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 | ||
afd46503 JP |
65 | #define pr_fmt(fmt) "IPv4: " fmt |
66 | ||
1da177e4 LT |
67 | #include <linux/module.h> |
68 | #include <asm/uaccess.h> | |
1da177e4 LT |
69 | #include <linux/bitops.h> |
70 | #include <linux/types.h> | |
71 | #include <linux/kernel.h> | |
1da177e4 LT |
72 | #include <linux/mm.h> |
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> | |
82 | #include <linux/skbuff.h> | |
1da177e4 LT |
83 | #include <linux/inetdevice.h> |
84 | #include <linux/igmp.h> | |
85 | #include <linux/pkt_sched.h> | |
86 | #include <linux/mroute.h> | |
87 | #include <linux/netfilter_ipv4.h> | |
88 | #include <linux/random.h> | |
1da177e4 LT |
89 | #include <linux/rcupdate.h> |
90 | #include <linux/times.h> | |
5a0e3ad6 | 91 | #include <linux/slab.h> |
73f156a6 | 92 | #include <linux/jhash.h> |
352e512c | 93 | #include <net/dst.h> |
1b7179d3 | 94 | #include <net/dst_metadata.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> | |
571e7226 | 106 | #include <net/lwtunnel.h> |
8d71740c | 107 | #include <net/netevent.h> |
63f3444f | 108 | #include <net/rtnetlink.h> |
1da177e4 LT |
109 | #ifdef CONFIG_SYSCTL |
110 | #include <linux/sysctl.h> | |
7426a564 | 111 | #include <linux/kmemleak.h> |
1da177e4 | 112 | #endif |
6e5714ea | 113 | #include <net/secure_seq.h> |
1b7179d3 | 114 | #include <net/ip_tunnels.h> |
385add90 | 115 | #include <net/l3mdev.h> |
1da177e4 | 116 | |
68a5e3dd | 117 | #define RT_FL_TOS(oldflp4) \ |
f61759e6 | 118 | ((oldflp4)->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK)) |
1da177e4 | 119 | |
1da177e4 LT |
120 | #define RT_GC_TIMEOUT (300*HZ) |
121 | ||
1da177e4 | 122 | static int ip_rt_max_size; |
817bc4db SH |
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; | |
817bc4db SH |
128 | static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ; |
129 | static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20; | |
130 | static int ip_rt_min_advmss __read_mostly = 256; | |
9f28a2fc | 131 | |
deed49df | 132 | static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT; |
1da177e4 LT |
133 | /* |
134 | * Interface to generic destination cache. | |
135 | */ | |
136 | ||
137 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie); | |
0dbaee3b | 138 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst); |
ebb762f2 | 139 | static unsigned int ipv4_mtu(const struct dst_entry *dst); |
1da177e4 LT |
140 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst); |
141 | static void ipv4_link_failure(struct sk_buff *skb); | |
6700c270 DM |
142 | static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk, |
143 | struct sk_buff *skb, u32 mtu); | |
144 | static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, | |
145 | struct sk_buff *skb); | |
caacf05e | 146 | static void ipv4_dst_destroy(struct dst_entry *dst); |
1da177e4 | 147 | |
62fa8a84 DM |
148 | static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old) |
149 | { | |
31248731 DM |
150 | WARN_ON(1); |
151 | return NULL; | |
62fa8a84 DM |
152 | } |
153 | ||
f894cbf8 DM |
154 | static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, |
155 | struct sk_buff *skb, | |
156 | const void *daddr); | |
d3aaeb38 | 157 | |
1da177e4 LT |
158 | static struct dst_ops ipv4_dst_ops = { |
159 | .family = AF_INET, | |
1da177e4 | 160 | .check = ipv4_dst_check, |
0dbaee3b | 161 | .default_advmss = ipv4_default_advmss, |
ebb762f2 | 162 | .mtu = ipv4_mtu, |
62fa8a84 | 163 | .cow_metrics = ipv4_cow_metrics, |
caacf05e | 164 | .destroy = ipv4_dst_destroy, |
1da177e4 LT |
165 | .negative_advice = ipv4_negative_advice, |
166 | .link_failure = ipv4_link_failure, | |
167 | .update_pmtu = ip_rt_update_pmtu, | |
e47a185b | 168 | .redirect = ip_do_redirect, |
b92dacd4 | 169 | .local_out = __ip_local_out, |
d3aaeb38 | 170 | .neigh_lookup = ipv4_neigh_lookup, |
1da177e4 LT |
171 | }; |
172 | ||
173 | #define ECN_OR_COST(class) TC_PRIO_##class | |
174 | ||
4839c52b | 175 | const __u8 ip_tos2prio[16] = { |
1da177e4 | 176 | TC_PRIO_BESTEFFORT, |
4a2b9c37 | 177 | ECN_OR_COST(BESTEFFORT), |
1da177e4 LT |
178 | TC_PRIO_BESTEFFORT, |
179 | ECN_OR_COST(BESTEFFORT), | |
180 | TC_PRIO_BULK, | |
181 | ECN_OR_COST(BULK), | |
182 | TC_PRIO_BULK, | |
183 | ECN_OR_COST(BULK), | |
184 | TC_PRIO_INTERACTIVE, | |
185 | ECN_OR_COST(INTERACTIVE), | |
186 | TC_PRIO_INTERACTIVE, | |
187 | ECN_OR_COST(INTERACTIVE), | |
188 | TC_PRIO_INTERACTIVE_BULK, | |
189 | ECN_OR_COST(INTERACTIVE_BULK), | |
190 | TC_PRIO_INTERACTIVE_BULK, | |
191 | ECN_OR_COST(INTERACTIVE_BULK) | |
192 | }; | |
d4a96865 | 193 | EXPORT_SYMBOL(ip_tos2prio); |
1da177e4 | 194 | |
2f970d83 | 195 | static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat); |
3ed66e91 | 196 | #define RT_CACHE_STAT_INC(field) raw_cpu_inc(rt_cache_stat.field) |
1da177e4 | 197 | |
1da177e4 | 198 | #ifdef CONFIG_PROC_FS |
1da177e4 LT |
199 | static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos) |
200 | { | |
29e75252 | 201 | if (*pos) |
89aef892 | 202 | return NULL; |
29e75252 | 203 | return SEQ_START_TOKEN; |
1da177e4 LT |
204 | } |
205 | ||
206 | static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
207 | { | |
1da177e4 | 208 | ++*pos; |
89aef892 | 209 | return NULL; |
1da177e4 LT |
210 | } |
211 | ||
212 | static void rt_cache_seq_stop(struct seq_file *seq, void *v) | |
213 | { | |
1da177e4 LT |
214 | } |
215 | ||
216 | static int rt_cache_seq_show(struct seq_file *seq, void *v) | |
217 | { | |
218 | if (v == SEQ_START_TOKEN) | |
219 | seq_printf(seq, "%-127s\n", | |
220 | "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t" | |
221 | "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t" | |
222 | "HHUptod\tSpecDst"); | |
e905a9ed | 223 | return 0; |
1da177e4 LT |
224 | } |
225 | ||
f690808e | 226 | static const struct seq_operations rt_cache_seq_ops = { |
1da177e4 LT |
227 | .start = rt_cache_seq_start, |
228 | .next = rt_cache_seq_next, | |
229 | .stop = rt_cache_seq_stop, | |
230 | .show = rt_cache_seq_show, | |
231 | }; | |
232 | ||
233 | static int rt_cache_seq_open(struct inode *inode, struct file *file) | |
234 | { | |
89aef892 | 235 | return seq_open(file, &rt_cache_seq_ops); |
1da177e4 LT |
236 | } |
237 | ||
9a32144e | 238 | static const struct file_operations rt_cache_seq_fops = { |
1da177e4 LT |
239 | .owner = THIS_MODULE, |
240 | .open = rt_cache_seq_open, | |
241 | .read = seq_read, | |
242 | .llseek = seq_lseek, | |
89aef892 | 243 | .release = seq_release, |
1da177e4 LT |
244 | }; |
245 | ||
246 | ||
247 | static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos) | |
248 | { | |
249 | int cpu; | |
250 | ||
251 | if (*pos == 0) | |
252 | return SEQ_START_TOKEN; | |
253 | ||
0f23174a | 254 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
255 | if (!cpu_possible(cpu)) |
256 | continue; | |
257 | *pos = cpu+1; | |
2f970d83 | 258 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
259 | } |
260 | return NULL; | |
261 | } | |
262 | ||
263 | static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
264 | { | |
265 | int cpu; | |
266 | ||
0f23174a | 267 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
268 | if (!cpu_possible(cpu)) |
269 | continue; | |
270 | *pos = cpu+1; | |
2f970d83 | 271 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
272 | } |
273 | return NULL; | |
e905a9ed | 274 | |
1da177e4 LT |
275 | } |
276 | ||
277 | static void rt_cpu_seq_stop(struct seq_file *seq, void *v) | |
278 | { | |
279 | ||
280 | } | |
281 | ||
282 | static int rt_cpu_seq_show(struct seq_file *seq, void *v) | |
283 | { | |
284 | struct rt_cache_stat *st = v; | |
285 | ||
286 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 287 | 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 |
288 | return 0; |
289 | } | |
e905a9ed | 290 | |
1da177e4 LT |
291 | seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x " |
292 | " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n", | |
fc66f95c | 293 | dst_entries_get_slow(&ipv4_dst_ops), |
0baf2b35 | 294 | 0, /* st->in_hit */ |
1da177e4 LT |
295 | st->in_slow_tot, |
296 | st->in_slow_mc, | |
297 | st->in_no_route, | |
298 | st->in_brd, | |
299 | st->in_martian_dst, | |
300 | st->in_martian_src, | |
301 | ||
0baf2b35 | 302 | 0, /* st->out_hit */ |
1da177e4 | 303 | st->out_slow_tot, |
e905a9ed | 304 | st->out_slow_mc, |
1da177e4 | 305 | |
0baf2b35 ED |
306 | 0, /* st->gc_total */ |
307 | 0, /* st->gc_ignored */ | |
308 | 0, /* st->gc_goal_miss */ | |
309 | 0, /* st->gc_dst_overflow */ | |
310 | 0, /* st->in_hlist_search */ | |
311 | 0 /* st->out_hlist_search */ | |
1da177e4 LT |
312 | ); |
313 | return 0; | |
314 | } | |
315 | ||
f690808e | 316 | static const struct seq_operations rt_cpu_seq_ops = { |
1da177e4 LT |
317 | .start = rt_cpu_seq_start, |
318 | .next = rt_cpu_seq_next, | |
319 | .stop = rt_cpu_seq_stop, | |
320 | .show = rt_cpu_seq_show, | |
321 | }; | |
322 | ||
323 | ||
324 | static int rt_cpu_seq_open(struct inode *inode, struct file *file) | |
325 | { | |
326 | return seq_open(file, &rt_cpu_seq_ops); | |
327 | } | |
328 | ||
9a32144e | 329 | static const struct file_operations rt_cpu_seq_fops = { |
1da177e4 LT |
330 | .owner = THIS_MODULE, |
331 | .open = rt_cpu_seq_open, | |
332 | .read = seq_read, | |
333 | .llseek = seq_lseek, | |
334 | .release = seq_release, | |
335 | }; | |
336 | ||
c7066f70 | 337 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 338 | static int rt_acct_proc_show(struct seq_file *m, void *v) |
78c686e9 | 339 | { |
a661c419 AD |
340 | struct ip_rt_acct *dst, *src; |
341 | unsigned int i, j; | |
342 | ||
343 | dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL); | |
344 | if (!dst) | |
345 | return -ENOMEM; | |
346 | ||
347 | for_each_possible_cpu(i) { | |
348 | src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i); | |
349 | for (j = 0; j < 256; j++) { | |
350 | dst[j].o_bytes += src[j].o_bytes; | |
351 | dst[j].o_packets += src[j].o_packets; | |
352 | dst[j].i_bytes += src[j].i_bytes; | |
353 | dst[j].i_packets += src[j].i_packets; | |
354 | } | |
78c686e9 PE |
355 | } |
356 | ||
a661c419 AD |
357 | seq_write(m, dst, 256 * sizeof(struct ip_rt_acct)); |
358 | kfree(dst); | |
359 | return 0; | |
360 | } | |
78c686e9 | 361 | |
a661c419 AD |
362 | static int rt_acct_proc_open(struct inode *inode, struct file *file) |
363 | { | |
364 | return single_open(file, rt_acct_proc_show, NULL); | |
78c686e9 | 365 | } |
a661c419 AD |
366 | |
367 | static const struct file_operations rt_acct_proc_fops = { | |
368 | .owner = THIS_MODULE, | |
369 | .open = rt_acct_proc_open, | |
370 | .read = seq_read, | |
371 | .llseek = seq_lseek, | |
372 | .release = single_release, | |
373 | }; | |
78c686e9 | 374 | #endif |
107f1634 | 375 | |
73b38711 | 376 | static int __net_init ip_rt_do_proc_init(struct net *net) |
107f1634 PE |
377 | { |
378 | struct proc_dir_entry *pde; | |
379 | ||
d4beaa66 G |
380 | pde = proc_create("rt_cache", S_IRUGO, net->proc_net, |
381 | &rt_cache_seq_fops); | |
107f1634 PE |
382 | if (!pde) |
383 | goto err1; | |
384 | ||
77020720 WC |
385 | pde = proc_create("rt_cache", S_IRUGO, |
386 | net->proc_net_stat, &rt_cpu_seq_fops); | |
107f1634 PE |
387 | if (!pde) |
388 | goto err2; | |
389 | ||
c7066f70 | 390 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 391 | pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops); |
107f1634 PE |
392 | if (!pde) |
393 | goto err3; | |
394 | #endif | |
395 | return 0; | |
396 | ||
c7066f70 | 397 | #ifdef CONFIG_IP_ROUTE_CLASSID |
107f1634 PE |
398 | err3: |
399 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
400 | #endif | |
401 | err2: | |
402 | remove_proc_entry("rt_cache", net->proc_net); | |
403 | err1: | |
404 | return -ENOMEM; | |
405 | } | |
73b38711 DL |
406 | |
407 | static void __net_exit ip_rt_do_proc_exit(struct net *net) | |
408 | { | |
409 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
410 | remove_proc_entry("rt_cache", net->proc_net); | |
c7066f70 | 411 | #ifdef CONFIG_IP_ROUTE_CLASSID |
73b38711 | 412 | remove_proc_entry("rt_acct", net->proc_net); |
0a931acf | 413 | #endif |
73b38711 DL |
414 | } |
415 | ||
416 | static struct pernet_operations ip_rt_proc_ops __net_initdata = { | |
417 | .init = ip_rt_do_proc_init, | |
418 | .exit = ip_rt_do_proc_exit, | |
419 | }; | |
420 | ||
421 | static int __init ip_rt_proc_init(void) | |
422 | { | |
423 | return register_pernet_subsys(&ip_rt_proc_ops); | |
424 | } | |
425 | ||
107f1634 | 426 | #else |
73b38711 | 427 | static inline int ip_rt_proc_init(void) |
107f1634 PE |
428 | { |
429 | return 0; | |
430 | } | |
1da177e4 | 431 | #endif /* CONFIG_PROC_FS */ |
e905a9ed | 432 | |
4331debc | 433 | static inline bool rt_is_expired(const struct rtable *rth) |
e84f84f2 | 434 | { |
ca4c3fc2 | 435 | return rth->rt_genid != rt_genid_ipv4(dev_net(rth->dst.dev)); |
e84f84f2 DL |
436 | } |
437 | ||
4ccfe6d4 | 438 | void rt_cache_flush(struct net *net) |
1da177e4 | 439 | { |
ca4c3fc2 | 440 | rt_genid_bump_ipv4(net); |
98376387 ED |
441 | } |
442 | ||
f894cbf8 DM |
443 | static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, |
444 | struct sk_buff *skb, | |
445 | const void *daddr) | |
3769cffb | 446 | { |
d3aaeb38 DM |
447 | struct net_device *dev = dst->dev; |
448 | const __be32 *pkey = daddr; | |
39232973 | 449 | const struct rtable *rt; |
3769cffb DM |
450 | struct neighbour *n; |
451 | ||
39232973 | 452 | rt = (const struct rtable *) dst; |
a263b309 | 453 | if (rt->rt_gateway) |
39232973 | 454 | pkey = (const __be32 *) &rt->rt_gateway; |
f894cbf8 DM |
455 | else if (skb) |
456 | pkey = &ip_hdr(skb)->daddr; | |
d3aaeb38 | 457 | |
80703d26 | 458 | n = __ipv4_neigh_lookup(dev, *(__force u32 *)pkey); |
d3aaeb38 DM |
459 | if (n) |
460 | return n; | |
32092ecf | 461 | return neigh_create(&arp_tbl, pkey, dev); |
d3aaeb38 DM |
462 | } |
463 | ||
04ca6973 | 464 | #define IP_IDENTS_SZ 2048u |
04ca6973 | 465 | |
355b590c ED |
466 | static atomic_t *ip_idents __read_mostly; |
467 | static u32 *ip_tstamps __read_mostly; | |
04ca6973 ED |
468 | |
469 | /* In order to protect privacy, we add a perturbation to identifiers | |
470 | * if one generator is seldom used. This makes hard for an attacker | |
471 | * to infer how many packets were sent between two points in time. | |
472 | */ | |
473 | u32 ip_idents_reserve(u32 hash, int segs) | |
474 | { | |
355b590c ED |
475 | u32 *p_tstamp = ip_tstamps + hash % IP_IDENTS_SZ; |
476 | atomic_t *p_id = ip_idents + hash % IP_IDENTS_SZ; | |
477 | u32 old = ACCESS_ONCE(*p_tstamp); | |
04ca6973 | 478 | u32 now = (u32)jiffies; |
adb03115 | 479 | u32 new, delta = 0; |
04ca6973 | 480 | |
355b590c | 481 | if (old != now && cmpxchg(p_tstamp, old, now) == old) |
04ca6973 ED |
482 | delta = prandom_u32_max(now - old); |
483 | ||
adb03115 ED |
484 | /* Do not use atomic_add_return() as it makes UBSAN unhappy */ |
485 | do { | |
486 | old = (u32)atomic_read(p_id); | |
487 | new = old + delta + segs; | |
488 | } while (atomic_cmpxchg(p_id, old, new) != old); | |
489 | ||
490 | return new - segs; | |
04ca6973 ED |
491 | } |
492 | EXPORT_SYMBOL(ip_idents_reserve); | |
1da177e4 | 493 | |
b6a7719a | 494 | void __ip_select_ident(struct net *net, struct iphdr *iph, int segs) |
1da177e4 | 495 | { |
73f156a6 ED |
496 | static u32 ip_idents_hashrnd __read_mostly; |
497 | u32 hash, id; | |
1da177e4 | 498 | |
73f156a6 | 499 | net_get_random_once(&ip_idents_hashrnd, sizeof(ip_idents_hashrnd)); |
1da177e4 | 500 | |
04ca6973 ED |
501 | hash = jhash_3words((__force u32)iph->daddr, |
502 | (__force u32)iph->saddr, | |
b6a7719a | 503 | iph->protocol ^ net_hash_mix(net), |
04ca6973 | 504 | ip_idents_hashrnd); |
73f156a6 ED |
505 | id = ip_idents_reserve(hash, segs); |
506 | iph->id = htons(id); | |
1da177e4 | 507 | } |
4bc2f18b | 508 | EXPORT_SYMBOL(__ip_select_ident); |
1da177e4 | 509 | |
5abf7f7e | 510 | static void __build_flow_key(struct flowi4 *fl4, const struct sock *sk, |
4895c771 DM |
511 | const struct iphdr *iph, |
512 | int oif, u8 tos, | |
513 | u8 prot, u32 mark, int flow_flags) | |
514 | { | |
515 | if (sk) { | |
516 | const struct inet_sock *inet = inet_sk(sk); | |
517 | ||
518 | oif = sk->sk_bound_dev_if; | |
519 | mark = sk->sk_mark; | |
520 | tos = RT_CONN_FLAGS(sk); | |
521 | prot = inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol; | |
522 | } | |
523 | flowi4_init_output(fl4, oif, mark, tos, | |
524 | RT_SCOPE_UNIVERSE, prot, | |
525 | flow_flags, | |
526 | iph->daddr, iph->saddr, 0, 0); | |
527 | } | |
528 | ||
5abf7f7e ED |
529 | static void build_skb_flow_key(struct flowi4 *fl4, const struct sk_buff *skb, |
530 | const struct sock *sk) | |
4895c771 DM |
531 | { |
532 | const struct iphdr *iph = ip_hdr(skb); | |
533 | int oif = skb->dev->ifindex; | |
534 | u8 tos = RT_TOS(iph->tos); | |
535 | u8 prot = iph->protocol; | |
536 | u32 mark = skb->mark; | |
537 | ||
538 | __build_flow_key(fl4, sk, iph, oif, tos, prot, mark, 0); | |
539 | } | |
540 | ||
5abf7f7e | 541 | static void build_sk_flow_key(struct flowi4 *fl4, const struct sock *sk) |
4895c771 DM |
542 | { |
543 | const struct inet_sock *inet = inet_sk(sk); | |
5abf7f7e | 544 | const struct ip_options_rcu *inet_opt; |
4895c771 DM |
545 | __be32 daddr = inet->inet_daddr; |
546 | ||
547 | rcu_read_lock(); | |
548 | inet_opt = rcu_dereference(inet->inet_opt); | |
549 | if (inet_opt && inet_opt->opt.srr) | |
550 | daddr = inet_opt->opt.faddr; | |
551 | flowi4_init_output(fl4, sk->sk_bound_dev_if, sk->sk_mark, | |
552 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, | |
553 | inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol, | |
554 | inet_sk_flowi_flags(sk), | |
555 | daddr, inet->inet_saddr, 0, 0); | |
556 | rcu_read_unlock(); | |
557 | } | |
558 | ||
5abf7f7e ED |
559 | static void ip_rt_build_flow_key(struct flowi4 *fl4, const struct sock *sk, |
560 | const struct sk_buff *skb) | |
4895c771 DM |
561 | { |
562 | if (skb) | |
563 | build_skb_flow_key(fl4, skb, sk); | |
564 | else | |
565 | build_sk_flow_key(fl4, sk); | |
566 | } | |
567 | ||
c5038a83 DM |
568 | static inline void rt_free(struct rtable *rt) |
569 | { | |
570 | call_rcu(&rt->dst.rcu_head, dst_rcu_free); | |
571 | } | |
572 | ||
573 | static DEFINE_SPINLOCK(fnhe_lock); | |
4895c771 | 574 | |
2ffae99d TT |
575 | static void fnhe_flush_routes(struct fib_nh_exception *fnhe) |
576 | { | |
577 | struct rtable *rt; | |
578 | ||
579 | rt = rcu_dereference(fnhe->fnhe_rth_input); | |
580 | if (rt) { | |
581 | RCU_INIT_POINTER(fnhe->fnhe_rth_input, NULL); | |
582 | rt_free(rt); | |
583 | } | |
584 | rt = rcu_dereference(fnhe->fnhe_rth_output); | |
585 | if (rt) { | |
586 | RCU_INIT_POINTER(fnhe->fnhe_rth_output, NULL); | |
587 | rt_free(rt); | |
588 | } | |
589 | } | |
590 | ||
aee06da6 | 591 | static struct fib_nh_exception *fnhe_oldest(struct fnhe_hash_bucket *hash) |
4895c771 DM |
592 | { |
593 | struct fib_nh_exception *fnhe, *oldest; | |
594 | ||
595 | oldest = rcu_dereference(hash->chain); | |
596 | for (fnhe = rcu_dereference(oldest->fnhe_next); fnhe; | |
597 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
598 | if (time_before(fnhe->fnhe_stamp, oldest->fnhe_stamp)) | |
599 | oldest = fnhe; | |
600 | } | |
2ffae99d | 601 | fnhe_flush_routes(oldest); |
4895c771 DM |
602 | return oldest; |
603 | } | |
604 | ||
d3a25c98 DM |
605 | static inline u32 fnhe_hashfun(__be32 daddr) |
606 | { | |
d546c621 | 607 | static u32 fnhe_hashrnd __read_mostly; |
d3a25c98 DM |
608 | u32 hval; |
609 | ||
d546c621 ED |
610 | net_get_random_once(&fnhe_hashrnd, sizeof(fnhe_hashrnd)); |
611 | hval = jhash_1word((__force u32) daddr, fnhe_hashrnd); | |
612 | return hash_32(hval, FNHE_HASH_SHIFT); | |
d3a25c98 DM |
613 | } |
614 | ||
387aa65a TT |
615 | static void fill_route_from_fnhe(struct rtable *rt, struct fib_nh_exception *fnhe) |
616 | { | |
617 | rt->rt_pmtu = fnhe->fnhe_pmtu; | |
618 | rt->dst.expires = fnhe->fnhe_expires; | |
619 | ||
620 | if (fnhe->fnhe_gw) { | |
621 | rt->rt_flags |= RTCF_REDIRECTED; | |
622 | rt->rt_gateway = fnhe->fnhe_gw; | |
623 | rt->rt_uses_gateway = 1; | |
624 | } | |
625 | } | |
626 | ||
aee06da6 JA |
627 | static void update_or_create_fnhe(struct fib_nh *nh, __be32 daddr, __be32 gw, |
628 | u32 pmtu, unsigned long expires) | |
4895c771 | 629 | { |
aee06da6 | 630 | struct fnhe_hash_bucket *hash; |
4895c771 | 631 | struct fib_nh_exception *fnhe; |
387aa65a TT |
632 | struct rtable *rt; |
633 | unsigned int i; | |
4895c771 | 634 | int depth; |
aee06da6 JA |
635 | u32 hval = fnhe_hashfun(daddr); |
636 | ||
c5038a83 | 637 | spin_lock_bh(&fnhe_lock); |
4895c771 | 638 | |
caa41527 | 639 | hash = rcu_dereference(nh->nh_exceptions); |
4895c771 | 640 | if (!hash) { |
aee06da6 | 641 | hash = kzalloc(FNHE_HASH_SIZE * sizeof(*hash), GFP_ATOMIC); |
4895c771 | 642 | if (!hash) |
aee06da6 | 643 | goto out_unlock; |
caa41527 | 644 | rcu_assign_pointer(nh->nh_exceptions, hash); |
4895c771 DM |
645 | } |
646 | ||
4895c771 DM |
647 | hash += hval; |
648 | ||
649 | depth = 0; | |
650 | for (fnhe = rcu_dereference(hash->chain); fnhe; | |
651 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
652 | if (fnhe->fnhe_daddr == daddr) | |
aee06da6 | 653 | break; |
4895c771 DM |
654 | depth++; |
655 | } | |
656 | ||
aee06da6 JA |
657 | if (fnhe) { |
658 | if (gw) | |
659 | fnhe->fnhe_gw = gw; | |
660 | if (pmtu) { | |
661 | fnhe->fnhe_pmtu = pmtu; | |
387aa65a | 662 | fnhe->fnhe_expires = max(1UL, expires); |
aee06da6 | 663 | } |
387aa65a | 664 | /* Update all cached dsts too */ |
2ffae99d TT |
665 | rt = rcu_dereference(fnhe->fnhe_rth_input); |
666 | if (rt) | |
667 | fill_route_from_fnhe(rt, fnhe); | |
668 | rt = rcu_dereference(fnhe->fnhe_rth_output); | |
387aa65a TT |
669 | if (rt) |
670 | fill_route_from_fnhe(rt, fnhe); | |
aee06da6 JA |
671 | } else { |
672 | if (depth > FNHE_RECLAIM_DEPTH) | |
673 | fnhe = fnhe_oldest(hash); | |
674 | else { | |
675 | fnhe = kzalloc(sizeof(*fnhe), GFP_ATOMIC); | |
676 | if (!fnhe) | |
677 | goto out_unlock; | |
678 | ||
679 | fnhe->fnhe_next = hash->chain; | |
680 | rcu_assign_pointer(hash->chain, fnhe); | |
681 | } | |
5aad1de5 | 682 | fnhe->fnhe_genid = fnhe_genid(dev_net(nh->nh_dev)); |
aee06da6 JA |
683 | fnhe->fnhe_daddr = daddr; |
684 | fnhe->fnhe_gw = gw; | |
685 | fnhe->fnhe_pmtu = pmtu; | |
686 | fnhe->fnhe_expires = expires; | |
387aa65a TT |
687 | |
688 | /* Exception created; mark the cached routes for the nexthop | |
689 | * stale, so anyone caching it rechecks if this exception | |
690 | * applies to them. | |
691 | */ | |
2ffae99d TT |
692 | rt = rcu_dereference(nh->nh_rth_input); |
693 | if (rt) | |
694 | rt->dst.obsolete = DST_OBSOLETE_KILL; | |
695 | ||
387aa65a TT |
696 | for_each_possible_cpu(i) { |
697 | struct rtable __rcu **prt; | |
698 | prt = per_cpu_ptr(nh->nh_pcpu_rth_output, i); | |
699 | rt = rcu_dereference(*prt); | |
700 | if (rt) | |
701 | rt->dst.obsolete = DST_OBSOLETE_KILL; | |
702 | } | |
4895c771 | 703 | } |
4895c771 | 704 | |
4895c771 | 705 | fnhe->fnhe_stamp = jiffies; |
aee06da6 JA |
706 | |
707 | out_unlock: | |
c5038a83 | 708 | spin_unlock_bh(&fnhe_lock); |
4895c771 DM |
709 | } |
710 | ||
ceb33206 DM |
711 | static void __ip_do_redirect(struct rtable *rt, struct sk_buff *skb, struct flowi4 *fl4, |
712 | bool kill_route) | |
1da177e4 | 713 | { |
e47a185b | 714 | __be32 new_gw = icmp_hdr(skb)->un.gateway; |
94206125 | 715 | __be32 old_gw = ip_hdr(skb)->saddr; |
e47a185b | 716 | struct net_device *dev = skb->dev; |
e47a185b | 717 | struct in_device *in_dev; |
4895c771 | 718 | struct fib_result res; |
e47a185b | 719 | struct neighbour *n; |
317805b8 | 720 | struct net *net; |
1da177e4 | 721 | |
94206125 DM |
722 | switch (icmp_hdr(skb)->code & 7) { |
723 | case ICMP_REDIR_NET: | |
724 | case ICMP_REDIR_NETTOS: | |
725 | case ICMP_REDIR_HOST: | |
726 | case ICMP_REDIR_HOSTTOS: | |
727 | break; | |
728 | ||
729 | default: | |
730 | return; | |
731 | } | |
732 | ||
e47a185b DM |
733 | if (rt->rt_gateway != old_gw) |
734 | return; | |
735 | ||
736 | in_dev = __in_dev_get_rcu(dev); | |
737 | if (!in_dev) | |
738 | return; | |
739 | ||
c346dca1 | 740 | net = dev_net(dev); |
9d4fb27d JP |
741 | if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) || |
742 | ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) || | |
743 | ipv4_is_zeronet(new_gw)) | |
1da177e4 LT |
744 | goto reject_redirect; |
745 | ||
746 | if (!IN_DEV_SHARED_MEDIA(in_dev)) { | |
747 | if (!inet_addr_onlink(in_dev, new_gw, old_gw)) | |
748 | goto reject_redirect; | |
749 | if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev)) | |
750 | goto reject_redirect; | |
751 | } else { | |
317805b8 | 752 | if (inet_addr_type(net, new_gw) != RTN_UNICAST) |
1da177e4 LT |
753 | goto reject_redirect; |
754 | } | |
755 | ||
4895c771 | 756 | n = ipv4_neigh_lookup(&rt->dst, NULL, &new_gw); |
2c1a4311 | 757 | if (!IS_ERR(n)) { |
e47a185b DM |
758 | if (!(n->nud_state & NUD_VALID)) { |
759 | neigh_event_send(n, NULL); | |
760 | } else { | |
0eeb075f | 761 | if (fib_lookup(net, fl4, &res, 0) == 0) { |
4895c771 | 762 | struct fib_nh *nh = &FIB_RES_NH(res); |
4895c771 | 763 | |
aee06da6 | 764 | update_or_create_fnhe(nh, fl4->daddr, new_gw, |
deed49df | 765 | 0, jiffies + ip_rt_gc_timeout); |
4895c771 | 766 | } |
ceb33206 DM |
767 | if (kill_route) |
768 | rt->dst.obsolete = DST_OBSOLETE_KILL; | |
e47a185b DM |
769 | call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, n); |
770 | } | |
771 | neigh_release(n); | |
772 | } | |
773 | return; | |
774 | ||
775 | reject_redirect: | |
776 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
99ee038d DM |
777 | if (IN_DEV_LOG_MARTIANS(in_dev)) { |
778 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
779 | __be32 daddr = iph->daddr; | |
780 | __be32 saddr = iph->saddr; | |
781 | ||
e47a185b DM |
782 | net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n" |
783 | " Advised path = %pI4 -> %pI4\n", | |
784 | &old_gw, dev->name, &new_gw, | |
785 | &saddr, &daddr); | |
99ee038d | 786 | } |
e47a185b DM |
787 | #endif |
788 | ; | |
789 | } | |
790 | ||
4895c771 DM |
791 | static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb) |
792 | { | |
793 | struct rtable *rt; | |
794 | struct flowi4 fl4; | |
f96ef988 MK |
795 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
796 | int oif = skb->dev->ifindex; | |
797 | u8 tos = RT_TOS(iph->tos); | |
798 | u8 prot = iph->protocol; | |
799 | u32 mark = skb->mark; | |
4895c771 DM |
800 | |
801 | rt = (struct rtable *) dst; | |
802 | ||
f96ef988 | 803 | __build_flow_key(&fl4, sk, iph, oif, tos, prot, mark, 0); |
ceb33206 | 804 | __ip_do_redirect(rt, skb, &fl4, true); |
4895c771 DM |
805 | } |
806 | ||
1da177e4 LT |
807 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst) |
808 | { | |
ee6b9673 | 809 | struct rtable *rt = (struct rtable *)dst; |
1da177e4 LT |
810 | struct dst_entry *ret = dst; |
811 | ||
812 | if (rt) { | |
d11a4dc1 | 813 | if (dst->obsolete > 0) { |
1da177e4 LT |
814 | ip_rt_put(rt); |
815 | ret = NULL; | |
5943634f DM |
816 | } else if ((rt->rt_flags & RTCF_REDIRECTED) || |
817 | rt->dst.expires) { | |
89aef892 | 818 | ip_rt_put(rt); |
1da177e4 LT |
819 | ret = NULL; |
820 | } | |
821 | } | |
822 | return ret; | |
823 | } | |
824 | ||
825 | /* | |
826 | * Algorithm: | |
827 | * 1. The first ip_rt_redirect_number redirects are sent | |
828 | * with exponential backoff, then we stop sending them at all, | |
829 | * assuming that the host ignores our redirects. | |
830 | * 2. If we did not see packets requiring redirects | |
831 | * during ip_rt_redirect_silence, we assume that the host | |
832 | * forgot redirected route and start to send redirects again. | |
833 | * | |
834 | * This algorithm is much cheaper and more intelligent than dumb load limiting | |
835 | * in icmp.c. | |
836 | * | |
837 | * NOTE. Do not forget to inhibit load limiting for redirects (redundant) | |
838 | * and "frag. need" (breaks PMTU discovery) in icmp.c. | |
839 | */ | |
840 | ||
841 | void ip_rt_send_redirect(struct sk_buff *skb) | |
842 | { | |
511c3f92 | 843 | struct rtable *rt = skb_rtable(skb); |
30038fc6 | 844 | struct in_device *in_dev; |
92d86829 | 845 | struct inet_peer *peer; |
1d861aa4 | 846 | struct net *net; |
30038fc6 | 847 | int log_martians; |
192132b9 | 848 | int vif; |
1da177e4 | 849 | |
30038fc6 | 850 | rcu_read_lock(); |
d8d1f30b | 851 | in_dev = __in_dev_get_rcu(rt->dst.dev); |
30038fc6 ED |
852 | if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) { |
853 | rcu_read_unlock(); | |
1da177e4 | 854 | return; |
30038fc6 ED |
855 | } |
856 | log_martians = IN_DEV_LOG_MARTIANS(in_dev); | |
385add90 | 857 | vif = l3mdev_master_ifindex_rcu(rt->dst.dev); |
30038fc6 | 858 | rcu_read_unlock(); |
1da177e4 | 859 | |
1d861aa4 | 860 | net = dev_net(rt->dst.dev); |
192132b9 | 861 | peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, vif, 1); |
92d86829 | 862 | if (!peer) { |
e81da0e1 JA |
863 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, |
864 | rt_nexthop(rt, ip_hdr(skb)->daddr)); | |
92d86829 DM |
865 | return; |
866 | } | |
867 | ||
1da177e4 LT |
868 | /* No redirected packets during ip_rt_redirect_silence; |
869 | * reset the algorithm. | |
870 | */ | |
92d86829 DM |
871 | if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence)) |
872 | peer->rate_tokens = 0; | |
1da177e4 LT |
873 | |
874 | /* Too many ignored redirects; do not send anything | |
d8d1f30b | 875 | * set dst.rate_last to the last seen redirected packet. |
1da177e4 | 876 | */ |
92d86829 DM |
877 | if (peer->rate_tokens >= ip_rt_redirect_number) { |
878 | peer->rate_last = jiffies; | |
1d861aa4 | 879 | goto out_put_peer; |
1da177e4 LT |
880 | } |
881 | ||
882 | /* Check for load limit; set rate_last to the latest sent | |
883 | * redirect. | |
884 | */ | |
92d86829 | 885 | if (peer->rate_tokens == 0 || |
14fb8a76 | 886 | time_after(jiffies, |
92d86829 DM |
887 | (peer->rate_last + |
888 | (ip_rt_redirect_load << peer->rate_tokens)))) { | |
e81da0e1 JA |
889 | __be32 gw = rt_nexthop(rt, ip_hdr(skb)->daddr); |
890 | ||
891 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, gw); | |
92d86829 DM |
892 | peer->rate_last = jiffies; |
893 | ++peer->rate_tokens; | |
1da177e4 | 894 | #ifdef CONFIG_IP_ROUTE_VERBOSE |
30038fc6 | 895 | if (log_martians && |
e87cc472 JP |
896 | peer->rate_tokens == ip_rt_redirect_number) |
897 | net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n", | |
92101b3b | 898 | &ip_hdr(skb)->saddr, inet_iif(skb), |
e81da0e1 | 899 | &ip_hdr(skb)->daddr, &gw); |
1da177e4 LT |
900 | #endif |
901 | } | |
1d861aa4 DM |
902 | out_put_peer: |
903 | inet_putpeer(peer); | |
1da177e4 LT |
904 | } |
905 | ||
906 | static int ip_error(struct sk_buff *skb) | |
907 | { | |
251da413 | 908 | struct in_device *in_dev = __in_dev_get_rcu(skb->dev); |
511c3f92 | 909 | struct rtable *rt = skb_rtable(skb); |
92d86829 | 910 | struct inet_peer *peer; |
1da177e4 | 911 | unsigned long now; |
251da413 | 912 | struct net *net; |
92d86829 | 913 | bool send; |
1da177e4 LT |
914 | int code; |
915 | ||
381c759d EB |
916 | /* IP on this device is disabled. */ |
917 | if (!in_dev) | |
918 | goto out; | |
919 | ||
251da413 DM |
920 | net = dev_net(rt->dst.dev); |
921 | if (!IN_DEV_FORWARD(in_dev)) { | |
922 | switch (rt->dst.error) { | |
923 | case EHOSTUNREACH: | |
b45386ef | 924 | __IP_INC_STATS(net, IPSTATS_MIB_INADDRERRORS); |
251da413 DM |
925 | break; |
926 | ||
927 | case ENETUNREACH: | |
b45386ef | 928 | __IP_INC_STATS(net, IPSTATS_MIB_INNOROUTES); |
251da413 DM |
929 | break; |
930 | } | |
931 | goto out; | |
932 | } | |
933 | ||
d8d1f30b | 934 | switch (rt->dst.error) { |
4500ebf8 JP |
935 | case EINVAL: |
936 | default: | |
937 | goto out; | |
938 | case EHOSTUNREACH: | |
939 | code = ICMP_HOST_UNREACH; | |
940 | break; | |
941 | case ENETUNREACH: | |
942 | code = ICMP_NET_UNREACH; | |
b45386ef | 943 | __IP_INC_STATS(net, IPSTATS_MIB_INNOROUTES); |
4500ebf8 JP |
944 | break; |
945 | case EACCES: | |
946 | code = ICMP_PKT_FILTERED; | |
947 | break; | |
1da177e4 LT |
948 | } |
949 | ||
192132b9 | 950 | peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, |
385add90 | 951 | l3mdev_master_ifindex(skb->dev), 1); |
92d86829 DM |
952 | |
953 | send = true; | |
954 | if (peer) { | |
955 | now = jiffies; | |
956 | peer->rate_tokens += now - peer->rate_last; | |
957 | if (peer->rate_tokens > ip_rt_error_burst) | |
958 | peer->rate_tokens = ip_rt_error_burst; | |
959 | peer->rate_last = now; | |
960 | if (peer->rate_tokens >= ip_rt_error_cost) | |
961 | peer->rate_tokens -= ip_rt_error_cost; | |
962 | else | |
963 | send = false; | |
1d861aa4 | 964 | inet_putpeer(peer); |
1da177e4 | 965 | } |
92d86829 DM |
966 | if (send) |
967 | icmp_send(skb, ICMP_DEST_UNREACH, code, 0); | |
1da177e4 LT |
968 | |
969 | out: kfree_skb(skb); | |
970 | return 0; | |
e905a9ed | 971 | } |
1da177e4 | 972 | |
d851c12b | 973 | static void __ip_rt_update_pmtu(struct rtable *rt, struct flowi4 *fl4, u32 mtu) |
1da177e4 | 974 | { |
d851c12b | 975 | struct dst_entry *dst = &rt->dst; |
4895c771 | 976 | struct fib_result res; |
2c8cec5c | 977 | |
fa1e492a SK |
978 | if (dst_metric_locked(dst, RTAX_MTU)) |
979 | return; | |
980 | ||
cb6ccf09 | 981 | if (ipv4_mtu(dst) < mtu) |
3cdaa5be LW |
982 | return; |
983 | ||
5943634f DM |
984 | if (mtu < ip_rt_min_pmtu) |
985 | mtu = ip_rt_min_pmtu; | |
2c8cec5c | 986 | |
f016229e TT |
987 | if (rt->rt_pmtu == mtu && |
988 | time_before(jiffies, dst->expires - ip_rt_mtu_expires / 2)) | |
989 | return; | |
990 | ||
c5ae7d41 | 991 | rcu_read_lock(); |
0eeb075f | 992 | if (fib_lookup(dev_net(dst->dev), fl4, &res, 0) == 0) { |
4895c771 | 993 | struct fib_nh *nh = &FIB_RES_NH(res); |
4895c771 | 994 | |
aee06da6 JA |
995 | update_or_create_fnhe(nh, fl4->daddr, 0, mtu, |
996 | jiffies + ip_rt_mtu_expires); | |
4895c771 | 997 | } |
c5ae7d41 | 998 | rcu_read_unlock(); |
1da177e4 LT |
999 | } |
1000 | ||
4895c771 DM |
1001 | static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk, |
1002 | struct sk_buff *skb, u32 mtu) | |
1003 | { | |
1004 | struct rtable *rt = (struct rtable *) dst; | |
1005 | struct flowi4 fl4; | |
1006 | ||
1007 | ip_rt_build_flow_key(&fl4, sk, skb); | |
d851c12b | 1008 | __ip_rt_update_pmtu(rt, &fl4, mtu); |
4895c771 DM |
1009 | } |
1010 | ||
36393395 DM |
1011 | void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, |
1012 | int oif, u32 mark, u8 protocol, int flow_flags) | |
1013 | { | |
4895c771 | 1014 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
36393395 DM |
1015 | struct flowi4 fl4; |
1016 | struct rtable *rt; | |
1017 | ||
1b3c61dc LC |
1018 | if (!mark) |
1019 | mark = IP4_REPLY_MARK(net, skb->mark); | |
1020 | ||
4895c771 DM |
1021 | __build_flow_key(&fl4, NULL, iph, oif, |
1022 | RT_TOS(iph->tos), protocol, mark, flow_flags); | |
36393395 DM |
1023 | rt = __ip_route_output_key(net, &fl4); |
1024 | if (!IS_ERR(rt)) { | |
4895c771 | 1025 | __ip_rt_update_pmtu(rt, &fl4, mtu); |
36393395 DM |
1026 | ip_rt_put(rt); |
1027 | } | |
1028 | } | |
1029 | EXPORT_SYMBOL_GPL(ipv4_update_pmtu); | |
1030 | ||
9cb3a50c | 1031 | static void __ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu) |
36393395 | 1032 | { |
4895c771 DM |
1033 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
1034 | struct flowi4 fl4; | |
1035 | struct rtable *rt; | |
36393395 | 1036 | |
4895c771 | 1037 | __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0); |
1b3c61dc LC |
1038 | |
1039 | if (!fl4.flowi4_mark) | |
1040 | fl4.flowi4_mark = IP4_REPLY_MARK(sock_net(sk), skb->mark); | |
1041 | ||
4895c771 DM |
1042 | rt = __ip_route_output_key(sock_net(sk), &fl4); |
1043 | if (!IS_ERR(rt)) { | |
1044 | __ip_rt_update_pmtu(rt, &fl4, mtu); | |
1045 | ip_rt_put(rt); | |
1046 | } | |
36393395 | 1047 | } |
9cb3a50c SK |
1048 | |
1049 | void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu) | |
1050 | { | |
1051 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
1052 | struct flowi4 fl4; | |
1053 | struct rtable *rt; | |
7f502361 | 1054 | struct dst_entry *odst = NULL; |
b44108db | 1055 | bool new = false; |
9cb3a50c SK |
1056 | |
1057 | bh_lock_sock(sk); | |
482fc609 HFS |
1058 | |
1059 | if (!ip_sk_accept_pmtu(sk)) | |
1060 | goto out; | |
1061 | ||
7f502361 | 1062 | odst = sk_dst_get(sk); |
9cb3a50c | 1063 | |
7f502361 | 1064 | if (sock_owned_by_user(sk) || !odst) { |
9cb3a50c SK |
1065 | __ipv4_sk_update_pmtu(skb, sk, mtu); |
1066 | goto out; | |
1067 | } | |
1068 | ||
1069 | __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0); | |
1070 | ||
7f502361 | 1071 | rt = (struct rtable *)odst; |
51456b29 | 1072 | if (odst->obsolete && !odst->ops->check(odst, 0)) { |
9cb3a50c SK |
1073 | rt = ip_route_output_flow(sock_net(sk), &fl4, sk); |
1074 | if (IS_ERR(rt)) | |
1075 | goto out; | |
b44108db SK |
1076 | |
1077 | new = true; | |
9cb3a50c SK |
1078 | } |
1079 | ||
1080 | __ip_rt_update_pmtu((struct rtable *) rt->dst.path, &fl4, mtu); | |
1081 | ||
7f502361 | 1082 | if (!dst_check(&rt->dst, 0)) { |
b44108db SK |
1083 | if (new) |
1084 | dst_release(&rt->dst); | |
1085 | ||
9cb3a50c SK |
1086 | rt = ip_route_output_flow(sock_net(sk), &fl4, sk); |
1087 | if (IS_ERR(rt)) | |
1088 | goto out; | |
1089 | ||
b44108db | 1090 | new = true; |
9cb3a50c SK |
1091 | } |
1092 | ||
b44108db | 1093 | if (new) |
7f502361 | 1094 | sk_dst_set(sk, &rt->dst); |
9cb3a50c SK |
1095 | |
1096 | out: | |
1097 | bh_unlock_sock(sk); | |
7f502361 | 1098 | dst_release(odst); |
9cb3a50c | 1099 | } |
36393395 | 1100 | EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu); |
f39925db | 1101 | |
b42597e2 DM |
1102 | void ipv4_redirect(struct sk_buff *skb, struct net *net, |
1103 | int oif, u32 mark, u8 protocol, int flow_flags) | |
1104 | { | |
4895c771 | 1105 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
b42597e2 DM |
1106 | struct flowi4 fl4; |
1107 | struct rtable *rt; | |
1108 | ||
4895c771 DM |
1109 | __build_flow_key(&fl4, NULL, iph, oif, |
1110 | RT_TOS(iph->tos), protocol, mark, flow_flags); | |
b42597e2 DM |
1111 | rt = __ip_route_output_key(net, &fl4); |
1112 | if (!IS_ERR(rt)) { | |
ceb33206 | 1113 | __ip_do_redirect(rt, skb, &fl4, false); |
b42597e2 DM |
1114 | ip_rt_put(rt); |
1115 | } | |
1116 | } | |
1117 | EXPORT_SYMBOL_GPL(ipv4_redirect); | |
1118 | ||
1119 | void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk) | |
1120 | { | |
4895c771 DM |
1121 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
1122 | struct flowi4 fl4; | |
1123 | struct rtable *rt; | |
b42597e2 | 1124 | |
4895c771 DM |
1125 | __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0); |
1126 | rt = __ip_route_output_key(sock_net(sk), &fl4); | |
1127 | if (!IS_ERR(rt)) { | |
ceb33206 | 1128 | __ip_do_redirect(rt, skb, &fl4, false); |
4895c771 DM |
1129 | ip_rt_put(rt); |
1130 | } | |
b42597e2 DM |
1131 | } |
1132 | EXPORT_SYMBOL_GPL(ipv4_sk_redirect); | |
1133 | ||
efbc368d DM |
1134 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie) |
1135 | { | |
1136 | struct rtable *rt = (struct rtable *) dst; | |
1137 | ||
ceb33206 DM |
1138 | /* All IPV4 dsts are created with ->obsolete set to the value |
1139 | * DST_OBSOLETE_FORCE_CHK which forces validation calls down | |
1140 | * into this function always. | |
1141 | * | |
387aa65a TT |
1142 | * When a PMTU/redirect information update invalidates a route, |
1143 | * this is indicated by setting obsolete to DST_OBSOLETE_KILL or | |
1144 | * DST_OBSOLETE_DEAD by dst_free(). | |
ceb33206 | 1145 | */ |
387aa65a | 1146 | if (dst->obsolete != DST_OBSOLETE_FORCE_CHK || rt_is_expired(rt)) |
efbc368d | 1147 | return NULL; |
d11a4dc1 | 1148 | return dst; |
1da177e4 LT |
1149 | } |
1150 | ||
1da177e4 LT |
1151 | static void ipv4_link_failure(struct sk_buff *skb) |
1152 | { | |
1153 | struct rtable *rt; | |
1154 | ||
1155 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0); | |
1156 | ||
511c3f92 | 1157 | rt = skb_rtable(skb); |
5943634f DM |
1158 | if (rt) |
1159 | dst_set_expires(&rt->dst, 0); | |
1da177e4 LT |
1160 | } |
1161 | ||
ede2059d | 1162 | static int ip_rt_bug(struct net *net, struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1163 | { |
91df42be JP |
1164 | pr_debug("%s: %pI4 -> %pI4, %s\n", |
1165 | __func__, &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
1166 | skb->dev ? skb->dev->name : "?"); | |
1da177e4 | 1167 | kfree_skb(skb); |
c378a9c0 | 1168 | WARN_ON(1); |
1da177e4 LT |
1169 | return 0; |
1170 | } | |
1171 | ||
1172 | /* | |
1173 | We do not cache source address of outgoing interface, | |
1174 | because it is used only by IP RR, TS and SRR options, | |
1175 | so that it out of fast path. | |
1176 | ||
1177 | BTW remember: "addr" is allowed to be not aligned | |
1178 | in IP options! | |
1179 | */ | |
1180 | ||
8e36360a | 1181 | void ip_rt_get_source(u8 *addr, struct sk_buff *skb, struct rtable *rt) |
1da177e4 | 1182 | { |
a61ced5d | 1183 | __be32 src; |
1da177e4 | 1184 | |
c7537967 | 1185 | if (rt_is_output_route(rt)) |
c5be24ff | 1186 | src = ip_hdr(skb)->saddr; |
ebc0ffae | 1187 | else { |
8e36360a DM |
1188 | struct fib_result res; |
1189 | struct flowi4 fl4; | |
1190 | struct iphdr *iph; | |
1191 | ||
1192 | iph = ip_hdr(skb); | |
1193 | ||
1194 | memset(&fl4, 0, sizeof(fl4)); | |
1195 | fl4.daddr = iph->daddr; | |
1196 | fl4.saddr = iph->saddr; | |
b0fe4a31 | 1197 | fl4.flowi4_tos = RT_TOS(iph->tos); |
8e36360a DM |
1198 | fl4.flowi4_oif = rt->dst.dev->ifindex; |
1199 | fl4.flowi4_iif = skb->dev->ifindex; | |
1200 | fl4.flowi4_mark = skb->mark; | |
5e2b61f7 | 1201 | |
ebc0ffae | 1202 | rcu_read_lock(); |
0eeb075f | 1203 | if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res, 0) == 0) |
436c3b66 | 1204 | src = FIB_RES_PREFSRC(dev_net(rt->dst.dev), res); |
ebc0ffae | 1205 | else |
f8126f1d DM |
1206 | src = inet_select_addr(rt->dst.dev, |
1207 | rt_nexthop(rt, iph->daddr), | |
1208 | RT_SCOPE_UNIVERSE); | |
ebc0ffae ED |
1209 | rcu_read_unlock(); |
1210 | } | |
1da177e4 LT |
1211 | memcpy(addr, &src, 4); |
1212 | } | |
1213 | ||
c7066f70 | 1214 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 LT |
1215 | static void set_class_tag(struct rtable *rt, u32 tag) |
1216 | { | |
d8d1f30b CG |
1217 | if (!(rt->dst.tclassid & 0xFFFF)) |
1218 | rt->dst.tclassid |= tag & 0xFFFF; | |
1219 | if (!(rt->dst.tclassid & 0xFFFF0000)) | |
1220 | rt->dst.tclassid |= tag & 0xFFFF0000; | |
1da177e4 LT |
1221 | } |
1222 | #endif | |
1223 | ||
0dbaee3b DM |
1224 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst) |
1225 | { | |
1226 | unsigned int advmss = dst_metric_raw(dst, RTAX_ADVMSS); | |
1227 | ||
1228 | if (advmss == 0) { | |
1229 | advmss = max_t(unsigned int, dst->dev->mtu - 40, | |
1230 | ip_rt_min_advmss); | |
1231 | if (advmss > 65535 - 40) | |
1232 | advmss = 65535 - 40; | |
1233 | } | |
1234 | return advmss; | |
1235 | } | |
1236 | ||
ebb762f2 | 1237 | static unsigned int ipv4_mtu(const struct dst_entry *dst) |
d33e4553 | 1238 | { |
261663b0 | 1239 | const struct rtable *rt = (const struct rtable *) dst; |
5943634f DM |
1240 | unsigned int mtu = rt->rt_pmtu; |
1241 | ||
98d75c37 | 1242 | if (!mtu || time_after_eq(jiffies, rt->dst.expires)) |
5943634f | 1243 | mtu = dst_metric_raw(dst, RTAX_MTU); |
618f9bc7 | 1244 | |
38d523e2 | 1245 | if (mtu) |
618f9bc7 SK |
1246 | return mtu; |
1247 | ||
1248 | mtu = dst->dev->mtu; | |
d33e4553 DM |
1249 | |
1250 | if (unlikely(dst_metric_locked(dst, RTAX_MTU))) { | |
155e8336 | 1251 | if (rt->rt_uses_gateway && mtu > 576) |
d33e4553 DM |
1252 | mtu = 576; |
1253 | } | |
1254 | ||
734d2725 | 1255 | return min_t(unsigned int, mtu, IP_MAX_MTU); |
d33e4553 DM |
1256 | } |
1257 | ||
f2bb4bed | 1258 | static struct fib_nh_exception *find_exception(struct fib_nh *nh, __be32 daddr) |
4895c771 | 1259 | { |
caa41527 | 1260 | struct fnhe_hash_bucket *hash = rcu_dereference(nh->nh_exceptions); |
4895c771 DM |
1261 | struct fib_nh_exception *fnhe; |
1262 | u32 hval; | |
1263 | ||
f2bb4bed DM |
1264 | if (!hash) |
1265 | return NULL; | |
1266 | ||
d3a25c98 | 1267 | hval = fnhe_hashfun(daddr); |
4895c771 DM |
1268 | |
1269 | for (fnhe = rcu_dereference(hash[hval].chain); fnhe; | |
1270 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
f2bb4bed DM |
1271 | if (fnhe->fnhe_daddr == daddr) |
1272 | return fnhe; | |
1273 | } | |
1274 | return NULL; | |
1275 | } | |
aee06da6 | 1276 | |
caacf05e | 1277 | static bool rt_bind_exception(struct rtable *rt, struct fib_nh_exception *fnhe, |
f2bb4bed DM |
1278 | __be32 daddr) |
1279 | { | |
caacf05e DM |
1280 | bool ret = false; |
1281 | ||
c5038a83 | 1282 | spin_lock_bh(&fnhe_lock); |
f2bb4bed | 1283 | |
c5038a83 | 1284 | if (daddr == fnhe->fnhe_daddr) { |
2ffae99d TT |
1285 | struct rtable __rcu **porig; |
1286 | struct rtable *orig; | |
5aad1de5 | 1287 | int genid = fnhe_genid(dev_net(rt->dst.dev)); |
2ffae99d TT |
1288 | |
1289 | if (rt_is_input_route(rt)) | |
1290 | porig = &fnhe->fnhe_rth_input; | |
1291 | else | |
1292 | porig = &fnhe->fnhe_rth_output; | |
1293 | orig = rcu_dereference(*porig); | |
5aad1de5 TT |
1294 | |
1295 | if (fnhe->fnhe_genid != genid) { | |
1296 | fnhe->fnhe_genid = genid; | |
13d82bf5 SK |
1297 | fnhe->fnhe_gw = 0; |
1298 | fnhe->fnhe_pmtu = 0; | |
1299 | fnhe->fnhe_expires = 0; | |
2ffae99d TT |
1300 | fnhe_flush_routes(fnhe); |
1301 | orig = NULL; | |
13d82bf5 | 1302 | } |
387aa65a TT |
1303 | fill_route_from_fnhe(rt, fnhe); |
1304 | if (!rt->rt_gateway) | |
155e8336 | 1305 | rt->rt_gateway = daddr; |
f2bb4bed | 1306 | |
2ffae99d TT |
1307 | if (!(rt->dst.flags & DST_NOCACHE)) { |
1308 | rcu_assign_pointer(*porig, rt); | |
1309 | if (orig) | |
1310 | rt_free(orig); | |
1311 | ret = true; | |
1312 | } | |
c5038a83 DM |
1313 | |
1314 | fnhe->fnhe_stamp = jiffies; | |
c5038a83 DM |
1315 | } |
1316 | spin_unlock_bh(&fnhe_lock); | |
caacf05e DM |
1317 | |
1318 | return ret; | |
54764bb6 ED |
1319 | } |
1320 | ||
caacf05e | 1321 | static bool rt_cache_route(struct fib_nh *nh, struct rtable *rt) |
f2bb4bed | 1322 | { |
d26b3a7c | 1323 | struct rtable *orig, *prev, **p; |
caacf05e | 1324 | bool ret = true; |
f2bb4bed | 1325 | |
d26b3a7c | 1326 | if (rt_is_input_route(rt)) { |
54764bb6 | 1327 | p = (struct rtable **)&nh->nh_rth_input; |
d26b3a7c | 1328 | } else { |
903ceff7 | 1329 | p = (struct rtable **)raw_cpu_ptr(nh->nh_pcpu_rth_output); |
d26b3a7c | 1330 | } |
f2bb4bed DM |
1331 | orig = *p; |
1332 | ||
1333 | prev = cmpxchg(p, orig, rt); | |
1334 | if (prev == orig) { | |
f2bb4bed | 1335 | if (orig) |
54764bb6 | 1336 | rt_free(orig); |
155e8336 | 1337 | } else |
caacf05e | 1338 | ret = false; |
caacf05e DM |
1339 | |
1340 | return ret; | |
1341 | } | |
1342 | ||
5055c371 ED |
1343 | struct uncached_list { |
1344 | spinlock_t lock; | |
1345 | struct list_head head; | |
1346 | }; | |
1347 | ||
1348 | static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt_uncached_list); | |
caacf05e DM |
1349 | |
1350 | static void rt_add_uncached_list(struct rtable *rt) | |
1351 | { | |
5055c371 ED |
1352 | struct uncached_list *ul = raw_cpu_ptr(&rt_uncached_list); |
1353 | ||
1354 | rt->rt_uncached_list = ul; | |
1355 | ||
1356 | spin_lock_bh(&ul->lock); | |
1357 | list_add_tail(&rt->rt_uncached, &ul->head); | |
1358 | spin_unlock_bh(&ul->lock); | |
caacf05e DM |
1359 | } |
1360 | ||
1361 | static void ipv4_dst_destroy(struct dst_entry *dst) | |
1362 | { | |
1363 | struct rtable *rt = (struct rtable *) dst; | |
1364 | ||
78df76a0 | 1365 | if (!list_empty(&rt->rt_uncached)) { |
5055c371 ED |
1366 | struct uncached_list *ul = rt->rt_uncached_list; |
1367 | ||
1368 | spin_lock_bh(&ul->lock); | |
caacf05e | 1369 | list_del(&rt->rt_uncached); |
5055c371 | 1370 | spin_unlock_bh(&ul->lock); |
caacf05e DM |
1371 | } |
1372 | } | |
1373 | ||
1374 | void rt_flush_dev(struct net_device *dev) | |
1375 | { | |
5055c371 ED |
1376 | struct net *net = dev_net(dev); |
1377 | struct rtable *rt; | |
1378 | int cpu; | |
1379 | ||
1380 | for_each_possible_cpu(cpu) { | |
1381 | struct uncached_list *ul = &per_cpu(rt_uncached_list, cpu); | |
caacf05e | 1382 | |
5055c371 ED |
1383 | spin_lock_bh(&ul->lock); |
1384 | list_for_each_entry(rt, &ul->head, rt_uncached) { | |
caacf05e DM |
1385 | if (rt->dst.dev != dev) |
1386 | continue; | |
1387 | rt->dst.dev = net->loopback_dev; | |
1388 | dev_hold(rt->dst.dev); | |
1389 | dev_put(dev); | |
1390 | } | |
5055c371 | 1391 | spin_unlock_bh(&ul->lock); |
4895c771 DM |
1392 | } |
1393 | } | |
1394 | ||
4331debc | 1395 | static bool rt_cache_valid(const struct rtable *rt) |
d2d68ba9 | 1396 | { |
4331debc ED |
1397 | return rt && |
1398 | rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK && | |
1399 | !rt_is_expired(rt); | |
d2d68ba9 DM |
1400 | } |
1401 | ||
f2bb4bed | 1402 | static void rt_set_nexthop(struct rtable *rt, __be32 daddr, |
5e2b61f7 | 1403 | const struct fib_result *res, |
f2bb4bed | 1404 | struct fib_nh_exception *fnhe, |
982721f3 | 1405 | struct fib_info *fi, u16 type, u32 itag) |
1da177e4 | 1406 | { |
caacf05e DM |
1407 | bool cached = false; |
1408 | ||
1da177e4 | 1409 | if (fi) { |
4895c771 DM |
1410 | struct fib_nh *nh = &FIB_RES_NH(*res); |
1411 | ||
155e8336 | 1412 | if (nh->nh_gw && nh->nh_scope == RT_SCOPE_LINK) { |
4895c771 | 1413 | rt->rt_gateway = nh->nh_gw; |
155e8336 JA |
1414 | rt->rt_uses_gateway = 1; |
1415 | } | |
2860583f | 1416 | dst_init_metrics(&rt->dst, fi->fib_metrics, true); |
c7066f70 | 1417 | #ifdef CONFIG_IP_ROUTE_CLASSID |
f2bb4bed | 1418 | rt->dst.tclassid = nh->nh_tclassid; |
1da177e4 | 1419 | #endif |
61adedf3 | 1420 | rt->dst.lwtstate = lwtstate_get(nh->nh_lwtstate); |
c5038a83 | 1421 | if (unlikely(fnhe)) |
caacf05e | 1422 | cached = rt_bind_exception(rt, fnhe, daddr); |
c5038a83 | 1423 | else if (!(rt->dst.flags & DST_NOCACHE)) |
caacf05e | 1424 | cached = rt_cache_route(nh, rt); |
155e8336 JA |
1425 | if (unlikely(!cached)) { |
1426 | /* Routes we intend to cache in nexthop exception or | |
1427 | * FIB nexthop have the DST_NOCACHE bit clear. | |
1428 | * However, if we are unsuccessful at storing this | |
1429 | * route into the cache we really need to set it. | |
1430 | */ | |
1431 | rt->dst.flags |= DST_NOCACHE; | |
1432 | if (!rt->rt_gateway) | |
1433 | rt->rt_gateway = daddr; | |
1434 | rt_add_uncached_list(rt); | |
1435 | } | |
1436 | } else | |
caacf05e | 1437 | rt_add_uncached_list(rt); |
defb3519 | 1438 | |
c7066f70 | 1439 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 | 1440 | #ifdef CONFIG_IP_MULTIPLE_TABLES |
85b91b03 | 1441 | set_class_tag(rt, res->tclassid); |
1da177e4 LT |
1442 | #endif |
1443 | set_class_tag(rt, itag); | |
1444 | #endif | |
1da177e4 LT |
1445 | } |
1446 | ||
9ab179d8 DA |
1447 | struct rtable *rt_dst_alloc(struct net_device *dev, |
1448 | unsigned int flags, u16 type, | |
1449 | bool nopolicy, bool noxfrm, bool will_cache) | |
0c4dcd58 | 1450 | { |
d08c4f35 DA |
1451 | struct rtable *rt; |
1452 | ||
1453 | rt = dst_alloc(&ipv4_dst_ops, dev, 1, DST_OBSOLETE_FORCE_CHK, | |
1454 | (will_cache ? 0 : (DST_HOST | DST_NOCACHE)) | | |
1455 | (nopolicy ? DST_NOPOLICY : 0) | | |
1456 | (noxfrm ? DST_NOXFRM : 0)); | |
1457 | ||
1458 | if (rt) { | |
1459 | rt->rt_genid = rt_genid_ipv4(dev_net(dev)); | |
1460 | rt->rt_flags = flags; | |
1461 | rt->rt_type = type; | |
1462 | rt->rt_is_input = 0; | |
1463 | rt->rt_iif = 0; | |
1464 | rt->rt_pmtu = 0; | |
1465 | rt->rt_gateway = 0; | |
1466 | rt->rt_uses_gateway = 0; | |
b7503e0c | 1467 | rt->rt_table_id = 0; |
d08c4f35 DA |
1468 | INIT_LIST_HEAD(&rt->rt_uncached); |
1469 | ||
1470 | rt->dst.output = ip_output; | |
1471 | if (flags & RTCF_LOCAL) | |
1472 | rt->dst.input = ip_local_deliver; | |
1473 | } | |
1474 | ||
1475 | return rt; | |
0c4dcd58 | 1476 | } |
9ab179d8 | 1477 | EXPORT_SYMBOL(rt_dst_alloc); |
0c4dcd58 | 1478 | |
96d36220 | 1479 | /* called in rcu_read_lock() section */ |
9e12bb22 | 1480 | static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
1481 | u8 tos, struct net_device *dev, int our) |
1482 | { | |
1da177e4 | 1483 | struct rtable *rth; |
96d36220 | 1484 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
d08c4f35 | 1485 | unsigned int flags = RTCF_MULTICAST; |
1da177e4 | 1486 | u32 itag = 0; |
b5f7e755 | 1487 | int err; |
1da177e4 LT |
1488 | |
1489 | /* Primary sanity checks. */ | |
1490 | ||
51456b29 | 1491 | if (!in_dev) |
1da177e4 LT |
1492 | return -EINVAL; |
1493 | ||
1e637c74 | 1494 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
d0daebc3 | 1495 | skb->protocol != htons(ETH_P_IP)) |
1da177e4 LT |
1496 | goto e_inval; |
1497 | ||
75fea73d AD |
1498 | if (ipv4_is_loopback(saddr) && !IN_DEV_ROUTE_LOCALNET(in_dev)) |
1499 | goto e_inval; | |
d0daebc3 | 1500 | |
f97c1e0c JP |
1501 | if (ipv4_is_zeronet(saddr)) { |
1502 | if (!ipv4_is_local_multicast(daddr)) | |
1da177e4 | 1503 | goto e_inval; |
b5f7e755 | 1504 | } else { |
9e56e380 DM |
1505 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, |
1506 | in_dev, &itag); | |
b5f7e755 ED |
1507 | if (err < 0) |
1508 | goto e_err; | |
1509 | } | |
d08c4f35 DA |
1510 | if (our) |
1511 | flags |= RTCF_LOCAL; | |
1512 | ||
1513 | rth = rt_dst_alloc(dev_net(dev)->loopback_dev, flags, RTN_MULTICAST, | |
f2bb4bed | 1514 | IN_DEV_CONF_GET(in_dev, NOPOLICY), false, false); |
1da177e4 LT |
1515 | if (!rth) |
1516 | goto e_nobufs; | |
1517 | ||
cf911662 DM |
1518 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1519 | rth->dst.tclassid = itag; | |
1520 | #endif | |
d8d1f30b | 1521 | rth->dst.output = ip_rt_bug; |
9917e1e8 | 1522 | rth->rt_is_input= 1; |
1da177e4 LT |
1523 | |
1524 | #ifdef CONFIG_IP_MROUTE | |
f97c1e0c | 1525 | if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev)) |
d8d1f30b | 1526 | rth->dst.input = ip_mr_input; |
1da177e4 LT |
1527 | #endif |
1528 | RT_CACHE_STAT_INC(in_slow_mc); | |
1529 | ||
89aef892 DM |
1530 | skb_dst_set(skb, &rth->dst); |
1531 | return 0; | |
1da177e4 LT |
1532 | |
1533 | e_nobufs: | |
1da177e4 | 1534 | return -ENOBUFS; |
1da177e4 | 1535 | e_inval: |
96d36220 | 1536 | return -EINVAL; |
b5f7e755 | 1537 | e_err: |
b5f7e755 | 1538 | return err; |
1da177e4 LT |
1539 | } |
1540 | ||
1541 | ||
1542 | static void ip_handle_martian_source(struct net_device *dev, | |
1543 | struct in_device *in_dev, | |
1544 | struct sk_buff *skb, | |
9e12bb22 AV |
1545 | __be32 daddr, |
1546 | __be32 saddr) | |
1da177e4 LT |
1547 | { |
1548 | RT_CACHE_STAT_INC(in_martian_src); | |
1549 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1550 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) { | |
1551 | /* | |
1552 | * RFC1812 recommendation, if source is martian, | |
1553 | * the only hint is MAC header. | |
1554 | */ | |
058bd4d2 | 1555 | pr_warn("martian source %pI4 from %pI4, on dev %s\n", |
673d57e7 | 1556 | &daddr, &saddr, dev->name); |
98e399f8 | 1557 | if (dev->hard_header_len && skb_mac_header_was_set(skb)) { |
058bd4d2 JP |
1558 | print_hex_dump(KERN_WARNING, "ll header: ", |
1559 | DUMP_PREFIX_OFFSET, 16, 1, | |
1560 | skb_mac_header(skb), | |
1561 | dev->hard_header_len, true); | |
1da177e4 LT |
1562 | } |
1563 | } | |
1564 | #endif | |
1565 | } | |
1566 | ||
deed49df XL |
1567 | static void ip_del_fnhe(struct fib_nh *nh, __be32 daddr) |
1568 | { | |
1569 | struct fnhe_hash_bucket *hash; | |
1570 | struct fib_nh_exception *fnhe, __rcu **fnhe_p; | |
1571 | u32 hval = fnhe_hashfun(daddr); | |
1572 | ||
1573 | spin_lock_bh(&fnhe_lock); | |
1574 | ||
1575 | hash = rcu_dereference_protected(nh->nh_exceptions, | |
1576 | lockdep_is_held(&fnhe_lock)); | |
1577 | hash += hval; | |
1578 | ||
1579 | fnhe_p = &hash->chain; | |
1580 | fnhe = rcu_dereference_protected(*fnhe_p, lockdep_is_held(&fnhe_lock)); | |
1581 | while (fnhe) { | |
1582 | if (fnhe->fnhe_daddr == daddr) { | |
1583 | rcu_assign_pointer(*fnhe_p, rcu_dereference_protected( | |
1584 | fnhe->fnhe_next, lockdep_is_held(&fnhe_lock))); | |
1585 | fnhe_flush_routes(fnhe); | |
1586 | kfree_rcu(fnhe, rcu); | |
1587 | break; | |
1588 | } | |
1589 | fnhe_p = &fnhe->fnhe_next; | |
1590 | fnhe = rcu_dereference_protected(fnhe->fnhe_next, | |
1591 | lockdep_is_held(&fnhe_lock)); | |
1592 | } | |
1593 | ||
1594 | spin_unlock_bh(&fnhe_lock); | |
1595 | } | |
1596 | ||
47360228 | 1597 | /* called in rcu_read_lock() section */ |
5969f71d | 1598 | static int __mkroute_input(struct sk_buff *skb, |
982721f3 | 1599 | const struct fib_result *res, |
5969f71d | 1600 | struct in_device *in_dev, |
c6cffba4 | 1601 | __be32 daddr, __be32 saddr, u32 tos) |
1da177e4 | 1602 | { |
2ffae99d | 1603 | struct fib_nh_exception *fnhe; |
1da177e4 LT |
1604 | struct rtable *rth; |
1605 | int err; | |
1606 | struct in_device *out_dev; | |
d2d68ba9 | 1607 | bool do_cache; |
fbdc0ad0 | 1608 | u32 itag = 0; |
1da177e4 LT |
1609 | |
1610 | /* get a working reference to the output device */ | |
47360228 | 1611 | out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res)); |
51456b29 | 1612 | if (!out_dev) { |
e87cc472 | 1613 | net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n"); |
1da177e4 LT |
1614 | return -EINVAL; |
1615 | } | |
1616 | ||
5c04c819 | 1617 | err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res), |
9e56e380 | 1618 | in_dev->dev, in_dev, &itag); |
1da177e4 | 1619 | if (err < 0) { |
e905a9ed | 1620 | ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr, |
1da177e4 | 1621 | saddr); |
e905a9ed | 1622 | |
1da177e4 LT |
1623 | goto cleanup; |
1624 | } | |
1625 | ||
e81da0e1 JA |
1626 | do_cache = res->fi && !itag; |
1627 | if (out_dev == in_dev && err && IN_DEV_TX_REDIRECTS(out_dev) && | |
df4d9254 | 1628 | skb->protocol == htons(ETH_P_IP) && |
1da177e4 | 1629 | (IN_DEV_SHARED_MEDIA(out_dev) || |
df4d9254 HFS |
1630 | inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) |
1631 | IPCB(skb)->flags |= IPSKB_DOREDIRECT; | |
1da177e4 LT |
1632 | |
1633 | if (skb->protocol != htons(ETH_P_IP)) { | |
1634 | /* Not IP (i.e. ARP). Do not create route, if it is | |
1635 | * invalid for proxy arp. DNAT routes are always valid. | |
65324144 JDB |
1636 | * |
1637 | * Proxy arp feature have been extended to allow, ARP | |
1638 | * replies back to the same interface, to support | |
1639 | * Private VLAN switch technologies. See arp.c. | |
1da177e4 | 1640 | */ |
65324144 JDB |
1641 | if (out_dev == in_dev && |
1642 | IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) { | |
1da177e4 LT |
1643 | err = -EINVAL; |
1644 | goto cleanup; | |
1645 | } | |
1646 | } | |
1647 | ||
2ffae99d | 1648 | fnhe = find_exception(&FIB_RES_NH(*res), daddr); |
e81da0e1 | 1649 | if (do_cache) { |
deed49df | 1650 | if (fnhe) { |
2ffae99d | 1651 | rth = rcu_dereference(fnhe->fnhe_rth_input); |
deed49df XL |
1652 | if (rth && rth->dst.expires && |
1653 | time_after(jiffies, rth->dst.expires)) { | |
1654 | ip_del_fnhe(&FIB_RES_NH(*res), daddr); | |
1655 | fnhe = NULL; | |
1656 | } else { | |
1657 | goto rt_cache; | |
1658 | } | |
1659 | } | |
1660 | ||
1661 | rth = rcu_dereference(FIB_RES_NH(*res).nh_rth_input); | |
2ffae99d | 1662 | |
deed49df | 1663 | rt_cache: |
e81da0e1 JA |
1664 | if (rt_cache_valid(rth)) { |
1665 | skb_dst_set_noref(skb, &rth->dst); | |
1666 | goto out; | |
d2d68ba9 DM |
1667 | } |
1668 | } | |
f2bb4bed | 1669 | |
d08c4f35 | 1670 | rth = rt_dst_alloc(out_dev->dev, 0, res->type, |
5c1e6aa3 | 1671 | IN_DEV_CONF_GET(in_dev, NOPOLICY), |
d2d68ba9 | 1672 | IN_DEV_CONF_GET(out_dev, NOXFRM), do_cache); |
1da177e4 LT |
1673 | if (!rth) { |
1674 | err = -ENOBUFS; | |
1675 | goto cleanup; | |
1676 | } | |
1677 | ||
9917e1e8 | 1678 | rth->rt_is_input = 1; |
b7503e0c DA |
1679 | if (res->table) |
1680 | rth->rt_table_id = res->table->tb_id; | |
a6254864 | 1681 | RT_CACHE_STAT_INC(in_slow_tot); |
1da177e4 | 1682 | |
d8d1f30b | 1683 | rth->dst.input = ip_forward; |
1da177e4 | 1684 | |
2ffae99d | 1685 | rt_set_nexthop(rth, daddr, res, fnhe, res->fi, res->type, itag); |
61adedf3 JB |
1686 | if (lwtunnel_output_redirect(rth->dst.lwtstate)) { |
1687 | rth->dst.lwtstate->orig_output = rth->dst.output; | |
8602a625 | 1688 | rth->dst.output = lwtunnel_output; |
25368623 | 1689 | } |
61adedf3 JB |
1690 | if (lwtunnel_input_redirect(rth->dst.lwtstate)) { |
1691 | rth->dst.lwtstate->orig_input = rth->dst.input; | |
25368623 TH |
1692 | rth->dst.input = lwtunnel_input; |
1693 | } | |
c6cffba4 | 1694 | skb_dst_set(skb, &rth->dst); |
d2d68ba9 | 1695 | out: |
1da177e4 LT |
1696 | err = 0; |
1697 | cleanup: | |
1da177e4 | 1698 | return err; |
e905a9ed | 1699 | } |
1da177e4 | 1700 | |
79a13159 PN |
1701 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
1702 | ||
1703 | /* To make ICMP packets follow the right flow, the multipath hash is | |
1704 | * calculated from the inner IP addresses in reverse order. | |
1705 | */ | |
1706 | static int ip_multipath_icmp_hash(struct sk_buff *skb) | |
1707 | { | |
1708 | const struct iphdr *outer_iph = ip_hdr(skb); | |
1709 | struct icmphdr _icmph; | |
1710 | const struct icmphdr *icmph; | |
1711 | struct iphdr _inner_iph; | |
1712 | const struct iphdr *inner_iph; | |
1713 | ||
1714 | if (unlikely((outer_iph->frag_off & htons(IP_OFFSET)) != 0)) | |
1715 | goto standard_hash; | |
1716 | ||
1717 | icmph = skb_header_pointer(skb, outer_iph->ihl * 4, sizeof(_icmph), | |
1718 | &_icmph); | |
1719 | if (!icmph) | |
1720 | goto standard_hash; | |
1721 | ||
1722 | if (icmph->type != ICMP_DEST_UNREACH && | |
1723 | icmph->type != ICMP_REDIRECT && | |
1724 | icmph->type != ICMP_TIME_EXCEEDED && | |
1725 | icmph->type != ICMP_PARAMETERPROB) { | |
1726 | goto standard_hash; | |
1727 | } | |
1728 | ||
1729 | inner_iph = skb_header_pointer(skb, | |
1730 | outer_iph->ihl * 4 + sizeof(_icmph), | |
1731 | sizeof(_inner_iph), &_inner_iph); | |
1732 | if (!inner_iph) | |
1733 | goto standard_hash; | |
1734 | ||
1735 | return fib_multipath_hash(inner_iph->daddr, inner_iph->saddr); | |
1736 | ||
1737 | standard_hash: | |
1738 | return fib_multipath_hash(outer_iph->saddr, outer_iph->daddr); | |
1739 | } | |
1740 | ||
1741 | #endif /* CONFIG_IP_ROUTE_MULTIPATH */ | |
1742 | ||
5969f71d SH |
1743 | static int ip_mkroute_input(struct sk_buff *skb, |
1744 | struct fib_result *res, | |
68a5e3dd | 1745 | const struct flowi4 *fl4, |
5969f71d SH |
1746 | struct in_device *in_dev, |
1747 | __be32 daddr, __be32 saddr, u32 tos) | |
1da177e4 | 1748 | { |
1da177e4 | 1749 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
0e884c78 PN |
1750 | if (res->fi && res->fi->fib_nhs > 1) { |
1751 | int h; | |
1752 | ||
79a13159 PN |
1753 | if (unlikely(ip_hdr(skb)->protocol == IPPROTO_ICMP)) |
1754 | h = ip_multipath_icmp_hash(skb); | |
1755 | else | |
1756 | h = fib_multipath_hash(saddr, daddr); | |
0e884c78 PN |
1757 | fib_select_multipath(res, h); |
1758 | } | |
1da177e4 LT |
1759 | #endif |
1760 | ||
1761 | /* create a routing cache entry */ | |
c6cffba4 | 1762 | return __mkroute_input(skb, res, in_dev, daddr, saddr, tos); |
1da177e4 LT |
1763 | } |
1764 | ||
1da177e4 LT |
1765 | /* |
1766 | * NOTE. We drop all the packets that has local source | |
1767 | * addresses, because every properly looped back packet | |
1768 | * must have correct destination already attached by output routine. | |
1769 | * | |
1770 | * Such approach solves two big problems: | |
1771 | * 1. Not simplex devices are handled properly. | |
1772 | * 2. IP spoofing attempts are filtered with 100% of guarantee. | |
ebc0ffae | 1773 | * called with rcu_read_lock() |
1da177e4 LT |
1774 | */ |
1775 | ||
9e12bb22 | 1776 | static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
c10237e0 | 1777 | u8 tos, struct net_device *dev) |
1da177e4 LT |
1778 | { |
1779 | struct fib_result res; | |
96d36220 | 1780 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1b7179d3 | 1781 | struct ip_tunnel_info *tun_info; |
68a5e3dd | 1782 | struct flowi4 fl4; |
95c96174 | 1783 | unsigned int flags = 0; |
1da177e4 | 1784 | u32 itag = 0; |
95c96174 | 1785 | struct rtable *rth; |
1da177e4 | 1786 | int err = -EINVAL; |
5e73ea1a | 1787 | struct net *net = dev_net(dev); |
d2d68ba9 | 1788 | bool do_cache; |
1da177e4 LT |
1789 | |
1790 | /* IP on this device is disabled. */ | |
1791 | ||
1792 | if (!in_dev) | |
1793 | goto out; | |
1794 | ||
1795 | /* Check for the most weird martians, which can be not detected | |
1796 | by fib_lookup. | |
1797 | */ | |
1798 | ||
61adedf3 | 1799 | tun_info = skb_tunnel_info(skb); |
46fa062a | 1800 | if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX)) |
1b7179d3 TG |
1801 | fl4.flowi4_tun_key.tun_id = tun_info->key.tun_id; |
1802 | else | |
1803 | fl4.flowi4_tun_key.tun_id = 0; | |
f38a9eb1 TG |
1804 | skb_dst_drop(skb); |
1805 | ||
d0daebc3 | 1806 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr)) |
1da177e4 LT |
1807 | goto martian_source; |
1808 | ||
d2d68ba9 | 1809 | res.fi = NULL; |
bde6f9de | 1810 | res.table = NULL; |
27a954bd | 1811 | if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0)) |
1da177e4 LT |
1812 | goto brd_input; |
1813 | ||
1814 | /* Accept zero addresses only to limited broadcast; | |
1815 | * I even do not know to fix it or not. Waiting for complains :-) | |
1816 | */ | |
f97c1e0c | 1817 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
1818 | goto martian_source; |
1819 | ||
d0daebc3 | 1820 | if (ipv4_is_zeronet(daddr)) |
1da177e4 LT |
1821 | goto martian_destination; |
1822 | ||
9eb43e76 ED |
1823 | /* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(), |
1824 | * and call it once if daddr or/and saddr are loopback addresses | |
1825 | */ | |
1826 | if (ipv4_is_loopback(daddr)) { | |
1827 | if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net)) | |
d0daebc3 | 1828 | goto martian_destination; |
9eb43e76 ED |
1829 | } else if (ipv4_is_loopback(saddr)) { |
1830 | if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net)) | |
d0daebc3 TG |
1831 | goto martian_source; |
1832 | } | |
1833 | ||
1da177e4 LT |
1834 | /* |
1835 | * Now we are ready to route packet. | |
1836 | */ | |
68a5e3dd | 1837 | fl4.flowi4_oif = 0; |
385add90 | 1838 | fl4.flowi4_iif = l3mdev_fib_oif_rcu(dev); |
68a5e3dd DM |
1839 | fl4.flowi4_mark = skb->mark; |
1840 | fl4.flowi4_tos = tos; | |
1841 | fl4.flowi4_scope = RT_SCOPE_UNIVERSE; | |
b84f7878 | 1842 | fl4.flowi4_flags = 0; |
68a5e3dd DM |
1843 | fl4.daddr = daddr; |
1844 | fl4.saddr = saddr; | |
0eeb075f | 1845 | err = fib_lookup(net, &fl4, &res, 0); |
cd0f0b95 DJ |
1846 | if (err != 0) { |
1847 | if (!IN_DEV_FORWARD(in_dev)) | |
1848 | err = -EHOSTUNREACH; | |
1da177e4 | 1849 | goto no_route; |
cd0f0b95 | 1850 | } |
1da177e4 LT |
1851 | |
1852 | if (res.type == RTN_BROADCAST) | |
1853 | goto brd_input; | |
1854 | ||
1855 | if (res.type == RTN_LOCAL) { | |
5c04c819 | 1856 | err = fib_validate_source(skb, saddr, daddr, tos, |
0d5edc68 | 1857 | 0, dev, in_dev, &itag); |
b5f7e755 | 1858 | if (err < 0) |
0d753960 | 1859 | goto martian_source; |
1da177e4 LT |
1860 | goto local_input; |
1861 | } | |
1862 | ||
cd0f0b95 DJ |
1863 | if (!IN_DEV_FORWARD(in_dev)) { |
1864 | err = -EHOSTUNREACH; | |
251da413 | 1865 | goto no_route; |
cd0f0b95 | 1866 | } |
1da177e4 LT |
1867 | if (res.type != RTN_UNICAST) |
1868 | goto martian_destination; | |
1869 | ||
68a5e3dd | 1870 | err = ip_mkroute_input(skb, &res, &fl4, in_dev, daddr, saddr, tos); |
1da177e4 LT |
1871 | out: return err; |
1872 | ||
1873 | brd_input: | |
1874 | if (skb->protocol != htons(ETH_P_IP)) | |
1875 | goto e_inval; | |
1876 | ||
41347dcd | 1877 | if (!ipv4_is_zeronet(saddr)) { |
9e56e380 DM |
1878 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, |
1879 | in_dev, &itag); | |
1da177e4 | 1880 | if (err < 0) |
0d753960 | 1881 | goto martian_source; |
1da177e4 LT |
1882 | } |
1883 | flags |= RTCF_BROADCAST; | |
1884 | res.type = RTN_BROADCAST; | |
1885 | RT_CACHE_STAT_INC(in_brd); | |
1886 | ||
1887 | local_input: | |
d2d68ba9 DM |
1888 | do_cache = false; |
1889 | if (res.fi) { | |
fe3edf45 | 1890 | if (!itag) { |
54764bb6 | 1891 | rth = rcu_dereference(FIB_RES_NH(res).nh_rth_input); |
d2d68ba9 | 1892 | if (rt_cache_valid(rth)) { |
c6cffba4 DM |
1893 | skb_dst_set_noref(skb, &rth->dst); |
1894 | err = 0; | |
1895 | goto out; | |
d2d68ba9 DM |
1896 | } |
1897 | do_cache = true; | |
1898 | } | |
1899 | } | |
1900 | ||
d08c4f35 | 1901 | rth = rt_dst_alloc(net->loopback_dev, flags | RTCF_LOCAL, res.type, |
d2d68ba9 | 1902 | IN_DEV_CONF_GET(in_dev, NOPOLICY), false, do_cache); |
1da177e4 LT |
1903 | if (!rth) |
1904 | goto e_nobufs; | |
1905 | ||
d8d1f30b | 1906 | rth->dst.output= ip_rt_bug; |
cf911662 DM |
1907 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1908 | rth->dst.tclassid = itag; | |
1909 | #endif | |
9917e1e8 | 1910 | rth->rt_is_input = 1; |
b7503e0c DA |
1911 | if (res.table) |
1912 | rth->rt_table_id = res.table->tb_id; | |
571e7226 | 1913 | |
a6254864 | 1914 | RT_CACHE_STAT_INC(in_slow_tot); |
1da177e4 | 1915 | if (res.type == RTN_UNREACHABLE) { |
d8d1f30b CG |
1916 | rth->dst.input= ip_error; |
1917 | rth->dst.error= -err; | |
1da177e4 LT |
1918 | rth->rt_flags &= ~RTCF_LOCAL; |
1919 | } | |
dcdfdf56 AS |
1920 | if (do_cache) { |
1921 | if (unlikely(!rt_cache_route(&FIB_RES_NH(res), rth))) { | |
1922 | rth->dst.flags |= DST_NOCACHE; | |
1923 | rt_add_uncached_list(rth); | |
1924 | } | |
1925 | } | |
89aef892 | 1926 | skb_dst_set(skb, &rth->dst); |
b23dd4fe | 1927 | err = 0; |
ebc0ffae | 1928 | goto out; |
1da177e4 LT |
1929 | |
1930 | no_route: | |
1931 | RT_CACHE_STAT_INC(in_no_route); | |
1da177e4 | 1932 | res.type = RTN_UNREACHABLE; |
fa19c2b0 | 1933 | res.fi = NULL; |
bde6f9de | 1934 | res.table = NULL; |
1da177e4 LT |
1935 | goto local_input; |
1936 | ||
1937 | /* | |
1938 | * Do not cache martian addresses: they should be logged (RFC1812) | |
1939 | */ | |
1940 | martian_destination: | |
1941 | RT_CACHE_STAT_INC(in_martian_dst); | |
1942 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
e87cc472 JP |
1943 | if (IN_DEV_LOG_MARTIANS(in_dev)) |
1944 | net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n", | |
1945 | &daddr, &saddr, dev->name); | |
1da177e4 | 1946 | #endif |
2c2910a4 | 1947 | |
1da177e4 LT |
1948 | e_inval: |
1949 | err = -EINVAL; | |
ebc0ffae | 1950 | goto out; |
1da177e4 LT |
1951 | |
1952 | e_nobufs: | |
1953 | err = -ENOBUFS; | |
ebc0ffae | 1954 | goto out; |
1da177e4 LT |
1955 | |
1956 | martian_source: | |
1957 | ip_handle_martian_source(dev, in_dev, skb, daddr, saddr); | |
ebc0ffae | 1958 | goto out; |
1da177e4 LT |
1959 | } |
1960 | ||
c6cffba4 DM |
1961 | int ip_route_input_noref(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1962 | u8 tos, struct net_device *dev) | |
1da177e4 | 1963 | { |
96d36220 | 1964 | int res; |
1da177e4 | 1965 | |
96d36220 ED |
1966 | rcu_read_lock(); |
1967 | ||
1da177e4 LT |
1968 | /* Multicast recognition logic is moved from route cache to here. |
1969 | The problem was that too many Ethernet cards have broken/missing | |
1970 | hardware multicast filters :-( As result the host on multicasting | |
1971 | network acquires a lot of useless route cache entries, sort of | |
1972 | SDR messages from all the world. Now we try to get rid of them. | |
1973 | Really, provided software IP multicast filter is organized | |
1974 | reasonably (at least, hashed), it does not result in a slowdown | |
1975 | comparing with route cache reject entries. | |
1976 | Note, that multicast routers are not affected, because | |
1977 | route cache entry is created eventually. | |
1978 | */ | |
f97c1e0c | 1979 | if (ipv4_is_multicast(daddr)) { |
96d36220 | 1980 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1981 | |
96d36220 | 1982 | if (in_dev) { |
dbdd9a52 DM |
1983 | int our = ip_check_mc_rcu(in_dev, daddr, saddr, |
1984 | ip_hdr(skb)->protocol); | |
1da177e4 LT |
1985 | if (our |
1986 | #ifdef CONFIG_IP_MROUTE | |
9d4fb27d JP |
1987 | || |
1988 | (!ipv4_is_local_multicast(daddr) && | |
1989 | IN_DEV_MFORWARD(in_dev)) | |
1da177e4 | 1990 | #endif |
9d4fb27d | 1991 | ) { |
96d36220 ED |
1992 | int res = ip_route_input_mc(skb, daddr, saddr, |
1993 | tos, dev, our); | |
1da177e4 | 1994 | rcu_read_unlock(); |
96d36220 | 1995 | return res; |
1da177e4 LT |
1996 | } |
1997 | } | |
1998 | rcu_read_unlock(); | |
1999 | return -EINVAL; | |
2000 | } | |
c10237e0 | 2001 | res = ip_route_input_slow(skb, daddr, saddr, tos, dev); |
96d36220 ED |
2002 | rcu_read_unlock(); |
2003 | return res; | |
1da177e4 | 2004 | } |
c6cffba4 | 2005 | EXPORT_SYMBOL(ip_route_input_noref); |
1da177e4 | 2006 | |
ebc0ffae | 2007 | /* called with rcu_read_lock() */ |
982721f3 | 2008 | static struct rtable *__mkroute_output(const struct fib_result *res, |
1a00fee4 | 2009 | const struct flowi4 *fl4, int orig_oif, |
f61759e6 | 2010 | struct net_device *dev_out, |
5ada5527 | 2011 | unsigned int flags) |
1da177e4 | 2012 | { |
982721f3 | 2013 | struct fib_info *fi = res->fi; |
f2bb4bed | 2014 | struct fib_nh_exception *fnhe; |
5ada5527 | 2015 | struct in_device *in_dev; |
982721f3 | 2016 | u16 type = res->type; |
5ada5527 | 2017 | struct rtable *rth; |
c92b9655 | 2018 | bool do_cache; |
1da177e4 | 2019 | |
d0daebc3 TG |
2020 | in_dev = __in_dev_get_rcu(dev_out); |
2021 | if (!in_dev) | |
5ada5527 | 2022 | return ERR_PTR(-EINVAL); |
1da177e4 | 2023 | |
d0daebc3 TG |
2024 | if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev))) |
2025 | if (ipv4_is_loopback(fl4->saddr) && !(dev_out->flags & IFF_LOOPBACK)) | |
2026 | return ERR_PTR(-EINVAL); | |
2027 | ||
68a5e3dd | 2028 | if (ipv4_is_lbcast(fl4->daddr)) |
982721f3 | 2029 | type = RTN_BROADCAST; |
68a5e3dd | 2030 | else if (ipv4_is_multicast(fl4->daddr)) |
982721f3 | 2031 | type = RTN_MULTICAST; |
68a5e3dd | 2032 | else if (ipv4_is_zeronet(fl4->daddr)) |
5ada5527 | 2033 | return ERR_PTR(-EINVAL); |
1da177e4 LT |
2034 | |
2035 | if (dev_out->flags & IFF_LOOPBACK) | |
2036 | flags |= RTCF_LOCAL; | |
2037 | ||
63617421 | 2038 | do_cache = true; |
982721f3 | 2039 | if (type == RTN_BROADCAST) { |
1da177e4 | 2040 | flags |= RTCF_BROADCAST | RTCF_LOCAL; |
982721f3 DM |
2041 | fi = NULL; |
2042 | } else if (type == RTN_MULTICAST) { | |
dd28d1a0 | 2043 | flags |= RTCF_MULTICAST | RTCF_LOCAL; |
813b3b5d DM |
2044 | if (!ip_check_mc_rcu(in_dev, fl4->daddr, fl4->saddr, |
2045 | fl4->flowi4_proto)) | |
1da177e4 | 2046 | flags &= ~RTCF_LOCAL; |
63617421 JA |
2047 | else |
2048 | do_cache = false; | |
1da177e4 | 2049 | /* If multicast route do not exist use |
dd28d1a0 ED |
2050 | * default one, but do not gateway in this case. |
2051 | * Yes, it is hack. | |
1da177e4 | 2052 | */ |
982721f3 DM |
2053 | if (fi && res->prefixlen < 4) |
2054 | fi = NULL; | |
d6d5e999 CF |
2055 | } else if ((type == RTN_LOCAL) && (orig_oif != 0) && |
2056 | (orig_oif != dev_out->ifindex)) { | |
2057 | /* For local routes that require a particular output interface | |
2058 | * we do not want to cache the result. Caching the result | |
2059 | * causes incorrect behaviour when there are multiple source | |
2060 | * addresses on the interface, the end result being that if the | |
2061 | * intended recipient is waiting on that interface for the | |
2062 | * packet he won't receive it because it will be delivered on | |
2063 | * the loopback interface and the IP_PKTINFO ipi_ifindex will | |
2064 | * be set to the loopback interface as well. | |
2065 | */ | |
2066 | fi = NULL; | |
1da177e4 LT |
2067 | } |
2068 | ||
f2bb4bed | 2069 | fnhe = NULL; |
63617421 JA |
2070 | do_cache &= fi != NULL; |
2071 | if (do_cache) { | |
c5038a83 | 2072 | struct rtable __rcu **prth; |
c92b9655 | 2073 | struct fib_nh *nh = &FIB_RES_NH(*res); |
d26b3a7c | 2074 | |
c92b9655 | 2075 | fnhe = find_exception(nh, fl4->daddr); |
deed49df | 2076 | if (fnhe) { |
2ffae99d | 2077 | prth = &fnhe->fnhe_rth_output; |
deed49df XL |
2078 | rth = rcu_dereference(*prth); |
2079 | if (rth && rth->dst.expires && | |
2080 | time_after(jiffies, rth->dst.expires)) { | |
2081 | ip_del_fnhe(nh, fl4->daddr); | |
2082 | fnhe = NULL; | |
2083 | } else { | |
2084 | goto rt_cache; | |
c92b9655 | 2085 | } |
c92b9655 | 2086 | } |
deed49df XL |
2087 | |
2088 | if (unlikely(fl4->flowi4_flags & | |
2089 | FLOWI_FLAG_KNOWN_NH && | |
2090 | !(nh->nh_gw && | |
2091 | nh->nh_scope == RT_SCOPE_LINK))) { | |
2092 | do_cache = false; | |
2093 | goto add; | |
2094 | } | |
2095 | prth = raw_cpu_ptr(nh->nh_pcpu_rth_output); | |
c5038a83 | 2096 | rth = rcu_dereference(*prth); |
deed49df XL |
2097 | |
2098 | rt_cache: | |
c5038a83 DM |
2099 | if (rt_cache_valid(rth)) { |
2100 | dst_hold(&rth->dst); | |
2101 | return rth; | |
f2bb4bed DM |
2102 | } |
2103 | } | |
c92b9655 JA |
2104 | |
2105 | add: | |
d08c4f35 | 2106 | rth = rt_dst_alloc(dev_out, flags, type, |
5c1e6aa3 | 2107 | IN_DEV_CONF_GET(in_dev, NOPOLICY), |
f2bb4bed | 2108 | IN_DEV_CONF_GET(in_dev, NOXFRM), |
c92b9655 | 2109 | do_cache); |
8391d07b | 2110 | if (!rth) |
5ada5527 | 2111 | return ERR_PTR(-ENOBUFS); |
8391d07b | 2112 | |
13378cad | 2113 | rth->rt_iif = orig_oif ? : 0; |
b7503e0c DA |
2114 | if (res->table) |
2115 | rth->rt_table_id = res->table->tb_id; | |
2116 | ||
1da177e4 LT |
2117 | RT_CACHE_STAT_INC(out_slow_tot); |
2118 | ||
1da177e4 | 2119 | if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { |
e905a9ed | 2120 | if (flags & RTCF_LOCAL && |
1da177e4 | 2121 | !(dev_out->flags & IFF_LOOPBACK)) { |
d8d1f30b | 2122 | rth->dst.output = ip_mc_output; |
1da177e4 LT |
2123 | RT_CACHE_STAT_INC(out_slow_mc); |
2124 | } | |
2125 | #ifdef CONFIG_IP_MROUTE | |
982721f3 | 2126 | if (type == RTN_MULTICAST) { |
1da177e4 | 2127 | if (IN_DEV_MFORWARD(in_dev) && |
813b3b5d | 2128 | !ipv4_is_local_multicast(fl4->daddr)) { |
d8d1f30b CG |
2129 | rth->dst.input = ip_mr_input; |
2130 | rth->dst.output = ip_mc_output; | |
1da177e4 LT |
2131 | } |
2132 | } | |
2133 | #endif | |
2134 | } | |
2135 | ||
f2bb4bed | 2136 | rt_set_nexthop(rth, fl4->daddr, res, fnhe, fi, type, 0); |
61adedf3 | 2137 | if (lwtunnel_output_redirect(rth->dst.lwtstate)) |
0335f5b5 | 2138 | rth->dst.output = lwtunnel_output; |
1da177e4 | 2139 | |
5ada5527 | 2140 | return rth; |
1da177e4 LT |
2141 | } |
2142 | ||
1da177e4 LT |
2143 | /* |
2144 | * Major route resolver routine. | |
2145 | */ | |
2146 | ||
79a13159 PN |
2147 | struct rtable *__ip_route_output_key_hash(struct net *net, struct flowi4 *fl4, |
2148 | int mp_hash) | |
1da177e4 | 2149 | { |
1da177e4 | 2150 | struct net_device *dev_out = NULL; |
f61759e6 | 2151 | __u8 tos = RT_FL_TOS(fl4); |
813b3b5d DM |
2152 | unsigned int flags = 0; |
2153 | struct fib_result res; | |
5ada5527 | 2154 | struct rtable *rth; |
1ff23bee | 2155 | int master_idx; |
813b3b5d | 2156 | int orig_oif; |
0315e382 | 2157 | int err = -ENETUNREACH; |
1da177e4 | 2158 | |
85b91b03 | 2159 | res.tclassid = 0; |
1da177e4 | 2160 | res.fi = NULL; |
8b96d22d | 2161 | res.table = NULL; |
1da177e4 | 2162 | |
813b3b5d DM |
2163 | orig_oif = fl4->flowi4_oif; |
2164 | ||
1ff23bee DA |
2165 | master_idx = l3mdev_master_ifindex_by_index(net, fl4->flowi4_oif); |
2166 | if (master_idx) | |
2167 | fl4->flowi4_oif = master_idx; | |
1fb9489b | 2168 | fl4->flowi4_iif = LOOPBACK_IFINDEX; |
813b3b5d DM |
2169 | fl4->flowi4_tos = tos & IPTOS_RT_MASK; |
2170 | fl4->flowi4_scope = ((tos & RTO_ONLINK) ? | |
2171 | RT_SCOPE_LINK : RT_SCOPE_UNIVERSE); | |
44713b67 | 2172 | |
010c2708 | 2173 | rcu_read_lock(); |
813b3b5d | 2174 | if (fl4->saddr) { |
b23dd4fe | 2175 | rth = ERR_PTR(-EINVAL); |
813b3b5d DM |
2176 | if (ipv4_is_multicast(fl4->saddr) || |
2177 | ipv4_is_lbcast(fl4->saddr) || | |
2178 | ipv4_is_zeronet(fl4->saddr)) | |
1da177e4 LT |
2179 | goto out; |
2180 | ||
1da177e4 LT |
2181 | /* I removed check for oif == dev_out->oif here. |
2182 | It was wrong for two reasons: | |
1ab35276 DL |
2183 | 1. ip_dev_find(net, saddr) can return wrong iface, if saddr |
2184 | is assigned to multiple interfaces. | |
1da177e4 LT |
2185 | 2. Moreover, we are allowed to send packets with saddr |
2186 | of another iface. --ANK | |
2187 | */ | |
2188 | ||
813b3b5d DM |
2189 | if (fl4->flowi4_oif == 0 && |
2190 | (ipv4_is_multicast(fl4->daddr) || | |
2191 | ipv4_is_lbcast(fl4->daddr))) { | |
a210d01a | 2192 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
813b3b5d | 2193 | dev_out = __ip_dev_find(net, fl4->saddr, false); |
51456b29 | 2194 | if (!dev_out) |
a210d01a JA |
2195 | goto out; |
2196 | ||
1da177e4 LT |
2197 | /* Special hack: user can direct multicasts |
2198 | and limited broadcast via necessary interface | |
2199 | without fiddling with IP_MULTICAST_IF or IP_PKTINFO. | |
2200 | This hack is not just for fun, it allows | |
2201 | vic,vat and friends to work. | |
2202 | They bind socket to loopback, set ttl to zero | |
2203 | and expect that it will work. | |
2204 | From the viewpoint of routing cache they are broken, | |
2205 | because we are not allowed to build multicast path | |
2206 | with loopback source addr (look, routing cache | |
2207 | cannot know, that ttl is zero, so that packet | |
2208 | will not leave this host and route is valid). | |
2209 | Luckily, this hack is good workaround. | |
2210 | */ | |
2211 | ||
813b3b5d | 2212 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2213 | goto make_route; |
2214 | } | |
a210d01a | 2215 | |
813b3b5d | 2216 | if (!(fl4->flowi4_flags & FLOWI_FLAG_ANYSRC)) { |
a210d01a | 2217 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
813b3b5d | 2218 | if (!__ip_dev_find(net, fl4->saddr, false)) |
a210d01a | 2219 | goto out; |
a210d01a | 2220 | } |
1da177e4 LT |
2221 | } |
2222 | ||
2223 | ||
813b3b5d DM |
2224 | if (fl4->flowi4_oif) { |
2225 | dev_out = dev_get_by_index_rcu(net, fl4->flowi4_oif); | |
b23dd4fe | 2226 | rth = ERR_PTR(-ENODEV); |
51456b29 | 2227 | if (!dev_out) |
1da177e4 | 2228 | goto out; |
e5ed6399 HX |
2229 | |
2230 | /* RACE: Check return value of inet_select_addr instead. */ | |
fc75fc83 | 2231 | if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) { |
b23dd4fe | 2232 | rth = ERR_PTR(-ENETUNREACH); |
fc75fc83 ED |
2233 | goto out; |
2234 | } | |
813b3b5d | 2235 | if (ipv4_is_local_multicast(fl4->daddr) || |
6a211654 AL |
2236 | ipv4_is_lbcast(fl4->daddr) || |
2237 | fl4->flowi4_proto == IPPROTO_IGMP) { | |
813b3b5d DM |
2238 | if (!fl4->saddr) |
2239 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2240 | RT_SCOPE_LINK); | |
1da177e4 LT |
2241 | goto make_route; |
2242 | } | |
0a7e2260 | 2243 | if (!fl4->saddr) { |
813b3b5d DM |
2244 | if (ipv4_is_multicast(fl4->daddr)) |
2245 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2246 | fl4->flowi4_scope); | |
2247 | else if (!fl4->daddr) | |
2248 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2249 | RT_SCOPE_HOST); | |
1da177e4 | 2250 | } |
8e1ed705 DA |
2251 | |
2252 | rth = l3mdev_get_rtable(dev_out, fl4); | |
2253 | if (rth) | |
613d09b3 | 2254 | goto out; |
1da177e4 LT |
2255 | } |
2256 | ||
813b3b5d DM |
2257 | if (!fl4->daddr) { |
2258 | fl4->daddr = fl4->saddr; | |
2259 | if (!fl4->daddr) | |
2260 | fl4->daddr = fl4->saddr = htonl(INADDR_LOOPBACK); | |
b40afd0e | 2261 | dev_out = net->loopback_dev; |
1fb9489b | 2262 | fl4->flowi4_oif = LOOPBACK_IFINDEX; |
1da177e4 LT |
2263 | res.type = RTN_LOCAL; |
2264 | flags |= RTCF_LOCAL; | |
2265 | goto make_route; | |
2266 | } | |
2267 | ||
0315e382 NF |
2268 | err = fib_lookup(net, fl4, &res, 0); |
2269 | if (err) { | |
1da177e4 | 2270 | res.fi = NULL; |
8b96d22d | 2271 | res.table = NULL; |
28335a74 DA |
2272 | if (fl4->flowi4_oif && |
2273 | !netif_index_is_l3_master(net, fl4->flowi4_oif)) { | |
1da177e4 LT |
2274 | /* Apparently, routing tables are wrong. Assume, |
2275 | that the destination is on link. | |
2276 | ||
2277 | WHY? DW. | |
2278 | Because we are allowed to send to iface | |
2279 | even if it has NO routes and NO assigned | |
2280 | addresses. When oif is specified, routing | |
2281 | tables are looked up with only one purpose: | |
2282 | to catch if destination is gatewayed, rather than | |
2283 | direct. Moreover, if MSG_DONTROUTE is set, | |
2284 | we send packet, ignoring both routing tables | |
2285 | and ifaddr state. --ANK | |
2286 | ||
2287 | ||
2288 | We could make it even if oif is unknown, | |
2289 | likely IPv6, but we do not. | |
2290 | */ | |
2291 | ||
813b3b5d DM |
2292 | if (fl4->saddr == 0) |
2293 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2294 | RT_SCOPE_LINK); | |
1da177e4 LT |
2295 | res.type = RTN_UNICAST; |
2296 | goto make_route; | |
2297 | } | |
0315e382 | 2298 | rth = ERR_PTR(err); |
1da177e4 LT |
2299 | goto out; |
2300 | } | |
1da177e4 LT |
2301 | |
2302 | if (res.type == RTN_LOCAL) { | |
813b3b5d | 2303 | if (!fl4->saddr) { |
9fc3bbb4 | 2304 | if (res.fi->fib_prefsrc) |
813b3b5d | 2305 | fl4->saddr = res.fi->fib_prefsrc; |
9fc3bbb4 | 2306 | else |
813b3b5d | 2307 | fl4->saddr = fl4->daddr; |
9fc3bbb4 | 2308 | } |
b40afd0e | 2309 | dev_out = net->loopback_dev; |
813b3b5d | 2310 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2311 | flags |= RTCF_LOCAL; |
2312 | goto make_route; | |
2313 | } | |
2314 | ||
3ce58d84 | 2315 | fib_select_path(net, &res, fl4, mp_hash); |
1da177e4 | 2316 | |
1da177e4 | 2317 | dev_out = FIB_RES_DEV(res); |
813b3b5d | 2318 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2319 | |
2320 | ||
2321 | make_route: | |
1a00fee4 | 2322 | rth = __mkroute_output(&res, fl4, orig_oif, dev_out, flags); |
1da177e4 | 2323 | |
010c2708 DM |
2324 | out: |
2325 | rcu_read_unlock(); | |
b23dd4fe | 2326 | return rth; |
1da177e4 | 2327 | } |
79a13159 | 2328 | EXPORT_SYMBOL_GPL(__ip_route_output_key_hash); |
d8c97a94 | 2329 | |
ae2688d5 JW |
2330 | static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie) |
2331 | { | |
2332 | return NULL; | |
2333 | } | |
2334 | ||
ebb762f2 | 2335 | static unsigned int ipv4_blackhole_mtu(const struct dst_entry *dst) |
ec831ea7 | 2336 | { |
618f9bc7 SK |
2337 | unsigned int mtu = dst_metric_raw(dst, RTAX_MTU); |
2338 | ||
2339 | return mtu ? : dst->dev->mtu; | |
ec831ea7 RD |
2340 | } |
2341 | ||
6700c270 DM |
2342 | static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk, |
2343 | struct sk_buff *skb, u32 mtu) | |
14e50e57 DM |
2344 | { |
2345 | } | |
2346 | ||
6700c270 DM |
2347 | static void ipv4_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk, |
2348 | struct sk_buff *skb) | |
b587ee3b DM |
2349 | { |
2350 | } | |
2351 | ||
0972ddb2 HB |
2352 | static u32 *ipv4_rt_blackhole_cow_metrics(struct dst_entry *dst, |
2353 | unsigned long old) | |
2354 | { | |
2355 | return NULL; | |
2356 | } | |
2357 | ||
14e50e57 DM |
2358 | static struct dst_ops ipv4_dst_blackhole_ops = { |
2359 | .family = AF_INET, | |
ae2688d5 | 2360 | .check = ipv4_blackhole_dst_check, |
ebb762f2 | 2361 | .mtu = ipv4_blackhole_mtu, |
214f45c9 | 2362 | .default_advmss = ipv4_default_advmss, |
14e50e57 | 2363 | .update_pmtu = ipv4_rt_blackhole_update_pmtu, |
b587ee3b | 2364 | .redirect = ipv4_rt_blackhole_redirect, |
0972ddb2 | 2365 | .cow_metrics = ipv4_rt_blackhole_cow_metrics, |
d3aaeb38 | 2366 | .neigh_lookup = ipv4_neigh_lookup, |
14e50e57 DM |
2367 | }; |
2368 | ||
2774c131 | 2369 | struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig) |
14e50e57 | 2370 | { |
2774c131 | 2371 | struct rtable *ort = (struct rtable *) dst_orig; |
f5b0a874 | 2372 | struct rtable *rt; |
14e50e57 | 2373 | |
f5b0a874 | 2374 | rt = dst_alloc(&ipv4_dst_blackhole_ops, NULL, 1, DST_OBSOLETE_NONE, 0); |
14e50e57 | 2375 | if (rt) { |
d8d1f30b | 2376 | struct dst_entry *new = &rt->dst; |
14e50e57 | 2377 | |
14e50e57 | 2378 | new->__use = 1; |
352e512c | 2379 | new->input = dst_discard; |
ede2059d | 2380 | new->output = dst_discard_out; |
14e50e57 | 2381 | |
d8d1f30b | 2382 | new->dev = ort->dst.dev; |
14e50e57 DM |
2383 | if (new->dev) |
2384 | dev_hold(new->dev); | |
2385 | ||
9917e1e8 | 2386 | rt->rt_is_input = ort->rt_is_input; |
5e2b61f7 | 2387 | rt->rt_iif = ort->rt_iif; |
5943634f | 2388 | rt->rt_pmtu = ort->rt_pmtu; |
14e50e57 | 2389 | |
ca4c3fc2 | 2390 | rt->rt_genid = rt_genid_ipv4(net); |
14e50e57 DM |
2391 | rt->rt_flags = ort->rt_flags; |
2392 | rt->rt_type = ort->rt_type; | |
14e50e57 | 2393 | rt->rt_gateway = ort->rt_gateway; |
155e8336 | 2394 | rt->rt_uses_gateway = ort->rt_uses_gateway; |
14e50e57 | 2395 | |
caacf05e | 2396 | INIT_LIST_HEAD(&rt->rt_uncached); |
14e50e57 DM |
2397 | dst_free(new); |
2398 | } | |
2399 | ||
2774c131 DM |
2400 | dst_release(dst_orig); |
2401 | ||
2402 | return rt ? &rt->dst : ERR_PTR(-ENOMEM); | |
14e50e57 DM |
2403 | } |
2404 | ||
9d6ec938 | 2405 | struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4, |
6f9c9615 | 2406 | const struct sock *sk) |
1da177e4 | 2407 | { |
9d6ec938 | 2408 | struct rtable *rt = __ip_route_output_key(net, flp4); |
1da177e4 | 2409 | |
b23dd4fe DM |
2410 | if (IS_ERR(rt)) |
2411 | return rt; | |
1da177e4 | 2412 | |
56157872 | 2413 | if (flp4->flowi4_proto) |
f92ee619 SK |
2414 | rt = (struct rtable *)xfrm_lookup_route(net, &rt->dst, |
2415 | flowi4_to_flowi(flp4), | |
2416 | sk, 0); | |
1da177e4 | 2417 | |
b23dd4fe | 2418 | return rt; |
1da177e4 | 2419 | } |
d8c97a94 ACM |
2420 | EXPORT_SYMBOL_GPL(ip_route_output_flow); |
2421 | ||
c36ba660 | 2422 | static int rt_fill_info(struct net *net, __be32 dst, __be32 src, u32 table_id, |
15e47304 | 2423 | struct flowi4 *fl4, struct sk_buff *skb, u32 portid, |
f1ce3062 | 2424 | u32 seq, int event, int nowait, unsigned int flags) |
1da177e4 | 2425 | { |
511c3f92 | 2426 | struct rtable *rt = skb_rtable(skb); |
1da177e4 | 2427 | struct rtmsg *r; |
be403ea1 | 2428 | struct nlmsghdr *nlh; |
2bc8ca40 | 2429 | unsigned long expires = 0; |
f185071d | 2430 | u32 error; |
521f5490 | 2431 | u32 metrics[RTAX_MAX]; |
be403ea1 | 2432 | |
15e47304 | 2433 | nlh = nlmsg_put(skb, portid, seq, event, sizeof(*r), flags); |
51456b29 | 2434 | if (!nlh) |
26932566 | 2435 | return -EMSGSIZE; |
be403ea1 TG |
2436 | |
2437 | r = nlmsg_data(nlh); | |
1da177e4 LT |
2438 | r->rtm_family = AF_INET; |
2439 | r->rtm_dst_len = 32; | |
2440 | r->rtm_src_len = 0; | |
d6c0a4f6 | 2441 | r->rtm_tos = fl4->flowi4_tos; |
c36ba660 DA |
2442 | r->rtm_table = table_id; |
2443 | if (nla_put_u32(skb, RTA_TABLE, table_id)) | |
f3756b79 | 2444 | goto nla_put_failure; |
1da177e4 LT |
2445 | r->rtm_type = rt->rt_type; |
2446 | r->rtm_scope = RT_SCOPE_UNIVERSE; | |
2447 | r->rtm_protocol = RTPROT_UNSPEC; | |
2448 | r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED; | |
2449 | if (rt->rt_flags & RTCF_NOTIFY) | |
2450 | r->rtm_flags |= RTM_F_NOTIFY; | |
df4d9254 HFS |
2451 | if (IPCB(skb)->flags & IPSKB_DOREDIRECT) |
2452 | r->rtm_flags |= RTCF_DOREDIRECT; | |
be403ea1 | 2453 | |
930345ea | 2454 | if (nla_put_in_addr(skb, RTA_DST, dst)) |
f3756b79 | 2455 | goto nla_put_failure; |
1a00fee4 | 2456 | if (src) { |
1da177e4 | 2457 | r->rtm_src_len = 32; |
930345ea | 2458 | if (nla_put_in_addr(skb, RTA_SRC, src)) |
f3756b79 | 2459 | goto nla_put_failure; |
1da177e4 | 2460 | } |
f3756b79 DM |
2461 | if (rt->dst.dev && |
2462 | nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex)) | |
2463 | goto nla_put_failure; | |
c7066f70 | 2464 | #ifdef CONFIG_IP_ROUTE_CLASSID |
f3756b79 DM |
2465 | if (rt->dst.tclassid && |
2466 | nla_put_u32(skb, RTA_FLOW, rt->dst.tclassid)) | |
2467 | goto nla_put_failure; | |
1da177e4 | 2468 | #endif |
41347dcd | 2469 | if (!rt_is_input_route(rt) && |
d6c0a4f6 | 2470 | fl4->saddr != src) { |
930345ea | 2471 | if (nla_put_in_addr(skb, RTA_PREFSRC, fl4->saddr)) |
f3756b79 DM |
2472 | goto nla_put_failure; |
2473 | } | |
155e8336 | 2474 | if (rt->rt_uses_gateway && |
930345ea | 2475 | nla_put_in_addr(skb, RTA_GATEWAY, rt->rt_gateway)) |
f3756b79 | 2476 | goto nla_put_failure; |
be403ea1 | 2477 | |
ee9a8f7a SK |
2478 | expires = rt->dst.expires; |
2479 | if (expires) { | |
2480 | unsigned long now = jiffies; | |
2481 | ||
2482 | if (time_before(now, expires)) | |
2483 | expires -= now; | |
2484 | else | |
2485 | expires = 0; | |
2486 | } | |
2487 | ||
521f5490 | 2488 | memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics)); |
ee9a8f7a | 2489 | if (rt->rt_pmtu && expires) |
521f5490 JA |
2490 | metrics[RTAX_MTU - 1] = rt->rt_pmtu; |
2491 | if (rtnetlink_put_metrics(skb, metrics) < 0) | |
be403ea1 TG |
2492 | goto nla_put_failure; |
2493 | ||
b4869889 | 2494 | if (fl4->flowi4_mark && |
68aaed54 | 2495 | nla_put_u32(skb, RTA_MARK, fl4->flowi4_mark)) |
f3756b79 | 2496 | goto nla_put_failure; |
963bfeee | 2497 | |
d8d1f30b | 2498 | error = rt->dst.error; |
be403ea1 | 2499 | |
c7537967 | 2500 | if (rt_is_input_route(rt)) { |
8caaf7b6 ND |
2501 | #ifdef CONFIG_IP_MROUTE |
2502 | if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) && | |
2503 | IPV4_DEVCONF_ALL(net, MC_FORWARDING)) { | |
2504 | int err = ipmr_get_route(net, skb, | |
2505 | fl4->saddr, fl4->daddr, | |
2cf75070 NA |
2506 | r, nowait, portid); |
2507 | ||
8caaf7b6 ND |
2508 | if (err <= 0) { |
2509 | if (!nowait) { | |
2510 | if (err == 0) | |
2511 | return 0; | |
2512 | goto nla_put_failure; | |
2513 | } else { | |
2514 | if (err == -EMSGSIZE) | |
2515 | goto nla_put_failure; | |
2516 | error = err; | |
2517 | } | |
2518 | } | |
2519 | } else | |
2520 | #endif | |
91146153 | 2521 | if (nla_put_u32(skb, RTA_IIF, skb->dev->ifindex)) |
8caaf7b6 | 2522 | goto nla_put_failure; |
1da177e4 LT |
2523 | } |
2524 | ||
f185071d | 2525 | if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, error) < 0) |
e3703b3d | 2526 | goto nla_put_failure; |
be403ea1 | 2527 | |
053c095a JB |
2528 | nlmsg_end(skb, nlh); |
2529 | return 0; | |
1da177e4 | 2530 | |
be403ea1 | 2531 | nla_put_failure: |
26932566 PM |
2532 | nlmsg_cancel(skb, nlh); |
2533 | return -EMSGSIZE; | |
1da177e4 LT |
2534 | } |
2535 | ||
661d2967 | 2536 | static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh) |
1da177e4 | 2537 | { |
3b1e0a65 | 2538 | struct net *net = sock_net(in_skb->sk); |
d889ce3b TG |
2539 | struct rtmsg *rtm; |
2540 | struct nlattr *tb[RTA_MAX+1]; | |
1da177e4 | 2541 | struct rtable *rt = NULL; |
d6c0a4f6 | 2542 | struct flowi4 fl4; |
9e12bb22 AV |
2543 | __be32 dst = 0; |
2544 | __be32 src = 0; | |
2545 | u32 iif; | |
d889ce3b | 2546 | int err; |
963bfeee | 2547 | int mark; |
1da177e4 | 2548 | struct sk_buff *skb; |
c36ba660 | 2549 | u32 table_id = RT_TABLE_MAIN; |
1da177e4 | 2550 | |
d889ce3b TG |
2551 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy); |
2552 | if (err < 0) | |
2553 | goto errout; | |
2554 | ||
2555 | rtm = nlmsg_data(nlh); | |
2556 | ||
1da177e4 | 2557 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
51456b29 | 2558 | if (!skb) { |
d889ce3b TG |
2559 | err = -ENOBUFS; |
2560 | goto errout; | |
2561 | } | |
1da177e4 LT |
2562 | |
2563 | /* Reserve room for dummy headers, this skb can pass | |
2564 | through good chunk of routing engine. | |
2565 | */ | |
459a98ed | 2566 | skb_reset_mac_header(skb); |
c1d2bbe1 | 2567 | skb_reset_network_header(skb); |
d2c962b8 SH |
2568 | |
2569 | /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */ | |
eddc9ec5 | 2570 | ip_hdr(skb)->protocol = IPPROTO_ICMP; |
1da177e4 LT |
2571 | skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr)); |
2572 | ||
67b61f6c JB |
2573 | src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0; |
2574 | dst = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0; | |
d889ce3b | 2575 | iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0; |
963bfeee | 2576 | mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0; |
1da177e4 | 2577 | |
d6c0a4f6 DM |
2578 | memset(&fl4, 0, sizeof(fl4)); |
2579 | fl4.daddr = dst; | |
2580 | fl4.saddr = src; | |
2581 | fl4.flowi4_tos = rtm->rtm_tos; | |
2582 | fl4.flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0; | |
2583 | fl4.flowi4_mark = mark; | |
2584 | ||
deaa0a6a DA |
2585 | if (netif_index_is_l3_master(net, fl4.flowi4_oif)) |
2586 | fl4.flowi4_flags = FLOWI_FLAG_L3MDEV_SRC | FLOWI_FLAG_SKIP_NH_OIF; | |
2587 | ||
1da177e4 | 2588 | if (iif) { |
d889ce3b TG |
2589 | struct net_device *dev; |
2590 | ||
1937504d | 2591 | dev = __dev_get_by_index(net, iif); |
51456b29 | 2592 | if (!dev) { |
d889ce3b TG |
2593 | err = -ENODEV; |
2594 | goto errout_free; | |
2595 | } | |
2596 | ||
1da177e4 LT |
2597 | skb->protocol = htons(ETH_P_IP); |
2598 | skb->dev = dev; | |
963bfeee | 2599 | skb->mark = mark; |
1da177e4 LT |
2600 | local_bh_disable(); |
2601 | err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev); | |
2602 | local_bh_enable(); | |
d889ce3b | 2603 | |
511c3f92 | 2604 | rt = skb_rtable(skb); |
d8d1f30b CG |
2605 | if (err == 0 && rt->dst.error) |
2606 | err = -rt->dst.error; | |
1da177e4 | 2607 | } else { |
9d6ec938 | 2608 | rt = ip_route_output_key(net, &fl4); |
b23dd4fe DM |
2609 | |
2610 | err = 0; | |
2611 | if (IS_ERR(rt)) | |
2612 | err = PTR_ERR(rt); | |
1da177e4 | 2613 | } |
d889ce3b | 2614 | |
1da177e4 | 2615 | if (err) |
d889ce3b | 2616 | goto errout_free; |
1da177e4 | 2617 | |
d8d1f30b | 2618 | skb_dst_set(skb, &rt->dst); |
1da177e4 LT |
2619 | if (rtm->rtm_flags & RTM_F_NOTIFY) |
2620 | rt->rt_flags |= RTCF_NOTIFY; | |
2621 | ||
c36ba660 DA |
2622 | if (rtm->rtm_flags & RTM_F_LOOKUP_TABLE) |
2623 | table_id = rt->rt_table_id; | |
2624 | ||
2625 | err = rt_fill_info(net, dst, src, table_id, &fl4, skb, | |
15e47304 | 2626 | NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
1937504d | 2627 | RTM_NEWROUTE, 0, 0); |
7b46a644 | 2628 | if (err < 0) |
d889ce3b | 2629 | goto errout_free; |
1da177e4 | 2630 | |
15e47304 | 2631 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); |
d889ce3b | 2632 | errout: |
2942e900 | 2633 | return err; |
1da177e4 | 2634 | |
d889ce3b | 2635 | errout_free: |
1da177e4 | 2636 | kfree_skb(skb); |
d889ce3b | 2637 | goto errout; |
1da177e4 LT |
2638 | } |
2639 | ||
1da177e4 LT |
2640 | void ip_rt_multicast_event(struct in_device *in_dev) |
2641 | { | |
4ccfe6d4 | 2642 | rt_cache_flush(dev_net(in_dev->dev)); |
1da177e4 LT |
2643 | } |
2644 | ||
2645 | #ifdef CONFIG_SYSCTL | |
082c7ca4 G |
2646 | static int ip_rt_gc_interval __read_mostly = 60 * HZ; |
2647 | static int ip_rt_gc_min_interval __read_mostly = HZ / 2; | |
2648 | static int ip_rt_gc_elasticity __read_mostly = 8; | |
2649 | ||
fe2c6338 | 2650 | static int ipv4_sysctl_rtcache_flush(struct ctl_table *__ctl, int write, |
8d65af78 | 2651 | void __user *buffer, |
1da177e4 LT |
2652 | size_t *lenp, loff_t *ppos) |
2653 | { | |
5aad1de5 TT |
2654 | struct net *net = (struct net *)__ctl->extra1; |
2655 | ||
1da177e4 | 2656 | if (write) { |
5aad1de5 TT |
2657 | rt_cache_flush(net); |
2658 | fnhe_genid_bump(net); | |
1da177e4 | 2659 | return 0; |
e905a9ed | 2660 | } |
1da177e4 LT |
2661 | |
2662 | return -EINVAL; | |
2663 | } | |
2664 | ||
fe2c6338 | 2665 | static struct ctl_table ipv4_route_table[] = { |
1da177e4 | 2666 | { |
1da177e4 LT |
2667 | .procname = "gc_thresh", |
2668 | .data = &ipv4_dst_ops.gc_thresh, | |
2669 | .maxlen = sizeof(int), | |
2670 | .mode = 0644, | |
6d9f239a | 2671 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2672 | }, |
2673 | { | |
1da177e4 LT |
2674 | .procname = "max_size", |
2675 | .data = &ip_rt_max_size, | |
2676 | .maxlen = sizeof(int), | |
2677 | .mode = 0644, | |
6d9f239a | 2678 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2679 | }, |
2680 | { | |
2681 | /* Deprecated. Use gc_min_interval_ms */ | |
e905a9ed | 2682 | |
1da177e4 LT |
2683 | .procname = "gc_min_interval", |
2684 | .data = &ip_rt_gc_min_interval, | |
2685 | .maxlen = sizeof(int), | |
2686 | .mode = 0644, | |
6d9f239a | 2687 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
2688 | }, |
2689 | { | |
1da177e4 LT |
2690 | .procname = "gc_min_interval_ms", |
2691 | .data = &ip_rt_gc_min_interval, | |
2692 | .maxlen = sizeof(int), | |
2693 | .mode = 0644, | |
6d9f239a | 2694 | .proc_handler = proc_dointvec_ms_jiffies, |
1da177e4 LT |
2695 | }, |
2696 | { | |
1da177e4 LT |
2697 | .procname = "gc_timeout", |
2698 | .data = &ip_rt_gc_timeout, | |
2699 | .maxlen = sizeof(int), | |
2700 | .mode = 0644, | |
6d9f239a | 2701 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 | 2702 | }, |
9f28a2fc ED |
2703 | { |
2704 | .procname = "gc_interval", | |
2705 | .data = &ip_rt_gc_interval, | |
2706 | .maxlen = sizeof(int), | |
2707 | .mode = 0644, | |
2708 | .proc_handler = proc_dointvec_jiffies, | |
2709 | }, | |
1da177e4 | 2710 | { |
1da177e4 LT |
2711 | .procname = "redirect_load", |
2712 | .data = &ip_rt_redirect_load, | |
2713 | .maxlen = sizeof(int), | |
2714 | .mode = 0644, | |
6d9f239a | 2715 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2716 | }, |
2717 | { | |
1da177e4 LT |
2718 | .procname = "redirect_number", |
2719 | .data = &ip_rt_redirect_number, | |
2720 | .maxlen = sizeof(int), | |
2721 | .mode = 0644, | |
6d9f239a | 2722 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2723 | }, |
2724 | { | |
1da177e4 LT |
2725 | .procname = "redirect_silence", |
2726 | .data = &ip_rt_redirect_silence, | |
2727 | .maxlen = sizeof(int), | |
2728 | .mode = 0644, | |
6d9f239a | 2729 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2730 | }, |
2731 | { | |
1da177e4 LT |
2732 | .procname = "error_cost", |
2733 | .data = &ip_rt_error_cost, | |
2734 | .maxlen = sizeof(int), | |
2735 | .mode = 0644, | |
6d9f239a | 2736 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2737 | }, |
2738 | { | |
1da177e4 LT |
2739 | .procname = "error_burst", |
2740 | .data = &ip_rt_error_burst, | |
2741 | .maxlen = sizeof(int), | |
2742 | .mode = 0644, | |
6d9f239a | 2743 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2744 | }, |
2745 | { | |
1da177e4 LT |
2746 | .procname = "gc_elasticity", |
2747 | .data = &ip_rt_gc_elasticity, | |
2748 | .maxlen = sizeof(int), | |
2749 | .mode = 0644, | |
6d9f239a | 2750 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2751 | }, |
2752 | { | |
1da177e4 LT |
2753 | .procname = "mtu_expires", |
2754 | .data = &ip_rt_mtu_expires, | |
2755 | .maxlen = sizeof(int), | |
2756 | .mode = 0644, | |
6d9f239a | 2757 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
2758 | }, |
2759 | { | |
1da177e4 LT |
2760 | .procname = "min_pmtu", |
2761 | .data = &ip_rt_min_pmtu, | |
2762 | .maxlen = sizeof(int), | |
2763 | .mode = 0644, | |
6d9f239a | 2764 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2765 | }, |
2766 | { | |
1da177e4 LT |
2767 | .procname = "min_adv_mss", |
2768 | .data = &ip_rt_min_advmss, | |
2769 | .maxlen = sizeof(int), | |
2770 | .mode = 0644, | |
6d9f239a | 2771 | .proc_handler = proc_dointvec, |
1da177e4 | 2772 | }, |
f8572d8f | 2773 | { } |
1da177e4 | 2774 | }; |
39a23e75 | 2775 | |
39a23e75 DL |
2776 | static struct ctl_table ipv4_route_flush_table[] = { |
2777 | { | |
39a23e75 DL |
2778 | .procname = "flush", |
2779 | .maxlen = sizeof(int), | |
2780 | .mode = 0200, | |
6d9f239a | 2781 | .proc_handler = ipv4_sysctl_rtcache_flush, |
39a23e75 | 2782 | }, |
f8572d8f | 2783 | { }, |
39a23e75 DL |
2784 | }; |
2785 | ||
2786 | static __net_init int sysctl_route_net_init(struct net *net) | |
2787 | { | |
2788 | struct ctl_table *tbl; | |
2789 | ||
2790 | tbl = ipv4_route_flush_table; | |
09ad9bc7 | 2791 | if (!net_eq(net, &init_net)) { |
39a23e75 | 2792 | tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL); |
51456b29 | 2793 | if (!tbl) |
39a23e75 | 2794 | goto err_dup; |
464dc801 EB |
2795 | |
2796 | /* Don't export sysctls to unprivileged users */ | |
2797 | if (net->user_ns != &init_user_ns) | |
2798 | tbl[0].procname = NULL; | |
39a23e75 DL |
2799 | } |
2800 | tbl[0].extra1 = net; | |
2801 | ||
ec8f23ce | 2802 | net->ipv4.route_hdr = register_net_sysctl(net, "net/ipv4/route", tbl); |
51456b29 | 2803 | if (!net->ipv4.route_hdr) |
39a23e75 DL |
2804 | goto err_reg; |
2805 | return 0; | |
2806 | ||
2807 | err_reg: | |
2808 | if (tbl != ipv4_route_flush_table) | |
2809 | kfree(tbl); | |
2810 | err_dup: | |
2811 | return -ENOMEM; | |
2812 | } | |
2813 | ||
2814 | static __net_exit void sysctl_route_net_exit(struct net *net) | |
2815 | { | |
2816 | struct ctl_table *tbl; | |
2817 | ||
2818 | tbl = net->ipv4.route_hdr->ctl_table_arg; | |
2819 | unregister_net_sysctl_table(net->ipv4.route_hdr); | |
2820 | BUG_ON(tbl == ipv4_route_flush_table); | |
2821 | kfree(tbl); | |
2822 | } | |
2823 | ||
2824 | static __net_initdata struct pernet_operations sysctl_route_ops = { | |
2825 | .init = sysctl_route_net_init, | |
2826 | .exit = sysctl_route_net_exit, | |
2827 | }; | |
1da177e4 LT |
2828 | #endif |
2829 | ||
3ee94372 | 2830 | static __net_init int rt_genid_init(struct net *net) |
9f5e97e5 | 2831 | { |
ca4c3fc2 | 2832 | atomic_set(&net->ipv4.rt_genid, 0); |
5aad1de5 | 2833 | atomic_set(&net->fnhe_genid, 0); |
436c3b66 DM |
2834 | get_random_bytes(&net->ipv4.dev_addr_genid, |
2835 | sizeof(net->ipv4.dev_addr_genid)); | |
9f5e97e5 DL |
2836 | return 0; |
2837 | } | |
2838 | ||
3ee94372 NH |
2839 | static __net_initdata struct pernet_operations rt_genid_ops = { |
2840 | .init = rt_genid_init, | |
9f5e97e5 DL |
2841 | }; |
2842 | ||
c3426b47 DM |
2843 | static int __net_init ipv4_inetpeer_init(struct net *net) |
2844 | { | |
2845 | struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL); | |
2846 | ||
2847 | if (!bp) | |
2848 | return -ENOMEM; | |
2849 | inet_peer_base_init(bp); | |
2850 | net->ipv4.peers = bp; | |
2851 | return 0; | |
2852 | } | |
2853 | ||
2854 | static void __net_exit ipv4_inetpeer_exit(struct net *net) | |
2855 | { | |
2856 | struct inet_peer_base *bp = net->ipv4.peers; | |
2857 | ||
2858 | net->ipv4.peers = NULL; | |
56a6b248 | 2859 | inetpeer_invalidate_tree(bp); |
c3426b47 DM |
2860 | kfree(bp); |
2861 | } | |
2862 | ||
2863 | static __net_initdata struct pernet_operations ipv4_inetpeer_ops = { | |
2864 | .init = ipv4_inetpeer_init, | |
2865 | .exit = ipv4_inetpeer_exit, | |
2866 | }; | |
9f5e97e5 | 2867 | |
c7066f70 | 2868 | #ifdef CONFIG_IP_ROUTE_CLASSID |
7d720c3e | 2869 | struct ip_rt_acct __percpu *ip_rt_acct __read_mostly; |
c7066f70 | 2870 | #endif /* CONFIG_IP_ROUTE_CLASSID */ |
1da177e4 | 2871 | |
1da177e4 LT |
2872 | int __init ip_rt_init(void) |
2873 | { | |
424c4b70 | 2874 | int rc = 0; |
5055c371 | 2875 | int cpu; |
1da177e4 | 2876 | |
73f156a6 ED |
2877 | ip_idents = kmalloc(IP_IDENTS_SZ * sizeof(*ip_idents), GFP_KERNEL); |
2878 | if (!ip_idents) | |
2879 | panic("IP: failed to allocate ip_idents\n"); | |
2880 | ||
2881 | prandom_bytes(ip_idents, IP_IDENTS_SZ * sizeof(*ip_idents)); | |
2882 | ||
355b590c ED |
2883 | ip_tstamps = kcalloc(IP_IDENTS_SZ, sizeof(*ip_tstamps), GFP_KERNEL); |
2884 | if (!ip_tstamps) | |
2885 | panic("IP: failed to allocate ip_tstamps\n"); | |
2886 | ||
5055c371 ED |
2887 | for_each_possible_cpu(cpu) { |
2888 | struct uncached_list *ul = &per_cpu(rt_uncached_list, cpu); | |
2889 | ||
2890 | INIT_LIST_HEAD(&ul->head); | |
2891 | spin_lock_init(&ul->lock); | |
2892 | } | |
c7066f70 | 2893 | #ifdef CONFIG_IP_ROUTE_CLASSID |
0dcec8c2 | 2894 | ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct)); |
1da177e4 LT |
2895 | if (!ip_rt_acct) |
2896 | panic("IP: failed to allocate ip_rt_acct\n"); | |
1da177e4 LT |
2897 | #endif |
2898 | ||
e5d679f3 AD |
2899 | ipv4_dst_ops.kmem_cachep = |
2900 | kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0, | |
20c2df83 | 2901 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 2902 | |
14e50e57 DM |
2903 | ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep; |
2904 | ||
fc66f95c ED |
2905 | if (dst_entries_init(&ipv4_dst_ops) < 0) |
2906 | panic("IP: failed to allocate ipv4_dst_ops counter\n"); | |
2907 | ||
2908 | if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0) | |
2909 | panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n"); | |
2910 | ||
89aef892 DM |
2911 | ipv4_dst_ops.gc_thresh = ~0; |
2912 | ip_rt_max_size = INT_MAX; | |
1da177e4 | 2913 | |
1da177e4 LT |
2914 | devinet_init(); |
2915 | ip_fib_init(); | |
2916 | ||
73b38711 | 2917 | if (ip_rt_proc_init()) |
058bd4d2 | 2918 | pr_err("Unable to create route proc files\n"); |
1da177e4 LT |
2919 | #ifdef CONFIG_XFRM |
2920 | xfrm_init(); | |
703fb94e | 2921 | xfrm4_init(); |
1da177e4 | 2922 | #endif |
c7ac8679 | 2923 | rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL, NULL); |
63f3444f | 2924 | |
39a23e75 DL |
2925 | #ifdef CONFIG_SYSCTL |
2926 | register_pernet_subsys(&sysctl_route_ops); | |
2927 | #endif | |
3ee94372 | 2928 | register_pernet_subsys(&rt_genid_ops); |
c3426b47 | 2929 | register_pernet_subsys(&ipv4_inetpeer_ops); |
1da177e4 LT |
2930 | return rc; |
2931 | } | |
2932 | ||
a1bc6eb4 | 2933 | #ifdef CONFIG_SYSCTL |
eeb61f71 AV |
2934 | /* |
2935 | * We really need to sanitize the damn ipv4 init order, then all | |
2936 | * this nonsense will go away. | |
2937 | */ | |
2938 | void __init ip_static_sysctl_init(void) | |
2939 | { | |
4e5ca785 | 2940 | register_net_sysctl(&init_net, "net/ipv4/route", ipv4_route_table); |
eeb61f71 | 2941 | } |
a1bc6eb4 | 2942 | #endif |