 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) | |
673d57e7 HH |
2246 | printk(KERN_WARNING "martian destination %pI4 from %pI4, dev %s\n", |
2247 | &daddr, &saddr, dev->name); | |
1da177e4 | 2248 | #endif |
2c2910a4 DE |
2249 | |
2250 | e_hostunreach: | |
e905a9ed | 2251 | err = -EHOSTUNREACH; |
ebc0ffae | 2252 | goto out; |
2c2910a4 | 2253 | |
1da177e4 LT |
2254 | e_inval: |
2255 | err = -EINVAL; | |
ebc0ffae | 2256 | goto out; |
1da177e4 LT |
2257 | |
2258 | e_nobufs: | |
2259 | err = -ENOBUFS; | |
ebc0ffae | 2260 | goto out; |
1da177e4 LT |
2261 | |
2262 | martian_source: | |
b5f7e755 ED |
2263 | err = -EINVAL; |
2264 | martian_source_keep_err: | |
1da177e4 | 2265 | ip_handle_martian_source(dev, in_dev, skb, daddr, saddr); |
ebc0ffae | 2266 | goto out; |
1da177e4 LT |
2267 | } |
2268 | ||
407eadd9 ED |
2269 | int ip_route_input_common(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
2270 | u8 tos, struct net_device *dev, bool noref) | |
1da177e4 LT |
2271 | { |
2272 | struct rtable * rth; | |
2273 | unsigned hash; | |
2274 | int iif = dev->ifindex; | |
b5921910 | 2275 | struct net *net; |
96d36220 | 2276 | int res; |
1da177e4 | 2277 | |
c346dca1 | 2278 | net = dev_net(dev); |
1080d709 | 2279 | |
96d36220 ED |
2280 | rcu_read_lock(); |
2281 | ||
1080d709 NH |
2282 | if (!rt_caching(net)) |
2283 | goto skip_cache; | |
2284 | ||
1da177e4 | 2285 | tos &= IPTOS_RT_MASK; |
e84f84f2 | 2286 | hash = rt_hash(daddr, saddr, iif, rt_genid(net)); |
1da177e4 | 2287 | |
1da177e4 | 2288 | for (rth = rcu_dereference(rt_hash_table[hash].chain); rth; |
d8d1f30b | 2289 | rth = rcu_dereference(rth->dst.rt_next)) { |
0eae88f3 ED |
2290 | if ((((__force u32)rth->fl.fl4_dst ^ (__force u32)daddr) | |
2291 | ((__force u32)rth->fl.fl4_src ^ (__force u32)saddr) | | |
c0b8c32b SH |
2292 | (rth->fl.iif ^ iif) | |
2293 | rth->fl.oif | | |
2294 | (rth->fl.fl4_tos ^ tos)) == 0 && | |
47dcf0cb | 2295 | rth->fl.mark == skb->mark && |
d8d1f30b | 2296 | net_eq(dev_net(rth->dst.dev), net) && |
e84f84f2 | 2297 | !rt_is_expired(rth)) { |
407eadd9 | 2298 | if (noref) { |
d8d1f30b CG |
2299 | dst_use_noref(&rth->dst, jiffies); |
2300 | skb_dst_set_noref(skb, &rth->dst); | |
407eadd9 | 2301 | } else { |
d8d1f30b CG |
2302 | dst_use(&rth->dst, jiffies); |
2303 | skb_dst_set(skb, &rth->dst); | |
407eadd9 | 2304 | } |
1da177e4 LT |
2305 | RT_CACHE_STAT_INC(in_hit); |
2306 | rcu_read_unlock(); | |
1da177e4 LT |
2307 | return 0; |
2308 | } | |
2309 | RT_CACHE_STAT_INC(in_hlist_search); | |
2310 | } | |
1da177e4 | 2311 | |
1080d709 | 2312 | skip_cache: |
1da177e4 LT |
2313 | /* Multicast recognition logic is moved from route cache to here. |
2314 | The problem was that too many Ethernet cards have broken/missing | |
2315 | hardware multicast filters :-( As result the host on multicasting | |
2316 | network acquires a lot of useless route cache entries, sort of | |
2317 | SDR messages from all the world. Now we try to get rid of them. | |
2318 | Really, provided software IP multicast filter is organized | |
2319 | reasonably (at least, hashed), it does not result in a slowdown | |
2320 | comparing with route cache reject entries. | |
2321 | Note, that multicast routers are not affected, because | |
2322 | route cache entry is created eventually. | |
2323 | */ | |
f97c1e0c | 2324 | if (ipv4_is_multicast(daddr)) { |
96d36220 | 2325 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 2326 | |
96d36220 | 2327 | if (in_dev) { |
1da177e4 | 2328 | int our = ip_check_mc(in_dev, daddr, saddr, |
96d36220 | 2329 | ip_hdr(skb)->protocol); |
1da177e4 LT |
2330 | if (our |
2331 | #ifdef CONFIG_IP_MROUTE | |
9d4fb27d JP |
2332 | || |
2333 | (!ipv4_is_local_multicast(daddr) && | |
2334 | IN_DEV_MFORWARD(in_dev)) | |
1da177e4 | 2335 | #endif |
9d4fb27d | 2336 | ) { |
96d36220 ED |
2337 | int res = ip_route_input_mc(skb, daddr, saddr, |
2338 | tos, dev, our); | |
1da177e4 | 2339 | rcu_read_unlock(); |
96d36220 | 2340 | return res; |
1da177e4 LT |
2341 | } |
2342 | } | |
2343 | rcu_read_unlock(); | |
2344 | return -EINVAL; | |
2345 | } | |
96d36220 ED |
2346 | res = ip_route_input_slow(skb, daddr, saddr, tos, dev); |
2347 | rcu_read_unlock(); | |
2348 | return res; | |
1da177e4 | 2349 | } |
407eadd9 | 2350 | EXPORT_SYMBOL(ip_route_input_common); |
1da177e4 | 2351 | |
ebc0ffae | 2352 | /* called with rcu_read_lock() */ |
5969f71d SH |
2353 | static int __mkroute_output(struct rtable **result, |
2354 | struct fib_result *res, | |
2355 | const struct flowi *fl, | |
2356 | const struct flowi *oldflp, | |
2357 | struct net_device *dev_out, | |
2358 | unsigned flags) | |
1da177e4 LT |
2359 | { |
2360 | struct rtable *rth; | |
2361 | struct in_device *in_dev; | |
2362 | u32 tos = RT_FL_TOS(oldflp); | |
1da177e4 | 2363 | |
dd28d1a0 | 2364 | if (ipv4_is_loopback(fl->fl4_src) && !(dev_out->flags & IFF_LOOPBACK)) |
1da177e4 LT |
2365 | return -EINVAL; |
2366 | ||
e448515c | 2367 | if (fl->fl4_dst == htonl(0xFFFFFFFF)) |
1da177e4 | 2368 | res->type = RTN_BROADCAST; |
f97c1e0c | 2369 | else if (ipv4_is_multicast(fl->fl4_dst)) |
1da177e4 | 2370 | res->type = RTN_MULTICAST; |
1e637c74 | 2371 | else if (ipv4_is_lbcast(fl->fl4_dst) || ipv4_is_zeronet(fl->fl4_dst)) |
1da177e4 LT |
2372 | return -EINVAL; |
2373 | ||
2374 | if (dev_out->flags & IFF_LOOPBACK) | |
2375 | flags |= RTCF_LOCAL; | |
2376 | ||
dd28d1a0 | 2377 | in_dev = __in_dev_get_rcu(dev_out); |
ebc0ffae | 2378 | if (!in_dev) |
1da177e4 | 2379 | return -EINVAL; |
ebc0ffae | 2380 | |
1da177e4 LT |
2381 | if (res->type == RTN_BROADCAST) { |
2382 | flags |= RTCF_BROADCAST | RTCF_LOCAL; | |
ebc0ffae | 2383 | res->fi = NULL; |
1da177e4 | 2384 | } else if (res->type == RTN_MULTICAST) { |
dd28d1a0 | 2385 | flags |= RTCF_MULTICAST | RTCF_LOCAL; |
e905a9ed | 2386 | if (!ip_check_mc(in_dev, oldflp->fl4_dst, oldflp->fl4_src, |
1da177e4 LT |
2387 | oldflp->proto)) |
2388 | flags &= ~RTCF_LOCAL; | |
2389 | /* If multicast route do not exist use | |
dd28d1a0 ED |
2390 | * default one, but do not gateway in this case. |
2391 | * Yes, it is hack. | |
1da177e4 | 2392 | */ |
ebc0ffae | 2393 | if (res->fi && res->prefixlen < 4) |
1da177e4 | 2394 | res->fi = NULL; |
1da177e4 LT |
2395 | } |
2396 | ||
2397 | ||
2398 | rth = dst_alloc(&ipv4_dst_ops); | |
8391d07b | 2399 | if (!rth) |
dd28d1a0 | 2400 | return -ENOBUFS; |
8391d07b | 2401 | |
dd28d1a0 | 2402 | in_dev_hold(in_dev); |
dd28d1a0 | 2403 | rth->idev = in_dev; |
1da177e4 | 2404 | |
d8d1f30b CG |
2405 | atomic_set(&rth->dst.__refcnt, 1); |
2406 | rth->dst.flags= DST_HOST; | |
42f811b8 | 2407 | if (IN_DEV_CONF_GET(in_dev, NOXFRM)) |
d8d1f30b | 2408 | rth->dst.flags |= DST_NOXFRM; |
42f811b8 | 2409 | if (IN_DEV_CONF_GET(in_dev, NOPOLICY)) |
d8d1f30b | 2410 | rth->dst.flags |= DST_NOPOLICY; |
1da177e4 LT |
2411 | |
2412 | rth->fl.fl4_dst = oldflp->fl4_dst; | |
2413 | rth->fl.fl4_tos = tos; | |
2414 | rth->fl.fl4_src = oldflp->fl4_src; | |
2415 | rth->fl.oif = oldflp->oif; | |
47dcf0cb | 2416 | rth->fl.mark = oldflp->mark; |
1da177e4 LT |
2417 | rth->rt_dst = fl->fl4_dst; |
2418 | rth->rt_src = fl->fl4_src; | |
2419 | rth->rt_iif = oldflp->oif ? : dev_out->ifindex; | |
e905a9ed | 2420 | /* get references to the devices that are to be hold by the routing |
1da177e4 | 2421 | cache entry */ |
d8d1f30b | 2422 | rth->dst.dev = dev_out; |
1da177e4 | 2423 | dev_hold(dev_out); |
1da177e4 LT |
2424 | rth->rt_gateway = fl->fl4_dst; |
2425 | rth->rt_spec_dst= fl->fl4_src; | |
2426 | ||
d8d1f30b CG |
2427 | rth->dst.output=ip_output; |
2428 | rth->dst.obsolete = -1; | |
e84f84f2 | 2429 | rth->rt_genid = rt_genid(dev_net(dev_out)); |
1da177e4 LT |
2430 | |
2431 | RT_CACHE_STAT_INC(out_slow_tot); | |
2432 | ||
2433 | if (flags & RTCF_LOCAL) { | |
d8d1f30b | 2434 | rth->dst.input = ip_local_deliver; |
1da177e4 LT |
2435 | rth->rt_spec_dst = fl->fl4_dst; |
2436 | } | |
2437 | if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { | |
2438 | rth->rt_spec_dst = fl->fl4_src; | |
e905a9ed | 2439 | if (flags & RTCF_LOCAL && |
1da177e4 | 2440 | !(dev_out->flags & IFF_LOOPBACK)) { |
d8d1f30b | 2441 | rth->dst.output = ip_mc_output; |
1da177e4 LT |
2442 | RT_CACHE_STAT_INC(out_slow_mc); |
2443 | } | |
2444 | #ifdef CONFIG_IP_MROUTE | |
2445 | if (res->type == RTN_MULTICAST) { | |
2446 | if (IN_DEV_MFORWARD(in_dev) && | |
f97c1e0c | 2447 | !ipv4_is_local_multicast(oldflp->fl4_dst)) { |
d8d1f30b CG |
2448 | rth->dst.input = ip_mr_input; |
2449 | rth->dst.output = ip_mc_output; | |
1da177e4 LT |
2450 | } |
2451 | } | |
2452 | #endif | |
2453 | } | |
2454 | ||
2455 | rt_set_nexthop(rth, res, 0); | |
2456 | ||
2457 | rth->rt_flags = flags; | |
1da177e4 | 2458 | *result = rth; |
dd28d1a0 | 2459 | return 0; |
1da177e4 LT |
2460 | } |
2461 | ||
ebc0ffae | 2462 | /* called with rcu_read_lock() */ |
5969f71d SH |
2463 | static int ip_mkroute_output(struct rtable **rp, |
2464 | struct fib_result *res, | |
2465 | const struct flowi *fl, | |
2466 | const struct flowi *oldflp, | |
2467 | struct net_device *dev_out, | |
2468 | unsigned flags) | |
1da177e4 | 2469 | { |
7abaa27c | 2470 | struct rtable *rth = NULL; |
1da177e4 LT |
2471 | int err = __mkroute_output(&rth, res, fl, oldflp, dev_out, flags); |
2472 | unsigned hash; | |
2473 | if (err == 0) { | |
b00180de | 2474 | hash = rt_hash(oldflp->fl4_dst, oldflp->fl4_src, oldflp->oif, |
e84f84f2 | 2475 | rt_genid(dev_net(dev_out))); |
6a2bad70 | 2476 | err = rt_intern_hash(hash, rth, rp, NULL, oldflp->oif); |
1da177e4 | 2477 | } |
e905a9ed | 2478 | |
1da177e4 LT |
2479 | return err; |
2480 | } | |
2481 | ||
1da177e4 LT |
2482 | /* |
2483 | * Major route resolver routine. | |
0197aa38 | 2484 | * called with rcu_read_lock(); |
1da177e4 LT |
2485 | */ |
2486 | ||
b40afd0e DL |
2487 | static int ip_route_output_slow(struct net *net, struct rtable **rp, |
2488 | const struct flowi *oldflp) | |
1da177e4 LT |
2489 | { |
2490 | u32 tos = RT_FL_TOS(oldflp); | |
2491 | struct flowi fl = { .nl_u = { .ip4_u = | |
2492 | { .daddr = oldflp->fl4_dst, | |
2493 | .saddr = oldflp->fl4_src, | |
2494 | .tos = tos & IPTOS_RT_MASK, | |
2495 | .scope = ((tos & RTO_ONLINK) ? | |
2496 | RT_SCOPE_LINK : | |
2497 | RT_SCOPE_UNIVERSE), | |
1da177e4 | 2498 | } }, |
47dcf0cb | 2499 | .mark = oldflp->mark, |
b40afd0e | 2500 | .iif = net->loopback_dev->ifindex, |
1da177e4 LT |
2501 | .oif = oldflp->oif }; |
2502 | struct fib_result res; | |
0197aa38 | 2503 | unsigned int flags = 0; |
1da177e4 | 2504 | struct net_device *dev_out = NULL; |
1da177e4 LT |
2505 | int err; |
2506 | ||
2507 | ||
2508 | res.fi = NULL; | |
2509 | #ifdef CONFIG_IP_MULTIPLE_TABLES | |
2510 | res.r = NULL; | |
2511 | #endif | |
2512 | ||
2513 | if (oldflp->fl4_src) { | |
2514 | err = -EINVAL; | |
f97c1e0c | 2515 | if (ipv4_is_multicast(oldflp->fl4_src) || |
1e637c74 | 2516 | ipv4_is_lbcast(oldflp->fl4_src) || |
f97c1e0c | 2517 | ipv4_is_zeronet(oldflp->fl4_src)) |
1da177e4 LT |
2518 | goto out; |
2519 | ||
1da177e4 LT |
2520 | /* I removed check for oif == dev_out->oif here. |
2521 | It was wrong for two reasons: | |
1ab35276 DL |
2522 | 1. ip_dev_find(net, saddr) can return wrong iface, if saddr |
2523 | is assigned to multiple interfaces. | |
1da177e4 LT |
2524 | 2. Moreover, we are allowed to send packets with saddr |
2525 | of another iface. --ANK | |
2526 | */ | |
2527 | ||
9d4fb27d JP |
2528 | if (oldflp->oif == 0 && |
2529 | (ipv4_is_multicast(oldflp->fl4_dst) || | |
2530 | oldflp->fl4_dst == htonl(0xFFFFFFFF))) { | |
a210d01a | 2531 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
0197aa38 | 2532 | dev_out = __ip_dev_find(net, oldflp->fl4_src, false); |
a210d01a JA |
2533 | if (dev_out == NULL) |
2534 | goto out; | |
2535 | ||
1da177e4 LT |
2536 | /* Special hack: user can direct multicasts |
2537 | and limited broadcast via necessary interface | |
2538 | without fiddling with IP_MULTICAST_IF or IP_PKTINFO. | |
2539 | This hack is not just for fun, it allows | |
2540 | vic,vat and friends to work. | |
2541 | They bind socket to loopback, set ttl to zero | |
2542 | and expect that it will work. | |
2543 | From the viewpoint of routing cache they are broken, | |
2544 | because we are not allowed to build multicast path | |
2545 | with loopback source addr (look, routing cache | |
2546 | cannot know, that ttl is zero, so that packet | |
2547 | will not leave this host and route is valid). | |
2548 | Luckily, this hack is good workaround. | |
2549 | */ | |
2550 | ||
2551 | fl.oif = dev_out->ifindex; | |
2552 | goto make_route; | |
2553 | } | |
a210d01a JA |
2554 | |
2555 | if (!(oldflp->flags & FLOWI_FLAG_ANYSRC)) { | |
2556 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ | |
0197aa38 | 2557 | if (!__ip_dev_find(net, oldflp->fl4_src, false)) |
a210d01a | 2558 | goto out; |
a210d01a | 2559 | } |
1da177e4 LT |
2560 | } |
2561 | ||
2562 | ||
2563 | if (oldflp->oif) { | |
0197aa38 | 2564 | dev_out = dev_get_by_index_rcu(net, oldflp->oif); |
1da177e4 LT |
2565 | err = -ENODEV; |
2566 | if (dev_out == NULL) | |
2567 | goto out; | |
e5ed6399 HX |
2568 | |
2569 | /* RACE: Check return value of inet_select_addr instead. */ | |
0197aa38 | 2570 | if (rcu_dereference(dev_out->ip_ptr) == NULL) |
1da177e4 | 2571 | goto out; /* Wrong error code */ |
1da177e4 | 2572 | |
f97c1e0c JP |
2573 | if (ipv4_is_local_multicast(oldflp->fl4_dst) || |
2574 | oldflp->fl4_dst == htonl(0xFFFFFFFF)) { | |
1da177e4 LT |
2575 | if (!fl.fl4_src) |
2576 | fl.fl4_src = inet_select_addr(dev_out, 0, | |
2577 | RT_SCOPE_LINK); | |
2578 | goto make_route; | |
2579 | } | |
2580 | if (!fl.fl4_src) { | |
f97c1e0c | 2581 | if (ipv4_is_multicast(oldflp->fl4_dst)) |
1da177e4 LT |
2582 | fl.fl4_src = inet_select_addr(dev_out, 0, |
2583 | fl.fl4_scope); | |
2584 | else if (!oldflp->fl4_dst) | |
2585 | fl.fl4_src = inet_select_addr(dev_out, 0, | |
2586 | RT_SCOPE_HOST); | |
2587 | } | |
2588 | } | |
2589 | ||
2590 | if (!fl.fl4_dst) { | |
2591 | fl.fl4_dst = fl.fl4_src; | |
2592 | if (!fl.fl4_dst) | |
2593 | fl.fl4_dst = fl.fl4_src = htonl(INADDR_LOOPBACK); | |
b40afd0e | 2594 | dev_out = net->loopback_dev; |
b40afd0e | 2595 | fl.oif = net->loopback_dev->ifindex; |
1da177e4 LT |
2596 | res.type = RTN_LOCAL; |
2597 | flags |= RTCF_LOCAL; | |
2598 | goto make_route; | |
2599 | } | |
2600 | ||
b40afd0e | 2601 | if (fib_lookup(net, &fl, &res)) { |
1da177e4 LT |
2602 | res.fi = NULL; |
2603 | if (oldflp->oif) { | |
2604 | /* Apparently, routing tables are wrong. Assume, | |
2605 | that the destination is on link. | |
2606 | ||
2607 | WHY? DW. | |
2608 | Because we are allowed to send to iface | |
2609 | even if it has NO routes and NO assigned | |
2610 | addresses. When oif is specified, routing | |
2611 | tables are looked up with only one purpose: | |
2612 | to catch if destination is gatewayed, rather than | |
2613 | direct. Moreover, if MSG_DONTROUTE is set, | |
2614 | we send packet, ignoring both routing tables | |
2615 | and ifaddr state. --ANK | |
2616 | ||
2617 | ||
2618 | We could make it even if oif is unknown, | |
2619 | likely IPv6, but we do not. | |
2620 | */ | |
2621 | ||
2622 | if (fl.fl4_src == 0) | |
2623 | fl.fl4_src = inet_select_addr(dev_out, 0, | |
2624 | RT_SCOPE_LINK); | |
2625 | res.type = RTN_UNICAST; | |
2626 | goto make_route; | |
2627 | } | |
1da177e4 LT |
2628 | err = -ENETUNREACH; |
2629 | goto out; | |
2630 | } | |
1da177e4 LT |
2631 | |
2632 | if (res.type == RTN_LOCAL) { | |
2633 | if (!fl.fl4_src) | |
2634 | fl.fl4_src = fl.fl4_dst; | |
b40afd0e | 2635 | dev_out = net->loopback_dev; |
1da177e4 | 2636 | fl.oif = dev_out->ifindex; |
1da177e4 LT |
2637 | res.fi = NULL; |
2638 | flags |= RTCF_LOCAL; | |
2639 | goto make_route; | |
2640 | } | |
2641 | ||
2642 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
2643 | if (res.fi->fib_nhs > 1 && fl.oif == 0) | |
2644 | fib_select_multipath(&fl, &res); | |
2645 | else | |
2646 | #endif | |
2647 | if (!res.prefixlen && res.type == RTN_UNICAST && !fl.oif) | |
b40afd0e | 2648 | fib_select_default(net, &fl, &res); |
1da177e4 LT |
2649 | |
2650 | if (!fl.fl4_src) | |
2651 | fl.fl4_src = FIB_RES_PREFSRC(res); | |
2652 | ||
1da177e4 | 2653 | dev_out = FIB_RES_DEV(res); |
1da177e4 LT |
2654 | fl.oif = dev_out->ifindex; |
2655 | ||
2656 | ||
2657 | make_route: | |
2658 | err = ip_mkroute_output(rp, &res, &fl, oldflp, dev_out, flags); | |
2659 | ||
1da177e4 LT |
2660 | out: return err; |
2661 | } | |
2662 | ||
611c183e DL |
2663 | int __ip_route_output_key(struct net *net, struct rtable **rp, |
2664 | const struct flowi *flp) | |
1da177e4 | 2665 | { |
0197aa38 ED |
2666 | unsigned int hash; |
2667 | int res; | |
1da177e4 LT |
2668 | struct rtable *rth; |
2669 | ||
1080d709 NH |
2670 | if (!rt_caching(net)) |
2671 | goto slow_output; | |
2672 | ||
e84f84f2 | 2673 | hash = rt_hash(flp->fl4_dst, flp->fl4_src, flp->oif, rt_genid(net)); |
1da177e4 LT |
2674 | |
2675 | rcu_read_lock_bh(); | |
a898def2 | 2676 | for (rth = rcu_dereference_bh(rt_hash_table[hash].chain); rth; |
d8d1f30b | 2677 | rth = rcu_dereference_bh(rth->dst.rt_next)) { |
1da177e4 LT |
2678 | if (rth->fl.fl4_dst == flp->fl4_dst && |
2679 | rth->fl.fl4_src == flp->fl4_src && | |
2680 | rth->fl.iif == 0 && | |
2681 | rth->fl.oif == flp->oif && | |
47dcf0cb | 2682 | rth->fl.mark == flp->mark && |
1da177e4 | 2683 | !((rth->fl.fl4_tos ^ flp->fl4_tos) & |
b5921910 | 2684 | (IPTOS_RT_MASK | RTO_ONLINK)) && |
d8d1f30b | 2685 | net_eq(dev_net(rth->dst.dev), net) && |
e84f84f2 | 2686 | !rt_is_expired(rth)) { |
d8d1f30b | 2687 | dst_use(&rth->dst, jiffies); |
1da177e4 LT |
2688 | RT_CACHE_STAT_INC(out_hit); |
2689 | rcu_read_unlock_bh(); | |
2690 | *rp = rth; | |
2691 | return 0; | |
2692 | } | |
2693 | RT_CACHE_STAT_INC(out_hlist_search); | |
2694 | } | |
2695 | rcu_read_unlock_bh(); | |
2696 | ||
1080d709 | 2697 | slow_output: |
0197aa38 ED |
2698 | rcu_read_lock(); |
2699 | res = ip_route_output_slow(net, rp, flp); | |
2700 | rcu_read_unlock(); | |
2701 | return res; | |
1da177e4 | 2702 | } |
d8c97a94 ACM |
2703 | EXPORT_SYMBOL_GPL(__ip_route_output_key); |
2704 | ||
ae2688d5 JW |
2705 | static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie) |
2706 | { | |
2707 | return NULL; | |
2708 | } | |
2709 | ||
14e50e57 DM |
2710 | static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu) |
2711 | { | |
2712 | } | |
2713 | ||
2714 | static struct dst_ops ipv4_dst_blackhole_ops = { | |
2715 | .family = AF_INET, | |
09640e63 | 2716 | .protocol = cpu_to_be16(ETH_P_IP), |
14e50e57 | 2717 | .destroy = ipv4_dst_destroy, |
ae2688d5 | 2718 | .check = ipv4_blackhole_dst_check, |
14e50e57 | 2719 | .update_pmtu = ipv4_rt_blackhole_update_pmtu, |
e2422970 | 2720 | .entries = ATOMIC_INIT(0), |
14e50e57 DM |
2721 | }; |
2722 | ||
2723 | ||
e84f84f2 | 2724 | static int ipv4_dst_blackhole(struct net *net, struct rtable **rp, struct flowi *flp) |
14e50e57 DM |
2725 | { |
2726 | struct rtable *ort = *rp; | |
2727 | struct rtable *rt = (struct rtable *) | |
2728 | dst_alloc(&ipv4_dst_blackhole_ops); | |
2729 | ||
2730 | if (rt) { | |
d8d1f30b | 2731 | struct dst_entry *new = &rt->dst; |
14e50e57 DM |
2732 | |
2733 | atomic_set(&new->__refcnt, 1); | |
2734 | new->__use = 1; | |
352e512c HX |
2735 | new->input = dst_discard; |
2736 | new->output = dst_discard; | |
d8d1f30b | 2737 | memcpy(new->metrics, ort->dst.metrics, RTAX_MAX*sizeof(u32)); |
14e50e57 | 2738 | |
d8d1f30b | 2739 | new->dev = ort->dst.dev; |
14e50e57 DM |
2740 | if (new->dev) |
2741 | dev_hold(new->dev); | |
2742 | ||
2743 | rt->fl = ort->fl; | |
2744 | ||
2745 | rt->idev = ort->idev; | |
2746 | if (rt->idev) | |
2747 | in_dev_hold(rt->idev); | |
e84f84f2 | 2748 | rt->rt_genid = rt_genid(net); |
14e50e57 DM |
2749 | rt->rt_flags = ort->rt_flags; |
2750 | rt->rt_type = ort->rt_type; | |
2751 | rt->rt_dst = ort->rt_dst; | |
2752 | rt->rt_src = ort->rt_src; | |
2753 | rt->rt_iif = ort->rt_iif; | |
2754 | rt->rt_gateway = ort->rt_gateway; | |
2755 | rt->rt_spec_dst = ort->rt_spec_dst; | |
2756 | rt->peer = ort->peer; | |
2757 | if (rt->peer) | |
2758 | atomic_inc(&rt->peer->refcnt); | |
2759 | ||
2760 | dst_free(new); | |
2761 | } | |
2762 | ||
d8d1f30b | 2763 | dst_release(&(*rp)->dst); |
14e50e57 | 2764 | *rp = rt; |
a02cec21 | 2765 | return rt ? 0 : -ENOMEM; |
14e50e57 DM |
2766 | } |
2767 | ||
f1b050bf DL |
2768 | int ip_route_output_flow(struct net *net, struct rtable **rp, struct flowi *flp, |
2769 | struct sock *sk, int flags) | |
1da177e4 LT |
2770 | { |
2771 | int err; | |
2772 | ||
f1b050bf | 2773 | if ((err = __ip_route_output_key(net, rp, flp)) != 0) |
1da177e4 LT |
2774 | return err; |
2775 | ||
2776 | if (flp->proto) { | |
2777 | if (!flp->fl4_src) | |
2778 | flp->fl4_src = (*rp)->rt_src; | |
2779 | if (!flp->fl4_dst) | |
2780 | flp->fl4_dst = (*rp)->rt_dst; | |
52479b62 | 2781 | err = __xfrm_lookup(net, (struct dst_entry **)rp, flp, sk, |
bb72845e | 2782 | flags ? XFRM_LOOKUP_WAIT : 0); |
14e50e57 | 2783 | if (err == -EREMOTE) |
e84f84f2 | 2784 | err = ipv4_dst_blackhole(net, rp, flp); |
14e50e57 DM |
2785 | |
2786 | return err; | |
1da177e4 LT |
2787 | } |
2788 | ||
2789 | return 0; | |
2790 | } | |
d8c97a94 ACM |
2791 | EXPORT_SYMBOL_GPL(ip_route_output_flow); |
2792 | ||
f206351a | 2793 | int ip_route_output_key(struct net *net, struct rtable **rp, struct flowi *flp) |
1da177e4 | 2794 | { |
f206351a | 2795 | return ip_route_output_flow(net, rp, flp, NULL, 0); |
1da177e4 | 2796 | } |
4bc2f18b | 2797 | EXPORT_SYMBOL(ip_route_output_key); |
1da177e4 | 2798 | |
4feb88e5 BT |
2799 | static int rt_fill_info(struct net *net, |
2800 | struct sk_buff *skb, u32 pid, u32 seq, int event, | |
b6544c0b | 2801 | int nowait, unsigned int flags) |
1da177e4 | 2802 | { |
511c3f92 | 2803 | struct rtable *rt = skb_rtable(skb); |
1da177e4 | 2804 | struct rtmsg *r; |
be403ea1 | 2805 | struct nlmsghdr *nlh; |
e3703b3d TG |
2806 | long expires; |
2807 | u32 id = 0, ts = 0, tsage = 0, error; | |
be403ea1 TG |
2808 | |
2809 | nlh = nlmsg_put(skb, pid, seq, event, sizeof(*r), flags); | |
2810 | if (nlh == NULL) | |
26932566 | 2811 | return -EMSGSIZE; |
be403ea1 TG |
2812 | |
2813 | r = nlmsg_data(nlh); | |
1da177e4 LT |
2814 | r->rtm_family = AF_INET; |
2815 | r->rtm_dst_len = 32; | |
2816 | r->rtm_src_len = 0; | |
2817 | r->rtm_tos = rt->fl.fl4_tos; | |
2818 | r->rtm_table = RT_TABLE_MAIN; | |
be403ea1 | 2819 | NLA_PUT_U32(skb, RTA_TABLE, RT_TABLE_MAIN); |
1da177e4 LT |
2820 | r->rtm_type = rt->rt_type; |
2821 | r->rtm_scope = RT_SCOPE_UNIVERSE; | |
2822 | r->rtm_protocol = RTPROT_UNSPEC; | |
2823 | r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED; | |
2824 | if (rt->rt_flags & RTCF_NOTIFY) | |
2825 | r->rtm_flags |= RTM_F_NOTIFY; | |
be403ea1 | 2826 | |
17fb2c64 | 2827 | NLA_PUT_BE32(skb, RTA_DST, rt->rt_dst); |
be403ea1 | 2828 | |
1da177e4 LT |
2829 | if (rt->fl.fl4_src) { |
2830 | r->rtm_src_len = 32; | |
17fb2c64 | 2831 | NLA_PUT_BE32(skb, RTA_SRC, rt->fl.fl4_src); |
1da177e4 | 2832 | } |
d8d1f30b CG |
2833 | if (rt->dst.dev) |
2834 | NLA_PUT_U32(skb, RTA_OIF, rt->dst.dev->ifindex); | |
1da177e4 | 2835 | #ifdef CONFIG_NET_CLS_ROUTE |
d8d1f30b CG |
2836 | if (rt->dst.tclassid) |
2837 | NLA_PUT_U32(skb, RTA_FLOW, rt->dst.tclassid); | |
1da177e4 LT |
2838 | #endif |
2839 | if (rt->fl.iif) | |
17fb2c64 | 2840 | NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_spec_dst); |
1da177e4 | 2841 | else if (rt->rt_src != rt->fl.fl4_src) |
17fb2c64 | 2842 | NLA_PUT_BE32(skb, RTA_PREFSRC, rt->rt_src); |
be403ea1 | 2843 | |
1da177e4 | 2844 | if (rt->rt_dst != rt->rt_gateway) |
17fb2c64 | 2845 | NLA_PUT_BE32(skb, RTA_GATEWAY, rt->rt_gateway); |
be403ea1 | 2846 | |
d8d1f30b | 2847 | if (rtnetlink_put_metrics(skb, rt->dst.metrics) < 0) |
be403ea1 TG |
2848 | goto nla_put_failure; |
2849 | ||
963bfeee ED |
2850 | if (rt->fl.mark) |
2851 | NLA_PUT_BE32(skb, RTA_MARK, rt->fl.mark); | |
2852 | ||
d8d1f30b CG |
2853 | error = rt->dst.error; |
2854 | expires = rt->dst.expires ? rt->dst.expires - jiffies : 0; | |
1da177e4 | 2855 | if (rt->peer) { |
317fe0e6 | 2856 | inet_peer_refcheck(rt->peer); |
2c1409a0 | 2857 | id = atomic_read(&rt->peer->ip_id_count) & 0xffff; |
1da177e4 | 2858 | if (rt->peer->tcp_ts_stamp) { |
e3703b3d | 2859 | ts = rt->peer->tcp_ts; |
9d729f72 | 2860 | tsage = get_seconds() - rt->peer->tcp_ts_stamp; |
1da177e4 LT |
2861 | } |
2862 | } | |
be403ea1 | 2863 | |
1da177e4 LT |
2864 | if (rt->fl.iif) { |
2865 | #ifdef CONFIG_IP_MROUTE | |
e448515c | 2866 | __be32 dst = rt->rt_dst; |
1da177e4 | 2867 | |
f97c1e0c | 2868 | if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) && |
4feb88e5 BT |
2869 | IPV4_DEVCONF_ALL(net, MC_FORWARDING)) { |
2870 | int err = ipmr_get_route(net, skb, r, nowait); | |
1da177e4 LT |
2871 | if (err <= 0) { |
2872 | if (!nowait) { | |
2873 | if (err == 0) | |
2874 | return 0; | |
be403ea1 | 2875 | goto nla_put_failure; |
1da177e4 LT |
2876 | } else { |
2877 | if (err == -EMSGSIZE) | |
be403ea1 | 2878 | goto nla_put_failure; |
e3703b3d | 2879 | error = err; |
1da177e4 LT |
2880 | } |
2881 | } | |
2882 | } else | |
2883 | #endif | |
be403ea1 | 2884 | NLA_PUT_U32(skb, RTA_IIF, rt->fl.iif); |
1da177e4 LT |
2885 | } |
2886 | ||
d8d1f30b | 2887 | if (rtnl_put_cacheinfo(skb, &rt->dst, id, ts, tsage, |
e3703b3d TG |
2888 | expires, error) < 0) |
2889 | goto nla_put_failure; | |
be403ea1 TG |
2890 | |
2891 | return nlmsg_end(skb, nlh); | |
1da177e4 | 2892 | |
be403ea1 | 2893 | nla_put_failure: |
26932566 PM |
2894 | nlmsg_cancel(skb, nlh); |
2895 | return -EMSGSIZE; | |
1da177e4 LT |
2896 | } |
2897 | ||
63f3444f | 2898 | static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg) |
1da177e4 | 2899 | { |
3b1e0a65 | 2900 | struct net *net = sock_net(in_skb->sk); |
d889ce3b TG |
2901 | struct rtmsg *rtm; |
2902 | struct nlattr *tb[RTA_MAX+1]; | |
1da177e4 | 2903 | struct rtable *rt = NULL; |
9e12bb22 AV |
2904 | __be32 dst = 0; |
2905 | __be32 src = 0; | |
2906 | u32 iif; | |
d889ce3b | 2907 | int err; |
963bfeee | 2908 | int mark; |
1da177e4 LT |
2909 | struct sk_buff *skb; |
2910 | ||
d889ce3b TG |
2911 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy); |
2912 | if (err < 0) | |
2913 | goto errout; | |
2914 | ||
2915 | rtm = nlmsg_data(nlh); | |
2916 | ||
1da177e4 | 2917 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
d889ce3b TG |
2918 | if (skb == NULL) { |
2919 | err = -ENOBUFS; | |
2920 | goto errout; | |
2921 | } | |
1da177e4 LT |
2922 | |
2923 | /* Reserve room for dummy headers, this skb can pass | |
2924 | through good chunk of routing engine. | |
2925 | */ | |
459a98ed | 2926 | skb_reset_mac_header(skb); |
c1d2bbe1 | 2927 | skb_reset_network_header(skb); |
d2c962b8 SH |
2928 | |
2929 | /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */ | |
eddc9ec5 | 2930 | ip_hdr(skb)->protocol = IPPROTO_ICMP; |
1da177e4 LT |
2931 | skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr)); |
2932 | ||
17fb2c64 AV |
2933 | src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0; |
2934 | dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0; | |
d889ce3b | 2935 | iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0; |
963bfeee | 2936 | mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0; |
1da177e4 LT |
2937 | |
2938 | if (iif) { | |
d889ce3b TG |
2939 | struct net_device *dev; |
2940 | ||
1937504d | 2941 | dev = __dev_get_by_index(net, iif); |
d889ce3b TG |
2942 | if (dev == NULL) { |
2943 | err = -ENODEV; | |
2944 | goto errout_free; | |
2945 | } | |
2946 | ||
1da177e4 LT |
2947 | skb->protocol = htons(ETH_P_IP); |
2948 | skb->dev = dev; | |
963bfeee | 2949 | skb->mark = mark; |
1da177e4 LT |
2950 | local_bh_disable(); |
2951 | err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev); | |
2952 | local_bh_enable(); | |
d889ce3b | 2953 | |
511c3f92 | 2954 | rt = skb_rtable(skb); |
d8d1f30b CG |
2955 | if (err == 0 && rt->dst.error) |
2956 | err = -rt->dst.error; | |
1da177e4 | 2957 | } else { |
d889ce3b TG |
2958 | struct flowi fl = { |
2959 | .nl_u = { | |
2960 | .ip4_u = { | |
2961 | .daddr = dst, | |
2962 | .saddr = src, | |
2963 | .tos = rtm->rtm_tos, | |
2964 | }, | |
2965 | }, | |
2966 | .oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0, | |
963bfeee | 2967 | .mark = mark, |
d889ce3b | 2968 | }; |
1937504d | 2969 | err = ip_route_output_key(net, &rt, &fl); |
1da177e4 | 2970 | } |
d889ce3b | 2971 | |
1da177e4 | 2972 | if (err) |
d889ce3b | 2973 | goto errout_free; |
1da177e4 | 2974 | |
d8d1f30b | 2975 | skb_dst_set(skb, &rt->dst); |
1da177e4 LT |
2976 | if (rtm->rtm_flags & RTM_F_NOTIFY) |
2977 | rt->rt_flags |= RTCF_NOTIFY; | |
2978 | ||
4feb88e5 | 2979 | err = rt_fill_info(net, skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, |
1937504d | 2980 | RTM_NEWROUTE, 0, 0); |
d889ce3b TG |
2981 | if (err <= 0) |
2982 | goto errout_free; | |
1da177e4 | 2983 | |
1937504d | 2984 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid); |
d889ce3b | 2985 | errout: |
2942e900 | 2986 | return err; |
1da177e4 | 2987 | |
d889ce3b | 2988 | errout_free: |
1da177e4 | 2989 | kfree_skb(skb); |
d889ce3b | 2990 | goto errout; |
1da177e4 LT |
2991 | } |
2992 | ||
2993 | int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb) | |
2994 | { | |
2995 | struct rtable *rt; | |
2996 | int h, s_h; | |
2997 | int idx, s_idx; | |
1937504d DL |
2998 | struct net *net; |
2999 | ||
3b1e0a65 | 3000 | net = sock_net(skb->sk); |
1da177e4 LT |
3001 | |
3002 | s_h = cb->args[0]; | |
d8c92830 ED |
3003 | if (s_h < 0) |
3004 | s_h = 0; | |
1da177e4 | 3005 | s_idx = idx = cb->args[1]; |
a6272665 ED |
3006 | for (h = s_h; h <= rt_hash_mask; h++, s_idx = 0) { |
3007 | if (!rt_hash_table[h].chain) | |
3008 | continue; | |
1da177e4 | 3009 | rcu_read_lock_bh(); |
a898def2 | 3010 | for (rt = rcu_dereference_bh(rt_hash_table[h].chain), idx = 0; rt; |
d8d1f30b CG |
3011 | rt = rcu_dereference_bh(rt->dst.rt_next), idx++) { |
3012 | if (!net_eq(dev_net(rt->dst.dev), net) || idx < s_idx) | |
1da177e4 | 3013 | continue; |
e84f84f2 | 3014 | if (rt_is_expired(rt)) |
29e75252 | 3015 | continue; |
d8d1f30b | 3016 | skb_dst_set_noref(skb, &rt->dst); |
4feb88e5 | 3017 | if (rt_fill_info(net, skb, NETLINK_CB(cb->skb).pid, |
e905a9ed | 3018 | cb->nlh->nlmsg_seq, RTM_NEWROUTE, |
b6544c0b | 3019 | 1, NLM_F_MULTI) <= 0) { |
adf30907 | 3020 | skb_dst_drop(skb); |
1da177e4 LT |
3021 | rcu_read_unlock_bh(); |
3022 | goto done; | |
3023 | } | |
adf30907 | 3024 | skb_dst_drop(skb); |
1da177e4 LT |
3025 | } |
3026 | rcu_read_unlock_bh(); | |
3027 | } | |
3028 | ||
3029 | done: | |
3030 | cb->args[0] = h; | |
3031 | cb->args[1] = idx; | |
3032 | return skb->len; | |
3033 | } | |
3034 | ||
3035 | void ip_rt_multicast_event(struct in_device *in_dev) | |
3036 | { | |
76e6ebfb | 3037 | rt_cache_flush(dev_net(in_dev->dev), 0); |
1da177e4 LT |
3038 | } |
3039 | ||
3040 | #ifdef CONFIG_SYSCTL | |
81c684d1 | 3041 | static int ipv4_sysctl_rtcache_flush(ctl_table *__ctl, int write, |
8d65af78 | 3042 | void __user *buffer, |
1da177e4 LT |
3043 | size_t *lenp, loff_t *ppos) |
3044 | { | |
3045 | if (write) { | |
639e104f | 3046 | int flush_delay; |
81c684d1 | 3047 | ctl_table ctl; |
39a23e75 | 3048 | struct net *net; |
639e104f | 3049 | |
81c684d1 DL |
3050 | memcpy(&ctl, __ctl, sizeof(ctl)); |
3051 | ctl.data = &flush_delay; | |
8d65af78 | 3052 | proc_dointvec(&ctl, write, buffer, lenp, ppos); |
639e104f | 3053 | |
81c684d1 | 3054 | net = (struct net *)__ctl->extra1; |
39a23e75 | 3055 | rt_cache_flush(net, flush_delay); |
1da177e4 | 3056 | return 0; |
e905a9ed | 3057 | } |
1da177e4 LT |
3058 | |
3059 | return -EINVAL; | |
3060 | } | |
3061 | ||
eeb61f71 | 3062 | static ctl_table ipv4_route_table[] = { |
1da177e4 | 3063 | { |
1da177e4 LT |
3064 | .procname = "gc_thresh", |
3065 | .data = &ipv4_dst_ops.gc_thresh, | |
3066 | .maxlen = sizeof(int), | |
3067 | .mode = 0644, | |
6d9f239a | 3068 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3069 | }, |
3070 | { | |
1da177e4 LT |
3071 | .procname = "max_size", |
3072 | .data = &ip_rt_max_size, | |
3073 | .maxlen = sizeof(int), | |
3074 | .mode = 0644, | |
6d9f239a | 3075 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3076 | }, |
3077 | { | |
3078 | /* Deprecated. Use gc_min_interval_ms */ | |
e905a9ed | 3079 | |
1da177e4 LT |
3080 | .procname = "gc_min_interval", |
3081 | .data = &ip_rt_gc_min_interval, | |
3082 | .maxlen = sizeof(int), | |
3083 | .mode = 0644, | |
6d9f239a | 3084 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3085 | }, |
3086 | { | |
1da177e4 LT |
3087 | .procname = "gc_min_interval_ms", |
3088 | .data = &ip_rt_gc_min_interval, | |
3089 | .maxlen = sizeof(int), | |
3090 | .mode = 0644, | |
6d9f239a | 3091 | .proc_handler = proc_dointvec_ms_jiffies, |
1da177e4 LT |
3092 | }, |
3093 | { | |
1da177e4 LT |
3094 | .procname = "gc_timeout", |
3095 | .data = &ip_rt_gc_timeout, | |
3096 | .maxlen = sizeof(int), | |
3097 | .mode = 0644, | |
6d9f239a | 3098 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3099 | }, |
3100 | { | |
1da177e4 LT |
3101 | .procname = "gc_interval", |
3102 | .data = &ip_rt_gc_interval, | |
3103 | .maxlen = sizeof(int), | |
3104 | .mode = 0644, | |
6d9f239a | 3105 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3106 | }, |
3107 | { | |
1da177e4 LT |
3108 | .procname = "redirect_load", |
3109 | .data = &ip_rt_redirect_load, | |
3110 | .maxlen = sizeof(int), | |
3111 | .mode = 0644, | |
6d9f239a | 3112 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3113 | }, |
3114 | { | |
1da177e4 LT |
3115 | .procname = "redirect_number", |
3116 | .data = &ip_rt_redirect_number, | |
3117 | .maxlen = sizeof(int), | |
3118 | .mode = 0644, | |
6d9f239a | 3119 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3120 | }, |
3121 | { | |
1da177e4 LT |
3122 | .procname = "redirect_silence", |
3123 | .data = &ip_rt_redirect_silence, | |
3124 | .maxlen = sizeof(int), | |
3125 | .mode = 0644, | |
6d9f239a | 3126 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3127 | }, |
3128 | { | |
1da177e4 LT |
3129 | .procname = "error_cost", |
3130 | .data = &ip_rt_error_cost, | |
3131 | .maxlen = sizeof(int), | |
3132 | .mode = 0644, | |
6d9f239a | 3133 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3134 | }, |
3135 | { | |
1da177e4 LT |
3136 | .procname = "error_burst", |
3137 | .data = &ip_rt_error_burst, | |
3138 | .maxlen = sizeof(int), | |
3139 | .mode = 0644, | |
6d9f239a | 3140 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3141 | }, |
3142 | { | |
1da177e4 LT |
3143 | .procname = "gc_elasticity", |
3144 | .data = &ip_rt_gc_elasticity, | |
3145 | .maxlen = sizeof(int), | |
3146 | .mode = 0644, | |
6d9f239a | 3147 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3148 | }, |
3149 | { | |
1da177e4 LT |
3150 | .procname = "mtu_expires", |
3151 | .data = &ip_rt_mtu_expires, | |
3152 | .maxlen = sizeof(int), | |
3153 | .mode = 0644, | |
6d9f239a | 3154 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
3155 | }, |
3156 | { | |
1da177e4 LT |
3157 | .procname = "min_pmtu", |
3158 | .data = &ip_rt_min_pmtu, | |
3159 | .maxlen = sizeof(int), | |
3160 | .mode = 0644, | |
6d9f239a | 3161 | .proc_handler = proc_dointvec, |
1da177e4 LT |
3162 | }, |
3163 | { | |
1da177e4 LT |
3164 | .procname = "min_adv_mss", |
3165 | .data = &ip_rt_min_advmss, | |
3166 | .maxlen = sizeof(int), | |
3167 | .mode = 0644, | |
6d9f239a | 3168 | .proc_handler = proc_dointvec, |
1da177e4 | 3169 | }, |
f8572d8f | 3170 | { } |
1da177e4 | 3171 | }; |
39a23e75 | 3172 | |
2f4520d3 AV |
3173 | static struct ctl_table empty[1]; |
3174 | ||
3175 | static struct ctl_table ipv4_skeleton[] = | |
3176 | { | |
f8572d8f | 3177 | { .procname = "route", |
d994af0d | 3178 | .mode = 0555, .child = ipv4_route_table}, |
f8572d8f | 3179 | { .procname = "neigh", |
d994af0d | 3180 | .mode = 0555, .child = empty}, |
2f4520d3 AV |
3181 | { } |
3182 | }; | |
3183 | ||
3184 | static __net_initdata struct ctl_path ipv4_path[] = { | |
f8572d8f EB |
3185 | { .procname = "net", }, |
3186 | { .procname = "ipv4", }, | |
39a23e75 DL |
3187 | { }, |
3188 | }; | |
3189 | ||
39a23e75 DL |
3190 | static struct ctl_table ipv4_route_flush_table[] = { |
3191 | { | |
39a23e75 DL |
3192 | .procname = "flush", |
3193 | .maxlen = sizeof(int), | |
3194 | .mode = 0200, | |
6d9f239a | 3195 | .proc_handler = ipv4_sysctl_rtcache_flush, |
39a23e75 | 3196 | }, |
f8572d8f | 3197 | { }, |
39a23e75 DL |
3198 | }; |
3199 | ||
2f4520d3 | 3200 | static __net_initdata struct ctl_path ipv4_route_path[] = { |
f8572d8f EB |
3201 | { .procname = "net", }, |
3202 | { .procname = "ipv4", }, | |
3203 | { .procname = "route", }, | |
2f4520d3 AV |
3204 | { }, |
3205 | }; | |
3206 | ||
39a23e75 DL |
3207 | static __net_init int sysctl_route_net_init(struct net *net) |
3208 | { | |
3209 | struct ctl_table *tbl; | |
3210 | ||
3211 | tbl = ipv4_route_flush_table; | |
09ad9bc7 | 3212 | if (!net_eq(net, &init_net)) { |
39a23e75 DL |
3213 | tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL); |
3214 | if (tbl == NULL) | |
3215 | goto err_dup; | |
3216 | } | |
3217 | tbl[0].extra1 = net; | |
3218 | ||
3219 | net->ipv4.route_hdr = | |
3220 | register_net_sysctl_table(net, ipv4_route_path, tbl); | |
3221 | if (net->ipv4.route_hdr == NULL) | |
3222 | goto err_reg; | |
3223 | return 0; | |
3224 | ||
3225 | err_reg: | |
3226 | if (tbl != ipv4_route_flush_table) | |
3227 | kfree(tbl); | |
3228 | err_dup: | |
3229 | return -ENOMEM; | |
3230 | } | |
3231 | ||
3232 | static __net_exit void sysctl_route_net_exit(struct net *net) | |
3233 | { | |
3234 | struct ctl_table *tbl; | |
3235 | ||
3236 | tbl = net->ipv4.route_hdr->ctl_table_arg; | |
3237 | unregister_net_sysctl_table(net->ipv4.route_hdr); | |
3238 | BUG_ON(tbl == ipv4_route_flush_table); | |
3239 | kfree(tbl); | |
3240 | } | |
3241 | ||
3242 | static __net_initdata struct pernet_operations sysctl_route_ops = { | |
3243 | .init = sysctl_route_net_init, | |
3244 | .exit = sysctl_route_net_exit, | |
3245 | }; | |
1da177e4 LT |
3246 | #endif |
3247 | ||
3ee94372 | 3248 | static __net_init int rt_genid_init(struct net *net) |
9f5e97e5 | 3249 | { |
3ee94372 NH |
3250 | get_random_bytes(&net->ipv4.rt_genid, |
3251 | sizeof(net->ipv4.rt_genid)); | |
9f5e97e5 DL |
3252 | return 0; |
3253 | } | |
3254 | ||
3ee94372 NH |
3255 | static __net_initdata struct pernet_operations rt_genid_ops = { |
3256 | .init = rt_genid_init, | |
9f5e97e5 DL |
3257 | }; |
3258 | ||
3259 | ||
1da177e4 | 3260 | #ifdef CONFIG_NET_CLS_ROUTE |
7d720c3e | 3261 | struct ip_rt_acct __percpu *ip_rt_acct __read_mostly; |
1da177e4 LT |
3262 | #endif /* CONFIG_NET_CLS_ROUTE */ |
3263 | ||
3264 | static __initdata unsigned long rhash_entries; | |
3265 | static int __init set_rhash_entries(char *str) | |
3266 | { | |
3267 | if (!str) | |
3268 | return 0; | |
3269 | rhash_entries = simple_strtoul(str, &str, 0); | |
3270 | return 1; | |
3271 | } | |
3272 | __setup("rhash_entries=", set_rhash_entries); | |
3273 | ||
3274 | int __init ip_rt_init(void) | |
3275 | { | |
424c4b70 | 3276 | int rc = 0; |
1da177e4 | 3277 | |
1da177e4 | 3278 | #ifdef CONFIG_NET_CLS_ROUTE |
0dcec8c2 | 3279 | ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct)); |
1da177e4 LT |
3280 | if (!ip_rt_acct) |
3281 | panic("IP: failed to allocate ip_rt_acct\n"); | |
1da177e4 LT |
3282 | #endif |
3283 | ||
e5d679f3 AD |
3284 | ipv4_dst_ops.kmem_cachep = |
3285 | kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0, | |
20c2df83 | 3286 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 3287 | |
14e50e57 DM |
3288 | ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep; |
3289 | ||
424c4b70 ED |
3290 | rt_hash_table = (struct rt_hash_bucket *) |
3291 | alloc_large_system_hash("IP route cache", | |
3292 | sizeof(struct rt_hash_bucket), | |
3293 | rhash_entries, | |
4481374c | 3294 | (totalram_pages >= 128 * 1024) ? |
18955cfc | 3295 | 15 : 17, |
8d1502de | 3296 | 0, |
424c4b70 ED |
3297 | &rt_hash_log, |
3298 | &rt_hash_mask, | |
c9503e0f | 3299 | rhash_entries ? 0 : 512 * 1024); |
22c047cc ED |
3300 | memset(rt_hash_table, 0, (rt_hash_mask + 1) * sizeof(struct rt_hash_bucket)); |
3301 | rt_hash_lock_init(); | |
1da177e4 LT |
3302 | |
3303 | ipv4_dst_ops.gc_thresh = (rt_hash_mask + 1); | |
3304 | ip_rt_max_size = (rt_hash_mask + 1) * 16; | |
3305 | ||
1da177e4 LT |
3306 | devinet_init(); |
3307 | ip_fib_init(); | |
3308 | ||
1da177e4 LT |
3309 | /* All the timers, started at system startup tend |
3310 | to synchronize. Perturb it a bit. | |
3311 | */ | |
125bb8f5 ED |
3312 | INIT_DELAYED_WORK_DEFERRABLE(&expires_work, rt_worker_func); |
3313 | expires_ljiffies = jiffies; | |
39c90ece ED |
3314 | schedule_delayed_work(&expires_work, |
3315 | net_random() % ip_rt_gc_interval + ip_rt_gc_interval); | |
1da177e4 | 3316 | |
73b38711 | 3317 | if (ip_rt_proc_init()) |
107f1634 | 3318 | printk(KERN_ERR "Unable to create route proc files\n"); |
1da177e4 LT |
3319 | #ifdef CONFIG_XFRM |
3320 | xfrm_init(); | |
a33bc5c1 | 3321 | xfrm4_init(ip_rt_max_size); |
1da177e4 | 3322 | #endif |
63f3444f TG |
3323 | rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL); |
3324 | ||
39a23e75 DL |
3325 | #ifdef CONFIG_SYSCTL |
3326 | register_pernet_subsys(&sysctl_route_ops); | |
3327 | #endif | |
3ee94372 | 3328 | register_pernet_subsys(&rt_genid_ops); |
1da177e4 LT |
3329 | return rc; |
3330 | } | |
3331 | ||
a1bc6eb4 | 3332 | #ifdef CONFIG_SYSCTL |
eeb61f71 AV |
3333 | /* |
3334 | * We really need to sanitize the damn ipv4 init order, then all | |
3335 | * this nonsense will go away. | |
3336 | */ | |
3337 | void __init ip_static_sysctl_init(void) | |
3338 | { | |
2f4520d3 | 3339 | register_sysctl_paths(ipv4_path, ipv4_skeleton); |
eeb61f71 | 3340 | } |
a1bc6eb4 | 3341 | #endif |