]>
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> |
352e512c | 92 | #include <net/dst.h> |
457c4cbc | 93 | #include <net/net_namespace.h> |
1da177e4 LT |
94 | #include <net/protocol.h> |
95 | #include <net/ip.h> | |
96 | #include <net/route.h> | |
97 | #include <net/inetpeer.h> | |
98 | #include <net/sock.h> | |
99 | #include <net/ip_fib.h> | |
100 | #include <net/arp.h> | |
101 | #include <net/tcp.h> | |
102 | #include <net/icmp.h> | |
103 | #include <net/xfrm.h> | |
8d71740c | 104 | #include <net/netevent.h> |
63f3444f | 105 | #include <net/rtnetlink.h> |
1da177e4 LT |
106 | #ifdef CONFIG_SYSCTL |
107 | #include <linux/sysctl.h> | |
7426a564 | 108 | #include <linux/kmemleak.h> |
1da177e4 | 109 | #endif |
6e5714ea | 110 | #include <net/secure_seq.h> |
1da177e4 | 111 | |
68a5e3dd | 112 | #define RT_FL_TOS(oldflp4) \ |
f61759e6 | 113 | ((oldflp4)->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK)) |
1da177e4 LT |
114 | |
115 | #define IP_MAX_MTU 0xFFF0 | |
116 | ||
117 | #define RT_GC_TIMEOUT (300*HZ) | |
118 | ||
1da177e4 | 119 | static int ip_rt_max_size; |
817bc4db SH |
120 | static int ip_rt_redirect_number __read_mostly = 9; |
121 | static int ip_rt_redirect_load __read_mostly = HZ / 50; | |
122 | static int ip_rt_redirect_silence __read_mostly = ((HZ / 50) << (9 + 1)); | |
123 | static int ip_rt_error_cost __read_mostly = HZ; | |
124 | static int ip_rt_error_burst __read_mostly = 5 * HZ; | |
817bc4db SH |
125 | static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ; |
126 | static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20; | |
127 | static int ip_rt_min_advmss __read_mostly = 256; | |
9f28a2fc | 128 | |
1da177e4 LT |
129 | /* |
130 | * Interface to generic destination cache. | |
131 | */ | |
132 | ||
133 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie); | |
0dbaee3b | 134 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst); |
ebb762f2 | 135 | static unsigned int ipv4_mtu(const struct dst_entry *dst); |
1da177e4 LT |
136 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst); |
137 | static void ipv4_link_failure(struct sk_buff *skb); | |
6700c270 DM |
138 | static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk, |
139 | struct sk_buff *skb, u32 mtu); | |
140 | static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, | |
141 | struct sk_buff *skb); | |
caacf05e | 142 | static void ipv4_dst_destroy(struct dst_entry *dst); |
1da177e4 | 143 | |
72cdd1d9 ED |
144 | static void ipv4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, |
145 | int how) | |
146 | { | |
147 | } | |
1da177e4 | 148 | |
62fa8a84 DM |
149 | static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old) |
150 | { | |
31248731 DM |
151 | WARN_ON(1); |
152 | return NULL; | |
62fa8a84 DM |
153 | } |
154 | ||
f894cbf8 DM |
155 | static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, |
156 | struct sk_buff *skb, | |
157 | const void *daddr); | |
d3aaeb38 | 158 | |
1da177e4 LT |
159 | static struct dst_ops ipv4_dst_ops = { |
160 | .family = AF_INET, | |
09640e63 | 161 | .protocol = cpu_to_be16(ETH_P_IP), |
1da177e4 | 162 | .check = ipv4_dst_check, |
0dbaee3b | 163 | .default_advmss = ipv4_default_advmss, |
ebb762f2 | 164 | .mtu = ipv4_mtu, |
62fa8a84 | 165 | .cow_metrics = ipv4_cow_metrics, |
caacf05e | 166 | .destroy = ipv4_dst_destroy, |
1da177e4 LT |
167 | .ifdown = ipv4_dst_ifdown, |
168 | .negative_advice = ipv4_negative_advice, | |
169 | .link_failure = ipv4_link_failure, | |
170 | .update_pmtu = ip_rt_update_pmtu, | |
e47a185b | 171 | .redirect = ip_do_redirect, |
1ac06e03 | 172 | .local_out = __ip_local_out, |
d3aaeb38 | 173 | .neigh_lookup = ipv4_neigh_lookup, |
1da177e4 LT |
174 | }; |
175 | ||
176 | #define ECN_OR_COST(class) TC_PRIO_##class | |
177 | ||
4839c52b | 178 | const __u8 ip_tos2prio[16] = { |
1da177e4 | 179 | TC_PRIO_BESTEFFORT, |
4a2b9c37 | 180 | ECN_OR_COST(BESTEFFORT), |
1da177e4 LT |
181 | TC_PRIO_BESTEFFORT, |
182 | ECN_OR_COST(BESTEFFORT), | |
183 | TC_PRIO_BULK, | |
184 | ECN_OR_COST(BULK), | |
185 | TC_PRIO_BULK, | |
186 | ECN_OR_COST(BULK), | |
187 | TC_PRIO_INTERACTIVE, | |
188 | ECN_OR_COST(INTERACTIVE), | |
189 | TC_PRIO_INTERACTIVE, | |
190 | ECN_OR_COST(INTERACTIVE), | |
191 | TC_PRIO_INTERACTIVE_BULK, | |
192 | ECN_OR_COST(INTERACTIVE_BULK), | |
193 | TC_PRIO_INTERACTIVE_BULK, | |
194 | ECN_OR_COST(INTERACTIVE_BULK) | |
195 | }; | |
d4a96865 | 196 | EXPORT_SYMBOL(ip_tos2prio); |
1da177e4 | 197 | |
2f970d83 | 198 | static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat); |
27f39c73 | 199 | #define RT_CACHE_STAT_INC(field) __this_cpu_inc(rt_cache_stat.field) |
1da177e4 | 200 | |
1da177e4 | 201 | #ifdef CONFIG_PROC_FS |
1da177e4 LT |
202 | static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos) |
203 | { | |
29e75252 | 204 | if (*pos) |
89aef892 | 205 | return NULL; |
29e75252 | 206 | return SEQ_START_TOKEN; |
1da177e4 LT |
207 | } |
208 | ||
209 | static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
210 | { | |
1da177e4 | 211 | ++*pos; |
89aef892 | 212 | return NULL; |
1da177e4 LT |
213 | } |
214 | ||
215 | static void rt_cache_seq_stop(struct seq_file *seq, void *v) | |
216 | { | |
1da177e4 LT |
217 | } |
218 | ||
219 | static int rt_cache_seq_show(struct seq_file *seq, void *v) | |
220 | { | |
221 | if (v == SEQ_START_TOKEN) | |
222 | seq_printf(seq, "%-127s\n", | |
223 | "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t" | |
224 | "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t" | |
225 | "HHUptod\tSpecDst"); | |
e905a9ed | 226 | return 0; |
1da177e4 LT |
227 | } |
228 | ||
f690808e | 229 | static const struct seq_operations rt_cache_seq_ops = { |
1da177e4 LT |
230 | .start = rt_cache_seq_start, |
231 | .next = rt_cache_seq_next, | |
232 | .stop = rt_cache_seq_stop, | |
233 | .show = rt_cache_seq_show, | |
234 | }; | |
235 | ||
236 | static int rt_cache_seq_open(struct inode *inode, struct file *file) | |
237 | { | |
89aef892 | 238 | return seq_open(file, &rt_cache_seq_ops); |
1da177e4 LT |
239 | } |
240 | ||
9a32144e | 241 | static const struct file_operations rt_cache_seq_fops = { |
1da177e4 LT |
242 | .owner = THIS_MODULE, |
243 | .open = rt_cache_seq_open, | |
244 | .read = seq_read, | |
245 | .llseek = seq_lseek, | |
89aef892 | 246 | .release = seq_release, |
1da177e4 LT |
247 | }; |
248 | ||
249 | ||
250 | static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos) | |
251 | { | |
252 | int cpu; | |
253 | ||
254 | if (*pos == 0) | |
255 | return SEQ_START_TOKEN; | |
256 | ||
0f23174a | 257 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
258 | if (!cpu_possible(cpu)) |
259 | continue; | |
260 | *pos = cpu+1; | |
2f970d83 | 261 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
262 | } |
263 | return NULL; | |
264 | } | |
265 | ||
266 | static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
267 | { | |
268 | int cpu; | |
269 | ||
0f23174a | 270 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
271 | if (!cpu_possible(cpu)) |
272 | continue; | |
273 | *pos = cpu+1; | |
2f970d83 | 274 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
275 | } |
276 | return NULL; | |
e905a9ed | 277 | |
1da177e4 LT |
278 | } |
279 | ||
280 | static void rt_cpu_seq_stop(struct seq_file *seq, void *v) | |
281 | { | |
282 | ||
283 | } | |
284 | ||
285 | static int rt_cpu_seq_show(struct seq_file *seq, void *v) | |
286 | { | |
287 | struct rt_cache_stat *st = v; | |
288 | ||
289 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 290 | 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 |
291 | return 0; |
292 | } | |
e905a9ed | 293 | |
1da177e4 LT |
294 | seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x " |
295 | " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n", | |
fc66f95c | 296 | dst_entries_get_slow(&ipv4_dst_ops), |
1da177e4 LT |
297 | st->in_hit, |
298 | st->in_slow_tot, | |
299 | st->in_slow_mc, | |
300 | st->in_no_route, | |
301 | st->in_brd, | |
302 | st->in_martian_dst, | |
303 | st->in_martian_src, | |
304 | ||
305 | st->out_hit, | |
306 | st->out_slow_tot, | |
e905a9ed | 307 | st->out_slow_mc, |
1da177e4 LT |
308 | |
309 | st->gc_total, | |
310 | st->gc_ignored, | |
311 | st->gc_goal_miss, | |
312 | st->gc_dst_overflow, | |
313 | st->in_hlist_search, | |
314 | st->out_hlist_search | |
315 | ); | |
316 | return 0; | |
317 | } | |
318 | ||
f690808e | 319 | static const struct seq_operations rt_cpu_seq_ops = { |
1da177e4 LT |
320 | .start = rt_cpu_seq_start, |
321 | .next = rt_cpu_seq_next, | |
322 | .stop = rt_cpu_seq_stop, | |
323 | .show = rt_cpu_seq_show, | |
324 | }; | |
325 | ||
326 | ||
327 | static int rt_cpu_seq_open(struct inode *inode, struct file *file) | |
328 | { | |
329 | return seq_open(file, &rt_cpu_seq_ops); | |
330 | } | |
331 | ||
9a32144e | 332 | static const struct file_operations rt_cpu_seq_fops = { |
1da177e4 LT |
333 | .owner = THIS_MODULE, |
334 | .open = rt_cpu_seq_open, | |
335 | .read = seq_read, | |
336 | .llseek = seq_lseek, | |
337 | .release = seq_release, | |
338 | }; | |
339 | ||
c7066f70 | 340 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 341 | static int rt_acct_proc_show(struct seq_file *m, void *v) |
78c686e9 | 342 | { |
a661c419 AD |
343 | struct ip_rt_acct *dst, *src; |
344 | unsigned int i, j; | |
345 | ||
346 | dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL); | |
347 | if (!dst) | |
348 | return -ENOMEM; | |
349 | ||
350 | for_each_possible_cpu(i) { | |
351 | src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i); | |
352 | for (j = 0; j < 256; j++) { | |
353 | dst[j].o_bytes += src[j].o_bytes; | |
354 | dst[j].o_packets += src[j].o_packets; | |
355 | dst[j].i_bytes += src[j].i_bytes; | |
356 | dst[j].i_packets += src[j].i_packets; | |
357 | } | |
78c686e9 PE |
358 | } |
359 | ||
a661c419 AD |
360 | seq_write(m, dst, 256 * sizeof(struct ip_rt_acct)); |
361 | kfree(dst); | |
362 | return 0; | |
363 | } | |
78c686e9 | 364 | |
a661c419 AD |
365 | static int rt_acct_proc_open(struct inode *inode, struct file *file) |
366 | { | |
367 | return single_open(file, rt_acct_proc_show, NULL); | |
78c686e9 | 368 | } |
a661c419 AD |
369 | |
370 | static const struct file_operations rt_acct_proc_fops = { | |
371 | .owner = THIS_MODULE, | |
372 | .open = rt_acct_proc_open, | |
373 | .read = seq_read, | |
374 | .llseek = seq_lseek, | |
375 | .release = single_release, | |
376 | }; | |
78c686e9 | 377 | #endif |
107f1634 | 378 | |
73b38711 | 379 | static int __net_init ip_rt_do_proc_init(struct net *net) |
107f1634 PE |
380 | { |
381 | struct proc_dir_entry *pde; | |
382 | ||
d4beaa66 G |
383 | pde = proc_create("rt_cache", S_IRUGO, net->proc_net, |
384 | &rt_cache_seq_fops); | |
107f1634 PE |
385 | if (!pde) |
386 | goto err1; | |
387 | ||
77020720 WC |
388 | pde = proc_create("rt_cache", S_IRUGO, |
389 | net->proc_net_stat, &rt_cpu_seq_fops); | |
107f1634 PE |
390 | if (!pde) |
391 | goto err2; | |
392 | ||
c7066f70 | 393 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 394 | pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops); |
107f1634 PE |
395 | if (!pde) |
396 | goto err3; | |
397 | #endif | |
398 | return 0; | |
399 | ||
c7066f70 | 400 | #ifdef CONFIG_IP_ROUTE_CLASSID |
107f1634 PE |
401 | err3: |
402 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
403 | #endif | |
404 | err2: | |
405 | remove_proc_entry("rt_cache", net->proc_net); | |
406 | err1: | |
407 | return -ENOMEM; | |
408 | } | |
73b38711 DL |
409 | |
410 | static void __net_exit ip_rt_do_proc_exit(struct net *net) | |
411 | { | |
412 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
413 | remove_proc_entry("rt_cache", net->proc_net); | |
c7066f70 | 414 | #ifdef CONFIG_IP_ROUTE_CLASSID |
73b38711 | 415 | remove_proc_entry("rt_acct", net->proc_net); |
0a931acf | 416 | #endif |
73b38711 DL |
417 | } |
418 | ||
419 | static struct pernet_operations ip_rt_proc_ops __net_initdata = { | |
420 | .init = ip_rt_do_proc_init, | |
421 | .exit = ip_rt_do_proc_exit, | |
422 | }; | |
423 | ||
424 | static int __init ip_rt_proc_init(void) | |
425 | { | |
426 | return register_pernet_subsys(&ip_rt_proc_ops); | |
427 | } | |
428 | ||
107f1634 | 429 | #else |
73b38711 | 430 | static inline int ip_rt_proc_init(void) |
107f1634 PE |
431 | { |
432 | return 0; | |
433 | } | |
1da177e4 | 434 | #endif /* CONFIG_PROC_FS */ |
e905a9ed | 435 | |
4331debc | 436 | static inline bool rt_is_expired(const struct rtable *rth) |
e84f84f2 | 437 | { |
d8d1f30b | 438 | return rth->rt_genid != rt_genid(dev_net(rth->dst.dev)); |
e84f84f2 DL |
439 | } |
440 | ||
4ccfe6d4 | 441 | void rt_cache_flush(struct net *net) |
1da177e4 | 442 | { |
b42664f8 | 443 | rt_genid_bump(net); |
98376387 ED |
444 | } |
445 | ||
f894cbf8 DM |
446 | static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, |
447 | struct sk_buff *skb, | |
448 | const void *daddr) | |
3769cffb | 449 | { |
d3aaeb38 DM |
450 | struct net_device *dev = dst->dev; |
451 | const __be32 *pkey = daddr; | |
39232973 | 452 | const struct rtable *rt; |
3769cffb DM |
453 | struct neighbour *n; |
454 | ||
39232973 | 455 | rt = (const struct rtable *) dst; |
a263b309 | 456 | if (rt->rt_gateway) |
39232973 | 457 | pkey = (const __be32 *) &rt->rt_gateway; |
f894cbf8 DM |
458 | else if (skb) |
459 | pkey = &ip_hdr(skb)->daddr; | |
d3aaeb38 | 460 | |
80703d26 | 461 | n = __ipv4_neigh_lookup(dev, *(__force u32 *)pkey); |
d3aaeb38 DM |
462 | if (n) |
463 | return n; | |
32092ecf | 464 | return neigh_create(&arp_tbl, pkey, dev); |
d3aaeb38 DM |
465 | } |
466 | ||
1da177e4 LT |
467 | /* |
468 | * Peer allocation may fail only in serious out-of-memory conditions. However | |
469 | * we still can generate some output. | |
470 | * Random ID selection looks a bit dangerous because we have no chances to | |
471 | * select ID being unique in a reasonable period of time. | |
472 | * But broken packet identifier may be better than no packet at all. | |
473 | */ | |
474 | static void ip_select_fb_ident(struct iphdr *iph) | |
475 | { | |
476 | static DEFINE_SPINLOCK(ip_fb_id_lock); | |
477 | static u32 ip_fallback_id; | |
478 | u32 salt; | |
479 | ||
480 | spin_lock_bh(&ip_fb_id_lock); | |
e448515c | 481 | salt = secure_ip_id((__force __be32)ip_fallback_id ^ iph->daddr); |
1da177e4 LT |
482 | iph->id = htons(salt & 0xFFFF); |
483 | ip_fallback_id = salt; | |
484 | spin_unlock_bh(&ip_fb_id_lock); | |
485 | } | |
486 | ||
487 | void __ip_select_ident(struct iphdr *iph, struct dst_entry *dst, int more) | |
488 | { | |
1d861aa4 DM |
489 | struct net *net = dev_net(dst->dev); |
490 | struct inet_peer *peer; | |
1da177e4 | 491 | |
1d861aa4 DM |
492 | peer = inet_getpeer_v4(net->ipv4.peers, iph->daddr, 1); |
493 | if (peer) { | |
494 | iph->id = htons(inet_getid(peer, more)); | |
495 | inet_putpeer(peer); | |
496 | return; | |
497 | } | |
1da177e4 LT |
498 | |
499 | ip_select_fb_ident(iph); | |
500 | } | |
4bc2f18b | 501 | EXPORT_SYMBOL(__ip_select_ident); |
1da177e4 | 502 | |
5abf7f7e | 503 | static void __build_flow_key(struct flowi4 *fl4, const struct sock *sk, |
4895c771 DM |
504 | const struct iphdr *iph, |
505 | int oif, u8 tos, | |
506 | u8 prot, u32 mark, int flow_flags) | |
507 | { | |
508 | if (sk) { | |
509 | const struct inet_sock *inet = inet_sk(sk); | |
510 | ||
511 | oif = sk->sk_bound_dev_if; | |
512 | mark = sk->sk_mark; | |
513 | tos = RT_CONN_FLAGS(sk); | |
514 | prot = inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol; | |
515 | } | |
516 | flowi4_init_output(fl4, oif, mark, tos, | |
517 | RT_SCOPE_UNIVERSE, prot, | |
518 | flow_flags, | |
519 | iph->daddr, iph->saddr, 0, 0); | |
520 | } | |
521 | ||
5abf7f7e ED |
522 | static void build_skb_flow_key(struct flowi4 *fl4, const struct sk_buff *skb, |
523 | const struct sock *sk) | |
4895c771 DM |
524 | { |
525 | const struct iphdr *iph = ip_hdr(skb); | |
526 | int oif = skb->dev->ifindex; | |
527 | u8 tos = RT_TOS(iph->tos); | |
528 | u8 prot = iph->protocol; | |
529 | u32 mark = skb->mark; | |
530 | ||
531 | __build_flow_key(fl4, sk, iph, oif, tos, prot, mark, 0); | |
532 | } | |
533 | ||
5abf7f7e | 534 | static void build_sk_flow_key(struct flowi4 *fl4, const struct sock *sk) |
4895c771 DM |
535 | { |
536 | const struct inet_sock *inet = inet_sk(sk); | |
5abf7f7e | 537 | const struct ip_options_rcu *inet_opt; |
4895c771 DM |
538 | __be32 daddr = inet->inet_daddr; |
539 | ||
540 | rcu_read_lock(); | |
541 | inet_opt = rcu_dereference(inet->inet_opt); | |
542 | if (inet_opt && inet_opt->opt.srr) | |
543 | daddr = inet_opt->opt.faddr; | |
544 | flowi4_init_output(fl4, sk->sk_bound_dev_if, sk->sk_mark, | |
545 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, | |
546 | inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol, | |
547 | inet_sk_flowi_flags(sk), | |
548 | daddr, inet->inet_saddr, 0, 0); | |
549 | rcu_read_unlock(); | |
550 | } | |
551 | ||
5abf7f7e ED |
552 | static void ip_rt_build_flow_key(struct flowi4 *fl4, const struct sock *sk, |
553 | const struct sk_buff *skb) | |
4895c771 DM |
554 | { |
555 | if (skb) | |
556 | build_skb_flow_key(fl4, skb, sk); | |
557 | else | |
558 | build_sk_flow_key(fl4, sk); | |
559 | } | |
560 | ||
c5038a83 DM |
561 | static inline void rt_free(struct rtable *rt) |
562 | { | |
563 | call_rcu(&rt->dst.rcu_head, dst_rcu_free); | |
564 | } | |
565 | ||
566 | static DEFINE_SPINLOCK(fnhe_lock); | |
4895c771 | 567 | |
aee06da6 | 568 | static struct fib_nh_exception *fnhe_oldest(struct fnhe_hash_bucket *hash) |
4895c771 DM |
569 | { |
570 | struct fib_nh_exception *fnhe, *oldest; | |
c5038a83 | 571 | struct rtable *orig; |
4895c771 DM |
572 | |
573 | oldest = rcu_dereference(hash->chain); | |
574 | for (fnhe = rcu_dereference(oldest->fnhe_next); fnhe; | |
575 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
576 | if (time_before(fnhe->fnhe_stamp, oldest->fnhe_stamp)) | |
577 | oldest = fnhe; | |
578 | } | |
c5038a83 DM |
579 | orig = rcu_dereference(oldest->fnhe_rth); |
580 | if (orig) { | |
581 | RCU_INIT_POINTER(oldest->fnhe_rth, NULL); | |
582 | rt_free(orig); | |
583 | } | |
4895c771 DM |
584 | return oldest; |
585 | } | |
586 | ||
d3a25c98 DM |
587 | static inline u32 fnhe_hashfun(__be32 daddr) |
588 | { | |
589 | u32 hval; | |
590 | ||
591 | hval = (__force u32) daddr; | |
592 | hval ^= (hval >> 11) ^ (hval >> 22); | |
593 | ||
594 | return hval & (FNHE_HASH_SIZE - 1); | |
595 | } | |
596 | ||
387aa65a TT |
597 | static void fill_route_from_fnhe(struct rtable *rt, struct fib_nh_exception *fnhe) |
598 | { | |
599 | rt->rt_pmtu = fnhe->fnhe_pmtu; | |
600 | rt->dst.expires = fnhe->fnhe_expires; | |
601 | ||
602 | if (fnhe->fnhe_gw) { | |
603 | rt->rt_flags |= RTCF_REDIRECTED; | |
604 | rt->rt_gateway = fnhe->fnhe_gw; | |
605 | rt->rt_uses_gateway = 1; | |
606 | } | |
607 | } | |
608 | ||
aee06da6 JA |
609 | static void update_or_create_fnhe(struct fib_nh *nh, __be32 daddr, __be32 gw, |
610 | u32 pmtu, unsigned long expires) | |
4895c771 | 611 | { |
aee06da6 | 612 | struct fnhe_hash_bucket *hash; |
4895c771 | 613 | struct fib_nh_exception *fnhe; |
387aa65a TT |
614 | struct rtable *rt; |
615 | unsigned int i; | |
4895c771 | 616 | int depth; |
aee06da6 JA |
617 | u32 hval = fnhe_hashfun(daddr); |
618 | ||
c5038a83 | 619 | spin_lock_bh(&fnhe_lock); |
4895c771 | 620 | |
aee06da6 | 621 | hash = nh->nh_exceptions; |
4895c771 | 622 | if (!hash) { |
aee06da6 | 623 | hash = kzalloc(FNHE_HASH_SIZE * sizeof(*hash), GFP_ATOMIC); |
4895c771 | 624 | if (!hash) |
aee06da6 JA |
625 | goto out_unlock; |
626 | nh->nh_exceptions = hash; | |
4895c771 DM |
627 | } |
628 | ||
4895c771 DM |
629 | hash += hval; |
630 | ||
631 | depth = 0; | |
632 | for (fnhe = rcu_dereference(hash->chain); fnhe; | |
633 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
634 | if (fnhe->fnhe_daddr == daddr) | |
aee06da6 | 635 | break; |
4895c771 DM |
636 | depth++; |
637 | } | |
638 | ||
aee06da6 JA |
639 | if (fnhe) { |
640 | if (gw) | |
641 | fnhe->fnhe_gw = gw; | |
642 | if (pmtu) { | |
643 | fnhe->fnhe_pmtu = pmtu; | |
387aa65a | 644 | fnhe->fnhe_expires = max(1UL, expires); |
aee06da6 | 645 | } |
387aa65a TT |
646 | /* Update all cached dsts too */ |
647 | rt = rcu_dereference(fnhe->fnhe_rth); | |
648 | if (rt) | |
649 | fill_route_from_fnhe(rt, fnhe); | |
aee06da6 JA |
650 | } else { |
651 | if (depth > FNHE_RECLAIM_DEPTH) | |
652 | fnhe = fnhe_oldest(hash); | |
653 | else { | |
654 | fnhe = kzalloc(sizeof(*fnhe), GFP_ATOMIC); | |
655 | if (!fnhe) | |
656 | goto out_unlock; | |
657 | ||
658 | fnhe->fnhe_next = hash->chain; | |
659 | rcu_assign_pointer(hash->chain, fnhe); | |
660 | } | |
5aad1de5 | 661 | fnhe->fnhe_genid = fnhe_genid(dev_net(nh->nh_dev)); |
aee06da6 JA |
662 | fnhe->fnhe_daddr = daddr; |
663 | fnhe->fnhe_gw = gw; | |
664 | fnhe->fnhe_pmtu = pmtu; | |
665 | fnhe->fnhe_expires = expires; | |
387aa65a TT |
666 | |
667 | /* Exception created; mark the cached routes for the nexthop | |
668 | * stale, so anyone caching it rechecks if this exception | |
669 | * applies to them. | |
670 | */ | |
671 | for_each_possible_cpu(i) { | |
672 | struct rtable __rcu **prt; | |
673 | prt = per_cpu_ptr(nh->nh_pcpu_rth_output, i); | |
674 | rt = rcu_dereference(*prt); | |
675 | if (rt) | |
676 | rt->dst.obsolete = DST_OBSOLETE_KILL; | |
677 | } | |
4895c771 | 678 | } |
4895c771 | 679 | |
4895c771 | 680 | fnhe->fnhe_stamp = jiffies; |
aee06da6 JA |
681 | |
682 | out_unlock: | |
c5038a83 | 683 | spin_unlock_bh(&fnhe_lock); |
aee06da6 | 684 | return; |
4895c771 DM |
685 | } |
686 | ||
ceb33206 DM |
687 | static void __ip_do_redirect(struct rtable *rt, struct sk_buff *skb, struct flowi4 *fl4, |
688 | bool kill_route) | |
1da177e4 | 689 | { |
e47a185b | 690 | __be32 new_gw = icmp_hdr(skb)->un.gateway; |
94206125 | 691 | __be32 old_gw = ip_hdr(skb)->saddr; |
e47a185b | 692 | struct net_device *dev = skb->dev; |
e47a185b | 693 | struct in_device *in_dev; |
4895c771 | 694 | struct fib_result res; |
e47a185b | 695 | struct neighbour *n; |
317805b8 | 696 | struct net *net; |
1da177e4 | 697 | |
94206125 DM |
698 | switch (icmp_hdr(skb)->code & 7) { |
699 | case ICMP_REDIR_NET: | |
700 | case ICMP_REDIR_NETTOS: | |
701 | case ICMP_REDIR_HOST: | |
702 | case ICMP_REDIR_HOSTTOS: | |
703 | break; | |
704 | ||
705 | default: | |
706 | return; | |
707 | } | |
708 | ||
e47a185b DM |
709 | if (rt->rt_gateway != old_gw) |
710 | return; | |
711 | ||
712 | in_dev = __in_dev_get_rcu(dev); | |
713 | if (!in_dev) | |
714 | return; | |
715 | ||
c346dca1 | 716 | net = dev_net(dev); |
9d4fb27d JP |
717 | if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) || |
718 | ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) || | |
719 | ipv4_is_zeronet(new_gw)) | |
1da177e4 LT |
720 | goto reject_redirect; |
721 | ||
722 | if (!IN_DEV_SHARED_MEDIA(in_dev)) { | |
723 | if (!inet_addr_onlink(in_dev, new_gw, old_gw)) | |
724 | goto reject_redirect; | |
725 | if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev)) | |
726 | goto reject_redirect; | |
727 | } else { | |
317805b8 | 728 | if (inet_addr_type(net, new_gw) != RTN_UNICAST) |
1da177e4 LT |
729 | goto reject_redirect; |
730 | } | |
731 | ||
4895c771 | 732 | n = ipv4_neigh_lookup(&rt->dst, NULL, &new_gw); |
e47a185b DM |
733 | if (n) { |
734 | if (!(n->nud_state & NUD_VALID)) { | |
735 | neigh_event_send(n, NULL); | |
736 | } else { | |
4895c771 DM |
737 | if (fib_lookup(net, fl4, &res) == 0) { |
738 | struct fib_nh *nh = &FIB_RES_NH(res); | |
4895c771 | 739 | |
aee06da6 JA |
740 | update_or_create_fnhe(nh, fl4->daddr, new_gw, |
741 | 0, 0); | |
4895c771 | 742 | } |
ceb33206 DM |
743 | if (kill_route) |
744 | rt->dst.obsolete = DST_OBSOLETE_KILL; | |
e47a185b DM |
745 | call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, n); |
746 | } | |
747 | neigh_release(n); | |
748 | } | |
749 | return; | |
750 | ||
751 | reject_redirect: | |
752 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
99ee038d DM |
753 | if (IN_DEV_LOG_MARTIANS(in_dev)) { |
754 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
755 | __be32 daddr = iph->daddr; | |
756 | __be32 saddr = iph->saddr; | |
757 | ||
e47a185b DM |
758 | net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n" |
759 | " Advised path = %pI4 -> %pI4\n", | |
760 | &old_gw, dev->name, &new_gw, | |
761 | &saddr, &daddr); | |
99ee038d | 762 | } |
e47a185b DM |
763 | #endif |
764 | ; | |
765 | } | |
766 | ||
4895c771 DM |
767 | static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb) |
768 | { | |
769 | struct rtable *rt; | |
770 | struct flowi4 fl4; | |
f96ef988 MK |
771 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
772 | int oif = skb->dev->ifindex; | |
773 | u8 tos = RT_TOS(iph->tos); | |
774 | u8 prot = iph->protocol; | |
775 | u32 mark = skb->mark; | |
4895c771 DM |
776 | |
777 | rt = (struct rtable *) dst; | |
778 | ||
f96ef988 | 779 | __build_flow_key(&fl4, sk, iph, oif, tos, prot, mark, 0); |
ceb33206 | 780 | __ip_do_redirect(rt, skb, &fl4, true); |
4895c771 DM |
781 | } |
782 | ||
1da177e4 LT |
783 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst) |
784 | { | |
ee6b9673 | 785 | struct rtable *rt = (struct rtable *)dst; |
1da177e4 LT |
786 | struct dst_entry *ret = dst; |
787 | ||
788 | if (rt) { | |
d11a4dc1 | 789 | if (dst->obsolete > 0) { |
1da177e4 LT |
790 | ip_rt_put(rt); |
791 | ret = NULL; | |
5943634f DM |
792 | } else if ((rt->rt_flags & RTCF_REDIRECTED) || |
793 | rt->dst.expires) { | |
89aef892 | 794 | ip_rt_put(rt); |
1da177e4 LT |
795 | ret = NULL; |
796 | } | |
797 | } | |
798 | return ret; | |
799 | } | |
800 | ||
801 | /* | |
802 | * Algorithm: | |
803 | * 1. The first ip_rt_redirect_number redirects are sent | |
804 | * with exponential backoff, then we stop sending them at all, | |
805 | * assuming that the host ignores our redirects. | |
806 | * 2. If we did not see packets requiring redirects | |
807 | * during ip_rt_redirect_silence, we assume that the host | |
808 | * forgot redirected route and start to send redirects again. | |
809 | * | |
810 | * This algorithm is much cheaper and more intelligent than dumb load limiting | |
811 | * in icmp.c. | |
812 | * | |
813 | * NOTE. Do not forget to inhibit load limiting for redirects (redundant) | |
814 | * and "frag. need" (breaks PMTU discovery) in icmp.c. | |
815 | */ | |
816 | ||
817 | void ip_rt_send_redirect(struct sk_buff *skb) | |
818 | { | |
511c3f92 | 819 | struct rtable *rt = skb_rtable(skb); |
30038fc6 | 820 | struct in_device *in_dev; |
92d86829 | 821 | struct inet_peer *peer; |
1d861aa4 | 822 | struct net *net; |
30038fc6 | 823 | int log_martians; |
1da177e4 | 824 | |
30038fc6 | 825 | rcu_read_lock(); |
d8d1f30b | 826 | in_dev = __in_dev_get_rcu(rt->dst.dev); |
30038fc6 ED |
827 | if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) { |
828 | rcu_read_unlock(); | |
1da177e4 | 829 | return; |
30038fc6 ED |
830 | } |
831 | log_martians = IN_DEV_LOG_MARTIANS(in_dev); | |
832 | rcu_read_unlock(); | |
1da177e4 | 833 | |
1d861aa4 DM |
834 | net = dev_net(rt->dst.dev); |
835 | peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1); | |
92d86829 | 836 | if (!peer) { |
e81da0e1 JA |
837 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, |
838 | rt_nexthop(rt, ip_hdr(skb)->daddr)); | |
92d86829 DM |
839 | return; |
840 | } | |
841 | ||
1da177e4 LT |
842 | /* No redirected packets during ip_rt_redirect_silence; |
843 | * reset the algorithm. | |
844 | */ | |
92d86829 DM |
845 | if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence)) |
846 | peer->rate_tokens = 0; | |
1da177e4 LT |
847 | |
848 | /* Too many ignored redirects; do not send anything | |
d8d1f30b | 849 | * set dst.rate_last to the last seen redirected packet. |
1da177e4 | 850 | */ |
92d86829 DM |
851 | if (peer->rate_tokens >= ip_rt_redirect_number) { |
852 | peer->rate_last = jiffies; | |
1d861aa4 | 853 | goto out_put_peer; |
1da177e4 LT |
854 | } |
855 | ||
856 | /* Check for load limit; set rate_last to the latest sent | |
857 | * redirect. | |
858 | */ | |
92d86829 | 859 | if (peer->rate_tokens == 0 || |
14fb8a76 | 860 | time_after(jiffies, |
92d86829 DM |
861 | (peer->rate_last + |
862 | (ip_rt_redirect_load << peer->rate_tokens)))) { | |
e81da0e1 JA |
863 | __be32 gw = rt_nexthop(rt, ip_hdr(skb)->daddr); |
864 | ||
865 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, gw); | |
92d86829 DM |
866 | peer->rate_last = jiffies; |
867 | ++peer->rate_tokens; | |
1da177e4 | 868 | #ifdef CONFIG_IP_ROUTE_VERBOSE |
30038fc6 | 869 | if (log_martians && |
e87cc472 JP |
870 | peer->rate_tokens == ip_rt_redirect_number) |
871 | net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n", | |
92101b3b | 872 | &ip_hdr(skb)->saddr, inet_iif(skb), |
e81da0e1 | 873 | &ip_hdr(skb)->daddr, &gw); |
1da177e4 LT |
874 | #endif |
875 | } | |
1d861aa4 DM |
876 | out_put_peer: |
877 | inet_putpeer(peer); | |
1da177e4 LT |
878 | } |
879 | ||
880 | static int ip_error(struct sk_buff *skb) | |
881 | { | |
251da413 | 882 | struct in_device *in_dev = __in_dev_get_rcu(skb->dev); |
511c3f92 | 883 | struct rtable *rt = skb_rtable(skb); |
92d86829 | 884 | struct inet_peer *peer; |
1da177e4 | 885 | unsigned long now; |
251da413 | 886 | struct net *net; |
92d86829 | 887 | bool send; |
1da177e4 LT |
888 | int code; |
889 | ||
251da413 DM |
890 | net = dev_net(rt->dst.dev); |
891 | if (!IN_DEV_FORWARD(in_dev)) { | |
892 | switch (rt->dst.error) { | |
893 | case EHOSTUNREACH: | |
894 | IP_INC_STATS_BH(net, IPSTATS_MIB_INADDRERRORS); | |
895 | break; | |
896 | ||
897 | case ENETUNREACH: | |
898 | IP_INC_STATS_BH(net, IPSTATS_MIB_INNOROUTES); | |
899 | break; | |
900 | } | |
901 | goto out; | |
902 | } | |
903 | ||
d8d1f30b | 904 | switch (rt->dst.error) { |
4500ebf8 JP |
905 | case EINVAL: |
906 | default: | |
907 | goto out; | |
908 | case EHOSTUNREACH: | |
909 | code = ICMP_HOST_UNREACH; | |
910 | break; | |
911 | case ENETUNREACH: | |
912 | code = ICMP_NET_UNREACH; | |
251da413 | 913 | IP_INC_STATS_BH(net, IPSTATS_MIB_INNOROUTES); |
4500ebf8 JP |
914 | break; |
915 | case EACCES: | |
916 | code = ICMP_PKT_FILTERED; | |
917 | break; | |
1da177e4 LT |
918 | } |
919 | ||
1d861aa4 | 920 | peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, 1); |
92d86829 DM |
921 | |
922 | send = true; | |
923 | if (peer) { | |
924 | now = jiffies; | |
925 | peer->rate_tokens += now - peer->rate_last; | |
926 | if (peer->rate_tokens > ip_rt_error_burst) | |
927 | peer->rate_tokens = ip_rt_error_burst; | |
928 | peer->rate_last = now; | |
929 | if (peer->rate_tokens >= ip_rt_error_cost) | |
930 | peer->rate_tokens -= ip_rt_error_cost; | |
931 | else | |
932 | send = false; | |
1d861aa4 | 933 | inet_putpeer(peer); |
1da177e4 | 934 | } |
92d86829 DM |
935 | if (send) |
936 | icmp_send(skb, ICMP_DEST_UNREACH, code, 0); | |
1da177e4 LT |
937 | |
938 | out: kfree_skb(skb); | |
939 | return 0; | |
e905a9ed | 940 | } |
1da177e4 | 941 | |
d851c12b | 942 | static void __ip_rt_update_pmtu(struct rtable *rt, struct flowi4 *fl4, u32 mtu) |
1da177e4 | 943 | { |
d851c12b | 944 | struct dst_entry *dst = &rt->dst; |
4895c771 | 945 | struct fib_result res; |
2c8cec5c | 946 | |
fa1e492a SK |
947 | if (dst_metric_locked(dst, RTAX_MTU)) |
948 | return; | |
949 | ||
7f92d334 SK |
950 | if (dst->dev->mtu < mtu) |
951 | return; | |
952 | ||
5943634f DM |
953 | if (mtu < ip_rt_min_pmtu) |
954 | mtu = ip_rt_min_pmtu; | |
2c8cec5c | 955 | |
f016229e TT |
956 | if (rt->rt_pmtu == mtu && |
957 | time_before(jiffies, dst->expires - ip_rt_mtu_expires / 2)) | |
958 | return; | |
959 | ||
c5ae7d41 | 960 | rcu_read_lock(); |
d851c12b | 961 | if (fib_lookup(dev_net(dst->dev), fl4, &res) == 0) { |
4895c771 | 962 | struct fib_nh *nh = &FIB_RES_NH(res); |
4895c771 | 963 | |
aee06da6 JA |
964 | update_or_create_fnhe(nh, fl4->daddr, 0, mtu, |
965 | jiffies + ip_rt_mtu_expires); | |
4895c771 | 966 | } |
c5ae7d41 | 967 | rcu_read_unlock(); |
1da177e4 LT |
968 | } |
969 | ||
4895c771 DM |
970 | static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk, |
971 | struct sk_buff *skb, u32 mtu) | |
972 | { | |
973 | struct rtable *rt = (struct rtable *) dst; | |
974 | struct flowi4 fl4; | |
975 | ||
976 | ip_rt_build_flow_key(&fl4, sk, skb); | |
d851c12b | 977 | __ip_rt_update_pmtu(rt, &fl4, mtu); |
4895c771 DM |
978 | } |
979 | ||
36393395 DM |
980 | void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, |
981 | int oif, u32 mark, u8 protocol, int flow_flags) | |
982 | { | |
4895c771 | 983 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
36393395 DM |
984 | struct flowi4 fl4; |
985 | struct rtable *rt; | |
986 | ||
4895c771 DM |
987 | __build_flow_key(&fl4, NULL, iph, oif, |
988 | RT_TOS(iph->tos), protocol, mark, flow_flags); | |
36393395 DM |
989 | rt = __ip_route_output_key(net, &fl4); |
990 | if (!IS_ERR(rt)) { | |
4895c771 | 991 | __ip_rt_update_pmtu(rt, &fl4, mtu); |
36393395 DM |
992 | ip_rt_put(rt); |
993 | } | |
994 | } | |
995 | EXPORT_SYMBOL_GPL(ipv4_update_pmtu); | |
996 | ||
9cb3a50c | 997 | static void __ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu) |
36393395 | 998 | { |
4895c771 DM |
999 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
1000 | struct flowi4 fl4; | |
1001 | struct rtable *rt; | |
36393395 | 1002 | |
4895c771 DM |
1003 | __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0); |
1004 | rt = __ip_route_output_key(sock_net(sk), &fl4); | |
1005 | if (!IS_ERR(rt)) { | |
1006 | __ip_rt_update_pmtu(rt, &fl4, mtu); | |
1007 | ip_rt_put(rt); | |
1008 | } | |
36393395 | 1009 | } |
9cb3a50c SK |
1010 | |
1011 | void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu) | |
1012 | { | |
1013 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
1014 | struct flowi4 fl4; | |
1015 | struct rtable *rt; | |
1016 | struct dst_entry *dst; | |
b44108db | 1017 | bool new = false; |
9cb3a50c SK |
1018 | |
1019 | bh_lock_sock(sk); | |
1020 | rt = (struct rtable *) __sk_dst_get(sk); | |
1021 | ||
1022 | if (sock_owned_by_user(sk) || !rt) { | |
1023 | __ipv4_sk_update_pmtu(skb, sk, mtu); | |
1024 | goto out; | |
1025 | } | |
1026 | ||
1027 | __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0); | |
1028 | ||
1029 | if (!__sk_dst_check(sk, 0)) { | |
1030 | rt = ip_route_output_flow(sock_net(sk), &fl4, sk); | |
1031 | if (IS_ERR(rt)) | |
1032 | goto out; | |
b44108db SK |
1033 | |
1034 | new = true; | |
9cb3a50c SK |
1035 | } |
1036 | ||
1037 | __ip_rt_update_pmtu((struct rtable *) rt->dst.path, &fl4, mtu); | |
1038 | ||
1039 | dst = dst_check(&rt->dst, 0); | |
1040 | if (!dst) { | |
b44108db SK |
1041 | if (new) |
1042 | dst_release(&rt->dst); | |
1043 | ||
9cb3a50c SK |
1044 | rt = ip_route_output_flow(sock_net(sk), &fl4, sk); |
1045 | if (IS_ERR(rt)) | |
1046 | goto out; | |
1047 | ||
b44108db | 1048 | new = true; |
9cb3a50c SK |
1049 | } |
1050 | ||
b44108db SK |
1051 | if (new) |
1052 | __sk_dst_set(sk, &rt->dst); | |
9cb3a50c SK |
1053 | |
1054 | out: | |
1055 | bh_unlock_sock(sk); | |
1056 | } | |
36393395 | 1057 | EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu); |
f39925db | 1058 | |
b42597e2 DM |
1059 | void ipv4_redirect(struct sk_buff *skb, struct net *net, |
1060 | int oif, u32 mark, u8 protocol, int flow_flags) | |
1061 | { | |
4895c771 | 1062 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
b42597e2 DM |
1063 | struct flowi4 fl4; |
1064 | struct rtable *rt; | |
1065 | ||
4895c771 DM |
1066 | __build_flow_key(&fl4, NULL, iph, oif, |
1067 | RT_TOS(iph->tos), protocol, mark, flow_flags); | |
b42597e2 DM |
1068 | rt = __ip_route_output_key(net, &fl4); |
1069 | if (!IS_ERR(rt)) { | |
ceb33206 | 1070 | __ip_do_redirect(rt, skb, &fl4, false); |
b42597e2 DM |
1071 | ip_rt_put(rt); |
1072 | } | |
1073 | } | |
1074 | EXPORT_SYMBOL_GPL(ipv4_redirect); | |
1075 | ||
1076 | void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk) | |
1077 | { | |
4895c771 DM |
1078 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
1079 | struct flowi4 fl4; | |
1080 | struct rtable *rt; | |
b42597e2 | 1081 | |
4895c771 DM |
1082 | __build_flow_key(&fl4, sk, iph, 0, 0, 0, 0, 0); |
1083 | rt = __ip_route_output_key(sock_net(sk), &fl4); | |
1084 | if (!IS_ERR(rt)) { | |
ceb33206 | 1085 | __ip_do_redirect(rt, skb, &fl4, false); |
4895c771 DM |
1086 | ip_rt_put(rt); |
1087 | } | |
b42597e2 DM |
1088 | } |
1089 | EXPORT_SYMBOL_GPL(ipv4_sk_redirect); | |
1090 | ||
efbc368d DM |
1091 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie) |
1092 | { | |
1093 | struct rtable *rt = (struct rtable *) dst; | |
1094 | ||
ceb33206 DM |
1095 | /* All IPV4 dsts are created with ->obsolete set to the value |
1096 | * DST_OBSOLETE_FORCE_CHK which forces validation calls down | |
1097 | * into this function always. | |
1098 | * | |
387aa65a TT |
1099 | * When a PMTU/redirect information update invalidates a route, |
1100 | * this is indicated by setting obsolete to DST_OBSOLETE_KILL or | |
1101 | * DST_OBSOLETE_DEAD by dst_free(). | |
ceb33206 | 1102 | */ |
387aa65a | 1103 | if (dst->obsolete != DST_OBSOLETE_FORCE_CHK || rt_is_expired(rt)) |
efbc368d | 1104 | return NULL; |
d11a4dc1 | 1105 | return dst; |
1da177e4 LT |
1106 | } |
1107 | ||
1da177e4 LT |
1108 | static void ipv4_link_failure(struct sk_buff *skb) |
1109 | { | |
1110 | struct rtable *rt; | |
1111 | ||
1112 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0); | |
1113 | ||
511c3f92 | 1114 | rt = skb_rtable(skb); |
5943634f DM |
1115 | if (rt) |
1116 | dst_set_expires(&rt->dst, 0); | |
1da177e4 LT |
1117 | } |
1118 | ||
1119 | static int ip_rt_bug(struct sk_buff *skb) | |
1120 | { | |
91df42be JP |
1121 | pr_debug("%s: %pI4 -> %pI4, %s\n", |
1122 | __func__, &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
1123 | skb->dev ? skb->dev->name : "?"); | |
1da177e4 | 1124 | kfree_skb(skb); |
c378a9c0 | 1125 | WARN_ON(1); |
1da177e4 LT |
1126 | return 0; |
1127 | } | |
1128 | ||
1129 | /* | |
1130 | We do not cache source address of outgoing interface, | |
1131 | because it is used only by IP RR, TS and SRR options, | |
1132 | so that it out of fast path. | |
1133 | ||
1134 | BTW remember: "addr" is allowed to be not aligned | |
1135 | in IP options! | |
1136 | */ | |
1137 | ||
8e36360a | 1138 | void ip_rt_get_source(u8 *addr, struct sk_buff *skb, struct rtable *rt) |
1da177e4 | 1139 | { |
a61ced5d | 1140 | __be32 src; |
1da177e4 | 1141 | |
c7537967 | 1142 | if (rt_is_output_route(rt)) |
c5be24ff | 1143 | src = ip_hdr(skb)->saddr; |
ebc0ffae | 1144 | else { |
8e36360a DM |
1145 | struct fib_result res; |
1146 | struct flowi4 fl4; | |
1147 | struct iphdr *iph; | |
1148 | ||
1149 | iph = ip_hdr(skb); | |
1150 | ||
1151 | memset(&fl4, 0, sizeof(fl4)); | |
1152 | fl4.daddr = iph->daddr; | |
1153 | fl4.saddr = iph->saddr; | |
b0fe4a31 | 1154 | fl4.flowi4_tos = RT_TOS(iph->tos); |
8e36360a DM |
1155 | fl4.flowi4_oif = rt->dst.dev->ifindex; |
1156 | fl4.flowi4_iif = skb->dev->ifindex; | |
1157 | fl4.flowi4_mark = skb->mark; | |
5e2b61f7 | 1158 | |
ebc0ffae | 1159 | rcu_read_lock(); |
68a5e3dd | 1160 | if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res) == 0) |
436c3b66 | 1161 | src = FIB_RES_PREFSRC(dev_net(rt->dst.dev), res); |
ebc0ffae | 1162 | else |
f8126f1d DM |
1163 | src = inet_select_addr(rt->dst.dev, |
1164 | rt_nexthop(rt, iph->daddr), | |
1165 | RT_SCOPE_UNIVERSE); | |
ebc0ffae ED |
1166 | rcu_read_unlock(); |
1167 | } | |
1da177e4 LT |
1168 | memcpy(addr, &src, 4); |
1169 | } | |
1170 | ||
c7066f70 | 1171 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 LT |
1172 | static void set_class_tag(struct rtable *rt, u32 tag) |
1173 | { | |
d8d1f30b CG |
1174 | if (!(rt->dst.tclassid & 0xFFFF)) |
1175 | rt->dst.tclassid |= tag & 0xFFFF; | |
1176 | if (!(rt->dst.tclassid & 0xFFFF0000)) | |
1177 | rt->dst.tclassid |= tag & 0xFFFF0000; | |
1da177e4 LT |
1178 | } |
1179 | #endif | |
1180 | ||
0dbaee3b DM |
1181 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst) |
1182 | { | |
1183 | unsigned int advmss = dst_metric_raw(dst, RTAX_ADVMSS); | |
1184 | ||
1185 | if (advmss == 0) { | |
1186 | advmss = max_t(unsigned int, dst->dev->mtu - 40, | |
1187 | ip_rt_min_advmss); | |
1188 | if (advmss > 65535 - 40) | |
1189 | advmss = 65535 - 40; | |
1190 | } | |
1191 | return advmss; | |
1192 | } | |
1193 | ||
ebb762f2 | 1194 | static unsigned int ipv4_mtu(const struct dst_entry *dst) |
d33e4553 | 1195 | { |
261663b0 | 1196 | const struct rtable *rt = (const struct rtable *) dst; |
5943634f DM |
1197 | unsigned int mtu = rt->rt_pmtu; |
1198 | ||
98d75c37 | 1199 | if (!mtu || time_after_eq(jiffies, rt->dst.expires)) |
5943634f | 1200 | mtu = dst_metric_raw(dst, RTAX_MTU); |
618f9bc7 | 1201 | |
38d523e2 | 1202 | if (mtu) |
618f9bc7 SK |
1203 | return mtu; |
1204 | ||
1205 | mtu = dst->dev->mtu; | |
d33e4553 DM |
1206 | |
1207 | if (unlikely(dst_metric_locked(dst, RTAX_MTU))) { | |
155e8336 | 1208 | if (rt->rt_uses_gateway && mtu > 576) |
d33e4553 DM |
1209 | mtu = 576; |
1210 | } | |
1211 | ||
1212 | if (mtu > IP_MAX_MTU) | |
1213 | mtu = IP_MAX_MTU; | |
1214 | ||
1215 | return mtu; | |
1216 | } | |
1217 | ||
f2bb4bed | 1218 | static struct fib_nh_exception *find_exception(struct fib_nh *nh, __be32 daddr) |
4895c771 DM |
1219 | { |
1220 | struct fnhe_hash_bucket *hash = nh->nh_exceptions; | |
1221 | struct fib_nh_exception *fnhe; | |
1222 | u32 hval; | |
1223 | ||
f2bb4bed DM |
1224 | if (!hash) |
1225 | return NULL; | |
1226 | ||
d3a25c98 | 1227 | hval = fnhe_hashfun(daddr); |
4895c771 DM |
1228 | |
1229 | for (fnhe = rcu_dereference(hash[hval].chain); fnhe; | |
1230 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
f2bb4bed DM |
1231 | if (fnhe->fnhe_daddr == daddr) |
1232 | return fnhe; | |
1233 | } | |
1234 | return NULL; | |
1235 | } | |
aee06da6 | 1236 | |
caacf05e | 1237 | static bool rt_bind_exception(struct rtable *rt, struct fib_nh_exception *fnhe, |
f2bb4bed DM |
1238 | __be32 daddr) |
1239 | { | |
caacf05e DM |
1240 | bool ret = false; |
1241 | ||
c5038a83 | 1242 | spin_lock_bh(&fnhe_lock); |
f2bb4bed | 1243 | |
c5038a83 | 1244 | if (daddr == fnhe->fnhe_daddr) { |
5aad1de5 | 1245 | int genid = fnhe_genid(dev_net(rt->dst.dev)); |
13d82bf5 | 1246 | struct rtable *orig = rcu_dereference(fnhe->fnhe_rth); |
5aad1de5 TT |
1247 | |
1248 | if (fnhe->fnhe_genid != genid) { | |
1249 | fnhe->fnhe_genid = genid; | |
13d82bf5 SK |
1250 | fnhe->fnhe_gw = 0; |
1251 | fnhe->fnhe_pmtu = 0; | |
1252 | fnhe->fnhe_expires = 0; | |
1253 | } | |
387aa65a TT |
1254 | fill_route_from_fnhe(rt, fnhe); |
1255 | if (!rt->rt_gateway) | |
155e8336 | 1256 | rt->rt_gateway = daddr; |
f2bb4bed | 1257 | |
c5038a83 DM |
1258 | rcu_assign_pointer(fnhe->fnhe_rth, rt); |
1259 | if (orig) | |
1260 | rt_free(orig); | |
1261 | ||
1262 | fnhe->fnhe_stamp = jiffies; | |
caacf05e | 1263 | ret = true; |
c5038a83 DM |
1264 | } |
1265 | spin_unlock_bh(&fnhe_lock); | |
caacf05e DM |
1266 | |
1267 | return ret; | |
54764bb6 ED |
1268 | } |
1269 | ||
caacf05e | 1270 | static bool rt_cache_route(struct fib_nh *nh, struct rtable *rt) |
f2bb4bed | 1271 | { |
d26b3a7c | 1272 | struct rtable *orig, *prev, **p; |
caacf05e | 1273 | bool ret = true; |
f2bb4bed | 1274 | |
d26b3a7c | 1275 | if (rt_is_input_route(rt)) { |
54764bb6 | 1276 | p = (struct rtable **)&nh->nh_rth_input; |
d26b3a7c | 1277 | } else { |
d26b3a7c ED |
1278 | p = (struct rtable **)__this_cpu_ptr(nh->nh_pcpu_rth_output); |
1279 | } | |
f2bb4bed DM |
1280 | orig = *p; |
1281 | ||
1282 | prev = cmpxchg(p, orig, rt); | |
1283 | if (prev == orig) { | |
f2bb4bed | 1284 | if (orig) |
54764bb6 | 1285 | rt_free(orig); |
155e8336 | 1286 | } else |
caacf05e | 1287 | ret = false; |
caacf05e DM |
1288 | |
1289 | return ret; | |
1290 | } | |
1291 | ||
1292 | static DEFINE_SPINLOCK(rt_uncached_lock); | |
1293 | static LIST_HEAD(rt_uncached_list); | |
1294 | ||
1295 | static void rt_add_uncached_list(struct rtable *rt) | |
1296 | { | |
1297 | spin_lock_bh(&rt_uncached_lock); | |
1298 | list_add_tail(&rt->rt_uncached, &rt_uncached_list); | |
1299 | spin_unlock_bh(&rt_uncached_lock); | |
1300 | } | |
1301 | ||
1302 | static void ipv4_dst_destroy(struct dst_entry *dst) | |
1303 | { | |
1304 | struct rtable *rt = (struct rtable *) dst; | |
1305 | ||
78df76a0 | 1306 | if (!list_empty(&rt->rt_uncached)) { |
caacf05e DM |
1307 | spin_lock_bh(&rt_uncached_lock); |
1308 | list_del(&rt->rt_uncached); | |
1309 | spin_unlock_bh(&rt_uncached_lock); | |
1310 | } | |
1311 | } | |
1312 | ||
1313 | void rt_flush_dev(struct net_device *dev) | |
1314 | { | |
1315 | if (!list_empty(&rt_uncached_list)) { | |
1316 | struct net *net = dev_net(dev); | |
1317 | struct rtable *rt; | |
1318 | ||
1319 | spin_lock_bh(&rt_uncached_lock); | |
1320 | list_for_each_entry(rt, &rt_uncached_list, rt_uncached) { | |
1321 | if (rt->dst.dev != dev) | |
1322 | continue; | |
1323 | rt->dst.dev = net->loopback_dev; | |
1324 | dev_hold(rt->dst.dev); | |
1325 | dev_put(dev); | |
1326 | } | |
1327 | spin_unlock_bh(&rt_uncached_lock); | |
4895c771 DM |
1328 | } |
1329 | } | |
1330 | ||
4331debc | 1331 | static bool rt_cache_valid(const struct rtable *rt) |
d2d68ba9 | 1332 | { |
4331debc ED |
1333 | return rt && |
1334 | rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK && | |
1335 | !rt_is_expired(rt); | |
d2d68ba9 DM |
1336 | } |
1337 | ||
f2bb4bed | 1338 | static void rt_set_nexthop(struct rtable *rt, __be32 daddr, |
5e2b61f7 | 1339 | const struct fib_result *res, |
f2bb4bed | 1340 | struct fib_nh_exception *fnhe, |
982721f3 | 1341 | struct fib_info *fi, u16 type, u32 itag) |
1da177e4 | 1342 | { |
caacf05e DM |
1343 | bool cached = false; |
1344 | ||
1da177e4 | 1345 | if (fi) { |
4895c771 DM |
1346 | struct fib_nh *nh = &FIB_RES_NH(*res); |
1347 | ||
155e8336 | 1348 | if (nh->nh_gw && nh->nh_scope == RT_SCOPE_LINK) { |
4895c771 | 1349 | rt->rt_gateway = nh->nh_gw; |
155e8336 JA |
1350 | rt->rt_uses_gateway = 1; |
1351 | } | |
2860583f | 1352 | dst_init_metrics(&rt->dst, fi->fib_metrics, true); |
c7066f70 | 1353 | #ifdef CONFIG_IP_ROUTE_CLASSID |
f2bb4bed | 1354 | rt->dst.tclassid = nh->nh_tclassid; |
1da177e4 | 1355 | #endif |
c5038a83 | 1356 | if (unlikely(fnhe)) |
caacf05e | 1357 | cached = rt_bind_exception(rt, fnhe, daddr); |
c5038a83 | 1358 | else if (!(rt->dst.flags & DST_NOCACHE)) |
caacf05e | 1359 | cached = rt_cache_route(nh, rt); |
155e8336 JA |
1360 | if (unlikely(!cached)) { |
1361 | /* Routes we intend to cache in nexthop exception or | |
1362 | * FIB nexthop have the DST_NOCACHE bit clear. | |
1363 | * However, if we are unsuccessful at storing this | |
1364 | * route into the cache we really need to set it. | |
1365 | */ | |
1366 | rt->dst.flags |= DST_NOCACHE; | |
1367 | if (!rt->rt_gateway) | |
1368 | rt->rt_gateway = daddr; | |
1369 | rt_add_uncached_list(rt); | |
1370 | } | |
1371 | } else | |
caacf05e | 1372 | rt_add_uncached_list(rt); |
defb3519 | 1373 | |
c7066f70 | 1374 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 | 1375 | #ifdef CONFIG_IP_MULTIPLE_TABLES |
85b91b03 | 1376 | set_class_tag(rt, res->tclassid); |
1da177e4 LT |
1377 | #endif |
1378 | set_class_tag(rt, itag); | |
1379 | #endif | |
1da177e4 LT |
1380 | } |
1381 | ||
5c1e6aa3 | 1382 | static struct rtable *rt_dst_alloc(struct net_device *dev, |
f2bb4bed | 1383 | bool nopolicy, bool noxfrm, bool will_cache) |
0c4dcd58 | 1384 | { |
f5b0a874 | 1385 | return dst_alloc(&ipv4_dst_ops, dev, 1, DST_OBSOLETE_FORCE_CHK, |
c6cffba4 | 1386 | (will_cache ? 0 : (DST_HOST | DST_NOCACHE)) | |
5c1e6aa3 DM |
1387 | (nopolicy ? DST_NOPOLICY : 0) | |
1388 | (noxfrm ? DST_NOXFRM : 0)); | |
0c4dcd58 DM |
1389 | } |
1390 | ||
96d36220 | 1391 | /* called in rcu_read_lock() section */ |
9e12bb22 | 1392 | static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
1393 | u8 tos, struct net_device *dev, int our) |
1394 | { | |
1da177e4 | 1395 | struct rtable *rth; |
96d36220 | 1396 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1397 | u32 itag = 0; |
b5f7e755 | 1398 | int err; |
1da177e4 LT |
1399 | |
1400 | /* Primary sanity checks. */ | |
1401 | ||
1402 | if (in_dev == NULL) | |
1403 | return -EINVAL; | |
1404 | ||
1e637c74 | 1405 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
d0daebc3 | 1406 | skb->protocol != htons(ETH_P_IP)) |
1da177e4 LT |
1407 | goto e_inval; |
1408 | ||
d0daebc3 TG |
1409 | if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev))) |
1410 | if (ipv4_is_loopback(saddr)) | |
1411 | goto e_inval; | |
1412 | ||
f97c1e0c JP |
1413 | if (ipv4_is_zeronet(saddr)) { |
1414 | if (!ipv4_is_local_multicast(daddr)) | |
1da177e4 | 1415 | goto e_inval; |
b5f7e755 | 1416 | } else { |
9e56e380 DM |
1417 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, |
1418 | in_dev, &itag); | |
b5f7e755 ED |
1419 | if (err < 0) |
1420 | goto e_err; | |
1421 | } | |
4e7b2f14 | 1422 | rth = rt_dst_alloc(dev_net(dev)->loopback_dev, |
f2bb4bed | 1423 | IN_DEV_CONF_GET(in_dev, NOPOLICY), false, false); |
1da177e4 LT |
1424 | if (!rth) |
1425 | goto e_nobufs; | |
1426 | ||
cf911662 DM |
1427 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1428 | rth->dst.tclassid = itag; | |
1429 | #endif | |
d8d1f30b | 1430 | rth->dst.output = ip_rt_bug; |
1da177e4 | 1431 | |
cf911662 DM |
1432 | rth->rt_genid = rt_genid(dev_net(dev)); |
1433 | rth->rt_flags = RTCF_MULTICAST; | |
1434 | rth->rt_type = RTN_MULTICAST; | |
9917e1e8 | 1435 | rth->rt_is_input= 1; |
13378cad | 1436 | rth->rt_iif = 0; |
5943634f | 1437 | rth->rt_pmtu = 0; |
f8126f1d | 1438 | rth->rt_gateway = 0; |
155e8336 | 1439 | rth->rt_uses_gateway = 0; |
caacf05e | 1440 | INIT_LIST_HEAD(&rth->rt_uncached); |
1da177e4 | 1441 | if (our) { |
d8d1f30b | 1442 | rth->dst.input= ip_local_deliver; |
1da177e4 LT |
1443 | rth->rt_flags |= RTCF_LOCAL; |
1444 | } | |
1445 | ||
1446 | #ifdef CONFIG_IP_MROUTE | |
f97c1e0c | 1447 | if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev)) |
d8d1f30b | 1448 | rth->dst.input = ip_mr_input; |
1da177e4 LT |
1449 | #endif |
1450 | RT_CACHE_STAT_INC(in_slow_mc); | |
1451 | ||
89aef892 DM |
1452 | skb_dst_set(skb, &rth->dst); |
1453 | return 0; | |
1da177e4 LT |
1454 | |
1455 | e_nobufs: | |
1da177e4 | 1456 | return -ENOBUFS; |
1da177e4 | 1457 | e_inval: |
96d36220 | 1458 | return -EINVAL; |
b5f7e755 | 1459 | e_err: |
b5f7e755 | 1460 | return err; |
1da177e4 LT |
1461 | } |
1462 | ||
1463 | ||
1464 | static void ip_handle_martian_source(struct net_device *dev, | |
1465 | struct in_device *in_dev, | |
1466 | struct sk_buff *skb, | |
9e12bb22 AV |
1467 | __be32 daddr, |
1468 | __be32 saddr) | |
1da177e4 LT |
1469 | { |
1470 | RT_CACHE_STAT_INC(in_martian_src); | |
1471 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1472 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) { | |
1473 | /* | |
1474 | * RFC1812 recommendation, if source is martian, | |
1475 | * the only hint is MAC header. | |
1476 | */ | |
058bd4d2 | 1477 | pr_warn("martian source %pI4 from %pI4, on dev %s\n", |
673d57e7 | 1478 | &daddr, &saddr, dev->name); |
98e399f8 | 1479 | if (dev->hard_header_len && skb_mac_header_was_set(skb)) { |
058bd4d2 JP |
1480 | print_hex_dump(KERN_WARNING, "ll header: ", |
1481 | DUMP_PREFIX_OFFSET, 16, 1, | |
1482 | skb_mac_header(skb), | |
1483 | dev->hard_header_len, true); | |
1da177e4 LT |
1484 | } |
1485 | } | |
1486 | #endif | |
1487 | } | |
1488 | ||
47360228 | 1489 | /* called in rcu_read_lock() section */ |
5969f71d | 1490 | static int __mkroute_input(struct sk_buff *skb, |
982721f3 | 1491 | const struct fib_result *res, |
5969f71d | 1492 | struct in_device *in_dev, |
c6cffba4 | 1493 | __be32 daddr, __be32 saddr, u32 tos) |
1da177e4 | 1494 | { |
1da177e4 LT |
1495 | struct rtable *rth; |
1496 | int err; | |
1497 | struct in_device *out_dev; | |
47360228 | 1498 | unsigned int flags = 0; |
d2d68ba9 | 1499 | bool do_cache; |
d9c9df8c | 1500 | u32 itag; |
1da177e4 LT |
1501 | |
1502 | /* get a working reference to the output device */ | |
47360228 | 1503 | out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res)); |
1da177e4 | 1504 | if (out_dev == NULL) { |
e87cc472 | 1505 | net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n"); |
1da177e4 LT |
1506 | return -EINVAL; |
1507 | } | |
1508 | ||
5c04c819 | 1509 | err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res), |
9e56e380 | 1510 | in_dev->dev, in_dev, &itag); |
1da177e4 | 1511 | if (err < 0) { |
e905a9ed | 1512 | ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr, |
1da177e4 | 1513 | saddr); |
e905a9ed | 1514 | |
1da177e4 LT |
1515 | goto cleanup; |
1516 | } | |
1517 | ||
e81da0e1 JA |
1518 | do_cache = res->fi && !itag; |
1519 | if (out_dev == in_dev && err && IN_DEV_TX_REDIRECTS(out_dev) && | |
1da177e4 | 1520 | (IN_DEV_SHARED_MEDIA(out_dev) || |
e81da0e1 | 1521 | inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) { |
1da177e4 | 1522 | flags |= RTCF_DOREDIRECT; |
e81da0e1 JA |
1523 | do_cache = false; |
1524 | } | |
1da177e4 LT |
1525 | |
1526 | if (skb->protocol != htons(ETH_P_IP)) { | |
1527 | /* Not IP (i.e. ARP). Do not create route, if it is | |
1528 | * invalid for proxy arp. DNAT routes are always valid. | |
65324144 JDB |
1529 | * |
1530 | * Proxy arp feature have been extended to allow, ARP | |
1531 | * replies back to the same interface, to support | |
1532 | * Private VLAN switch technologies. See arp.c. | |
1da177e4 | 1533 | */ |
65324144 JDB |
1534 | if (out_dev == in_dev && |
1535 | IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) { | |
1da177e4 LT |
1536 | err = -EINVAL; |
1537 | goto cleanup; | |
1538 | } | |
1539 | } | |
1540 | ||
e81da0e1 JA |
1541 | if (do_cache) { |
1542 | rth = rcu_dereference(FIB_RES_NH(*res).nh_rth_input); | |
1543 | if (rt_cache_valid(rth)) { | |
1544 | skb_dst_set_noref(skb, &rth->dst); | |
1545 | goto out; | |
d2d68ba9 DM |
1546 | } |
1547 | } | |
f2bb4bed | 1548 | |
5c1e6aa3 DM |
1549 | rth = rt_dst_alloc(out_dev->dev, |
1550 | IN_DEV_CONF_GET(in_dev, NOPOLICY), | |
d2d68ba9 | 1551 | IN_DEV_CONF_GET(out_dev, NOXFRM), do_cache); |
1da177e4 LT |
1552 | if (!rth) { |
1553 | err = -ENOBUFS; | |
1554 | goto cleanup; | |
1555 | } | |
1556 | ||
cf911662 DM |
1557 | rth->rt_genid = rt_genid(dev_net(rth->dst.dev)); |
1558 | rth->rt_flags = flags; | |
1559 | rth->rt_type = res->type; | |
9917e1e8 | 1560 | rth->rt_is_input = 1; |
13378cad | 1561 | rth->rt_iif = 0; |
5943634f | 1562 | rth->rt_pmtu = 0; |
f8126f1d | 1563 | rth->rt_gateway = 0; |
155e8336 | 1564 | rth->rt_uses_gateway = 0; |
caacf05e | 1565 | INIT_LIST_HEAD(&rth->rt_uncached); |
1da177e4 | 1566 | |
d8d1f30b CG |
1567 | rth->dst.input = ip_forward; |
1568 | rth->dst.output = ip_output; | |
1da177e4 | 1569 | |
d2d68ba9 | 1570 | rt_set_nexthop(rth, daddr, res, NULL, res->fi, res->type, itag); |
c6cffba4 | 1571 | skb_dst_set(skb, &rth->dst); |
d2d68ba9 | 1572 | out: |
1da177e4 LT |
1573 | err = 0; |
1574 | cleanup: | |
1da177e4 | 1575 | return err; |
e905a9ed | 1576 | } |
1da177e4 | 1577 | |
5969f71d SH |
1578 | static int ip_mkroute_input(struct sk_buff *skb, |
1579 | struct fib_result *res, | |
68a5e3dd | 1580 | const struct flowi4 *fl4, |
5969f71d SH |
1581 | struct in_device *in_dev, |
1582 | __be32 daddr, __be32 saddr, u32 tos) | |
1da177e4 | 1583 | { |
1da177e4 | 1584 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
ff3fccb3 | 1585 | if (res->fi && res->fi->fib_nhs > 1) |
1b7fe593 | 1586 | fib_select_multipath(res); |
1da177e4 LT |
1587 | #endif |
1588 | ||
1589 | /* create a routing cache entry */ | |
c6cffba4 | 1590 | return __mkroute_input(skb, res, in_dev, daddr, saddr, tos); |
1da177e4 LT |
1591 | } |
1592 | ||
1da177e4 LT |
1593 | /* |
1594 | * NOTE. We drop all the packets that has local source | |
1595 | * addresses, because every properly looped back packet | |
1596 | * must have correct destination already attached by output routine. | |
1597 | * | |
1598 | * Such approach solves two big problems: | |
1599 | * 1. Not simplex devices are handled properly. | |
1600 | * 2. IP spoofing attempts are filtered with 100% of guarantee. | |
ebc0ffae | 1601 | * called with rcu_read_lock() |
1da177e4 LT |
1602 | */ |
1603 | ||
9e12bb22 | 1604 | static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
c10237e0 | 1605 | u8 tos, struct net_device *dev) |
1da177e4 LT |
1606 | { |
1607 | struct fib_result res; | |
96d36220 | 1608 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
68a5e3dd | 1609 | struct flowi4 fl4; |
95c96174 | 1610 | unsigned int flags = 0; |
1da177e4 | 1611 | u32 itag = 0; |
95c96174 | 1612 | struct rtable *rth; |
1da177e4 | 1613 | int err = -EINVAL; |
5e73ea1a | 1614 | struct net *net = dev_net(dev); |
d2d68ba9 | 1615 | bool do_cache; |
1da177e4 LT |
1616 | |
1617 | /* IP on this device is disabled. */ | |
1618 | ||
1619 | if (!in_dev) | |
1620 | goto out; | |
1621 | ||
1622 | /* Check for the most weird martians, which can be not detected | |
1623 | by fib_lookup. | |
1624 | */ | |
1625 | ||
d0daebc3 | 1626 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr)) |
1da177e4 LT |
1627 | goto martian_source; |
1628 | ||
d2d68ba9 | 1629 | res.fi = NULL; |
27a954bd | 1630 | if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0)) |
1da177e4 LT |
1631 | goto brd_input; |
1632 | ||
1633 | /* Accept zero addresses only to limited broadcast; | |
1634 | * I even do not know to fix it or not. Waiting for complains :-) | |
1635 | */ | |
f97c1e0c | 1636 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
1637 | goto martian_source; |
1638 | ||
d0daebc3 | 1639 | if (ipv4_is_zeronet(daddr)) |
1da177e4 LT |
1640 | goto martian_destination; |
1641 | ||
9eb43e76 ED |
1642 | /* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(), |
1643 | * and call it once if daddr or/and saddr are loopback addresses | |
1644 | */ | |
1645 | if (ipv4_is_loopback(daddr)) { | |
1646 | if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net)) | |
d0daebc3 | 1647 | goto martian_destination; |
9eb43e76 ED |
1648 | } else if (ipv4_is_loopback(saddr)) { |
1649 | if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net)) | |
d0daebc3 TG |
1650 | goto martian_source; |
1651 | } | |
1652 | ||
1da177e4 LT |
1653 | /* |
1654 | * Now we are ready to route packet. | |
1655 | */ | |
68a5e3dd DM |
1656 | fl4.flowi4_oif = 0; |
1657 | fl4.flowi4_iif = dev->ifindex; | |
1658 | fl4.flowi4_mark = skb->mark; | |
1659 | fl4.flowi4_tos = tos; | |
1660 | fl4.flowi4_scope = RT_SCOPE_UNIVERSE; | |
1661 | fl4.daddr = daddr; | |
1662 | fl4.saddr = saddr; | |
1663 | err = fib_lookup(net, &fl4, &res); | |
251da413 | 1664 | if (err != 0) |
1da177e4 | 1665 | goto no_route; |
1da177e4 LT |
1666 | |
1667 | RT_CACHE_STAT_INC(in_slow_tot); | |
1668 | ||
1669 | if (res.type == RTN_BROADCAST) | |
1670 | goto brd_input; | |
1671 | ||
1672 | if (res.type == RTN_LOCAL) { | |
5c04c819 | 1673 | err = fib_validate_source(skb, saddr, daddr, tos, |
1fb9489b | 1674 | LOOPBACK_IFINDEX, |
9e56e380 | 1675 | dev, in_dev, &itag); |
b5f7e755 ED |
1676 | if (err < 0) |
1677 | goto martian_source_keep_err; | |
1da177e4 LT |
1678 | goto local_input; |
1679 | } | |
1680 | ||
1681 | if (!IN_DEV_FORWARD(in_dev)) | |
251da413 | 1682 | goto no_route; |
1da177e4 LT |
1683 | if (res.type != RTN_UNICAST) |
1684 | goto martian_destination; | |
1685 | ||
68a5e3dd | 1686 | err = ip_mkroute_input(skb, &res, &fl4, in_dev, daddr, saddr, tos); |
1da177e4 LT |
1687 | out: return err; |
1688 | ||
1689 | brd_input: | |
1690 | if (skb->protocol != htons(ETH_P_IP)) | |
1691 | goto e_inval; | |
1692 | ||
41347dcd | 1693 | if (!ipv4_is_zeronet(saddr)) { |
9e56e380 DM |
1694 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, |
1695 | in_dev, &itag); | |
1da177e4 | 1696 | if (err < 0) |
b5f7e755 | 1697 | goto martian_source_keep_err; |
1da177e4 LT |
1698 | } |
1699 | flags |= RTCF_BROADCAST; | |
1700 | res.type = RTN_BROADCAST; | |
1701 | RT_CACHE_STAT_INC(in_brd); | |
1702 | ||
1703 | local_input: | |
d2d68ba9 DM |
1704 | do_cache = false; |
1705 | if (res.fi) { | |
fe3edf45 | 1706 | if (!itag) { |
54764bb6 | 1707 | rth = rcu_dereference(FIB_RES_NH(res).nh_rth_input); |
d2d68ba9 | 1708 | if (rt_cache_valid(rth)) { |
c6cffba4 DM |
1709 | skb_dst_set_noref(skb, &rth->dst); |
1710 | err = 0; | |
1711 | goto out; | |
d2d68ba9 DM |
1712 | } |
1713 | do_cache = true; | |
1714 | } | |
1715 | } | |
1716 | ||
5c1e6aa3 | 1717 | rth = rt_dst_alloc(net->loopback_dev, |
d2d68ba9 | 1718 | IN_DEV_CONF_GET(in_dev, NOPOLICY), false, do_cache); |
1da177e4 LT |
1719 | if (!rth) |
1720 | goto e_nobufs; | |
1721 | ||
cf911662 | 1722 | rth->dst.input= ip_local_deliver; |
d8d1f30b | 1723 | rth->dst.output= ip_rt_bug; |
cf911662 DM |
1724 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1725 | rth->dst.tclassid = itag; | |
1726 | #endif | |
1da177e4 | 1727 | |
cf911662 DM |
1728 | rth->rt_genid = rt_genid(net); |
1729 | rth->rt_flags = flags|RTCF_LOCAL; | |
1730 | rth->rt_type = res.type; | |
9917e1e8 | 1731 | rth->rt_is_input = 1; |
13378cad | 1732 | rth->rt_iif = 0; |
5943634f | 1733 | rth->rt_pmtu = 0; |
f8126f1d | 1734 | rth->rt_gateway = 0; |
155e8336 | 1735 | rth->rt_uses_gateway = 0; |
caacf05e | 1736 | INIT_LIST_HEAD(&rth->rt_uncached); |
1da177e4 | 1737 | if (res.type == RTN_UNREACHABLE) { |
d8d1f30b CG |
1738 | rth->dst.input= ip_error; |
1739 | rth->dst.error= -err; | |
1da177e4 LT |
1740 | rth->rt_flags &= ~RTCF_LOCAL; |
1741 | } | |
d2d68ba9 DM |
1742 | if (do_cache) |
1743 | rt_cache_route(&FIB_RES_NH(res), rth); | |
89aef892 | 1744 | skb_dst_set(skb, &rth->dst); |
b23dd4fe | 1745 | err = 0; |
ebc0ffae | 1746 | goto out; |
1da177e4 LT |
1747 | |
1748 | no_route: | |
1749 | RT_CACHE_STAT_INC(in_no_route); | |
1da177e4 | 1750 | res.type = RTN_UNREACHABLE; |
7f53878d MC |
1751 | if (err == -ESRCH) |
1752 | err = -ENETUNREACH; | |
1da177e4 LT |
1753 | goto local_input; |
1754 | ||
1755 | /* | |
1756 | * Do not cache martian addresses: they should be logged (RFC1812) | |
1757 | */ | |
1758 | martian_destination: | |
1759 | RT_CACHE_STAT_INC(in_martian_dst); | |
1760 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
e87cc472 JP |
1761 | if (IN_DEV_LOG_MARTIANS(in_dev)) |
1762 | net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n", | |
1763 | &daddr, &saddr, dev->name); | |
1da177e4 | 1764 | #endif |
2c2910a4 | 1765 | |
1da177e4 LT |
1766 | e_inval: |
1767 | err = -EINVAL; | |
ebc0ffae | 1768 | goto out; |
1da177e4 LT |
1769 | |
1770 | e_nobufs: | |
1771 | err = -ENOBUFS; | |
ebc0ffae | 1772 | goto out; |
1da177e4 LT |
1773 | |
1774 | martian_source: | |
b5f7e755 ED |
1775 | err = -EINVAL; |
1776 | martian_source_keep_err: | |
1da177e4 | 1777 | ip_handle_martian_source(dev, in_dev, skb, daddr, saddr); |
ebc0ffae | 1778 | goto out; |
1da177e4 LT |
1779 | } |
1780 | ||
c6cffba4 DM |
1781 | int ip_route_input_noref(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1782 | u8 tos, struct net_device *dev) | |
1da177e4 | 1783 | { |
96d36220 | 1784 | int res; |
1da177e4 | 1785 | |
96d36220 ED |
1786 | rcu_read_lock(); |
1787 | ||
1da177e4 LT |
1788 | /* Multicast recognition logic is moved from route cache to here. |
1789 | The problem was that too many Ethernet cards have broken/missing | |
1790 | hardware multicast filters :-( As result the host on multicasting | |
1791 | network acquires a lot of useless route cache entries, sort of | |
1792 | SDR messages from all the world. Now we try to get rid of them. | |
1793 | Really, provided software IP multicast filter is organized | |
1794 | reasonably (at least, hashed), it does not result in a slowdown | |
1795 | comparing with route cache reject entries. | |
1796 | Note, that multicast routers are not affected, because | |
1797 | route cache entry is created eventually. | |
1798 | */ | |
f97c1e0c | 1799 | if (ipv4_is_multicast(daddr)) { |
96d36220 | 1800 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1da177e4 | 1801 | |
96d36220 | 1802 | if (in_dev) { |
dbdd9a52 DM |
1803 | int our = ip_check_mc_rcu(in_dev, daddr, saddr, |
1804 | ip_hdr(skb)->protocol); | |
1da177e4 LT |
1805 | if (our |
1806 | #ifdef CONFIG_IP_MROUTE | |
9d4fb27d JP |
1807 | || |
1808 | (!ipv4_is_local_multicast(daddr) && | |
1809 | IN_DEV_MFORWARD(in_dev)) | |
1da177e4 | 1810 | #endif |
9d4fb27d | 1811 | ) { |
96d36220 ED |
1812 | int res = ip_route_input_mc(skb, daddr, saddr, |
1813 | tos, dev, our); | |
1da177e4 | 1814 | rcu_read_unlock(); |
96d36220 | 1815 | return res; |
1da177e4 LT |
1816 | } |
1817 | } | |
1818 | rcu_read_unlock(); | |
1819 | return -EINVAL; | |
1820 | } | |
c10237e0 | 1821 | res = ip_route_input_slow(skb, daddr, saddr, tos, dev); |
96d36220 ED |
1822 | rcu_read_unlock(); |
1823 | return res; | |
1da177e4 | 1824 | } |
c6cffba4 | 1825 | EXPORT_SYMBOL(ip_route_input_noref); |
1da177e4 | 1826 | |
ebc0ffae | 1827 | /* called with rcu_read_lock() */ |
982721f3 | 1828 | static struct rtable *__mkroute_output(const struct fib_result *res, |
1a00fee4 | 1829 | const struct flowi4 *fl4, int orig_oif, |
f61759e6 | 1830 | struct net_device *dev_out, |
5ada5527 | 1831 | unsigned int flags) |
1da177e4 | 1832 | { |
982721f3 | 1833 | struct fib_info *fi = res->fi; |
f2bb4bed | 1834 | struct fib_nh_exception *fnhe; |
5ada5527 | 1835 | struct in_device *in_dev; |
982721f3 | 1836 | u16 type = res->type; |
5ada5527 | 1837 | struct rtable *rth; |
c92b9655 | 1838 | bool do_cache; |
1da177e4 | 1839 | |
d0daebc3 TG |
1840 | in_dev = __in_dev_get_rcu(dev_out); |
1841 | if (!in_dev) | |
5ada5527 | 1842 | return ERR_PTR(-EINVAL); |
1da177e4 | 1843 | |
d0daebc3 TG |
1844 | if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev))) |
1845 | if (ipv4_is_loopback(fl4->saddr) && !(dev_out->flags & IFF_LOOPBACK)) | |
1846 | return ERR_PTR(-EINVAL); | |
1847 | ||
68a5e3dd | 1848 | if (ipv4_is_lbcast(fl4->daddr)) |
982721f3 | 1849 | type = RTN_BROADCAST; |
68a5e3dd | 1850 | else if (ipv4_is_multicast(fl4->daddr)) |
982721f3 | 1851 | type = RTN_MULTICAST; |
68a5e3dd | 1852 | else if (ipv4_is_zeronet(fl4->daddr)) |
5ada5527 | 1853 | return ERR_PTR(-EINVAL); |
1da177e4 LT |
1854 | |
1855 | if (dev_out->flags & IFF_LOOPBACK) | |
1856 | flags |= RTCF_LOCAL; | |
1857 | ||
63617421 | 1858 | do_cache = true; |
982721f3 | 1859 | if (type == RTN_BROADCAST) { |
1da177e4 | 1860 | flags |= RTCF_BROADCAST | RTCF_LOCAL; |
982721f3 DM |
1861 | fi = NULL; |
1862 | } else if (type == RTN_MULTICAST) { | |
dd28d1a0 | 1863 | flags |= RTCF_MULTICAST | RTCF_LOCAL; |
813b3b5d DM |
1864 | if (!ip_check_mc_rcu(in_dev, fl4->daddr, fl4->saddr, |
1865 | fl4->flowi4_proto)) | |
1da177e4 | 1866 | flags &= ~RTCF_LOCAL; |
63617421 JA |
1867 | else |
1868 | do_cache = false; | |
1da177e4 | 1869 | /* If multicast route do not exist use |
dd28d1a0 ED |
1870 | * default one, but do not gateway in this case. |
1871 | * Yes, it is hack. | |
1da177e4 | 1872 | */ |
982721f3 DM |
1873 | if (fi && res->prefixlen < 4) |
1874 | fi = NULL; | |
1da177e4 LT |
1875 | } |
1876 | ||
f2bb4bed | 1877 | fnhe = NULL; |
63617421 JA |
1878 | do_cache &= fi != NULL; |
1879 | if (do_cache) { | |
c5038a83 | 1880 | struct rtable __rcu **prth; |
c92b9655 | 1881 | struct fib_nh *nh = &FIB_RES_NH(*res); |
d26b3a7c | 1882 | |
c92b9655 | 1883 | fnhe = find_exception(nh, fl4->daddr); |
c5038a83 DM |
1884 | if (fnhe) |
1885 | prth = &fnhe->fnhe_rth; | |
c92b9655 JA |
1886 | else { |
1887 | if (unlikely(fl4->flowi4_flags & | |
1888 | FLOWI_FLAG_KNOWN_NH && | |
1889 | !(nh->nh_gw && | |
1890 | nh->nh_scope == RT_SCOPE_LINK))) { | |
1891 | do_cache = false; | |
1892 | goto add; | |
1893 | } | |
1894 | prth = __this_cpu_ptr(nh->nh_pcpu_rth_output); | |
1895 | } | |
c5038a83 DM |
1896 | rth = rcu_dereference(*prth); |
1897 | if (rt_cache_valid(rth)) { | |
1898 | dst_hold(&rth->dst); | |
1899 | return rth; | |
f2bb4bed DM |
1900 | } |
1901 | } | |
c92b9655 JA |
1902 | |
1903 | add: | |
5c1e6aa3 DM |
1904 | rth = rt_dst_alloc(dev_out, |
1905 | IN_DEV_CONF_GET(in_dev, NOPOLICY), | |
f2bb4bed | 1906 | IN_DEV_CONF_GET(in_dev, NOXFRM), |
c92b9655 | 1907 | do_cache); |
8391d07b | 1908 | if (!rth) |
5ada5527 | 1909 | return ERR_PTR(-ENOBUFS); |
8391d07b | 1910 | |
cf911662 DM |
1911 | rth->dst.output = ip_output; |
1912 | ||
cf911662 DM |
1913 | rth->rt_genid = rt_genid(dev_net(dev_out)); |
1914 | rth->rt_flags = flags; | |
1915 | rth->rt_type = type; | |
9917e1e8 | 1916 | rth->rt_is_input = 0; |
13378cad | 1917 | rth->rt_iif = orig_oif ? : 0; |
5943634f | 1918 | rth->rt_pmtu = 0; |
f8126f1d | 1919 | rth->rt_gateway = 0; |
155e8336 | 1920 | rth->rt_uses_gateway = 0; |
caacf05e | 1921 | INIT_LIST_HEAD(&rth->rt_uncached); |
1da177e4 LT |
1922 | |
1923 | RT_CACHE_STAT_INC(out_slow_tot); | |
1924 | ||
41347dcd | 1925 | if (flags & RTCF_LOCAL) |
d8d1f30b | 1926 | rth->dst.input = ip_local_deliver; |
1da177e4 | 1927 | if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { |
e905a9ed | 1928 | if (flags & RTCF_LOCAL && |
1da177e4 | 1929 | !(dev_out->flags & IFF_LOOPBACK)) { |
d8d1f30b | 1930 | rth->dst.output = ip_mc_output; |
1da177e4 LT |
1931 | RT_CACHE_STAT_INC(out_slow_mc); |
1932 | } | |
1933 | #ifdef CONFIG_IP_MROUTE | |
982721f3 | 1934 | if (type == RTN_MULTICAST) { |
1da177e4 | 1935 | if (IN_DEV_MFORWARD(in_dev) && |
813b3b5d | 1936 | !ipv4_is_local_multicast(fl4->daddr)) { |
d8d1f30b CG |
1937 | rth->dst.input = ip_mr_input; |
1938 | rth->dst.output = ip_mc_output; | |
1da177e4 LT |
1939 | } |
1940 | } | |
1941 | #endif | |
1942 | } | |
1943 | ||
f2bb4bed | 1944 | rt_set_nexthop(rth, fl4->daddr, res, fnhe, fi, type, 0); |
1da177e4 | 1945 | |
5ada5527 | 1946 | return rth; |
1da177e4 LT |
1947 | } |
1948 | ||
1da177e4 LT |
1949 | /* |
1950 | * Major route resolver routine. | |
1951 | */ | |
1952 | ||
89aef892 | 1953 | struct rtable *__ip_route_output_key(struct net *net, struct flowi4 *fl4) |
1da177e4 | 1954 | { |
1da177e4 | 1955 | struct net_device *dev_out = NULL; |
f61759e6 | 1956 | __u8 tos = RT_FL_TOS(fl4); |
813b3b5d DM |
1957 | unsigned int flags = 0; |
1958 | struct fib_result res; | |
5ada5527 | 1959 | struct rtable *rth; |
813b3b5d | 1960 | int orig_oif; |
1da177e4 | 1961 | |
85b91b03 | 1962 | res.tclassid = 0; |
1da177e4 | 1963 | res.fi = NULL; |
8b96d22d | 1964 | res.table = NULL; |
1da177e4 | 1965 | |
813b3b5d DM |
1966 | orig_oif = fl4->flowi4_oif; |
1967 | ||
1fb9489b | 1968 | fl4->flowi4_iif = LOOPBACK_IFINDEX; |
813b3b5d DM |
1969 | fl4->flowi4_tos = tos & IPTOS_RT_MASK; |
1970 | fl4->flowi4_scope = ((tos & RTO_ONLINK) ? | |
1971 | RT_SCOPE_LINK : RT_SCOPE_UNIVERSE); | |
44713b67 | 1972 | |
010c2708 | 1973 | rcu_read_lock(); |
813b3b5d | 1974 | if (fl4->saddr) { |
b23dd4fe | 1975 | rth = ERR_PTR(-EINVAL); |
813b3b5d DM |
1976 | if (ipv4_is_multicast(fl4->saddr) || |
1977 | ipv4_is_lbcast(fl4->saddr) || | |
1978 | ipv4_is_zeronet(fl4->saddr)) | |
1da177e4 LT |
1979 | goto out; |
1980 | ||
1da177e4 LT |
1981 | /* I removed check for oif == dev_out->oif here. |
1982 | It was wrong for two reasons: | |
1ab35276 DL |
1983 | 1. ip_dev_find(net, saddr) can return wrong iface, if saddr |
1984 | is assigned to multiple interfaces. | |
1da177e4 LT |
1985 | 2. Moreover, we are allowed to send packets with saddr |
1986 | of another iface. --ANK | |
1987 | */ | |
1988 | ||
813b3b5d DM |
1989 | if (fl4->flowi4_oif == 0 && |
1990 | (ipv4_is_multicast(fl4->daddr) || | |
1991 | ipv4_is_lbcast(fl4->daddr))) { | |
a210d01a | 1992 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
813b3b5d | 1993 | dev_out = __ip_dev_find(net, fl4->saddr, false); |
a210d01a JA |
1994 | if (dev_out == NULL) |
1995 | goto out; | |
1996 | ||
1da177e4 LT |
1997 | /* Special hack: user can direct multicasts |
1998 | and limited broadcast via necessary interface | |
1999 | without fiddling with IP_MULTICAST_IF or IP_PKTINFO. | |
2000 | This hack is not just for fun, it allows | |
2001 | vic,vat and friends to work. | |
2002 | They bind socket to loopback, set ttl to zero | |
2003 | and expect that it will work. | |
2004 | From the viewpoint of routing cache they are broken, | |
2005 | because we are not allowed to build multicast path | |
2006 | with loopback source addr (look, routing cache | |
2007 | cannot know, that ttl is zero, so that packet | |
2008 | will not leave this host and route is valid). | |
2009 | Luckily, this hack is good workaround. | |
2010 | */ | |
2011 | ||
813b3b5d | 2012 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2013 | goto make_route; |
2014 | } | |
a210d01a | 2015 | |
813b3b5d | 2016 | if (!(fl4->flowi4_flags & FLOWI_FLAG_ANYSRC)) { |
a210d01a | 2017 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
813b3b5d | 2018 | if (!__ip_dev_find(net, fl4->saddr, false)) |
a210d01a | 2019 | goto out; |
a210d01a | 2020 | } |
1da177e4 LT |
2021 | } |
2022 | ||
2023 | ||
813b3b5d DM |
2024 | if (fl4->flowi4_oif) { |
2025 | dev_out = dev_get_by_index_rcu(net, fl4->flowi4_oif); | |
b23dd4fe | 2026 | rth = ERR_PTR(-ENODEV); |
1da177e4 LT |
2027 | if (dev_out == NULL) |
2028 | goto out; | |
e5ed6399 HX |
2029 | |
2030 | /* RACE: Check return value of inet_select_addr instead. */ | |
fc75fc83 | 2031 | if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) { |
b23dd4fe | 2032 | rth = ERR_PTR(-ENETUNREACH); |
fc75fc83 ED |
2033 | goto out; |
2034 | } | |
813b3b5d DM |
2035 | if (ipv4_is_local_multicast(fl4->daddr) || |
2036 | ipv4_is_lbcast(fl4->daddr)) { | |
2037 | if (!fl4->saddr) | |
2038 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2039 | RT_SCOPE_LINK); | |
1da177e4 LT |
2040 | goto make_route; |
2041 | } | |
813b3b5d DM |
2042 | if (fl4->saddr) { |
2043 | if (ipv4_is_multicast(fl4->daddr)) | |
2044 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2045 | fl4->flowi4_scope); | |
2046 | else if (!fl4->daddr) | |
2047 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2048 | RT_SCOPE_HOST); | |
1da177e4 LT |
2049 | } |
2050 | } | |
2051 | ||
813b3b5d DM |
2052 | if (!fl4->daddr) { |
2053 | fl4->daddr = fl4->saddr; | |
2054 | if (!fl4->daddr) | |
2055 | fl4->daddr = fl4->saddr = htonl(INADDR_LOOPBACK); | |
b40afd0e | 2056 | dev_out = net->loopback_dev; |
1fb9489b | 2057 | fl4->flowi4_oif = LOOPBACK_IFINDEX; |
1da177e4 LT |
2058 | res.type = RTN_LOCAL; |
2059 | flags |= RTCF_LOCAL; | |
2060 | goto make_route; | |
2061 | } | |
2062 | ||
813b3b5d | 2063 | if (fib_lookup(net, fl4, &res)) { |
1da177e4 | 2064 | res.fi = NULL; |
8b96d22d | 2065 | res.table = NULL; |
813b3b5d | 2066 | if (fl4->flowi4_oif) { |
1da177e4 LT |
2067 | /* Apparently, routing tables are wrong. Assume, |
2068 | that the destination is on link. | |
2069 | ||
2070 | WHY? DW. | |
2071 | Because we are allowed to send to iface | |
2072 | even if it has NO routes and NO assigned | |
2073 | addresses. When oif is specified, routing | |
2074 | tables are looked up with only one purpose: | |
2075 | to catch if destination is gatewayed, rather than | |
2076 | direct. Moreover, if MSG_DONTROUTE is set, | |
2077 | we send packet, ignoring both routing tables | |
2078 | and ifaddr state. --ANK | |
2079 | ||
2080 | ||
2081 | We could make it even if oif is unknown, | |
2082 | likely IPv6, but we do not. | |
2083 | */ | |
2084 | ||
813b3b5d DM |
2085 | if (fl4->saddr == 0) |
2086 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2087 | RT_SCOPE_LINK); | |
1da177e4 LT |
2088 | res.type = RTN_UNICAST; |
2089 | goto make_route; | |
2090 | } | |
b23dd4fe | 2091 | rth = ERR_PTR(-ENETUNREACH); |
1da177e4 LT |
2092 | goto out; |
2093 | } | |
1da177e4 LT |
2094 | |
2095 | if (res.type == RTN_LOCAL) { | |
813b3b5d | 2096 | if (!fl4->saddr) { |
9fc3bbb4 | 2097 | if (res.fi->fib_prefsrc) |
813b3b5d | 2098 | fl4->saddr = res.fi->fib_prefsrc; |
9fc3bbb4 | 2099 | else |
813b3b5d | 2100 | fl4->saddr = fl4->daddr; |
9fc3bbb4 | 2101 | } |
b40afd0e | 2102 | dev_out = net->loopback_dev; |
813b3b5d | 2103 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2104 | flags |= RTCF_LOCAL; |
2105 | goto make_route; | |
2106 | } | |
2107 | ||
2108 | #ifdef CONFIG_IP_ROUTE_MULTIPATH | |
813b3b5d | 2109 | if (res.fi->fib_nhs > 1 && fl4->flowi4_oif == 0) |
1b7fe593 | 2110 | fib_select_multipath(&res); |
1da177e4 LT |
2111 | else |
2112 | #endif | |
21d8c49e DM |
2113 | if (!res.prefixlen && |
2114 | res.table->tb_num_default > 1 && | |
813b3b5d | 2115 | res.type == RTN_UNICAST && !fl4->flowi4_oif) |
0c838ff1 | 2116 | fib_select_default(&res); |
1da177e4 | 2117 | |
813b3b5d DM |
2118 | if (!fl4->saddr) |
2119 | fl4->saddr = FIB_RES_PREFSRC(net, res); | |
1da177e4 | 2120 | |
1da177e4 | 2121 | dev_out = FIB_RES_DEV(res); |
813b3b5d | 2122 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2123 | |
2124 | ||
2125 | make_route: | |
1a00fee4 | 2126 | rth = __mkroute_output(&res, fl4, orig_oif, dev_out, flags); |
1da177e4 | 2127 | |
010c2708 DM |
2128 | out: |
2129 | rcu_read_unlock(); | |
b23dd4fe | 2130 | return rth; |
1da177e4 | 2131 | } |
d8c97a94 ACM |
2132 | EXPORT_SYMBOL_GPL(__ip_route_output_key); |
2133 | ||
ae2688d5 JW |
2134 | static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie) |
2135 | { | |
2136 | return NULL; | |
2137 | } | |
2138 | ||
ebb762f2 | 2139 | static unsigned int ipv4_blackhole_mtu(const struct dst_entry *dst) |
ec831ea7 | 2140 | { |
618f9bc7 SK |
2141 | unsigned int mtu = dst_metric_raw(dst, RTAX_MTU); |
2142 | ||
2143 | return mtu ? : dst->dev->mtu; | |
ec831ea7 RD |
2144 | } |
2145 | ||
6700c270 DM |
2146 | static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk, |
2147 | struct sk_buff *skb, u32 mtu) | |
14e50e57 DM |
2148 | { |
2149 | } | |
2150 | ||
6700c270 DM |
2151 | static void ipv4_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk, |
2152 | struct sk_buff *skb) | |
b587ee3b DM |
2153 | { |
2154 | } | |
2155 | ||
0972ddb2 HB |
2156 | static u32 *ipv4_rt_blackhole_cow_metrics(struct dst_entry *dst, |
2157 | unsigned long old) | |
2158 | { | |
2159 | return NULL; | |
2160 | } | |
2161 | ||
14e50e57 DM |
2162 | static struct dst_ops ipv4_dst_blackhole_ops = { |
2163 | .family = AF_INET, | |
09640e63 | 2164 | .protocol = cpu_to_be16(ETH_P_IP), |
ae2688d5 | 2165 | .check = ipv4_blackhole_dst_check, |
ebb762f2 | 2166 | .mtu = ipv4_blackhole_mtu, |
214f45c9 | 2167 | .default_advmss = ipv4_default_advmss, |
14e50e57 | 2168 | .update_pmtu = ipv4_rt_blackhole_update_pmtu, |
b587ee3b | 2169 | .redirect = ipv4_rt_blackhole_redirect, |
0972ddb2 | 2170 | .cow_metrics = ipv4_rt_blackhole_cow_metrics, |
d3aaeb38 | 2171 | .neigh_lookup = ipv4_neigh_lookup, |
14e50e57 DM |
2172 | }; |
2173 | ||
2774c131 | 2174 | struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig) |
14e50e57 | 2175 | { |
2774c131 | 2176 | struct rtable *ort = (struct rtable *) dst_orig; |
f5b0a874 | 2177 | struct rtable *rt; |
14e50e57 | 2178 | |
f5b0a874 | 2179 | rt = dst_alloc(&ipv4_dst_blackhole_ops, NULL, 1, DST_OBSOLETE_NONE, 0); |
14e50e57 | 2180 | if (rt) { |
d8d1f30b | 2181 | struct dst_entry *new = &rt->dst; |
14e50e57 | 2182 | |
14e50e57 | 2183 | new->__use = 1; |
352e512c HX |
2184 | new->input = dst_discard; |
2185 | new->output = dst_discard; | |
14e50e57 | 2186 | |
d8d1f30b | 2187 | new->dev = ort->dst.dev; |
14e50e57 DM |
2188 | if (new->dev) |
2189 | dev_hold(new->dev); | |
2190 | ||
9917e1e8 | 2191 | rt->rt_is_input = ort->rt_is_input; |
5e2b61f7 | 2192 | rt->rt_iif = ort->rt_iif; |
5943634f | 2193 | rt->rt_pmtu = ort->rt_pmtu; |
14e50e57 | 2194 | |
e84f84f2 | 2195 | rt->rt_genid = rt_genid(net); |
14e50e57 DM |
2196 | rt->rt_flags = ort->rt_flags; |
2197 | rt->rt_type = ort->rt_type; | |
14e50e57 | 2198 | rt->rt_gateway = ort->rt_gateway; |
155e8336 | 2199 | rt->rt_uses_gateway = ort->rt_uses_gateway; |
14e50e57 | 2200 | |
caacf05e DM |
2201 | INIT_LIST_HEAD(&rt->rt_uncached); |
2202 | ||
14e50e57 DM |
2203 | dst_free(new); |
2204 | } | |
2205 | ||
2774c131 DM |
2206 | dst_release(dst_orig); |
2207 | ||
2208 | return rt ? &rt->dst : ERR_PTR(-ENOMEM); | |
14e50e57 DM |
2209 | } |
2210 | ||
9d6ec938 | 2211 | struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4, |
b23dd4fe | 2212 | struct sock *sk) |
1da177e4 | 2213 | { |
9d6ec938 | 2214 | struct rtable *rt = __ip_route_output_key(net, flp4); |
1da177e4 | 2215 | |
b23dd4fe DM |
2216 | if (IS_ERR(rt)) |
2217 | return rt; | |
1da177e4 | 2218 | |
56157872 | 2219 | if (flp4->flowi4_proto) |
9d6ec938 DM |
2220 | rt = (struct rtable *) xfrm_lookup(net, &rt->dst, |
2221 | flowi4_to_flowi(flp4), | |
2222 | sk, 0); | |
1da177e4 | 2223 | |
b23dd4fe | 2224 | return rt; |
1da177e4 | 2225 | } |
d8c97a94 ACM |
2226 | EXPORT_SYMBOL_GPL(ip_route_output_flow); |
2227 | ||
f1ce3062 | 2228 | static int rt_fill_info(struct net *net, __be32 dst, __be32 src, |
15e47304 | 2229 | struct flowi4 *fl4, struct sk_buff *skb, u32 portid, |
f1ce3062 | 2230 | u32 seq, int event, int nowait, unsigned int flags) |
1da177e4 | 2231 | { |
511c3f92 | 2232 | struct rtable *rt = skb_rtable(skb); |
1da177e4 | 2233 | struct rtmsg *r; |
be403ea1 | 2234 | struct nlmsghdr *nlh; |
2bc8ca40 | 2235 | unsigned long expires = 0; |
f185071d | 2236 | u32 error; |
521f5490 | 2237 | u32 metrics[RTAX_MAX]; |
be403ea1 | 2238 | |
15e47304 | 2239 | nlh = nlmsg_put(skb, portid, seq, event, sizeof(*r), flags); |
be403ea1 | 2240 | if (nlh == NULL) |
26932566 | 2241 | return -EMSGSIZE; |
be403ea1 TG |
2242 | |
2243 | r = nlmsg_data(nlh); | |
1da177e4 LT |
2244 | r->rtm_family = AF_INET; |
2245 | r->rtm_dst_len = 32; | |
2246 | r->rtm_src_len = 0; | |
d6c0a4f6 | 2247 | r->rtm_tos = fl4->flowi4_tos; |
1da177e4 | 2248 | r->rtm_table = RT_TABLE_MAIN; |
f3756b79 DM |
2249 | if (nla_put_u32(skb, RTA_TABLE, RT_TABLE_MAIN)) |
2250 | goto nla_put_failure; | |
1da177e4 LT |
2251 | r->rtm_type = rt->rt_type; |
2252 | r->rtm_scope = RT_SCOPE_UNIVERSE; | |
2253 | r->rtm_protocol = RTPROT_UNSPEC; | |
2254 | r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED; | |
2255 | if (rt->rt_flags & RTCF_NOTIFY) | |
2256 | r->rtm_flags |= RTM_F_NOTIFY; | |
be403ea1 | 2257 | |
f1ce3062 | 2258 | if (nla_put_be32(skb, RTA_DST, dst)) |
f3756b79 | 2259 | goto nla_put_failure; |
1a00fee4 | 2260 | if (src) { |
1da177e4 | 2261 | r->rtm_src_len = 32; |
1a00fee4 | 2262 | if (nla_put_be32(skb, RTA_SRC, src)) |
f3756b79 | 2263 | goto nla_put_failure; |
1da177e4 | 2264 | } |
f3756b79 DM |
2265 | if (rt->dst.dev && |
2266 | nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex)) | |
2267 | goto nla_put_failure; | |
c7066f70 | 2268 | #ifdef CONFIG_IP_ROUTE_CLASSID |
f3756b79 DM |
2269 | if (rt->dst.tclassid && |
2270 | nla_put_u32(skb, RTA_FLOW, rt->dst.tclassid)) | |
2271 | goto nla_put_failure; | |
1da177e4 | 2272 | #endif |
41347dcd | 2273 | if (!rt_is_input_route(rt) && |
d6c0a4f6 DM |
2274 | fl4->saddr != src) { |
2275 | if (nla_put_be32(skb, RTA_PREFSRC, fl4->saddr)) | |
f3756b79 DM |
2276 | goto nla_put_failure; |
2277 | } | |
155e8336 | 2278 | if (rt->rt_uses_gateway && |
f3756b79 DM |
2279 | nla_put_be32(skb, RTA_GATEWAY, rt->rt_gateway)) |
2280 | goto nla_put_failure; | |
be403ea1 | 2281 | |
ee9a8f7a SK |
2282 | expires = rt->dst.expires; |
2283 | if (expires) { | |
2284 | unsigned long now = jiffies; | |
2285 | ||
2286 | if (time_before(now, expires)) | |
2287 | expires -= now; | |
2288 | else | |
2289 | expires = 0; | |
2290 | } | |
2291 | ||
521f5490 | 2292 | memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics)); |
ee9a8f7a | 2293 | if (rt->rt_pmtu && expires) |
521f5490 JA |
2294 | metrics[RTAX_MTU - 1] = rt->rt_pmtu; |
2295 | if (rtnetlink_put_metrics(skb, metrics) < 0) | |
be403ea1 TG |
2296 | goto nla_put_failure; |
2297 | ||
b4869889 | 2298 | if (fl4->flowi4_mark && |
68aaed54 | 2299 | nla_put_u32(skb, RTA_MARK, fl4->flowi4_mark)) |
f3756b79 | 2300 | goto nla_put_failure; |
963bfeee | 2301 | |
d8d1f30b | 2302 | error = rt->dst.error; |
be403ea1 | 2303 | |
c7537967 | 2304 | if (rt_is_input_route(rt)) { |
8caaf7b6 ND |
2305 | #ifdef CONFIG_IP_MROUTE |
2306 | if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) && | |
2307 | IPV4_DEVCONF_ALL(net, MC_FORWARDING)) { | |
2308 | int err = ipmr_get_route(net, skb, | |
2309 | fl4->saddr, fl4->daddr, | |
2310 | r, nowait); | |
2311 | if (err <= 0) { | |
2312 | if (!nowait) { | |
2313 | if (err == 0) | |
2314 | return 0; | |
2315 | goto nla_put_failure; | |
2316 | } else { | |
2317 | if (err == -EMSGSIZE) | |
2318 | goto nla_put_failure; | |
2319 | error = err; | |
2320 | } | |
2321 | } | |
2322 | } else | |
2323 | #endif | |
2324 | if (nla_put_u32(skb, RTA_IIF, rt->rt_iif)) | |
2325 | goto nla_put_failure; | |
1da177e4 LT |
2326 | } |
2327 | ||
f185071d | 2328 | if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, error) < 0) |
e3703b3d | 2329 | goto nla_put_failure; |
be403ea1 TG |
2330 | |
2331 | return nlmsg_end(skb, nlh); | |
1da177e4 | 2332 | |
be403ea1 | 2333 | nla_put_failure: |
26932566 PM |
2334 | nlmsg_cancel(skb, nlh); |
2335 | return -EMSGSIZE; | |
1da177e4 LT |
2336 | } |
2337 | ||
661d2967 | 2338 | static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh) |
1da177e4 | 2339 | { |
3b1e0a65 | 2340 | struct net *net = sock_net(in_skb->sk); |
d889ce3b TG |
2341 | struct rtmsg *rtm; |
2342 | struct nlattr *tb[RTA_MAX+1]; | |
1da177e4 | 2343 | struct rtable *rt = NULL; |
d6c0a4f6 | 2344 | struct flowi4 fl4; |
9e12bb22 AV |
2345 | __be32 dst = 0; |
2346 | __be32 src = 0; | |
2347 | u32 iif; | |
d889ce3b | 2348 | int err; |
963bfeee | 2349 | int mark; |
1da177e4 LT |
2350 | struct sk_buff *skb; |
2351 | ||
d889ce3b TG |
2352 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy); |
2353 | if (err < 0) | |
2354 | goto errout; | |
2355 | ||
2356 | rtm = nlmsg_data(nlh); | |
2357 | ||
1da177e4 | 2358 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
d889ce3b TG |
2359 | if (skb == NULL) { |
2360 | err = -ENOBUFS; | |
2361 | goto errout; | |
2362 | } | |
1da177e4 LT |
2363 | |
2364 | /* Reserve room for dummy headers, this skb can pass | |
2365 | through good chunk of routing engine. | |
2366 | */ | |
459a98ed | 2367 | skb_reset_mac_header(skb); |
c1d2bbe1 | 2368 | skb_reset_network_header(skb); |
d2c962b8 SH |
2369 | |
2370 | /* Bugfix: need to give ip_route_input enough of an IP header to not gag. */ | |
eddc9ec5 | 2371 | ip_hdr(skb)->protocol = IPPROTO_ICMP; |
1da177e4 LT |
2372 | skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr)); |
2373 | ||
17fb2c64 AV |
2374 | src = tb[RTA_SRC] ? nla_get_be32(tb[RTA_SRC]) : 0; |
2375 | dst = tb[RTA_DST] ? nla_get_be32(tb[RTA_DST]) : 0; | |
d889ce3b | 2376 | iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0; |
963bfeee | 2377 | mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0; |
1da177e4 | 2378 | |
d6c0a4f6 DM |
2379 | memset(&fl4, 0, sizeof(fl4)); |
2380 | fl4.daddr = dst; | |
2381 | fl4.saddr = src; | |
2382 | fl4.flowi4_tos = rtm->rtm_tos; | |
2383 | fl4.flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0; | |
2384 | fl4.flowi4_mark = mark; | |
2385 | ||
1da177e4 | 2386 | if (iif) { |
d889ce3b TG |
2387 | struct net_device *dev; |
2388 | ||
1937504d | 2389 | dev = __dev_get_by_index(net, iif); |
d889ce3b TG |
2390 | if (dev == NULL) { |
2391 | err = -ENODEV; | |
2392 | goto errout_free; | |
2393 | } | |
2394 | ||
1da177e4 LT |
2395 | skb->protocol = htons(ETH_P_IP); |
2396 | skb->dev = dev; | |
963bfeee | 2397 | skb->mark = mark; |
1da177e4 LT |
2398 | local_bh_disable(); |
2399 | err = ip_route_input(skb, dst, src, rtm->rtm_tos, dev); | |
2400 | local_bh_enable(); | |
d889ce3b | 2401 | |
511c3f92 | 2402 | rt = skb_rtable(skb); |
d8d1f30b CG |
2403 | if (err == 0 && rt->dst.error) |
2404 | err = -rt->dst.error; | |
1da177e4 | 2405 | } else { |
9d6ec938 | 2406 | rt = ip_route_output_key(net, &fl4); |
b23dd4fe DM |
2407 | |
2408 | err = 0; | |
2409 | if (IS_ERR(rt)) | |
2410 | err = PTR_ERR(rt); | |
1da177e4 | 2411 | } |
d889ce3b | 2412 | |
1da177e4 | 2413 | if (err) |
d889ce3b | 2414 | goto errout_free; |
1da177e4 | 2415 | |
d8d1f30b | 2416 | skb_dst_set(skb, &rt->dst); |
1da177e4 LT |
2417 | if (rtm->rtm_flags & RTM_F_NOTIFY) |
2418 | rt->rt_flags |= RTCF_NOTIFY; | |
2419 | ||
f1ce3062 | 2420 | err = rt_fill_info(net, dst, src, &fl4, skb, |
15e47304 | 2421 | NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
1937504d | 2422 | RTM_NEWROUTE, 0, 0); |
d889ce3b TG |
2423 | if (err <= 0) |
2424 | goto errout_free; | |
1da177e4 | 2425 | |
15e47304 | 2426 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); |
d889ce3b | 2427 | errout: |
2942e900 | 2428 | return err; |
1da177e4 | 2429 | |
d889ce3b | 2430 | errout_free: |
1da177e4 | 2431 | kfree_skb(skb); |
d889ce3b | 2432 | goto errout; |
1da177e4 LT |
2433 | } |
2434 | ||
2435 | int ip_rt_dump(struct sk_buff *skb, struct netlink_callback *cb) | |
2436 | { | |
1da177e4 LT |
2437 | return skb->len; |
2438 | } | |
2439 | ||
2440 | void ip_rt_multicast_event(struct in_device *in_dev) | |
2441 | { | |
4ccfe6d4 | 2442 | rt_cache_flush(dev_net(in_dev->dev)); |
1da177e4 LT |
2443 | } |
2444 | ||
2445 | #ifdef CONFIG_SYSCTL | |
082c7ca4 G |
2446 | static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT; |
2447 | static int ip_rt_gc_interval __read_mostly = 60 * HZ; | |
2448 | static int ip_rt_gc_min_interval __read_mostly = HZ / 2; | |
2449 | static int ip_rt_gc_elasticity __read_mostly = 8; | |
2450 | ||
fe2c6338 | 2451 | static int ipv4_sysctl_rtcache_flush(struct ctl_table *__ctl, int write, |
8d65af78 | 2452 | void __user *buffer, |
1da177e4 LT |
2453 | size_t *lenp, loff_t *ppos) |
2454 | { | |
5aad1de5 TT |
2455 | struct net *net = (struct net *)__ctl->extra1; |
2456 | ||
1da177e4 | 2457 | if (write) { |
5aad1de5 TT |
2458 | rt_cache_flush(net); |
2459 | fnhe_genid_bump(net); | |
1da177e4 | 2460 | return 0; |
e905a9ed | 2461 | } |
1da177e4 LT |
2462 | |
2463 | return -EINVAL; | |
2464 | } | |
2465 | ||
fe2c6338 | 2466 | static struct ctl_table ipv4_route_table[] = { |
1da177e4 | 2467 | { |
1da177e4 LT |
2468 | .procname = "gc_thresh", |
2469 | .data = &ipv4_dst_ops.gc_thresh, | |
2470 | .maxlen = sizeof(int), | |
2471 | .mode = 0644, | |
6d9f239a | 2472 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2473 | }, |
2474 | { | |
1da177e4 LT |
2475 | .procname = "max_size", |
2476 | .data = &ip_rt_max_size, | |
2477 | .maxlen = sizeof(int), | |
2478 | .mode = 0644, | |
6d9f239a | 2479 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2480 | }, |
2481 | { | |
2482 | /* Deprecated. Use gc_min_interval_ms */ | |
e905a9ed | 2483 | |
1da177e4 LT |
2484 | .procname = "gc_min_interval", |
2485 | .data = &ip_rt_gc_min_interval, | |
2486 | .maxlen = sizeof(int), | |
2487 | .mode = 0644, | |
6d9f239a | 2488 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
2489 | }, |
2490 | { | |
1da177e4 LT |
2491 | .procname = "gc_min_interval_ms", |
2492 | .data = &ip_rt_gc_min_interval, | |
2493 | .maxlen = sizeof(int), | |
2494 | .mode = 0644, | |
6d9f239a | 2495 | .proc_handler = proc_dointvec_ms_jiffies, |
1da177e4 LT |
2496 | }, |
2497 | { | |
1da177e4 LT |
2498 | .procname = "gc_timeout", |
2499 | .data = &ip_rt_gc_timeout, | |
2500 | .maxlen = sizeof(int), | |
2501 | .mode = 0644, | |
6d9f239a | 2502 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 | 2503 | }, |
9f28a2fc ED |
2504 | { |
2505 | .procname = "gc_interval", | |
2506 | .data = &ip_rt_gc_interval, | |
2507 | .maxlen = sizeof(int), | |
2508 | .mode = 0644, | |
2509 | .proc_handler = proc_dointvec_jiffies, | |
2510 | }, | |
1da177e4 | 2511 | { |
1da177e4 LT |
2512 | .procname = "redirect_load", |
2513 | .data = &ip_rt_redirect_load, | |
2514 | .maxlen = sizeof(int), | |
2515 | .mode = 0644, | |
6d9f239a | 2516 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2517 | }, |
2518 | { | |
1da177e4 LT |
2519 | .procname = "redirect_number", |
2520 | .data = &ip_rt_redirect_number, | |
2521 | .maxlen = sizeof(int), | |
2522 | .mode = 0644, | |
6d9f239a | 2523 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2524 | }, |
2525 | { | |
1da177e4 LT |
2526 | .procname = "redirect_silence", |
2527 | .data = &ip_rt_redirect_silence, | |
2528 | .maxlen = sizeof(int), | |
2529 | .mode = 0644, | |
6d9f239a | 2530 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2531 | }, |
2532 | { | |
1da177e4 LT |
2533 | .procname = "error_cost", |
2534 | .data = &ip_rt_error_cost, | |
2535 | .maxlen = sizeof(int), | |
2536 | .mode = 0644, | |
6d9f239a | 2537 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2538 | }, |
2539 | { | |
1da177e4 LT |
2540 | .procname = "error_burst", |
2541 | .data = &ip_rt_error_burst, | |
2542 | .maxlen = sizeof(int), | |
2543 | .mode = 0644, | |
6d9f239a | 2544 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2545 | }, |
2546 | { | |
1da177e4 LT |
2547 | .procname = "gc_elasticity", |
2548 | .data = &ip_rt_gc_elasticity, | |
2549 | .maxlen = sizeof(int), | |
2550 | .mode = 0644, | |
6d9f239a | 2551 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2552 | }, |
2553 | { | |
1da177e4 LT |
2554 | .procname = "mtu_expires", |
2555 | .data = &ip_rt_mtu_expires, | |
2556 | .maxlen = sizeof(int), | |
2557 | .mode = 0644, | |
6d9f239a | 2558 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
2559 | }, |
2560 | { | |
1da177e4 LT |
2561 | .procname = "min_pmtu", |
2562 | .data = &ip_rt_min_pmtu, | |
2563 | .maxlen = sizeof(int), | |
2564 | .mode = 0644, | |
6d9f239a | 2565 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2566 | }, |
2567 | { | |
1da177e4 LT |
2568 | .procname = "min_adv_mss", |
2569 | .data = &ip_rt_min_advmss, | |
2570 | .maxlen = sizeof(int), | |
2571 | .mode = 0644, | |
6d9f239a | 2572 | .proc_handler = proc_dointvec, |
1da177e4 | 2573 | }, |
f8572d8f | 2574 | { } |
1da177e4 | 2575 | }; |
39a23e75 | 2576 | |
39a23e75 DL |
2577 | static struct ctl_table ipv4_route_flush_table[] = { |
2578 | { | |
39a23e75 DL |
2579 | .procname = "flush", |
2580 | .maxlen = sizeof(int), | |
2581 | .mode = 0200, | |
6d9f239a | 2582 | .proc_handler = ipv4_sysctl_rtcache_flush, |
39a23e75 | 2583 | }, |
f8572d8f | 2584 | { }, |
39a23e75 DL |
2585 | }; |
2586 | ||
2587 | static __net_init int sysctl_route_net_init(struct net *net) | |
2588 | { | |
2589 | struct ctl_table *tbl; | |
2590 | ||
2591 | tbl = ipv4_route_flush_table; | |
09ad9bc7 | 2592 | if (!net_eq(net, &init_net)) { |
39a23e75 DL |
2593 | tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL); |
2594 | if (tbl == NULL) | |
2595 | goto err_dup; | |
464dc801 EB |
2596 | |
2597 | /* Don't export sysctls to unprivileged users */ | |
2598 | if (net->user_ns != &init_user_ns) | |
2599 | tbl[0].procname = NULL; | |
39a23e75 DL |
2600 | } |
2601 | tbl[0].extra1 = net; | |
2602 | ||
ec8f23ce | 2603 | net->ipv4.route_hdr = register_net_sysctl(net, "net/ipv4/route", tbl); |
39a23e75 DL |
2604 | if (net->ipv4.route_hdr == NULL) |
2605 | goto err_reg; | |
2606 | return 0; | |
2607 | ||
2608 | err_reg: | |
2609 | if (tbl != ipv4_route_flush_table) | |
2610 | kfree(tbl); | |
2611 | err_dup: | |
2612 | return -ENOMEM; | |
2613 | } | |
2614 | ||
2615 | static __net_exit void sysctl_route_net_exit(struct net *net) | |
2616 | { | |
2617 | struct ctl_table *tbl; | |
2618 | ||
2619 | tbl = net->ipv4.route_hdr->ctl_table_arg; | |
2620 | unregister_net_sysctl_table(net->ipv4.route_hdr); | |
2621 | BUG_ON(tbl == ipv4_route_flush_table); | |
2622 | kfree(tbl); | |
2623 | } | |
2624 | ||
2625 | static __net_initdata struct pernet_operations sysctl_route_ops = { | |
2626 | .init = sysctl_route_net_init, | |
2627 | .exit = sysctl_route_net_exit, | |
2628 | }; | |
1da177e4 LT |
2629 | #endif |
2630 | ||
3ee94372 | 2631 | static __net_init int rt_genid_init(struct net *net) |
9f5e97e5 | 2632 | { |
b42664f8 | 2633 | atomic_set(&net->rt_genid, 0); |
5aad1de5 | 2634 | atomic_set(&net->fnhe_genid, 0); |
436c3b66 DM |
2635 | get_random_bytes(&net->ipv4.dev_addr_genid, |
2636 | sizeof(net->ipv4.dev_addr_genid)); | |
9f5e97e5 DL |
2637 | return 0; |
2638 | } | |
2639 | ||
3ee94372 NH |
2640 | static __net_initdata struct pernet_operations rt_genid_ops = { |
2641 | .init = rt_genid_init, | |
9f5e97e5 DL |
2642 | }; |
2643 | ||
c3426b47 DM |
2644 | static int __net_init ipv4_inetpeer_init(struct net *net) |
2645 | { | |
2646 | struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL); | |
2647 | ||
2648 | if (!bp) | |
2649 | return -ENOMEM; | |
2650 | inet_peer_base_init(bp); | |
2651 | net->ipv4.peers = bp; | |
2652 | return 0; | |
2653 | } | |
2654 | ||
2655 | static void __net_exit ipv4_inetpeer_exit(struct net *net) | |
2656 | { | |
2657 | struct inet_peer_base *bp = net->ipv4.peers; | |
2658 | ||
2659 | net->ipv4.peers = NULL; | |
56a6b248 | 2660 | inetpeer_invalidate_tree(bp); |
c3426b47 DM |
2661 | kfree(bp); |
2662 | } | |
2663 | ||
2664 | static __net_initdata struct pernet_operations ipv4_inetpeer_ops = { | |
2665 | .init = ipv4_inetpeer_init, | |
2666 | .exit = ipv4_inetpeer_exit, | |
2667 | }; | |
9f5e97e5 | 2668 | |
c7066f70 | 2669 | #ifdef CONFIG_IP_ROUTE_CLASSID |
7d720c3e | 2670 | struct ip_rt_acct __percpu *ip_rt_acct __read_mostly; |
c7066f70 | 2671 | #endif /* CONFIG_IP_ROUTE_CLASSID */ |
1da177e4 | 2672 | |
1da177e4 LT |
2673 | int __init ip_rt_init(void) |
2674 | { | |
424c4b70 | 2675 | int rc = 0; |
1da177e4 | 2676 | |
c7066f70 | 2677 | #ifdef CONFIG_IP_ROUTE_CLASSID |
0dcec8c2 | 2678 | ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct)); |
1da177e4 LT |
2679 | if (!ip_rt_acct) |
2680 | panic("IP: failed to allocate ip_rt_acct\n"); | |
1da177e4 LT |
2681 | #endif |
2682 | ||
e5d679f3 AD |
2683 | ipv4_dst_ops.kmem_cachep = |
2684 | kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0, | |
20c2df83 | 2685 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 2686 | |
14e50e57 DM |
2687 | ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep; |
2688 | ||
fc66f95c ED |
2689 | if (dst_entries_init(&ipv4_dst_ops) < 0) |
2690 | panic("IP: failed to allocate ipv4_dst_ops counter\n"); | |
2691 | ||
2692 | if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0) | |
2693 | panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n"); | |
2694 | ||
89aef892 DM |
2695 | ipv4_dst_ops.gc_thresh = ~0; |
2696 | ip_rt_max_size = INT_MAX; | |
1da177e4 | 2697 | |
1da177e4 LT |
2698 | devinet_init(); |
2699 | ip_fib_init(); | |
2700 | ||
73b38711 | 2701 | if (ip_rt_proc_init()) |
058bd4d2 | 2702 | pr_err("Unable to create route proc files\n"); |
1da177e4 LT |
2703 | #ifdef CONFIG_XFRM |
2704 | xfrm_init(); | |
703fb94e | 2705 | xfrm4_init(); |
1da177e4 | 2706 | #endif |
c7ac8679 | 2707 | rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL, NULL); |
63f3444f | 2708 | |
39a23e75 DL |
2709 | #ifdef CONFIG_SYSCTL |
2710 | register_pernet_subsys(&sysctl_route_ops); | |
2711 | #endif | |
3ee94372 | 2712 | register_pernet_subsys(&rt_genid_ops); |
c3426b47 | 2713 | register_pernet_subsys(&ipv4_inetpeer_ops); |
1da177e4 LT |
2714 | return rc; |
2715 | } | |
2716 | ||
a1bc6eb4 | 2717 | #ifdef CONFIG_SYSCTL |
eeb61f71 AV |
2718 | /* |
2719 | * We really need to sanitize the damn ipv4 init order, then all | |
2720 | * this nonsense will go away. | |
2721 | */ | |
2722 | void __init ip_static_sysctl_init(void) | |
2723 | { | |
4e5ca785 | 2724 | register_net_sysctl(&init_net, "net/ipv4/route", ipv4_route_table); |
eeb61f71 | 2725 | } |
a1bc6eb4 | 2726 | #endif |