]>
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 | 67 | #include <linux/module.h> |
7c0f6ba6 | 68 | #include <linux/uaccess.h> |
1da177e4 LT |
69 | #include <linux/bitops.h> |
70 | #include <linux/types.h> | |
71 | #include <linux/kernel.h> | |
1da177e4 LT |
72 | #include <linux/mm.h> |
73 | #include <linux/string.h> | |
74 | #include <linux/socket.h> | |
75 | #include <linux/sockios.h> | |
76 | #include <linux/errno.h> | |
77 | #include <linux/in.h> | |
78 | #include <linux/inet.h> | |
79 | #include <linux/netdevice.h> | |
80 | #include <linux/proc_fs.h> | |
81 | #include <linux/init.h> | |
82 | #include <linux/skbuff.h> | |
1da177e4 LT |
83 | #include <linux/inetdevice.h> |
84 | #include <linux/igmp.h> | |
85 | #include <linux/pkt_sched.h> | |
86 | #include <linux/mroute.h> | |
87 | #include <linux/netfilter_ipv4.h> | |
88 | #include <linux/random.h> | |
1da177e4 LT |
89 | #include <linux/rcupdate.h> |
90 | #include <linux/times.h> | |
5a0e3ad6 | 91 | #include <linux/slab.h> |
73f156a6 | 92 | #include <linux/jhash.h> |
352e512c | 93 | #include <net/dst.h> |
1b7179d3 | 94 | #include <net/dst_metadata.h> |
457c4cbc | 95 | #include <net/net_namespace.h> |
1da177e4 LT |
96 | #include <net/protocol.h> |
97 | #include <net/ip.h> | |
98 | #include <net/route.h> | |
99 | #include <net/inetpeer.h> | |
100 | #include <net/sock.h> | |
101 | #include <net/ip_fib.h> | |
102 | #include <net/arp.h> | |
103 | #include <net/tcp.h> | |
104 | #include <net/icmp.h> | |
105 | #include <net/xfrm.h> | |
571e7226 | 106 | #include <net/lwtunnel.h> |
8d71740c | 107 | #include <net/netevent.h> |
63f3444f | 108 | #include <net/rtnetlink.h> |
1da177e4 LT |
109 | #ifdef CONFIG_SYSCTL |
110 | #include <linux/sysctl.h> | |
7426a564 | 111 | #include <linux/kmemleak.h> |
1da177e4 | 112 | #endif |
6e5714ea | 113 | #include <net/secure_seq.h> |
1b7179d3 | 114 | #include <net/ip_tunnels.h> |
385add90 | 115 | #include <net/l3mdev.h> |
1da177e4 | 116 | |
b6179813 RP |
117 | #include "fib_lookup.h" |
118 | ||
68a5e3dd | 119 | #define RT_FL_TOS(oldflp4) \ |
f61759e6 | 120 | ((oldflp4)->flowi4_tos & (IPTOS_RT_MASK | RTO_ONLINK)) |
1da177e4 | 121 | |
1da177e4 LT |
122 | #define RT_GC_TIMEOUT (300*HZ) |
123 | ||
1da177e4 | 124 | static int ip_rt_max_size; |
817bc4db SH |
125 | static int ip_rt_redirect_number __read_mostly = 9; |
126 | static int ip_rt_redirect_load __read_mostly = HZ / 50; | |
127 | static int ip_rt_redirect_silence __read_mostly = ((HZ / 50) << (9 + 1)); | |
128 | static int ip_rt_error_cost __read_mostly = HZ; | |
129 | static int ip_rt_error_burst __read_mostly = 5 * HZ; | |
817bc4db SH |
130 | static int ip_rt_mtu_expires __read_mostly = 10 * 60 * HZ; |
131 | static int ip_rt_min_pmtu __read_mostly = 512 + 20 + 20; | |
132 | static int ip_rt_min_advmss __read_mostly = 256; | |
9f28a2fc | 133 | |
deed49df | 134 | static int ip_rt_gc_timeout __read_mostly = RT_GC_TIMEOUT; |
1da177e4 LT |
135 | /* |
136 | * Interface to generic destination cache. | |
137 | */ | |
138 | ||
139 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie); | |
0dbaee3b | 140 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst); |
ebb762f2 | 141 | static unsigned int ipv4_mtu(const struct dst_entry *dst); |
1da177e4 LT |
142 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst); |
143 | static void ipv4_link_failure(struct sk_buff *skb); | |
6700c270 DM |
144 | static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk, |
145 | struct sk_buff *skb, u32 mtu); | |
146 | static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, | |
147 | struct sk_buff *skb); | |
caacf05e | 148 | static void ipv4_dst_destroy(struct dst_entry *dst); |
1da177e4 | 149 | |
62fa8a84 DM |
150 | static u32 *ipv4_cow_metrics(struct dst_entry *dst, unsigned long old) |
151 | { | |
31248731 DM |
152 | WARN_ON(1); |
153 | return NULL; | |
62fa8a84 DM |
154 | } |
155 | ||
f894cbf8 DM |
156 | static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, |
157 | struct sk_buff *skb, | |
158 | const void *daddr); | |
63fca65d | 159 | static void ipv4_confirm_neigh(const struct dst_entry *dst, const void *daddr); |
d3aaeb38 | 160 | |
1da177e4 LT |
161 | static struct dst_ops ipv4_dst_ops = { |
162 | .family = AF_INET, | |
1da177e4 | 163 | .check = ipv4_dst_check, |
0dbaee3b | 164 | .default_advmss = ipv4_default_advmss, |
ebb762f2 | 165 | .mtu = ipv4_mtu, |
62fa8a84 | 166 | .cow_metrics = ipv4_cow_metrics, |
caacf05e | 167 | .destroy = ipv4_dst_destroy, |
1da177e4 LT |
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, |
b92dacd4 | 172 | .local_out = __ip_local_out, |
d3aaeb38 | 173 | .neigh_lookup = ipv4_neigh_lookup, |
63fca65d | 174 | .confirm_neigh = ipv4_confirm_neigh, |
1da177e4 LT |
175 | }; |
176 | ||
177 | #define ECN_OR_COST(class) TC_PRIO_##class | |
178 | ||
4839c52b | 179 | const __u8 ip_tos2prio[16] = { |
1da177e4 | 180 | TC_PRIO_BESTEFFORT, |
4a2b9c37 | 181 | ECN_OR_COST(BESTEFFORT), |
1da177e4 LT |
182 | TC_PRIO_BESTEFFORT, |
183 | ECN_OR_COST(BESTEFFORT), | |
184 | TC_PRIO_BULK, | |
185 | ECN_OR_COST(BULK), | |
186 | TC_PRIO_BULK, | |
187 | ECN_OR_COST(BULK), | |
188 | TC_PRIO_INTERACTIVE, | |
189 | ECN_OR_COST(INTERACTIVE), | |
190 | TC_PRIO_INTERACTIVE, | |
191 | ECN_OR_COST(INTERACTIVE), | |
192 | TC_PRIO_INTERACTIVE_BULK, | |
193 | ECN_OR_COST(INTERACTIVE_BULK), | |
194 | TC_PRIO_INTERACTIVE_BULK, | |
195 | ECN_OR_COST(INTERACTIVE_BULK) | |
196 | }; | |
d4a96865 | 197 | EXPORT_SYMBOL(ip_tos2prio); |
1da177e4 | 198 | |
2f970d83 | 199 | static DEFINE_PER_CPU(struct rt_cache_stat, rt_cache_stat); |
3ed66e91 | 200 | #define RT_CACHE_STAT_INC(field) raw_cpu_inc(rt_cache_stat.field) |
1da177e4 | 201 | |
1da177e4 | 202 | #ifdef CONFIG_PROC_FS |
1da177e4 LT |
203 | static void *rt_cache_seq_start(struct seq_file *seq, loff_t *pos) |
204 | { | |
29e75252 | 205 | if (*pos) |
89aef892 | 206 | return NULL; |
29e75252 | 207 | return SEQ_START_TOKEN; |
1da177e4 LT |
208 | } |
209 | ||
210 | static void *rt_cache_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
211 | { | |
1da177e4 | 212 | ++*pos; |
89aef892 | 213 | return NULL; |
1da177e4 LT |
214 | } |
215 | ||
216 | static void rt_cache_seq_stop(struct seq_file *seq, void *v) | |
217 | { | |
1da177e4 LT |
218 | } |
219 | ||
220 | static int rt_cache_seq_show(struct seq_file *seq, void *v) | |
221 | { | |
222 | if (v == SEQ_START_TOKEN) | |
223 | seq_printf(seq, "%-127s\n", | |
224 | "Iface\tDestination\tGateway \tFlags\t\tRefCnt\tUse\t" | |
225 | "Metric\tSource\t\tMTU\tWindow\tIRTT\tTOS\tHHRef\t" | |
226 | "HHUptod\tSpecDst"); | |
e905a9ed | 227 | return 0; |
1da177e4 LT |
228 | } |
229 | ||
f690808e | 230 | static const struct seq_operations rt_cache_seq_ops = { |
1da177e4 LT |
231 | .start = rt_cache_seq_start, |
232 | .next = rt_cache_seq_next, | |
233 | .stop = rt_cache_seq_stop, | |
234 | .show = rt_cache_seq_show, | |
235 | }; | |
236 | ||
237 | static int rt_cache_seq_open(struct inode *inode, struct file *file) | |
238 | { | |
89aef892 | 239 | return seq_open(file, &rt_cache_seq_ops); |
1da177e4 LT |
240 | } |
241 | ||
9a32144e | 242 | static const struct file_operations rt_cache_seq_fops = { |
1da177e4 LT |
243 | .owner = THIS_MODULE, |
244 | .open = rt_cache_seq_open, | |
245 | .read = seq_read, | |
246 | .llseek = seq_lseek, | |
89aef892 | 247 | .release = seq_release, |
1da177e4 LT |
248 | }; |
249 | ||
250 | ||
251 | static void *rt_cpu_seq_start(struct seq_file *seq, loff_t *pos) | |
252 | { | |
253 | int cpu; | |
254 | ||
255 | if (*pos == 0) | |
256 | return SEQ_START_TOKEN; | |
257 | ||
0f23174a | 258 | for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
259 | if (!cpu_possible(cpu)) |
260 | continue; | |
261 | *pos = cpu+1; | |
2f970d83 | 262 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
263 | } |
264 | return NULL; | |
265 | } | |
266 | ||
267 | static void *rt_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
268 | { | |
269 | int cpu; | |
270 | ||
0f23174a | 271 | for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) { |
1da177e4 LT |
272 | if (!cpu_possible(cpu)) |
273 | continue; | |
274 | *pos = cpu+1; | |
2f970d83 | 275 | return &per_cpu(rt_cache_stat, cpu); |
1da177e4 LT |
276 | } |
277 | return NULL; | |
e905a9ed | 278 | |
1da177e4 LT |
279 | } |
280 | ||
281 | static void rt_cpu_seq_stop(struct seq_file *seq, void *v) | |
282 | { | |
283 | ||
284 | } | |
285 | ||
286 | static int rt_cpu_seq_show(struct seq_file *seq, void *v) | |
287 | { | |
288 | struct rt_cache_stat *st = v; | |
289 | ||
290 | if (v == SEQ_START_TOKEN) { | |
5bec0039 | 291 | 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 |
292 | return 0; |
293 | } | |
e905a9ed | 294 | |
1da177e4 LT |
295 | seq_printf(seq,"%08x %08x %08x %08x %08x %08x %08x %08x " |
296 | " %08x %08x %08x %08x %08x %08x %08x %08x %08x \n", | |
fc66f95c | 297 | dst_entries_get_slow(&ipv4_dst_ops), |
0baf2b35 | 298 | 0, /* st->in_hit */ |
1da177e4 LT |
299 | st->in_slow_tot, |
300 | st->in_slow_mc, | |
301 | st->in_no_route, | |
302 | st->in_brd, | |
303 | st->in_martian_dst, | |
304 | st->in_martian_src, | |
305 | ||
0baf2b35 | 306 | 0, /* st->out_hit */ |
1da177e4 | 307 | st->out_slow_tot, |
e905a9ed | 308 | st->out_slow_mc, |
1da177e4 | 309 | |
0baf2b35 ED |
310 | 0, /* st->gc_total */ |
311 | 0, /* st->gc_ignored */ | |
312 | 0, /* st->gc_goal_miss */ | |
313 | 0, /* st->gc_dst_overflow */ | |
314 | 0, /* st->in_hlist_search */ | |
315 | 0 /* st->out_hlist_search */ | |
1da177e4 LT |
316 | ); |
317 | return 0; | |
318 | } | |
319 | ||
f690808e | 320 | static const struct seq_operations rt_cpu_seq_ops = { |
1da177e4 LT |
321 | .start = rt_cpu_seq_start, |
322 | .next = rt_cpu_seq_next, | |
323 | .stop = rt_cpu_seq_stop, | |
324 | .show = rt_cpu_seq_show, | |
325 | }; | |
326 | ||
327 | ||
328 | static int rt_cpu_seq_open(struct inode *inode, struct file *file) | |
329 | { | |
330 | return seq_open(file, &rt_cpu_seq_ops); | |
331 | } | |
332 | ||
9a32144e | 333 | static const struct file_operations rt_cpu_seq_fops = { |
1da177e4 LT |
334 | .owner = THIS_MODULE, |
335 | .open = rt_cpu_seq_open, | |
336 | .read = seq_read, | |
337 | .llseek = seq_lseek, | |
338 | .release = seq_release, | |
339 | }; | |
340 | ||
c7066f70 | 341 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 342 | static int rt_acct_proc_show(struct seq_file *m, void *v) |
78c686e9 | 343 | { |
a661c419 AD |
344 | struct ip_rt_acct *dst, *src; |
345 | unsigned int i, j; | |
346 | ||
347 | dst = kcalloc(256, sizeof(struct ip_rt_acct), GFP_KERNEL); | |
348 | if (!dst) | |
349 | return -ENOMEM; | |
350 | ||
351 | for_each_possible_cpu(i) { | |
352 | src = (struct ip_rt_acct *)per_cpu_ptr(ip_rt_acct, i); | |
353 | for (j = 0; j < 256; j++) { | |
354 | dst[j].o_bytes += src[j].o_bytes; | |
355 | dst[j].o_packets += src[j].o_packets; | |
356 | dst[j].i_bytes += src[j].i_bytes; | |
357 | dst[j].i_packets += src[j].i_packets; | |
358 | } | |
78c686e9 PE |
359 | } |
360 | ||
a661c419 AD |
361 | seq_write(m, dst, 256 * sizeof(struct ip_rt_acct)); |
362 | kfree(dst); | |
363 | return 0; | |
364 | } | |
78c686e9 | 365 | |
a661c419 AD |
366 | static int rt_acct_proc_open(struct inode *inode, struct file *file) |
367 | { | |
368 | return single_open(file, rt_acct_proc_show, NULL); | |
78c686e9 | 369 | } |
a661c419 AD |
370 | |
371 | static const struct file_operations rt_acct_proc_fops = { | |
372 | .owner = THIS_MODULE, | |
373 | .open = rt_acct_proc_open, | |
374 | .read = seq_read, | |
375 | .llseek = seq_lseek, | |
376 | .release = single_release, | |
377 | }; | |
78c686e9 | 378 | #endif |
107f1634 | 379 | |
73b38711 | 380 | static int __net_init ip_rt_do_proc_init(struct net *net) |
107f1634 PE |
381 | { |
382 | struct proc_dir_entry *pde; | |
383 | ||
d4beaa66 G |
384 | pde = proc_create("rt_cache", S_IRUGO, net->proc_net, |
385 | &rt_cache_seq_fops); | |
107f1634 PE |
386 | if (!pde) |
387 | goto err1; | |
388 | ||
77020720 WC |
389 | pde = proc_create("rt_cache", S_IRUGO, |
390 | net->proc_net_stat, &rt_cpu_seq_fops); | |
107f1634 PE |
391 | if (!pde) |
392 | goto err2; | |
393 | ||
c7066f70 | 394 | #ifdef CONFIG_IP_ROUTE_CLASSID |
a661c419 | 395 | pde = proc_create("rt_acct", 0, net->proc_net, &rt_acct_proc_fops); |
107f1634 PE |
396 | if (!pde) |
397 | goto err3; | |
398 | #endif | |
399 | return 0; | |
400 | ||
c7066f70 | 401 | #ifdef CONFIG_IP_ROUTE_CLASSID |
107f1634 PE |
402 | err3: |
403 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
404 | #endif | |
405 | err2: | |
406 | remove_proc_entry("rt_cache", net->proc_net); | |
407 | err1: | |
408 | return -ENOMEM; | |
409 | } | |
73b38711 DL |
410 | |
411 | static void __net_exit ip_rt_do_proc_exit(struct net *net) | |
412 | { | |
413 | remove_proc_entry("rt_cache", net->proc_net_stat); | |
414 | remove_proc_entry("rt_cache", net->proc_net); | |
c7066f70 | 415 | #ifdef CONFIG_IP_ROUTE_CLASSID |
73b38711 | 416 | remove_proc_entry("rt_acct", net->proc_net); |
0a931acf | 417 | #endif |
73b38711 DL |
418 | } |
419 | ||
420 | static struct pernet_operations ip_rt_proc_ops __net_initdata = { | |
421 | .init = ip_rt_do_proc_init, | |
422 | .exit = ip_rt_do_proc_exit, | |
423 | }; | |
424 | ||
425 | static int __init ip_rt_proc_init(void) | |
426 | { | |
427 | return register_pernet_subsys(&ip_rt_proc_ops); | |
428 | } | |
429 | ||
107f1634 | 430 | #else |
73b38711 | 431 | static inline int ip_rt_proc_init(void) |
107f1634 PE |
432 | { |
433 | return 0; | |
434 | } | |
1da177e4 | 435 | #endif /* CONFIG_PROC_FS */ |
e905a9ed | 436 | |
4331debc | 437 | static inline bool rt_is_expired(const struct rtable *rth) |
e84f84f2 | 438 | { |
ca4c3fc2 | 439 | return rth->rt_genid != rt_genid_ipv4(dev_net(rth->dst.dev)); |
e84f84f2 DL |
440 | } |
441 | ||
4ccfe6d4 | 442 | void rt_cache_flush(struct net *net) |
1da177e4 | 443 | { |
ca4c3fc2 | 444 | rt_genid_bump_ipv4(net); |
98376387 ED |
445 | } |
446 | ||
f894cbf8 DM |
447 | static struct neighbour *ipv4_neigh_lookup(const struct dst_entry *dst, |
448 | struct sk_buff *skb, | |
449 | const void *daddr) | |
3769cffb | 450 | { |
d3aaeb38 DM |
451 | struct net_device *dev = dst->dev; |
452 | const __be32 *pkey = daddr; | |
39232973 | 453 | const struct rtable *rt; |
3769cffb DM |
454 | struct neighbour *n; |
455 | ||
39232973 | 456 | rt = (const struct rtable *) dst; |
a263b309 | 457 | if (rt->rt_gateway) |
39232973 | 458 | pkey = (const __be32 *) &rt->rt_gateway; |
f894cbf8 DM |
459 | else if (skb) |
460 | pkey = &ip_hdr(skb)->daddr; | |
d3aaeb38 | 461 | |
80703d26 | 462 | n = __ipv4_neigh_lookup(dev, *(__force u32 *)pkey); |
d3aaeb38 DM |
463 | if (n) |
464 | return n; | |
32092ecf | 465 | return neigh_create(&arp_tbl, pkey, dev); |
d3aaeb38 DM |
466 | } |
467 | ||
63fca65d JA |
468 | static void ipv4_confirm_neigh(const struct dst_entry *dst, const void *daddr) |
469 | { | |
470 | struct net_device *dev = dst->dev; | |
471 | const __be32 *pkey = daddr; | |
472 | const struct rtable *rt; | |
473 | ||
474 | rt = (const struct rtable *)dst; | |
475 | if (rt->rt_gateway) | |
476 | pkey = (const __be32 *)&rt->rt_gateway; | |
477 | else if (!daddr || | |
478 | (rt->rt_flags & | |
479 | (RTCF_MULTICAST | RTCF_BROADCAST | RTCF_LOCAL))) | |
480 | return; | |
481 | ||
482 | __ipv4_confirm_neigh(dev, *(__force u32 *)pkey); | |
483 | } | |
484 | ||
04ca6973 | 485 | #define IP_IDENTS_SZ 2048u |
04ca6973 | 486 | |
355b590c ED |
487 | static atomic_t *ip_idents __read_mostly; |
488 | static u32 *ip_tstamps __read_mostly; | |
04ca6973 ED |
489 | |
490 | /* In order to protect privacy, we add a perturbation to identifiers | |
491 | * if one generator is seldom used. This makes hard for an attacker | |
492 | * to infer how many packets were sent between two points in time. | |
493 | */ | |
494 | u32 ip_idents_reserve(u32 hash, int segs) | |
495 | { | |
355b590c ED |
496 | u32 *p_tstamp = ip_tstamps + hash % IP_IDENTS_SZ; |
497 | atomic_t *p_id = ip_idents + hash % IP_IDENTS_SZ; | |
498 | u32 old = ACCESS_ONCE(*p_tstamp); | |
04ca6973 | 499 | u32 now = (u32)jiffies; |
adb03115 | 500 | u32 new, delta = 0; |
04ca6973 | 501 | |
355b590c | 502 | if (old != now && cmpxchg(p_tstamp, old, now) == old) |
04ca6973 ED |
503 | delta = prandom_u32_max(now - old); |
504 | ||
adb03115 ED |
505 | /* Do not use atomic_add_return() as it makes UBSAN unhappy */ |
506 | do { | |
507 | old = (u32)atomic_read(p_id); | |
508 | new = old + delta + segs; | |
509 | } while (atomic_cmpxchg(p_id, old, new) != old); | |
510 | ||
511 | return new - segs; | |
04ca6973 ED |
512 | } |
513 | EXPORT_SYMBOL(ip_idents_reserve); | |
1da177e4 | 514 | |
b6a7719a | 515 | void __ip_select_ident(struct net *net, struct iphdr *iph, int segs) |
1da177e4 | 516 | { |
73f156a6 ED |
517 | static u32 ip_idents_hashrnd __read_mostly; |
518 | u32 hash, id; | |
1da177e4 | 519 | |
73f156a6 | 520 | net_get_random_once(&ip_idents_hashrnd, sizeof(ip_idents_hashrnd)); |
1da177e4 | 521 | |
04ca6973 ED |
522 | hash = jhash_3words((__force u32)iph->daddr, |
523 | (__force u32)iph->saddr, | |
b6a7719a | 524 | iph->protocol ^ net_hash_mix(net), |
04ca6973 | 525 | ip_idents_hashrnd); |
73f156a6 ED |
526 | id = ip_idents_reserve(hash, segs); |
527 | iph->id = htons(id); | |
1da177e4 | 528 | } |
4bc2f18b | 529 | EXPORT_SYMBOL(__ip_select_ident); |
1da177e4 | 530 | |
e2d118a1 LC |
531 | static void __build_flow_key(const struct net *net, struct flowi4 *fl4, |
532 | const struct sock *sk, | |
4895c771 DM |
533 | const struct iphdr *iph, |
534 | int oif, u8 tos, | |
535 | u8 prot, u32 mark, int flow_flags) | |
536 | { | |
537 | if (sk) { | |
538 | const struct inet_sock *inet = inet_sk(sk); | |
539 | ||
540 | oif = sk->sk_bound_dev_if; | |
541 | mark = sk->sk_mark; | |
542 | tos = RT_CONN_FLAGS(sk); | |
543 | prot = inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol; | |
544 | } | |
545 | flowi4_init_output(fl4, oif, mark, tos, | |
546 | RT_SCOPE_UNIVERSE, prot, | |
547 | flow_flags, | |
e2d118a1 LC |
548 | iph->daddr, iph->saddr, 0, 0, |
549 | sock_net_uid(net, sk)); | |
4895c771 DM |
550 | } |
551 | ||
5abf7f7e ED |
552 | static void build_skb_flow_key(struct flowi4 *fl4, const struct sk_buff *skb, |
553 | const struct sock *sk) | |
4895c771 | 554 | { |
d109e61b | 555 | const struct net *net = dev_net(skb->dev); |
4895c771 DM |
556 | const struct iphdr *iph = ip_hdr(skb); |
557 | int oif = skb->dev->ifindex; | |
558 | u8 tos = RT_TOS(iph->tos); | |
559 | u8 prot = iph->protocol; | |
560 | u32 mark = skb->mark; | |
561 | ||
d109e61b | 562 | __build_flow_key(net, fl4, sk, iph, oif, tos, prot, mark, 0); |
4895c771 DM |
563 | } |
564 | ||
5abf7f7e | 565 | static void build_sk_flow_key(struct flowi4 *fl4, const struct sock *sk) |
4895c771 DM |
566 | { |
567 | const struct inet_sock *inet = inet_sk(sk); | |
5abf7f7e | 568 | const struct ip_options_rcu *inet_opt; |
4895c771 DM |
569 | __be32 daddr = inet->inet_daddr; |
570 | ||
571 | rcu_read_lock(); | |
572 | inet_opt = rcu_dereference(inet->inet_opt); | |
573 | if (inet_opt && inet_opt->opt.srr) | |
574 | daddr = inet_opt->opt.faddr; | |
575 | flowi4_init_output(fl4, sk->sk_bound_dev_if, sk->sk_mark, | |
576 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, | |
577 | inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol, | |
578 | inet_sk_flowi_flags(sk), | |
e2d118a1 | 579 | daddr, inet->inet_saddr, 0, 0, sk->sk_uid); |
4895c771 DM |
580 | rcu_read_unlock(); |
581 | } | |
582 | ||
5abf7f7e ED |
583 | static void ip_rt_build_flow_key(struct flowi4 *fl4, const struct sock *sk, |
584 | const struct sk_buff *skb) | |
4895c771 DM |
585 | { |
586 | if (skb) | |
587 | build_skb_flow_key(fl4, skb, sk); | |
588 | else | |
589 | build_sk_flow_key(fl4, sk); | |
590 | } | |
591 | ||
c5038a83 DM |
592 | static inline void rt_free(struct rtable *rt) |
593 | { | |
594 | call_rcu(&rt->dst.rcu_head, dst_rcu_free); | |
595 | } | |
596 | ||
597 | static DEFINE_SPINLOCK(fnhe_lock); | |
4895c771 | 598 | |
2ffae99d TT |
599 | static void fnhe_flush_routes(struct fib_nh_exception *fnhe) |
600 | { | |
601 | struct rtable *rt; | |
602 | ||
603 | rt = rcu_dereference(fnhe->fnhe_rth_input); | |
604 | if (rt) { | |
605 | RCU_INIT_POINTER(fnhe->fnhe_rth_input, NULL); | |
606 | rt_free(rt); | |
607 | } | |
608 | rt = rcu_dereference(fnhe->fnhe_rth_output); | |
609 | if (rt) { | |
610 | RCU_INIT_POINTER(fnhe->fnhe_rth_output, NULL); | |
611 | rt_free(rt); | |
612 | } | |
613 | } | |
614 | ||
aee06da6 | 615 | static struct fib_nh_exception *fnhe_oldest(struct fnhe_hash_bucket *hash) |
4895c771 DM |
616 | { |
617 | struct fib_nh_exception *fnhe, *oldest; | |
618 | ||
619 | oldest = rcu_dereference(hash->chain); | |
620 | for (fnhe = rcu_dereference(oldest->fnhe_next); fnhe; | |
621 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
622 | if (time_before(fnhe->fnhe_stamp, oldest->fnhe_stamp)) | |
623 | oldest = fnhe; | |
624 | } | |
2ffae99d | 625 | fnhe_flush_routes(oldest); |
4895c771 DM |
626 | return oldest; |
627 | } | |
628 | ||
d3a25c98 DM |
629 | static inline u32 fnhe_hashfun(__be32 daddr) |
630 | { | |
d546c621 | 631 | static u32 fnhe_hashrnd __read_mostly; |
d3a25c98 DM |
632 | u32 hval; |
633 | ||
d546c621 ED |
634 | net_get_random_once(&fnhe_hashrnd, sizeof(fnhe_hashrnd)); |
635 | hval = jhash_1word((__force u32) daddr, fnhe_hashrnd); | |
636 | return hash_32(hval, FNHE_HASH_SHIFT); | |
d3a25c98 DM |
637 | } |
638 | ||
387aa65a TT |
639 | static void fill_route_from_fnhe(struct rtable *rt, struct fib_nh_exception *fnhe) |
640 | { | |
641 | rt->rt_pmtu = fnhe->fnhe_pmtu; | |
642 | rt->dst.expires = fnhe->fnhe_expires; | |
643 | ||
644 | if (fnhe->fnhe_gw) { | |
645 | rt->rt_flags |= RTCF_REDIRECTED; | |
646 | rt->rt_gateway = fnhe->fnhe_gw; | |
647 | rt->rt_uses_gateway = 1; | |
648 | } | |
649 | } | |
650 | ||
aee06da6 JA |
651 | static void update_or_create_fnhe(struct fib_nh *nh, __be32 daddr, __be32 gw, |
652 | u32 pmtu, unsigned long expires) | |
4895c771 | 653 | { |
aee06da6 | 654 | struct fnhe_hash_bucket *hash; |
4895c771 | 655 | struct fib_nh_exception *fnhe; |
387aa65a TT |
656 | struct rtable *rt; |
657 | unsigned int i; | |
4895c771 | 658 | int depth; |
aee06da6 JA |
659 | u32 hval = fnhe_hashfun(daddr); |
660 | ||
c5038a83 | 661 | spin_lock_bh(&fnhe_lock); |
4895c771 | 662 | |
caa41527 | 663 | hash = rcu_dereference(nh->nh_exceptions); |
4895c771 | 664 | if (!hash) { |
aee06da6 | 665 | hash = kzalloc(FNHE_HASH_SIZE * sizeof(*hash), GFP_ATOMIC); |
4895c771 | 666 | if (!hash) |
aee06da6 | 667 | goto out_unlock; |
caa41527 | 668 | rcu_assign_pointer(nh->nh_exceptions, hash); |
4895c771 DM |
669 | } |
670 | ||
4895c771 DM |
671 | hash += hval; |
672 | ||
673 | depth = 0; | |
674 | for (fnhe = rcu_dereference(hash->chain); fnhe; | |
675 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
676 | if (fnhe->fnhe_daddr == daddr) | |
aee06da6 | 677 | break; |
4895c771 DM |
678 | depth++; |
679 | } | |
680 | ||
aee06da6 JA |
681 | if (fnhe) { |
682 | if (gw) | |
683 | fnhe->fnhe_gw = gw; | |
684 | if (pmtu) { | |
685 | fnhe->fnhe_pmtu = pmtu; | |
387aa65a | 686 | fnhe->fnhe_expires = max(1UL, expires); |
aee06da6 | 687 | } |
387aa65a | 688 | /* Update all cached dsts too */ |
2ffae99d TT |
689 | rt = rcu_dereference(fnhe->fnhe_rth_input); |
690 | if (rt) | |
691 | fill_route_from_fnhe(rt, fnhe); | |
692 | rt = rcu_dereference(fnhe->fnhe_rth_output); | |
387aa65a TT |
693 | if (rt) |
694 | fill_route_from_fnhe(rt, fnhe); | |
aee06da6 JA |
695 | } else { |
696 | if (depth > FNHE_RECLAIM_DEPTH) | |
697 | fnhe = fnhe_oldest(hash); | |
698 | else { | |
699 | fnhe = kzalloc(sizeof(*fnhe), GFP_ATOMIC); | |
700 | if (!fnhe) | |
701 | goto out_unlock; | |
702 | ||
703 | fnhe->fnhe_next = hash->chain; | |
704 | rcu_assign_pointer(hash->chain, fnhe); | |
705 | } | |
5aad1de5 | 706 | fnhe->fnhe_genid = fnhe_genid(dev_net(nh->nh_dev)); |
aee06da6 JA |
707 | fnhe->fnhe_daddr = daddr; |
708 | fnhe->fnhe_gw = gw; | |
709 | fnhe->fnhe_pmtu = pmtu; | |
710 | fnhe->fnhe_expires = expires; | |
387aa65a TT |
711 | |
712 | /* Exception created; mark the cached routes for the nexthop | |
713 | * stale, so anyone caching it rechecks if this exception | |
714 | * applies to them. | |
715 | */ | |
2ffae99d TT |
716 | rt = rcu_dereference(nh->nh_rth_input); |
717 | if (rt) | |
718 | rt->dst.obsolete = DST_OBSOLETE_KILL; | |
719 | ||
387aa65a TT |
720 | for_each_possible_cpu(i) { |
721 | struct rtable __rcu **prt; | |
722 | prt = per_cpu_ptr(nh->nh_pcpu_rth_output, i); | |
723 | rt = rcu_dereference(*prt); | |
724 | if (rt) | |
725 | rt->dst.obsolete = DST_OBSOLETE_KILL; | |
726 | } | |
4895c771 | 727 | } |
4895c771 | 728 | |
4895c771 | 729 | fnhe->fnhe_stamp = jiffies; |
aee06da6 JA |
730 | |
731 | out_unlock: | |
c5038a83 | 732 | spin_unlock_bh(&fnhe_lock); |
4895c771 DM |
733 | } |
734 | ||
ceb33206 DM |
735 | static void __ip_do_redirect(struct rtable *rt, struct sk_buff *skb, struct flowi4 *fl4, |
736 | bool kill_route) | |
1da177e4 | 737 | { |
e47a185b | 738 | __be32 new_gw = icmp_hdr(skb)->un.gateway; |
94206125 | 739 | __be32 old_gw = ip_hdr(skb)->saddr; |
e47a185b | 740 | struct net_device *dev = skb->dev; |
e47a185b | 741 | struct in_device *in_dev; |
4895c771 | 742 | struct fib_result res; |
e47a185b | 743 | struct neighbour *n; |
317805b8 | 744 | struct net *net; |
1da177e4 | 745 | |
94206125 DM |
746 | switch (icmp_hdr(skb)->code & 7) { |
747 | case ICMP_REDIR_NET: | |
748 | case ICMP_REDIR_NETTOS: | |
749 | case ICMP_REDIR_HOST: | |
750 | case ICMP_REDIR_HOSTTOS: | |
751 | break; | |
752 | ||
753 | default: | |
754 | return; | |
755 | } | |
756 | ||
e47a185b DM |
757 | if (rt->rt_gateway != old_gw) |
758 | return; | |
759 | ||
760 | in_dev = __in_dev_get_rcu(dev); | |
761 | if (!in_dev) | |
762 | return; | |
763 | ||
c346dca1 | 764 | net = dev_net(dev); |
9d4fb27d JP |
765 | if (new_gw == old_gw || !IN_DEV_RX_REDIRECTS(in_dev) || |
766 | ipv4_is_multicast(new_gw) || ipv4_is_lbcast(new_gw) || | |
767 | ipv4_is_zeronet(new_gw)) | |
1da177e4 LT |
768 | goto reject_redirect; |
769 | ||
770 | if (!IN_DEV_SHARED_MEDIA(in_dev)) { | |
771 | if (!inet_addr_onlink(in_dev, new_gw, old_gw)) | |
772 | goto reject_redirect; | |
773 | if (IN_DEV_SEC_REDIRECTS(in_dev) && ip_fib_check_default(new_gw, dev)) | |
774 | goto reject_redirect; | |
775 | } else { | |
317805b8 | 776 | if (inet_addr_type(net, new_gw) != RTN_UNICAST) |
1da177e4 LT |
777 | goto reject_redirect; |
778 | } | |
779 | ||
969447f2 SSL |
780 | n = __ipv4_neigh_lookup(rt->dst.dev, new_gw); |
781 | if (!n) | |
782 | n = neigh_create(&arp_tbl, &new_gw, rt->dst.dev); | |
2c1a4311 | 783 | if (!IS_ERR(n)) { |
e47a185b DM |
784 | if (!(n->nud_state & NUD_VALID)) { |
785 | neigh_event_send(n, NULL); | |
786 | } else { | |
0eeb075f | 787 | if (fib_lookup(net, fl4, &res, 0) == 0) { |
4895c771 | 788 | struct fib_nh *nh = &FIB_RES_NH(res); |
4895c771 | 789 | |
aee06da6 | 790 | update_or_create_fnhe(nh, fl4->daddr, new_gw, |
deed49df | 791 | 0, jiffies + ip_rt_gc_timeout); |
4895c771 | 792 | } |
ceb33206 DM |
793 | if (kill_route) |
794 | rt->dst.obsolete = DST_OBSOLETE_KILL; | |
e47a185b DM |
795 | call_netevent_notifiers(NETEVENT_NEIGH_UPDATE, n); |
796 | } | |
797 | neigh_release(n); | |
798 | } | |
799 | return; | |
800 | ||
801 | reject_redirect: | |
802 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
99ee038d DM |
803 | if (IN_DEV_LOG_MARTIANS(in_dev)) { |
804 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
805 | __be32 daddr = iph->daddr; | |
806 | __be32 saddr = iph->saddr; | |
807 | ||
e47a185b DM |
808 | net_info_ratelimited("Redirect from %pI4 on %s about %pI4 ignored\n" |
809 | " Advised path = %pI4 -> %pI4\n", | |
810 | &old_gw, dev->name, &new_gw, | |
811 | &saddr, &daddr); | |
99ee038d | 812 | } |
e47a185b DM |
813 | #endif |
814 | ; | |
815 | } | |
816 | ||
4895c771 DM |
817 | static void ip_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb) |
818 | { | |
819 | struct rtable *rt; | |
820 | struct flowi4 fl4; | |
f96ef988 | 821 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
7d995694 | 822 | struct net *net = dev_net(skb->dev); |
f96ef988 MK |
823 | int oif = skb->dev->ifindex; |
824 | u8 tos = RT_TOS(iph->tos); | |
825 | u8 prot = iph->protocol; | |
826 | u32 mark = skb->mark; | |
4895c771 DM |
827 | |
828 | rt = (struct rtable *) dst; | |
829 | ||
7d995694 | 830 | __build_flow_key(net, &fl4, sk, iph, oif, tos, prot, mark, 0); |
ceb33206 | 831 | __ip_do_redirect(rt, skb, &fl4, true); |
4895c771 DM |
832 | } |
833 | ||
1da177e4 LT |
834 | static struct dst_entry *ipv4_negative_advice(struct dst_entry *dst) |
835 | { | |
ee6b9673 | 836 | struct rtable *rt = (struct rtable *)dst; |
1da177e4 LT |
837 | struct dst_entry *ret = dst; |
838 | ||
839 | if (rt) { | |
d11a4dc1 | 840 | if (dst->obsolete > 0) { |
1da177e4 LT |
841 | ip_rt_put(rt); |
842 | ret = NULL; | |
5943634f DM |
843 | } else if ((rt->rt_flags & RTCF_REDIRECTED) || |
844 | rt->dst.expires) { | |
89aef892 | 845 | ip_rt_put(rt); |
1da177e4 LT |
846 | ret = NULL; |
847 | } | |
848 | } | |
849 | return ret; | |
850 | } | |
851 | ||
852 | /* | |
853 | * Algorithm: | |
854 | * 1. The first ip_rt_redirect_number redirects are sent | |
855 | * with exponential backoff, then we stop sending them at all, | |
856 | * assuming that the host ignores our redirects. | |
857 | * 2. If we did not see packets requiring redirects | |
858 | * during ip_rt_redirect_silence, we assume that the host | |
859 | * forgot redirected route and start to send redirects again. | |
860 | * | |
861 | * This algorithm is much cheaper and more intelligent than dumb load limiting | |
862 | * in icmp.c. | |
863 | * | |
864 | * NOTE. Do not forget to inhibit load limiting for redirects (redundant) | |
865 | * and "frag. need" (breaks PMTU discovery) in icmp.c. | |
866 | */ | |
867 | ||
868 | void ip_rt_send_redirect(struct sk_buff *skb) | |
869 | { | |
511c3f92 | 870 | struct rtable *rt = skb_rtable(skb); |
30038fc6 | 871 | struct in_device *in_dev; |
92d86829 | 872 | struct inet_peer *peer; |
1d861aa4 | 873 | struct net *net; |
30038fc6 | 874 | int log_martians; |
192132b9 | 875 | int vif; |
1da177e4 | 876 | |
30038fc6 | 877 | rcu_read_lock(); |
d8d1f30b | 878 | in_dev = __in_dev_get_rcu(rt->dst.dev); |
30038fc6 ED |
879 | if (!in_dev || !IN_DEV_TX_REDIRECTS(in_dev)) { |
880 | rcu_read_unlock(); | |
1da177e4 | 881 | return; |
30038fc6 ED |
882 | } |
883 | log_martians = IN_DEV_LOG_MARTIANS(in_dev); | |
385add90 | 884 | vif = l3mdev_master_ifindex_rcu(rt->dst.dev); |
30038fc6 | 885 | rcu_read_unlock(); |
1da177e4 | 886 | |
1d861aa4 | 887 | net = dev_net(rt->dst.dev); |
192132b9 | 888 | peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, vif, 1); |
92d86829 | 889 | if (!peer) { |
e81da0e1 JA |
890 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, |
891 | rt_nexthop(rt, ip_hdr(skb)->daddr)); | |
92d86829 DM |
892 | return; |
893 | } | |
894 | ||
1da177e4 LT |
895 | /* No redirected packets during ip_rt_redirect_silence; |
896 | * reset the algorithm. | |
897 | */ | |
92d86829 DM |
898 | if (time_after(jiffies, peer->rate_last + ip_rt_redirect_silence)) |
899 | peer->rate_tokens = 0; | |
1da177e4 LT |
900 | |
901 | /* Too many ignored redirects; do not send anything | |
d8d1f30b | 902 | * set dst.rate_last to the last seen redirected packet. |
1da177e4 | 903 | */ |
92d86829 DM |
904 | if (peer->rate_tokens >= ip_rt_redirect_number) { |
905 | peer->rate_last = jiffies; | |
1d861aa4 | 906 | goto out_put_peer; |
1da177e4 LT |
907 | } |
908 | ||
909 | /* Check for load limit; set rate_last to the latest sent | |
910 | * redirect. | |
911 | */ | |
92d86829 | 912 | if (peer->rate_tokens == 0 || |
14fb8a76 | 913 | time_after(jiffies, |
92d86829 DM |
914 | (peer->rate_last + |
915 | (ip_rt_redirect_load << peer->rate_tokens)))) { | |
e81da0e1 JA |
916 | __be32 gw = rt_nexthop(rt, ip_hdr(skb)->daddr); |
917 | ||
918 | icmp_send(skb, ICMP_REDIRECT, ICMP_REDIR_HOST, gw); | |
92d86829 DM |
919 | peer->rate_last = jiffies; |
920 | ++peer->rate_tokens; | |
1da177e4 | 921 | #ifdef CONFIG_IP_ROUTE_VERBOSE |
30038fc6 | 922 | if (log_martians && |
e87cc472 JP |
923 | peer->rate_tokens == ip_rt_redirect_number) |
924 | net_warn_ratelimited("host %pI4/if%d ignores redirects for %pI4 to %pI4\n", | |
92101b3b | 925 | &ip_hdr(skb)->saddr, inet_iif(skb), |
e81da0e1 | 926 | &ip_hdr(skb)->daddr, &gw); |
1da177e4 LT |
927 | #endif |
928 | } | |
1d861aa4 DM |
929 | out_put_peer: |
930 | inet_putpeer(peer); | |
1da177e4 LT |
931 | } |
932 | ||
933 | static int ip_error(struct sk_buff *skb) | |
934 | { | |
251da413 | 935 | struct in_device *in_dev = __in_dev_get_rcu(skb->dev); |
511c3f92 | 936 | struct rtable *rt = skb_rtable(skb); |
92d86829 | 937 | struct inet_peer *peer; |
1da177e4 | 938 | unsigned long now; |
251da413 | 939 | struct net *net; |
92d86829 | 940 | bool send; |
1da177e4 LT |
941 | int code; |
942 | ||
381c759d EB |
943 | /* IP on this device is disabled. */ |
944 | if (!in_dev) | |
945 | goto out; | |
946 | ||
251da413 DM |
947 | net = dev_net(rt->dst.dev); |
948 | if (!IN_DEV_FORWARD(in_dev)) { | |
949 | switch (rt->dst.error) { | |
950 | case EHOSTUNREACH: | |
b45386ef | 951 | __IP_INC_STATS(net, IPSTATS_MIB_INADDRERRORS); |
251da413 DM |
952 | break; |
953 | ||
954 | case ENETUNREACH: | |
b45386ef | 955 | __IP_INC_STATS(net, IPSTATS_MIB_INNOROUTES); |
251da413 DM |
956 | break; |
957 | } | |
958 | goto out; | |
959 | } | |
960 | ||
d8d1f30b | 961 | switch (rt->dst.error) { |
4500ebf8 JP |
962 | case EINVAL: |
963 | default: | |
964 | goto out; | |
965 | case EHOSTUNREACH: | |
966 | code = ICMP_HOST_UNREACH; | |
967 | break; | |
968 | case ENETUNREACH: | |
969 | code = ICMP_NET_UNREACH; | |
b45386ef | 970 | __IP_INC_STATS(net, IPSTATS_MIB_INNOROUTES); |
4500ebf8 JP |
971 | break; |
972 | case EACCES: | |
973 | code = ICMP_PKT_FILTERED; | |
974 | break; | |
1da177e4 LT |
975 | } |
976 | ||
192132b9 | 977 | peer = inet_getpeer_v4(net->ipv4.peers, ip_hdr(skb)->saddr, |
385add90 | 978 | l3mdev_master_ifindex(skb->dev), 1); |
92d86829 DM |
979 | |
980 | send = true; | |
981 | if (peer) { | |
982 | now = jiffies; | |
983 | peer->rate_tokens += now - peer->rate_last; | |
984 | if (peer->rate_tokens > ip_rt_error_burst) | |
985 | peer->rate_tokens = ip_rt_error_burst; | |
986 | peer->rate_last = now; | |
987 | if (peer->rate_tokens >= ip_rt_error_cost) | |
988 | peer->rate_tokens -= ip_rt_error_cost; | |
989 | else | |
990 | send = false; | |
1d861aa4 | 991 | inet_putpeer(peer); |
1da177e4 | 992 | } |
92d86829 DM |
993 | if (send) |
994 | icmp_send(skb, ICMP_DEST_UNREACH, code, 0); | |
1da177e4 LT |
995 | |
996 | out: kfree_skb(skb); | |
997 | return 0; | |
e905a9ed | 998 | } |
1da177e4 | 999 | |
d851c12b | 1000 | static void __ip_rt_update_pmtu(struct rtable *rt, struct flowi4 *fl4, u32 mtu) |
1da177e4 | 1001 | { |
d851c12b | 1002 | struct dst_entry *dst = &rt->dst; |
4895c771 | 1003 | struct fib_result res; |
2c8cec5c | 1004 | |
fa1e492a SK |
1005 | if (dst_metric_locked(dst, RTAX_MTU)) |
1006 | return; | |
1007 | ||
cb6ccf09 | 1008 | if (ipv4_mtu(dst) < mtu) |
3cdaa5be LW |
1009 | return; |
1010 | ||
5943634f DM |
1011 | if (mtu < ip_rt_min_pmtu) |
1012 | mtu = ip_rt_min_pmtu; | |
2c8cec5c | 1013 | |
f016229e TT |
1014 | if (rt->rt_pmtu == mtu && |
1015 | time_before(jiffies, dst->expires - ip_rt_mtu_expires / 2)) | |
1016 | return; | |
1017 | ||
c5ae7d41 | 1018 | rcu_read_lock(); |
0eeb075f | 1019 | if (fib_lookup(dev_net(dst->dev), fl4, &res, 0) == 0) { |
4895c771 | 1020 | struct fib_nh *nh = &FIB_RES_NH(res); |
4895c771 | 1021 | |
aee06da6 JA |
1022 | update_or_create_fnhe(nh, fl4->daddr, 0, mtu, |
1023 | jiffies + ip_rt_mtu_expires); | |
4895c771 | 1024 | } |
c5ae7d41 | 1025 | rcu_read_unlock(); |
1da177e4 LT |
1026 | } |
1027 | ||
4895c771 DM |
1028 | static void ip_rt_update_pmtu(struct dst_entry *dst, struct sock *sk, |
1029 | struct sk_buff *skb, u32 mtu) | |
1030 | { | |
1031 | struct rtable *rt = (struct rtable *) dst; | |
1032 | struct flowi4 fl4; | |
1033 | ||
1034 | ip_rt_build_flow_key(&fl4, sk, skb); | |
d851c12b | 1035 | __ip_rt_update_pmtu(rt, &fl4, mtu); |
4895c771 DM |
1036 | } |
1037 | ||
36393395 DM |
1038 | void ipv4_update_pmtu(struct sk_buff *skb, struct net *net, u32 mtu, |
1039 | int oif, u32 mark, u8 protocol, int flow_flags) | |
1040 | { | |
4895c771 | 1041 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
36393395 DM |
1042 | struct flowi4 fl4; |
1043 | struct rtable *rt; | |
1044 | ||
1b3c61dc LC |
1045 | if (!mark) |
1046 | mark = IP4_REPLY_MARK(net, skb->mark); | |
1047 | ||
e2d118a1 | 1048 | __build_flow_key(net, &fl4, NULL, iph, oif, |
4895c771 | 1049 | RT_TOS(iph->tos), protocol, mark, flow_flags); |
36393395 DM |
1050 | rt = __ip_route_output_key(net, &fl4); |
1051 | if (!IS_ERR(rt)) { | |
4895c771 | 1052 | __ip_rt_update_pmtu(rt, &fl4, mtu); |
36393395 DM |
1053 | ip_rt_put(rt); |
1054 | } | |
1055 | } | |
1056 | EXPORT_SYMBOL_GPL(ipv4_update_pmtu); | |
1057 | ||
9cb3a50c | 1058 | static void __ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu) |
36393395 | 1059 | { |
4895c771 DM |
1060 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
1061 | struct flowi4 fl4; | |
1062 | struct rtable *rt; | |
36393395 | 1063 | |
e2d118a1 | 1064 | __build_flow_key(sock_net(sk), &fl4, sk, iph, 0, 0, 0, 0, 0); |
1b3c61dc LC |
1065 | |
1066 | if (!fl4.flowi4_mark) | |
1067 | fl4.flowi4_mark = IP4_REPLY_MARK(sock_net(sk), skb->mark); | |
1068 | ||
4895c771 DM |
1069 | rt = __ip_route_output_key(sock_net(sk), &fl4); |
1070 | if (!IS_ERR(rt)) { | |
1071 | __ip_rt_update_pmtu(rt, &fl4, mtu); | |
1072 | ip_rt_put(rt); | |
1073 | } | |
36393395 | 1074 | } |
9cb3a50c SK |
1075 | |
1076 | void ipv4_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, u32 mtu) | |
1077 | { | |
1078 | const struct iphdr *iph = (const struct iphdr *) skb->data; | |
1079 | struct flowi4 fl4; | |
1080 | struct rtable *rt; | |
7f502361 | 1081 | struct dst_entry *odst = NULL; |
b44108db | 1082 | bool new = false; |
e2d118a1 | 1083 | struct net *net = sock_net(sk); |
9cb3a50c SK |
1084 | |
1085 | bh_lock_sock(sk); | |
482fc609 HFS |
1086 | |
1087 | if (!ip_sk_accept_pmtu(sk)) | |
1088 | goto out; | |
1089 | ||
7f502361 | 1090 | odst = sk_dst_get(sk); |
9cb3a50c | 1091 | |
7f502361 | 1092 | if (sock_owned_by_user(sk) || !odst) { |
9cb3a50c SK |
1093 | __ipv4_sk_update_pmtu(skb, sk, mtu); |
1094 | goto out; | |
1095 | } | |
1096 | ||
e2d118a1 | 1097 | __build_flow_key(net, &fl4, sk, iph, 0, 0, 0, 0, 0); |
9cb3a50c | 1098 | |
7f502361 | 1099 | rt = (struct rtable *)odst; |
51456b29 | 1100 | if (odst->obsolete && !odst->ops->check(odst, 0)) { |
9cb3a50c SK |
1101 | rt = ip_route_output_flow(sock_net(sk), &fl4, sk); |
1102 | if (IS_ERR(rt)) | |
1103 | goto out; | |
b44108db SK |
1104 | |
1105 | new = true; | |
9cb3a50c SK |
1106 | } |
1107 | ||
1108 | __ip_rt_update_pmtu((struct rtable *) rt->dst.path, &fl4, mtu); | |
1109 | ||
7f502361 | 1110 | if (!dst_check(&rt->dst, 0)) { |
b44108db SK |
1111 | if (new) |
1112 | dst_release(&rt->dst); | |
1113 | ||
9cb3a50c SK |
1114 | rt = ip_route_output_flow(sock_net(sk), &fl4, sk); |
1115 | if (IS_ERR(rt)) | |
1116 | goto out; | |
1117 | ||
b44108db | 1118 | new = true; |
9cb3a50c SK |
1119 | } |
1120 | ||
b44108db | 1121 | if (new) |
7f502361 | 1122 | sk_dst_set(sk, &rt->dst); |
9cb3a50c SK |
1123 | |
1124 | out: | |
1125 | bh_unlock_sock(sk); | |
7f502361 | 1126 | dst_release(odst); |
9cb3a50c | 1127 | } |
36393395 | 1128 | EXPORT_SYMBOL_GPL(ipv4_sk_update_pmtu); |
f39925db | 1129 | |
b42597e2 DM |
1130 | void ipv4_redirect(struct sk_buff *skb, struct net *net, |
1131 | int oif, u32 mark, u8 protocol, int flow_flags) | |
1132 | { | |
4895c771 | 1133 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
b42597e2 DM |
1134 | struct flowi4 fl4; |
1135 | struct rtable *rt; | |
1136 | ||
e2d118a1 | 1137 | __build_flow_key(net, &fl4, NULL, iph, oif, |
4895c771 | 1138 | RT_TOS(iph->tos), protocol, mark, flow_flags); |
b42597e2 DM |
1139 | rt = __ip_route_output_key(net, &fl4); |
1140 | if (!IS_ERR(rt)) { | |
ceb33206 | 1141 | __ip_do_redirect(rt, skb, &fl4, false); |
b42597e2 DM |
1142 | ip_rt_put(rt); |
1143 | } | |
1144 | } | |
1145 | EXPORT_SYMBOL_GPL(ipv4_redirect); | |
1146 | ||
1147 | void ipv4_sk_redirect(struct sk_buff *skb, struct sock *sk) | |
1148 | { | |
4895c771 DM |
1149 | const struct iphdr *iph = (const struct iphdr *) skb->data; |
1150 | struct flowi4 fl4; | |
1151 | struct rtable *rt; | |
e2d118a1 | 1152 | struct net *net = sock_net(sk); |
b42597e2 | 1153 | |
e2d118a1 LC |
1154 | __build_flow_key(net, &fl4, sk, iph, 0, 0, 0, 0, 0); |
1155 | rt = __ip_route_output_key(net, &fl4); | |
4895c771 | 1156 | if (!IS_ERR(rt)) { |
ceb33206 | 1157 | __ip_do_redirect(rt, skb, &fl4, false); |
4895c771 DM |
1158 | ip_rt_put(rt); |
1159 | } | |
b42597e2 DM |
1160 | } |
1161 | EXPORT_SYMBOL_GPL(ipv4_sk_redirect); | |
1162 | ||
efbc368d DM |
1163 | static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie) |
1164 | { | |
1165 | struct rtable *rt = (struct rtable *) dst; | |
1166 | ||
ceb33206 DM |
1167 | /* All IPV4 dsts are created with ->obsolete set to the value |
1168 | * DST_OBSOLETE_FORCE_CHK which forces validation calls down | |
1169 | * into this function always. | |
1170 | * | |
387aa65a TT |
1171 | * When a PMTU/redirect information update invalidates a route, |
1172 | * this is indicated by setting obsolete to DST_OBSOLETE_KILL or | |
1173 | * DST_OBSOLETE_DEAD by dst_free(). | |
ceb33206 | 1174 | */ |
387aa65a | 1175 | if (dst->obsolete != DST_OBSOLETE_FORCE_CHK || rt_is_expired(rt)) |
efbc368d | 1176 | return NULL; |
d11a4dc1 | 1177 | return dst; |
1da177e4 LT |
1178 | } |
1179 | ||
1da177e4 LT |
1180 | static void ipv4_link_failure(struct sk_buff *skb) |
1181 | { | |
1182 | struct rtable *rt; | |
1183 | ||
1184 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0); | |
1185 | ||
511c3f92 | 1186 | rt = skb_rtable(skb); |
5943634f DM |
1187 | if (rt) |
1188 | dst_set_expires(&rt->dst, 0); | |
1da177e4 LT |
1189 | } |
1190 | ||
ede2059d | 1191 | static int ip_rt_bug(struct net *net, struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1192 | { |
91df42be JP |
1193 | pr_debug("%s: %pI4 -> %pI4, %s\n", |
1194 | __func__, &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
1195 | skb->dev ? skb->dev->name : "?"); | |
1da177e4 | 1196 | kfree_skb(skb); |
c378a9c0 | 1197 | WARN_ON(1); |
1da177e4 LT |
1198 | return 0; |
1199 | } | |
1200 | ||
1201 | /* | |
1202 | We do not cache source address of outgoing interface, | |
1203 | because it is used only by IP RR, TS and SRR options, | |
1204 | so that it out of fast path. | |
1205 | ||
1206 | BTW remember: "addr" is allowed to be not aligned | |
1207 | in IP options! | |
1208 | */ | |
1209 | ||
8e36360a | 1210 | void ip_rt_get_source(u8 *addr, struct sk_buff *skb, struct rtable *rt) |
1da177e4 | 1211 | { |
a61ced5d | 1212 | __be32 src; |
1da177e4 | 1213 | |
c7537967 | 1214 | if (rt_is_output_route(rt)) |
c5be24ff | 1215 | src = ip_hdr(skb)->saddr; |
ebc0ffae | 1216 | else { |
8e36360a DM |
1217 | struct fib_result res; |
1218 | struct flowi4 fl4; | |
1219 | struct iphdr *iph; | |
1220 | ||
1221 | iph = ip_hdr(skb); | |
1222 | ||
1223 | memset(&fl4, 0, sizeof(fl4)); | |
1224 | fl4.daddr = iph->daddr; | |
1225 | fl4.saddr = iph->saddr; | |
b0fe4a31 | 1226 | fl4.flowi4_tos = RT_TOS(iph->tos); |
8e36360a DM |
1227 | fl4.flowi4_oif = rt->dst.dev->ifindex; |
1228 | fl4.flowi4_iif = skb->dev->ifindex; | |
1229 | fl4.flowi4_mark = skb->mark; | |
5e2b61f7 | 1230 | |
ebc0ffae | 1231 | rcu_read_lock(); |
0eeb075f | 1232 | if (fib_lookup(dev_net(rt->dst.dev), &fl4, &res, 0) == 0) |
436c3b66 | 1233 | src = FIB_RES_PREFSRC(dev_net(rt->dst.dev), res); |
ebc0ffae | 1234 | else |
f8126f1d DM |
1235 | src = inet_select_addr(rt->dst.dev, |
1236 | rt_nexthop(rt, iph->daddr), | |
1237 | RT_SCOPE_UNIVERSE); | |
ebc0ffae ED |
1238 | rcu_read_unlock(); |
1239 | } | |
1da177e4 LT |
1240 | memcpy(addr, &src, 4); |
1241 | } | |
1242 | ||
c7066f70 | 1243 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 LT |
1244 | static void set_class_tag(struct rtable *rt, u32 tag) |
1245 | { | |
d8d1f30b CG |
1246 | if (!(rt->dst.tclassid & 0xFFFF)) |
1247 | rt->dst.tclassid |= tag & 0xFFFF; | |
1248 | if (!(rt->dst.tclassid & 0xFFFF0000)) | |
1249 | rt->dst.tclassid |= tag & 0xFFFF0000; | |
1da177e4 LT |
1250 | } |
1251 | #endif | |
1252 | ||
0dbaee3b DM |
1253 | static unsigned int ipv4_default_advmss(const struct dst_entry *dst) |
1254 | { | |
7ed14d97 GF |
1255 | unsigned int header_size = sizeof(struct tcphdr) + sizeof(struct iphdr); |
1256 | unsigned int advmss = max_t(unsigned int, dst->dev->mtu - header_size, | |
1257 | ip_rt_min_advmss); | |
0dbaee3b | 1258 | |
7ed14d97 | 1259 | return min(advmss, IPV4_MAX_PMTU - header_size); |
0dbaee3b DM |
1260 | } |
1261 | ||
ebb762f2 | 1262 | static unsigned int ipv4_mtu(const struct dst_entry *dst) |
d33e4553 | 1263 | { |
261663b0 | 1264 | const struct rtable *rt = (const struct rtable *) dst; |
5943634f DM |
1265 | unsigned int mtu = rt->rt_pmtu; |
1266 | ||
98d75c37 | 1267 | if (!mtu || time_after_eq(jiffies, rt->dst.expires)) |
5943634f | 1268 | mtu = dst_metric_raw(dst, RTAX_MTU); |
618f9bc7 | 1269 | |
38d523e2 | 1270 | if (mtu) |
618f9bc7 SK |
1271 | return mtu; |
1272 | ||
1273 | mtu = dst->dev->mtu; | |
d33e4553 DM |
1274 | |
1275 | if (unlikely(dst_metric_locked(dst, RTAX_MTU))) { | |
155e8336 | 1276 | if (rt->rt_uses_gateway && mtu > 576) |
d33e4553 DM |
1277 | mtu = 576; |
1278 | } | |
1279 | ||
14972cbd RP |
1280 | mtu = min_t(unsigned int, mtu, IP_MAX_MTU); |
1281 | ||
1282 | return mtu - lwtunnel_headroom(dst->lwtstate, mtu); | |
d33e4553 DM |
1283 | } |
1284 | ||
f2bb4bed | 1285 | static struct fib_nh_exception *find_exception(struct fib_nh *nh, __be32 daddr) |
4895c771 | 1286 | { |
caa41527 | 1287 | struct fnhe_hash_bucket *hash = rcu_dereference(nh->nh_exceptions); |
4895c771 DM |
1288 | struct fib_nh_exception *fnhe; |
1289 | u32 hval; | |
1290 | ||
f2bb4bed DM |
1291 | if (!hash) |
1292 | return NULL; | |
1293 | ||
d3a25c98 | 1294 | hval = fnhe_hashfun(daddr); |
4895c771 DM |
1295 | |
1296 | for (fnhe = rcu_dereference(hash[hval].chain); fnhe; | |
1297 | fnhe = rcu_dereference(fnhe->fnhe_next)) { | |
f2bb4bed DM |
1298 | if (fnhe->fnhe_daddr == daddr) |
1299 | return fnhe; | |
1300 | } | |
1301 | return NULL; | |
1302 | } | |
aee06da6 | 1303 | |
caacf05e | 1304 | static bool rt_bind_exception(struct rtable *rt, struct fib_nh_exception *fnhe, |
f2bb4bed DM |
1305 | __be32 daddr) |
1306 | { | |
caacf05e DM |
1307 | bool ret = false; |
1308 | ||
c5038a83 | 1309 | spin_lock_bh(&fnhe_lock); |
f2bb4bed | 1310 | |
c5038a83 | 1311 | if (daddr == fnhe->fnhe_daddr) { |
2ffae99d TT |
1312 | struct rtable __rcu **porig; |
1313 | struct rtable *orig; | |
5aad1de5 | 1314 | int genid = fnhe_genid(dev_net(rt->dst.dev)); |
2ffae99d TT |
1315 | |
1316 | if (rt_is_input_route(rt)) | |
1317 | porig = &fnhe->fnhe_rth_input; | |
1318 | else | |
1319 | porig = &fnhe->fnhe_rth_output; | |
1320 | orig = rcu_dereference(*porig); | |
5aad1de5 TT |
1321 | |
1322 | if (fnhe->fnhe_genid != genid) { | |
1323 | fnhe->fnhe_genid = genid; | |
13d82bf5 SK |
1324 | fnhe->fnhe_gw = 0; |
1325 | fnhe->fnhe_pmtu = 0; | |
1326 | fnhe->fnhe_expires = 0; | |
2ffae99d TT |
1327 | fnhe_flush_routes(fnhe); |
1328 | orig = NULL; | |
13d82bf5 | 1329 | } |
387aa65a TT |
1330 | fill_route_from_fnhe(rt, fnhe); |
1331 | if (!rt->rt_gateway) | |
155e8336 | 1332 | rt->rt_gateway = daddr; |
f2bb4bed | 1333 | |
2ffae99d TT |
1334 | if (!(rt->dst.flags & DST_NOCACHE)) { |
1335 | rcu_assign_pointer(*porig, rt); | |
1336 | if (orig) | |
1337 | rt_free(orig); | |
1338 | ret = true; | |
1339 | } | |
c5038a83 DM |
1340 | |
1341 | fnhe->fnhe_stamp = jiffies; | |
c5038a83 DM |
1342 | } |
1343 | spin_unlock_bh(&fnhe_lock); | |
caacf05e DM |
1344 | |
1345 | return ret; | |
54764bb6 ED |
1346 | } |
1347 | ||
caacf05e | 1348 | static bool rt_cache_route(struct fib_nh *nh, struct rtable *rt) |
f2bb4bed | 1349 | { |
d26b3a7c | 1350 | struct rtable *orig, *prev, **p; |
caacf05e | 1351 | bool ret = true; |
f2bb4bed | 1352 | |
d26b3a7c | 1353 | if (rt_is_input_route(rt)) { |
54764bb6 | 1354 | p = (struct rtable **)&nh->nh_rth_input; |
d26b3a7c | 1355 | } else { |
903ceff7 | 1356 | p = (struct rtable **)raw_cpu_ptr(nh->nh_pcpu_rth_output); |
d26b3a7c | 1357 | } |
f2bb4bed DM |
1358 | orig = *p; |
1359 | ||
1360 | prev = cmpxchg(p, orig, rt); | |
1361 | if (prev == orig) { | |
f2bb4bed | 1362 | if (orig) |
54764bb6 | 1363 | rt_free(orig); |
155e8336 | 1364 | } else |
caacf05e | 1365 | ret = false; |
caacf05e DM |
1366 | |
1367 | return ret; | |
1368 | } | |
1369 | ||
5055c371 ED |
1370 | struct uncached_list { |
1371 | spinlock_t lock; | |
1372 | struct list_head head; | |
1373 | }; | |
1374 | ||
1375 | static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt_uncached_list); | |
caacf05e DM |
1376 | |
1377 | static void rt_add_uncached_list(struct rtable *rt) | |
1378 | { | |
5055c371 ED |
1379 | struct uncached_list *ul = raw_cpu_ptr(&rt_uncached_list); |
1380 | ||
1381 | rt->rt_uncached_list = ul; | |
1382 | ||
1383 | spin_lock_bh(&ul->lock); | |
1384 | list_add_tail(&rt->rt_uncached, &ul->head); | |
1385 | spin_unlock_bh(&ul->lock); | |
caacf05e DM |
1386 | } |
1387 | ||
1388 | static void ipv4_dst_destroy(struct dst_entry *dst) | |
1389 | { | |
3fb07daf | 1390 | struct dst_metrics *p = (struct dst_metrics *)DST_METRICS_PTR(dst); |
caacf05e DM |
1391 | struct rtable *rt = (struct rtable *) dst; |
1392 | ||
3fb07daf ED |
1393 | if (p != &dst_default_metrics && atomic_dec_and_test(&p->refcnt)) |
1394 | kfree(p); | |
1395 | ||
78df76a0 | 1396 | if (!list_empty(&rt->rt_uncached)) { |
5055c371 ED |
1397 | struct uncached_list *ul = rt->rt_uncached_list; |
1398 | ||
1399 | spin_lock_bh(&ul->lock); | |
caacf05e | 1400 | list_del(&rt->rt_uncached); |
5055c371 | 1401 | spin_unlock_bh(&ul->lock); |
caacf05e DM |
1402 | } |
1403 | } | |
1404 | ||
1405 | void rt_flush_dev(struct net_device *dev) | |
1406 | { | |
5055c371 ED |
1407 | struct net *net = dev_net(dev); |
1408 | struct rtable *rt; | |
1409 | int cpu; | |
1410 | ||
1411 | for_each_possible_cpu(cpu) { | |
1412 | struct uncached_list *ul = &per_cpu(rt_uncached_list, cpu); | |
caacf05e | 1413 | |
5055c371 ED |
1414 | spin_lock_bh(&ul->lock); |
1415 | list_for_each_entry(rt, &ul->head, rt_uncached) { | |
caacf05e DM |
1416 | if (rt->dst.dev != dev) |
1417 | continue; | |
1418 | rt->dst.dev = net->loopback_dev; | |
1419 | dev_hold(rt->dst.dev); | |
1420 | dev_put(dev); | |
1421 | } | |
5055c371 | 1422 | spin_unlock_bh(&ul->lock); |
4895c771 DM |
1423 | } |
1424 | } | |
1425 | ||
4331debc | 1426 | static bool rt_cache_valid(const struct rtable *rt) |
d2d68ba9 | 1427 | { |
4331debc ED |
1428 | return rt && |
1429 | rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK && | |
1430 | !rt_is_expired(rt); | |
d2d68ba9 DM |
1431 | } |
1432 | ||
f2bb4bed | 1433 | static void rt_set_nexthop(struct rtable *rt, __be32 daddr, |
5e2b61f7 | 1434 | const struct fib_result *res, |
f2bb4bed | 1435 | struct fib_nh_exception *fnhe, |
982721f3 | 1436 | struct fib_info *fi, u16 type, u32 itag) |
1da177e4 | 1437 | { |
caacf05e DM |
1438 | bool cached = false; |
1439 | ||
1da177e4 | 1440 | if (fi) { |
4895c771 DM |
1441 | struct fib_nh *nh = &FIB_RES_NH(*res); |
1442 | ||
155e8336 | 1443 | if (nh->nh_gw && nh->nh_scope == RT_SCOPE_LINK) { |
4895c771 | 1444 | rt->rt_gateway = nh->nh_gw; |
155e8336 JA |
1445 | rt->rt_uses_gateway = 1; |
1446 | } | |
3fb07daf ED |
1447 | dst_init_metrics(&rt->dst, fi->fib_metrics->metrics, true); |
1448 | if (fi->fib_metrics != &dst_default_metrics) { | |
1449 | rt->dst._metrics |= DST_METRICS_REFCOUNTED; | |
1450 | atomic_inc(&fi->fib_metrics->refcnt); | |
1451 | } | |
c7066f70 | 1452 | #ifdef CONFIG_IP_ROUTE_CLASSID |
f2bb4bed | 1453 | rt->dst.tclassid = nh->nh_tclassid; |
1da177e4 | 1454 | #endif |
61adedf3 | 1455 | rt->dst.lwtstate = lwtstate_get(nh->nh_lwtstate); |
c5038a83 | 1456 | if (unlikely(fnhe)) |
caacf05e | 1457 | cached = rt_bind_exception(rt, fnhe, daddr); |
c5038a83 | 1458 | else if (!(rt->dst.flags & DST_NOCACHE)) |
caacf05e | 1459 | cached = rt_cache_route(nh, rt); |
155e8336 JA |
1460 | if (unlikely(!cached)) { |
1461 | /* Routes we intend to cache in nexthop exception or | |
1462 | * FIB nexthop have the DST_NOCACHE bit clear. | |
1463 | * However, if we are unsuccessful at storing this | |
1464 | * route into the cache we really need to set it. | |
1465 | */ | |
1466 | rt->dst.flags |= DST_NOCACHE; | |
1467 | if (!rt->rt_gateway) | |
1468 | rt->rt_gateway = daddr; | |
1469 | rt_add_uncached_list(rt); | |
1470 | } | |
1471 | } else | |
caacf05e | 1472 | rt_add_uncached_list(rt); |
defb3519 | 1473 | |
c7066f70 | 1474 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1da177e4 | 1475 | #ifdef CONFIG_IP_MULTIPLE_TABLES |
85b91b03 | 1476 | set_class_tag(rt, res->tclassid); |
1da177e4 LT |
1477 | #endif |
1478 | set_class_tag(rt, itag); | |
1479 | #endif | |
1da177e4 LT |
1480 | } |
1481 | ||
9ab179d8 DA |
1482 | struct rtable *rt_dst_alloc(struct net_device *dev, |
1483 | unsigned int flags, u16 type, | |
1484 | bool nopolicy, bool noxfrm, bool will_cache) | |
0c4dcd58 | 1485 | { |
d08c4f35 DA |
1486 | struct rtable *rt; |
1487 | ||
1488 | rt = dst_alloc(&ipv4_dst_ops, dev, 1, DST_OBSOLETE_FORCE_CHK, | |
1489 | (will_cache ? 0 : (DST_HOST | DST_NOCACHE)) | | |
1490 | (nopolicy ? DST_NOPOLICY : 0) | | |
1491 | (noxfrm ? DST_NOXFRM : 0)); | |
1492 | ||
1493 | if (rt) { | |
1494 | rt->rt_genid = rt_genid_ipv4(dev_net(dev)); | |
1495 | rt->rt_flags = flags; | |
1496 | rt->rt_type = type; | |
1497 | rt->rt_is_input = 0; | |
1498 | rt->rt_iif = 0; | |
1499 | rt->rt_pmtu = 0; | |
1500 | rt->rt_gateway = 0; | |
1501 | rt->rt_uses_gateway = 0; | |
b7503e0c | 1502 | rt->rt_table_id = 0; |
d08c4f35 DA |
1503 | INIT_LIST_HEAD(&rt->rt_uncached); |
1504 | ||
1505 | rt->dst.output = ip_output; | |
1506 | if (flags & RTCF_LOCAL) | |
1507 | rt->dst.input = ip_local_deliver; | |
1508 | } | |
1509 | ||
1510 | return rt; | |
0c4dcd58 | 1511 | } |
9ab179d8 | 1512 | EXPORT_SYMBOL(rt_dst_alloc); |
0c4dcd58 | 1513 | |
96d36220 | 1514 | /* called in rcu_read_lock() section */ |
9e12bb22 | 1515 | static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
1da177e4 LT |
1516 | u8 tos, struct net_device *dev, int our) |
1517 | { | |
1da177e4 | 1518 | struct rtable *rth; |
96d36220 | 1519 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
d08c4f35 | 1520 | unsigned int flags = RTCF_MULTICAST; |
1da177e4 | 1521 | u32 itag = 0; |
b5f7e755 | 1522 | int err; |
1da177e4 LT |
1523 | |
1524 | /* Primary sanity checks. */ | |
1525 | ||
51456b29 | 1526 | if (!in_dev) |
1da177e4 LT |
1527 | return -EINVAL; |
1528 | ||
1e637c74 | 1529 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || |
d0daebc3 | 1530 | skb->protocol != htons(ETH_P_IP)) |
1da177e4 LT |
1531 | goto e_inval; |
1532 | ||
75fea73d AD |
1533 | if (ipv4_is_loopback(saddr) && !IN_DEV_ROUTE_LOCALNET(in_dev)) |
1534 | goto e_inval; | |
d0daebc3 | 1535 | |
f97c1e0c JP |
1536 | if (ipv4_is_zeronet(saddr)) { |
1537 | if (!ipv4_is_local_multicast(daddr)) | |
1da177e4 | 1538 | goto e_inval; |
b5f7e755 | 1539 | } else { |
9e56e380 DM |
1540 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, |
1541 | in_dev, &itag); | |
b5f7e755 ED |
1542 | if (err < 0) |
1543 | goto e_err; | |
1544 | } | |
d08c4f35 DA |
1545 | if (our) |
1546 | flags |= RTCF_LOCAL; | |
1547 | ||
1548 | rth = rt_dst_alloc(dev_net(dev)->loopback_dev, flags, RTN_MULTICAST, | |
f2bb4bed | 1549 | IN_DEV_CONF_GET(in_dev, NOPOLICY), false, false); |
1da177e4 LT |
1550 | if (!rth) |
1551 | goto e_nobufs; | |
1552 | ||
cf911662 DM |
1553 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1554 | rth->dst.tclassid = itag; | |
1555 | #endif | |
d8d1f30b | 1556 | rth->dst.output = ip_rt_bug; |
9917e1e8 | 1557 | rth->rt_is_input= 1; |
1da177e4 LT |
1558 | |
1559 | #ifdef CONFIG_IP_MROUTE | |
f97c1e0c | 1560 | if (!ipv4_is_local_multicast(daddr) && IN_DEV_MFORWARD(in_dev)) |
d8d1f30b | 1561 | rth->dst.input = ip_mr_input; |
1da177e4 LT |
1562 | #endif |
1563 | RT_CACHE_STAT_INC(in_slow_mc); | |
1564 | ||
89aef892 DM |
1565 | skb_dst_set(skb, &rth->dst); |
1566 | return 0; | |
1da177e4 LT |
1567 | |
1568 | e_nobufs: | |
1da177e4 | 1569 | return -ENOBUFS; |
1da177e4 | 1570 | e_inval: |
96d36220 | 1571 | return -EINVAL; |
b5f7e755 | 1572 | e_err: |
b5f7e755 | 1573 | return err; |
1da177e4 LT |
1574 | } |
1575 | ||
1576 | ||
1577 | static void ip_handle_martian_source(struct net_device *dev, | |
1578 | struct in_device *in_dev, | |
1579 | struct sk_buff *skb, | |
9e12bb22 AV |
1580 | __be32 daddr, |
1581 | __be32 saddr) | |
1da177e4 LT |
1582 | { |
1583 | RT_CACHE_STAT_INC(in_martian_src); | |
1584 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
1585 | if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit()) { | |
1586 | /* | |
1587 | * RFC1812 recommendation, if source is martian, | |
1588 | * the only hint is MAC header. | |
1589 | */ | |
058bd4d2 | 1590 | pr_warn("martian source %pI4 from %pI4, on dev %s\n", |
673d57e7 | 1591 | &daddr, &saddr, dev->name); |
98e399f8 | 1592 | if (dev->hard_header_len && skb_mac_header_was_set(skb)) { |
058bd4d2 JP |
1593 | print_hex_dump(KERN_WARNING, "ll header: ", |
1594 | DUMP_PREFIX_OFFSET, 16, 1, | |
1595 | skb_mac_header(skb), | |
1596 | dev->hard_header_len, true); | |
1da177e4 LT |
1597 | } |
1598 | } | |
1599 | #endif | |
1600 | } | |
1601 | ||
deed49df XL |
1602 | static void ip_del_fnhe(struct fib_nh *nh, __be32 daddr) |
1603 | { | |
1604 | struct fnhe_hash_bucket *hash; | |
1605 | struct fib_nh_exception *fnhe, __rcu **fnhe_p; | |
1606 | u32 hval = fnhe_hashfun(daddr); | |
1607 | ||
1608 | spin_lock_bh(&fnhe_lock); | |
1609 | ||
1610 | hash = rcu_dereference_protected(nh->nh_exceptions, | |
1611 | lockdep_is_held(&fnhe_lock)); | |
1612 | hash += hval; | |
1613 | ||
1614 | fnhe_p = &hash->chain; | |
1615 | fnhe = rcu_dereference_protected(*fnhe_p, lockdep_is_held(&fnhe_lock)); | |
1616 | while (fnhe) { | |
1617 | if (fnhe->fnhe_daddr == daddr) { | |
1618 | rcu_assign_pointer(*fnhe_p, rcu_dereference_protected( | |
1619 | fnhe->fnhe_next, lockdep_is_held(&fnhe_lock))); | |
1620 | fnhe_flush_routes(fnhe); | |
1621 | kfree_rcu(fnhe, rcu); | |
1622 | break; | |
1623 | } | |
1624 | fnhe_p = &fnhe->fnhe_next; | |
1625 | fnhe = rcu_dereference_protected(fnhe->fnhe_next, | |
1626 | lockdep_is_held(&fnhe_lock)); | |
1627 | } | |
1628 | ||
1629 | spin_unlock_bh(&fnhe_lock); | |
1630 | } | |
1631 | ||
efd85700 TG |
1632 | static void set_lwt_redirect(struct rtable *rth) |
1633 | { | |
1634 | if (lwtunnel_output_redirect(rth->dst.lwtstate)) { | |
1635 | rth->dst.lwtstate->orig_output = rth->dst.output; | |
1636 | rth->dst.output = lwtunnel_output; | |
1637 | } | |
1638 | ||
1639 | if (lwtunnel_input_redirect(rth->dst.lwtstate)) { | |
1640 | rth->dst.lwtstate->orig_input = rth->dst.input; | |
1641 | rth->dst.input = lwtunnel_input; | |
1642 | } | |
1643 | } | |
1644 | ||
47360228 | 1645 | /* called in rcu_read_lock() section */ |
5969f71d | 1646 | static int __mkroute_input(struct sk_buff *skb, |
982721f3 | 1647 | const struct fib_result *res, |
5969f71d | 1648 | struct in_device *in_dev, |
c6cffba4 | 1649 | __be32 daddr, __be32 saddr, u32 tos) |
1da177e4 | 1650 | { |
2ffae99d | 1651 | struct fib_nh_exception *fnhe; |
1da177e4 LT |
1652 | struct rtable *rth; |
1653 | int err; | |
1654 | struct in_device *out_dev; | |
d2d68ba9 | 1655 | bool do_cache; |
fbdc0ad0 | 1656 | u32 itag = 0; |
1da177e4 LT |
1657 | |
1658 | /* get a working reference to the output device */ | |
47360228 | 1659 | out_dev = __in_dev_get_rcu(FIB_RES_DEV(*res)); |
51456b29 | 1660 | if (!out_dev) { |
e87cc472 | 1661 | net_crit_ratelimited("Bug in ip_route_input_slow(). Please report.\n"); |
1da177e4 LT |
1662 | return -EINVAL; |
1663 | } | |
1664 | ||
5c04c819 | 1665 | err = fib_validate_source(skb, saddr, daddr, tos, FIB_RES_OIF(*res), |
9e56e380 | 1666 | in_dev->dev, in_dev, &itag); |
1da177e4 | 1667 | if (err < 0) { |
e905a9ed | 1668 | ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr, |
1da177e4 | 1669 | saddr); |
e905a9ed | 1670 | |
1da177e4 LT |
1671 | goto cleanup; |
1672 | } | |
1673 | ||
e81da0e1 JA |
1674 | do_cache = res->fi && !itag; |
1675 | if (out_dev == in_dev && err && IN_DEV_TX_REDIRECTS(out_dev) && | |
df4d9254 | 1676 | skb->protocol == htons(ETH_P_IP) && |
1da177e4 | 1677 | (IN_DEV_SHARED_MEDIA(out_dev) || |
df4d9254 HFS |
1678 | inet_addr_onlink(out_dev, saddr, FIB_RES_GW(*res)))) |
1679 | IPCB(skb)->flags |= IPSKB_DOREDIRECT; | |
1da177e4 LT |
1680 | |
1681 | if (skb->protocol != htons(ETH_P_IP)) { | |
1682 | /* Not IP (i.e. ARP). Do not create route, if it is | |
1683 | * invalid for proxy arp. DNAT routes are always valid. | |
65324144 JDB |
1684 | * |
1685 | * Proxy arp feature have been extended to allow, ARP | |
1686 | * replies back to the same interface, to support | |
1687 | * Private VLAN switch technologies. See arp.c. | |
1da177e4 | 1688 | */ |
65324144 JDB |
1689 | if (out_dev == in_dev && |
1690 | IN_DEV_PROXY_ARP_PVLAN(in_dev) == 0) { | |
1da177e4 LT |
1691 | err = -EINVAL; |
1692 | goto cleanup; | |
1693 | } | |
1694 | } | |
1695 | ||
2ffae99d | 1696 | fnhe = find_exception(&FIB_RES_NH(*res), daddr); |
e81da0e1 | 1697 | if (do_cache) { |
deed49df | 1698 | if (fnhe) { |
2ffae99d | 1699 | rth = rcu_dereference(fnhe->fnhe_rth_input); |
deed49df XL |
1700 | if (rth && rth->dst.expires && |
1701 | time_after(jiffies, rth->dst.expires)) { | |
1702 | ip_del_fnhe(&FIB_RES_NH(*res), daddr); | |
1703 | fnhe = NULL; | |
1704 | } else { | |
1705 | goto rt_cache; | |
1706 | } | |
1707 | } | |
1708 | ||
1709 | rth = rcu_dereference(FIB_RES_NH(*res).nh_rth_input); | |
2ffae99d | 1710 | |
deed49df | 1711 | rt_cache: |
e81da0e1 JA |
1712 | if (rt_cache_valid(rth)) { |
1713 | skb_dst_set_noref(skb, &rth->dst); | |
1714 | goto out; | |
d2d68ba9 DM |
1715 | } |
1716 | } | |
f2bb4bed | 1717 | |
d08c4f35 | 1718 | rth = rt_dst_alloc(out_dev->dev, 0, res->type, |
5c1e6aa3 | 1719 | IN_DEV_CONF_GET(in_dev, NOPOLICY), |
d2d68ba9 | 1720 | IN_DEV_CONF_GET(out_dev, NOXFRM), do_cache); |
1da177e4 LT |
1721 | if (!rth) { |
1722 | err = -ENOBUFS; | |
1723 | goto cleanup; | |
1724 | } | |
1725 | ||
9917e1e8 | 1726 | rth->rt_is_input = 1; |
b7503e0c DA |
1727 | if (res->table) |
1728 | rth->rt_table_id = res->table->tb_id; | |
a6254864 | 1729 | RT_CACHE_STAT_INC(in_slow_tot); |
1da177e4 | 1730 | |
d8d1f30b | 1731 | rth->dst.input = ip_forward; |
1da177e4 | 1732 | |
2ffae99d | 1733 | rt_set_nexthop(rth, daddr, res, fnhe, res->fi, res->type, itag); |
efd85700 | 1734 | set_lwt_redirect(rth); |
c6cffba4 | 1735 | skb_dst_set(skb, &rth->dst); |
d2d68ba9 | 1736 | out: |
1da177e4 LT |
1737 | err = 0; |
1738 | cleanup: | |
1da177e4 | 1739 | return err; |
e905a9ed | 1740 | } |
1da177e4 | 1741 | |
79a13159 | 1742 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
79a13159 | 1743 | /* To make ICMP packets follow the right flow, the multipath hash is |
bf4e0a3d | 1744 | * calculated from the inner IP addresses. |
79a13159 | 1745 | */ |
bf4e0a3d NA |
1746 | static void ip_multipath_l3_keys(const struct sk_buff *skb, |
1747 | struct flow_keys *hash_keys) | |
79a13159 PN |
1748 | { |
1749 | const struct iphdr *outer_iph = ip_hdr(skb); | |
bf4e0a3d | 1750 | const struct iphdr *inner_iph; |
79a13159 PN |
1751 | const struct icmphdr *icmph; |
1752 | struct iphdr _inner_iph; | |
bf4e0a3d NA |
1753 | struct icmphdr _icmph; |
1754 | ||
1755 | hash_keys->addrs.v4addrs.src = outer_iph->saddr; | |
1756 | hash_keys->addrs.v4addrs.dst = outer_iph->daddr; | |
1757 | if (likely(outer_iph->protocol != IPPROTO_ICMP)) | |
1758 | return; | |
79a13159 PN |
1759 | |
1760 | if (unlikely((outer_iph->frag_off & htons(IP_OFFSET)) != 0)) | |
bf4e0a3d | 1761 | return; |
79a13159 PN |
1762 | |
1763 | icmph = skb_header_pointer(skb, outer_iph->ihl * 4, sizeof(_icmph), | |
1764 | &_icmph); | |
1765 | if (!icmph) | |
bf4e0a3d | 1766 | return; |
79a13159 PN |
1767 | |
1768 | if (icmph->type != ICMP_DEST_UNREACH && | |
1769 | icmph->type != ICMP_REDIRECT && | |
1770 | icmph->type != ICMP_TIME_EXCEEDED && | |
bf4e0a3d NA |
1771 | icmph->type != ICMP_PARAMETERPROB) |
1772 | return; | |
79a13159 PN |
1773 | |
1774 | inner_iph = skb_header_pointer(skb, | |
1775 | outer_iph->ihl * 4 + sizeof(_icmph), | |
1776 | sizeof(_inner_iph), &_inner_iph); | |
1777 | if (!inner_iph) | |
bf4e0a3d NA |
1778 | return; |
1779 | hash_keys->addrs.v4addrs.src = inner_iph->saddr; | |
1780 | hash_keys->addrs.v4addrs.dst = inner_iph->daddr; | |
1781 | } | |
79a13159 | 1782 | |
bf4e0a3d NA |
1783 | /* if skb is set it will be used and fl4 can be NULL */ |
1784 | int fib_multipath_hash(const struct fib_info *fi, const struct flowi4 *fl4, | |
1785 | const struct sk_buff *skb) | |
1786 | { | |
1787 | struct net *net = fi->fib_net; | |
1788 | struct flow_keys hash_keys; | |
1789 | u32 mhash; | |
79a13159 | 1790 | |
bf4e0a3d NA |
1791 | switch (net->ipv4.sysctl_fib_multipath_hash_policy) { |
1792 | case 0: | |
1793 | memset(&hash_keys, 0, sizeof(hash_keys)); | |
1794 | hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; | |
1795 | if (skb) { | |
1796 | ip_multipath_l3_keys(skb, &hash_keys); | |
1797 | } else { | |
1798 | hash_keys.addrs.v4addrs.src = fl4->saddr; | |
1799 | hash_keys.addrs.v4addrs.dst = fl4->daddr; | |
1800 | } | |
1801 | break; | |
1802 | case 1: | |
1803 | /* skb is currently provided only when forwarding */ | |
1804 | if (skb) { | |
1805 | unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP; | |
1806 | struct flow_keys keys; | |
1807 | ||
1808 | /* short-circuit if we already have L4 hash present */ | |
1809 | if (skb->l4_hash) | |
1810 | return skb_get_hash_raw(skb) >> 1; | |
1811 | memset(&hash_keys, 0, sizeof(hash_keys)); | |
1812 | skb_flow_dissect_flow_keys(skb, &keys, flag); | |
1813 | hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src; | |
1814 | hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst; | |
1815 | hash_keys.ports.src = keys.ports.src; | |
1816 | hash_keys.ports.dst = keys.ports.dst; | |
1817 | hash_keys.basic.ip_proto = keys.basic.ip_proto; | |
1818 | } else { | |
1819 | memset(&hash_keys, 0, sizeof(hash_keys)); | |
1820 | hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; | |
1821 | hash_keys.addrs.v4addrs.src = fl4->saddr; | |
1822 | hash_keys.addrs.v4addrs.dst = fl4->daddr; | |
1823 | hash_keys.ports.src = fl4->fl4_sport; | |
1824 | hash_keys.ports.dst = fl4->fl4_dport; | |
1825 | hash_keys.basic.ip_proto = fl4->flowi4_proto; | |
1826 | } | |
1827 | break; | |
1828 | } | |
1829 | mhash = flow_hash_from_keys(&hash_keys); | |
79a13159 | 1830 | |
bf4e0a3d NA |
1831 | return mhash >> 1; |
1832 | } | |
1833 | EXPORT_SYMBOL_GPL(fib_multipath_hash); | |
79a13159 PN |
1834 | #endif /* CONFIG_IP_ROUTE_MULTIPATH */ |
1835 | ||
5969f71d SH |
1836 | static int ip_mkroute_input(struct sk_buff *skb, |
1837 | struct fib_result *res, | |
5969f71d SH |
1838 | struct in_device *in_dev, |
1839 | __be32 daddr, __be32 saddr, u32 tos) | |
1da177e4 | 1840 | { |
1da177e4 | 1841 | #ifdef CONFIG_IP_ROUTE_MULTIPATH |
0e884c78 | 1842 | if (res->fi && res->fi->fib_nhs > 1) { |
bf4e0a3d | 1843 | int h = fib_multipath_hash(res->fi, NULL, skb); |
0e884c78 | 1844 | |
0e884c78 PN |
1845 | fib_select_multipath(res, h); |
1846 | } | |
1da177e4 LT |
1847 | #endif |
1848 | ||
1849 | /* create a routing cache entry */ | |
c6cffba4 | 1850 | return __mkroute_input(skb, res, in_dev, daddr, saddr, tos); |
1da177e4 LT |
1851 | } |
1852 | ||
1da177e4 LT |
1853 | /* |
1854 | * NOTE. We drop all the packets that has local source | |
1855 | * addresses, because every properly looped back packet | |
1856 | * must have correct destination already attached by output routine. | |
1857 | * | |
1858 | * Such approach solves two big problems: | |
1859 | * 1. Not simplex devices are handled properly. | |
1860 | * 2. IP spoofing attempts are filtered with 100% of guarantee. | |
ebc0ffae | 1861 | * called with rcu_read_lock() |
1da177e4 LT |
1862 | */ |
1863 | ||
9e12bb22 | 1864 | static int ip_route_input_slow(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
5510cdf7 DA |
1865 | u8 tos, struct net_device *dev, |
1866 | struct fib_result *res) | |
1da177e4 | 1867 | { |
96d36220 | 1868 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
1b7179d3 | 1869 | struct ip_tunnel_info *tun_info; |
68a5e3dd | 1870 | struct flowi4 fl4; |
95c96174 | 1871 | unsigned int flags = 0; |
1da177e4 | 1872 | u32 itag = 0; |
95c96174 | 1873 | struct rtable *rth; |
1da177e4 | 1874 | int err = -EINVAL; |
5e73ea1a | 1875 | struct net *net = dev_net(dev); |
d2d68ba9 | 1876 | bool do_cache; |
1da177e4 LT |
1877 | |
1878 | /* IP on this device is disabled. */ | |
1879 | ||
1880 | if (!in_dev) | |
1881 | goto out; | |
1882 | ||
1883 | /* Check for the most weird martians, which can be not detected | |
1884 | by fib_lookup. | |
1885 | */ | |
1886 | ||
61adedf3 | 1887 | tun_info = skb_tunnel_info(skb); |
46fa062a | 1888 | if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX)) |
1b7179d3 TG |
1889 | fl4.flowi4_tun_key.tun_id = tun_info->key.tun_id; |
1890 | else | |
1891 | fl4.flowi4_tun_key.tun_id = 0; | |
f38a9eb1 TG |
1892 | skb_dst_drop(skb); |
1893 | ||
d0daebc3 | 1894 | if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr)) |
1da177e4 LT |
1895 | goto martian_source; |
1896 | ||
5510cdf7 DA |
1897 | res->fi = NULL; |
1898 | res->table = NULL; | |
27a954bd | 1899 | if (ipv4_is_lbcast(daddr) || (saddr == 0 && daddr == 0)) |
1da177e4 LT |
1900 | goto brd_input; |
1901 | ||
1902 | /* Accept zero addresses only to limited broadcast; | |
1903 | * I even do not know to fix it or not. Waiting for complains :-) | |
1904 | */ | |
f97c1e0c | 1905 | if (ipv4_is_zeronet(saddr)) |
1da177e4 LT |
1906 | goto martian_source; |
1907 | ||
d0daebc3 | 1908 | if (ipv4_is_zeronet(daddr)) |
1da177e4 LT |
1909 | goto martian_destination; |
1910 | ||
9eb43e76 ED |
1911 | /* Following code try to avoid calling IN_DEV_NET_ROUTE_LOCALNET(), |
1912 | * and call it once if daddr or/and saddr are loopback addresses | |
1913 | */ | |
1914 | if (ipv4_is_loopback(daddr)) { | |
1915 | if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net)) | |
d0daebc3 | 1916 | goto martian_destination; |
9eb43e76 ED |
1917 | } else if (ipv4_is_loopback(saddr)) { |
1918 | if (!IN_DEV_NET_ROUTE_LOCALNET(in_dev, net)) | |
d0daebc3 TG |
1919 | goto martian_source; |
1920 | } | |
1921 | ||
1da177e4 LT |
1922 | /* |
1923 | * Now we are ready to route packet. | |
1924 | */ | |
68a5e3dd | 1925 | fl4.flowi4_oif = 0; |
e0d56fdd | 1926 | fl4.flowi4_iif = dev->ifindex; |
68a5e3dd DM |
1927 | fl4.flowi4_mark = skb->mark; |
1928 | fl4.flowi4_tos = tos; | |
1929 | fl4.flowi4_scope = RT_SCOPE_UNIVERSE; | |
b84f7878 | 1930 | fl4.flowi4_flags = 0; |
68a5e3dd DM |
1931 | fl4.daddr = daddr; |
1932 | fl4.saddr = saddr; | |
8bcfd092 | 1933 | fl4.flowi4_uid = sock_net_uid(net, NULL); |
5510cdf7 | 1934 | err = fib_lookup(net, &fl4, res, 0); |
cd0f0b95 DJ |
1935 | if (err != 0) { |
1936 | if (!IN_DEV_FORWARD(in_dev)) | |
1937 | err = -EHOSTUNREACH; | |
1da177e4 | 1938 | goto no_route; |
cd0f0b95 | 1939 | } |
1da177e4 | 1940 | |
5510cdf7 | 1941 | if (res->type == RTN_BROADCAST) |
1da177e4 LT |
1942 | goto brd_input; |
1943 | ||
5510cdf7 | 1944 | if (res->type == RTN_LOCAL) { |
5c04c819 | 1945 | err = fib_validate_source(skb, saddr, daddr, tos, |
0d5edc68 | 1946 | 0, dev, in_dev, &itag); |
b5f7e755 | 1947 | if (err < 0) |
0d753960 | 1948 | goto martian_source; |
1da177e4 LT |
1949 | goto local_input; |
1950 | } | |
1951 | ||
cd0f0b95 DJ |
1952 | if (!IN_DEV_FORWARD(in_dev)) { |
1953 | err = -EHOSTUNREACH; | |
251da413 | 1954 | goto no_route; |
cd0f0b95 | 1955 | } |
5510cdf7 | 1956 | if (res->type != RTN_UNICAST) |
1da177e4 LT |
1957 | goto martian_destination; |
1958 | ||
5510cdf7 | 1959 | err = ip_mkroute_input(skb, res, in_dev, daddr, saddr, tos); |
1da177e4 LT |
1960 | out: return err; |
1961 | ||
1962 | brd_input: | |
1963 | if (skb->protocol != htons(ETH_P_IP)) | |
1964 | goto e_inval; | |
1965 | ||
41347dcd | 1966 | if (!ipv4_is_zeronet(saddr)) { |
9e56e380 DM |
1967 | err = fib_validate_source(skb, saddr, 0, tos, 0, dev, |
1968 | in_dev, &itag); | |
1da177e4 | 1969 | if (err < 0) |
0d753960 | 1970 | goto martian_source; |
1da177e4 LT |
1971 | } |
1972 | flags |= RTCF_BROADCAST; | |
5510cdf7 | 1973 | res->type = RTN_BROADCAST; |
1da177e4 LT |
1974 | RT_CACHE_STAT_INC(in_brd); |
1975 | ||
1976 | local_input: | |
d2d68ba9 | 1977 | do_cache = false; |
5510cdf7 | 1978 | if (res->fi) { |
fe3edf45 | 1979 | if (!itag) { |
5510cdf7 | 1980 | rth = rcu_dereference(FIB_RES_NH(*res).nh_rth_input); |
d2d68ba9 | 1981 | if (rt_cache_valid(rth)) { |
c6cffba4 DM |
1982 | skb_dst_set_noref(skb, &rth->dst); |
1983 | err = 0; | |
1984 | goto out; | |
d2d68ba9 DM |
1985 | } |
1986 | do_cache = true; | |
1987 | } | |
1988 | } | |
1989 | ||
f5a0aab8 | 1990 | rth = rt_dst_alloc(l3mdev_master_dev_rcu(dev) ? : net->loopback_dev, |
5510cdf7 | 1991 | flags | RTCF_LOCAL, res->type, |
d2d68ba9 | 1992 | IN_DEV_CONF_GET(in_dev, NOPOLICY), false, do_cache); |
1da177e4 LT |
1993 | if (!rth) |
1994 | goto e_nobufs; | |
1995 | ||
d8d1f30b | 1996 | rth->dst.output= ip_rt_bug; |
cf911662 DM |
1997 | #ifdef CONFIG_IP_ROUTE_CLASSID |
1998 | rth->dst.tclassid = itag; | |
1999 | #endif | |
9917e1e8 | 2000 | rth->rt_is_input = 1; |
5510cdf7 DA |
2001 | if (res->table) |
2002 | rth->rt_table_id = res->table->tb_id; | |
571e7226 | 2003 | |
a6254864 | 2004 | RT_CACHE_STAT_INC(in_slow_tot); |
5510cdf7 | 2005 | if (res->type == RTN_UNREACHABLE) { |
d8d1f30b CG |
2006 | rth->dst.input= ip_error; |
2007 | rth->dst.error= -err; | |
1da177e4 LT |
2008 | rth->rt_flags &= ~RTCF_LOCAL; |
2009 | } | |
efd85700 | 2010 | |
dcdfdf56 | 2011 | if (do_cache) { |
5510cdf7 | 2012 | struct fib_nh *nh = &FIB_RES_NH(*res); |
efd85700 TG |
2013 | |
2014 | rth->dst.lwtstate = lwtstate_get(nh->nh_lwtstate); | |
2015 | if (lwtunnel_input_redirect(rth->dst.lwtstate)) { | |
2016 | WARN_ON(rth->dst.input == lwtunnel_input); | |
2017 | rth->dst.lwtstate->orig_input = rth->dst.input; | |
2018 | rth->dst.input = lwtunnel_input; | |
2019 | } | |
2020 | ||
2021 | if (unlikely(!rt_cache_route(nh, rth))) { | |
dcdfdf56 AS |
2022 | rth->dst.flags |= DST_NOCACHE; |
2023 | rt_add_uncached_list(rth); | |
2024 | } | |
2025 | } | |
89aef892 | 2026 | skb_dst_set(skb, &rth->dst); |
b23dd4fe | 2027 | err = 0; |
ebc0ffae | 2028 | goto out; |
1da177e4 LT |
2029 | |
2030 | no_route: | |
2031 | RT_CACHE_STAT_INC(in_no_route); | |
5510cdf7 DA |
2032 | res->type = RTN_UNREACHABLE; |
2033 | res->fi = NULL; | |
2034 | res->table = NULL; | |
1da177e4 LT |
2035 | goto local_input; |
2036 | ||
2037 | /* | |
2038 | * Do not cache martian addresses: they should be logged (RFC1812) | |
2039 | */ | |
2040 | martian_destination: | |
2041 | RT_CACHE_STAT_INC(in_martian_dst); | |
2042 | #ifdef CONFIG_IP_ROUTE_VERBOSE | |
e87cc472 JP |
2043 | if (IN_DEV_LOG_MARTIANS(in_dev)) |
2044 | net_warn_ratelimited("martian destination %pI4 from %pI4, dev %s\n", | |
2045 | &daddr, &saddr, dev->name); | |
1da177e4 | 2046 | #endif |
2c2910a4 | 2047 | |
1da177e4 LT |
2048 | e_inval: |
2049 | err = -EINVAL; | |
ebc0ffae | 2050 | goto out; |
1da177e4 LT |
2051 | |
2052 | e_nobufs: | |
2053 | err = -ENOBUFS; | |
ebc0ffae | 2054 | goto out; |
1da177e4 LT |
2055 | |
2056 | martian_source: | |
2057 | ip_handle_martian_source(dev, in_dev, skb, daddr, saddr); | |
ebc0ffae | 2058 | goto out; |
1da177e4 LT |
2059 | } |
2060 | ||
c6cffba4 DM |
2061 | int ip_route_input_noref(struct sk_buff *skb, __be32 daddr, __be32 saddr, |
2062 | u8 tos, struct net_device *dev) | |
1da177e4 | 2063 | { |
5510cdf7 DA |
2064 | struct fib_result res; |
2065 | int err; | |
1da177e4 | 2066 | |
6e28099d | 2067 | tos &= IPTOS_RT_MASK; |
96d36220 | 2068 | rcu_read_lock(); |
5510cdf7 DA |
2069 | err = ip_route_input_rcu(skb, daddr, saddr, tos, dev, &res); |
2070 | rcu_read_unlock(); | |
96d36220 | 2071 | |
5510cdf7 DA |
2072 | return err; |
2073 | } | |
2074 | EXPORT_SYMBOL(ip_route_input_noref); | |
2075 | ||
2076 | /* called with rcu_read_lock held */ | |
2077 | int ip_route_input_rcu(struct sk_buff *skb, __be32 daddr, __be32 saddr, | |
2078 | u8 tos, struct net_device *dev, struct fib_result *res) | |
2079 | { | |
1da177e4 LT |
2080 | /* Multicast recognition logic is moved from route cache to here. |
2081 | The problem was that too many Ethernet cards have broken/missing | |
2082 | hardware multicast filters :-( As result the host on multicasting | |
2083 | network acquires a lot of useless route cache entries, sort of | |
2084 | SDR messages from all the world. Now we try to get rid of them. | |
2085 | Really, provided software IP multicast filter is organized | |
2086 | reasonably (at least, hashed), it does not result in a slowdown | |
2087 | comparing with route cache reject entries. | |
2088 | Note, that multicast routers are not affected, because | |
2089 | route cache entry is created eventually. | |
2090 | */ | |
f97c1e0c | 2091 | if (ipv4_is_multicast(daddr)) { |
96d36220 | 2092 | struct in_device *in_dev = __in_dev_get_rcu(dev); |
e58e4159 | 2093 | int our = 0; |
5510cdf7 | 2094 | int err = -EINVAL; |
1da177e4 | 2095 | |
e58e4159 DA |
2096 | if (in_dev) |
2097 | our = ip_check_mc_rcu(in_dev, daddr, saddr, | |
2098 | ip_hdr(skb)->protocol); | |
2099 | ||
2100 | /* check l3 master if no match yet */ | |
2101 | if ((!in_dev || !our) && netif_is_l3_slave(dev)) { | |
2102 | struct in_device *l3_in_dev; | |
2103 | ||
2104 | l3_in_dev = __in_dev_get_rcu(skb->dev); | |
2105 | if (l3_in_dev) | |
2106 | our = ip_check_mc_rcu(l3_in_dev, daddr, saddr, | |
2107 | ip_hdr(skb)->protocol); | |
2108 | } | |
2109 | ||
e58e4159 | 2110 | if (our |
1da177e4 | 2111 | #ifdef CONFIG_IP_MROUTE |
e58e4159 DA |
2112 | || |
2113 | (!ipv4_is_local_multicast(daddr) && | |
2114 | IN_DEV_MFORWARD(in_dev)) | |
1da177e4 | 2115 | #endif |
e58e4159 | 2116 | ) { |
5510cdf7 | 2117 | err = ip_route_input_mc(skb, daddr, saddr, |
e58e4159 | 2118 | tos, dev, our); |
1da177e4 | 2119 | } |
5510cdf7 | 2120 | return err; |
1da177e4 | 2121 | } |
5510cdf7 DA |
2122 | |
2123 | return ip_route_input_slow(skb, daddr, saddr, tos, dev, res); | |
1da177e4 LT |
2124 | } |
2125 | ||
ebc0ffae | 2126 | /* called with rcu_read_lock() */ |
982721f3 | 2127 | static struct rtable *__mkroute_output(const struct fib_result *res, |
1a00fee4 | 2128 | const struct flowi4 *fl4, int orig_oif, |
f61759e6 | 2129 | struct net_device *dev_out, |
5ada5527 | 2130 | unsigned int flags) |
1da177e4 | 2131 | { |
982721f3 | 2132 | struct fib_info *fi = res->fi; |
f2bb4bed | 2133 | struct fib_nh_exception *fnhe; |
5ada5527 | 2134 | struct in_device *in_dev; |
982721f3 | 2135 | u16 type = res->type; |
5ada5527 | 2136 | struct rtable *rth; |
c92b9655 | 2137 | bool do_cache; |
1da177e4 | 2138 | |
d0daebc3 TG |
2139 | in_dev = __in_dev_get_rcu(dev_out); |
2140 | if (!in_dev) | |
5ada5527 | 2141 | return ERR_PTR(-EINVAL); |
1da177e4 | 2142 | |
d0daebc3 | 2143 | if (likely(!IN_DEV_ROUTE_LOCALNET(in_dev))) |
5f02ce24 DA |
2144 | if (ipv4_is_loopback(fl4->saddr) && |
2145 | !(dev_out->flags & IFF_LOOPBACK) && | |
2146 | !netif_is_l3_master(dev_out)) | |
d0daebc3 TG |
2147 | return ERR_PTR(-EINVAL); |
2148 | ||
68a5e3dd | 2149 | if (ipv4_is_lbcast(fl4->daddr)) |
982721f3 | 2150 | type = RTN_BROADCAST; |
68a5e3dd | 2151 | else if (ipv4_is_multicast(fl4->daddr)) |
982721f3 | 2152 | type = RTN_MULTICAST; |
68a5e3dd | 2153 | else if (ipv4_is_zeronet(fl4->daddr)) |
5ada5527 | 2154 | return ERR_PTR(-EINVAL); |
1da177e4 LT |
2155 | |
2156 | if (dev_out->flags & IFF_LOOPBACK) | |
2157 | flags |= RTCF_LOCAL; | |
2158 | ||
63617421 | 2159 | do_cache = true; |
982721f3 | 2160 | if (type == RTN_BROADCAST) { |
1da177e4 | 2161 | flags |= RTCF_BROADCAST | RTCF_LOCAL; |
982721f3 DM |
2162 | fi = NULL; |
2163 | } else if (type == RTN_MULTICAST) { | |
dd28d1a0 | 2164 | flags |= RTCF_MULTICAST | RTCF_LOCAL; |
813b3b5d DM |
2165 | if (!ip_check_mc_rcu(in_dev, fl4->daddr, fl4->saddr, |
2166 | fl4->flowi4_proto)) | |
1da177e4 | 2167 | flags &= ~RTCF_LOCAL; |
63617421 JA |
2168 | else |
2169 | do_cache = false; | |
1da177e4 | 2170 | /* If multicast route do not exist use |
dd28d1a0 ED |
2171 | * default one, but do not gateway in this case. |
2172 | * Yes, it is hack. | |
1da177e4 | 2173 | */ |
982721f3 DM |
2174 | if (fi && res->prefixlen < 4) |
2175 | fi = NULL; | |
d6d5e999 CF |
2176 | } else if ((type == RTN_LOCAL) && (orig_oif != 0) && |
2177 | (orig_oif != dev_out->ifindex)) { | |
2178 | /* For local routes that require a particular output interface | |
2179 | * we do not want to cache the result. Caching the result | |
2180 | * causes incorrect behaviour when there are multiple source | |
2181 | * addresses on the interface, the end result being that if the | |
2182 | * intended recipient is waiting on that interface for the | |
2183 | * packet he won't receive it because it will be delivered on | |
2184 | * the loopback interface and the IP_PKTINFO ipi_ifindex will | |
2185 | * be set to the loopback interface as well. | |
2186 | */ | |
2187 | fi = NULL; | |
1da177e4 LT |
2188 | } |
2189 | ||
f2bb4bed | 2190 | fnhe = NULL; |
63617421 JA |
2191 | do_cache &= fi != NULL; |
2192 | if (do_cache) { | |
c5038a83 | 2193 | struct rtable __rcu **prth; |
c92b9655 | 2194 | struct fib_nh *nh = &FIB_RES_NH(*res); |
d26b3a7c | 2195 | |
c92b9655 | 2196 | fnhe = find_exception(nh, fl4->daddr); |
deed49df | 2197 | if (fnhe) { |
2ffae99d | 2198 | prth = &fnhe->fnhe_rth_output; |
deed49df XL |
2199 | rth = rcu_dereference(*prth); |
2200 | if (rth && rth->dst.expires && | |
2201 | time_after(jiffies, rth->dst.expires)) { | |
2202 | ip_del_fnhe(nh, fl4->daddr); | |
2203 | fnhe = NULL; | |
2204 | } else { | |
2205 | goto rt_cache; | |
c92b9655 | 2206 | } |
c92b9655 | 2207 | } |
deed49df XL |
2208 | |
2209 | if (unlikely(fl4->flowi4_flags & | |
2210 | FLOWI_FLAG_KNOWN_NH && | |
2211 | !(nh->nh_gw && | |
2212 | nh->nh_scope == RT_SCOPE_LINK))) { | |
2213 | do_cache = false; | |
2214 | goto add; | |
2215 | } | |
2216 | prth = raw_cpu_ptr(nh->nh_pcpu_rth_output); | |
c5038a83 | 2217 | rth = rcu_dereference(*prth); |
deed49df XL |
2218 | |
2219 | rt_cache: | |
c5038a83 DM |
2220 | if (rt_cache_valid(rth)) { |
2221 | dst_hold(&rth->dst); | |
2222 | return rth; | |
f2bb4bed DM |
2223 | } |
2224 | } | |
c92b9655 JA |
2225 | |
2226 | add: | |
d08c4f35 | 2227 | rth = rt_dst_alloc(dev_out, flags, type, |
5c1e6aa3 | 2228 | IN_DEV_CONF_GET(in_dev, NOPOLICY), |
f2bb4bed | 2229 | IN_DEV_CONF_GET(in_dev, NOXFRM), |
c92b9655 | 2230 | do_cache); |
8391d07b | 2231 | if (!rth) |
5ada5527 | 2232 | return ERR_PTR(-ENOBUFS); |
8391d07b | 2233 | |
13378cad | 2234 | rth->rt_iif = orig_oif ? : 0; |
b7503e0c DA |
2235 | if (res->table) |
2236 | rth->rt_table_id = res->table->tb_id; | |
2237 | ||
1da177e4 LT |
2238 | RT_CACHE_STAT_INC(out_slow_tot); |
2239 | ||
1da177e4 | 2240 | if (flags & (RTCF_BROADCAST | RTCF_MULTICAST)) { |
e905a9ed | 2241 | if (flags & RTCF_LOCAL && |
1da177e4 | 2242 | !(dev_out->flags & IFF_LOOPBACK)) { |
d8d1f30b | 2243 | rth->dst.output = ip_mc_output; |
1da177e4 LT |
2244 | RT_CACHE_STAT_INC(out_slow_mc); |
2245 | } | |
2246 | #ifdef CONFIG_IP_MROUTE | |
982721f3 | 2247 | if (type == RTN_MULTICAST) { |
1da177e4 | 2248 | if (IN_DEV_MFORWARD(in_dev) && |
813b3b5d | 2249 | !ipv4_is_local_multicast(fl4->daddr)) { |
d8d1f30b CG |
2250 | rth->dst.input = ip_mr_input; |
2251 | rth->dst.output = ip_mc_output; | |
1da177e4 LT |
2252 | } |
2253 | } | |
2254 | #endif | |
2255 | } | |
2256 | ||
f2bb4bed | 2257 | rt_set_nexthop(rth, fl4->daddr, res, fnhe, fi, type, 0); |
efd85700 | 2258 | set_lwt_redirect(rth); |
1da177e4 | 2259 | |
5ada5527 | 2260 | return rth; |
1da177e4 LT |
2261 | } |
2262 | ||
1da177e4 LT |
2263 | /* |
2264 | * Major route resolver routine. | |
2265 | */ | |
2266 | ||
3abd1ade DA |
2267 | struct rtable *ip_route_output_key_hash(struct net *net, struct flowi4 *fl4, |
2268 | const struct sk_buff *skb) | |
1da177e4 | 2269 | { |
f61759e6 | 2270 | __u8 tos = RT_FL_TOS(fl4); |
813b3b5d | 2271 | struct fib_result res; |
5ada5527 | 2272 | struct rtable *rth; |
1da177e4 | 2273 | |
85b91b03 | 2274 | res.tclassid = 0; |
1da177e4 | 2275 | res.fi = NULL; |
8b96d22d | 2276 | res.table = NULL; |
1da177e4 | 2277 | |
1fb9489b | 2278 | fl4->flowi4_iif = LOOPBACK_IFINDEX; |
813b3b5d DM |
2279 | fl4->flowi4_tos = tos & IPTOS_RT_MASK; |
2280 | fl4->flowi4_scope = ((tos & RTO_ONLINK) ? | |
2281 | RT_SCOPE_LINK : RT_SCOPE_UNIVERSE); | |
44713b67 | 2282 | |
010c2708 | 2283 | rcu_read_lock(); |
3abd1ade DA |
2284 | rth = ip_route_output_key_hash_rcu(net, fl4, &res, skb); |
2285 | rcu_read_unlock(); | |
2286 | ||
2287 | return rth; | |
2288 | } | |
2289 | EXPORT_SYMBOL_GPL(ip_route_output_key_hash); | |
2290 | ||
2291 | struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *fl4, | |
2292 | struct fib_result *res, | |
2293 | const struct sk_buff *skb) | |
2294 | { | |
2295 | struct net_device *dev_out = NULL; | |
2296 | int orig_oif = fl4->flowi4_oif; | |
2297 | unsigned int flags = 0; | |
2298 | struct rtable *rth; | |
2299 | int err = -ENETUNREACH; | |
2300 | ||
813b3b5d | 2301 | if (fl4->saddr) { |
b23dd4fe | 2302 | rth = ERR_PTR(-EINVAL); |
813b3b5d DM |
2303 | if (ipv4_is_multicast(fl4->saddr) || |
2304 | ipv4_is_lbcast(fl4->saddr) || | |
2305 | ipv4_is_zeronet(fl4->saddr)) | |
1da177e4 LT |
2306 | goto out; |
2307 | ||
1da177e4 LT |
2308 | /* I removed check for oif == dev_out->oif here. |
2309 | It was wrong for two reasons: | |
1ab35276 DL |
2310 | 1. ip_dev_find(net, saddr) can return wrong iface, if saddr |
2311 | is assigned to multiple interfaces. | |
1da177e4 LT |
2312 | 2. Moreover, we are allowed to send packets with saddr |
2313 | of another iface. --ANK | |
2314 | */ | |
2315 | ||
813b3b5d DM |
2316 | if (fl4->flowi4_oif == 0 && |
2317 | (ipv4_is_multicast(fl4->daddr) || | |
2318 | ipv4_is_lbcast(fl4->daddr))) { | |
a210d01a | 2319 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
813b3b5d | 2320 | dev_out = __ip_dev_find(net, fl4->saddr, false); |
51456b29 | 2321 | if (!dev_out) |
a210d01a JA |
2322 | goto out; |
2323 | ||
1da177e4 LT |
2324 | /* Special hack: user can direct multicasts |
2325 | and limited broadcast via necessary interface | |
2326 | without fiddling with IP_MULTICAST_IF or IP_PKTINFO. | |
2327 | This hack is not just for fun, it allows | |
2328 | vic,vat and friends to work. | |
2329 | They bind socket to loopback, set ttl to zero | |
2330 | and expect that it will work. | |
2331 | From the viewpoint of routing cache they are broken, | |
2332 | because we are not allowed to build multicast path | |
2333 | with loopback source addr (look, routing cache | |
2334 | cannot know, that ttl is zero, so that packet | |
2335 | will not leave this host and route is valid). | |
2336 | Luckily, this hack is good workaround. | |
2337 | */ | |
2338 | ||
813b3b5d | 2339 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2340 | goto make_route; |
2341 | } | |
a210d01a | 2342 | |
813b3b5d | 2343 | if (!(fl4->flowi4_flags & FLOWI_FLAG_ANYSRC)) { |
a210d01a | 2344 | /* It is equivalent to inet_addr_type(saddr) == RTN_LOCAL */ |
813b3b5d | 2345 | if (!__ip_dev_find(net, fl4->saddr, false)) |
a210d01a | 2346 | goto out; |
a210d01a | 2347 | } |
1da177e4 LT |
2348 | } |
2349 | ||
2350 | ||
813b3b5d DM |
2351 | if (fl4->flowi4_oif) { |
2352 | dev_out = dev_get_by_index_rcu(net, fl4->flowi4_oif); | |
b23dd4fe | 2353 | rth = ERR_PTR(-ENODEV); |
51456b29 | 2354 | if (!dev_out) |
1da177e4 | 2355 | goto out; |
e5ed6399 HX |
2356 | |
2357 | /* RACE: Check return value of inet_select_addr instead. */ | |
fc75fc83 | 2358 | if (!(dev_out->flags & IFF_UP) || !__in_dev_get_rcu(dev_out)) { |
b23dd4fe | 2359 | rth = ERR_PTR(-ENETUNREACH); |
fc75fc83 ED |
2360 | goto out; |
2361 | } | |
813b3b5d | 2362 | if (ipv4_is_local_multicast(fl4->daddr) || |
6a211654 AL |
2363 | ipv4_is_lbcast(fl4->daddr) || |
2364 | fl4->flowi4_proto == IPPROTO_IGMP) { | |
813b3b5d DM |
2365 | if (!fl4->saddr) |
2366 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2367 | RT_SCOPE_LINK); | |
1da177e4 LT |
2368 | goto make_route; |
2369 | } | |
0a7e2260 | 2370 | if (!fl4->saddr) { |
813b3b5d DM |
2371 | if (ipv4_is_multicast(fl4->daddr)) |
2372 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2373 | fl4->flowi4_scope); | |
2374 | else if (!fl4->daddr) | |
2375 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2376 | RT_SCOPE_HOST); | |
1da177e4 LT |
2377 | } |
2378 | } | |
2379 | ||
813b3b5d DM |
2380 | if (!fl4->daddr) { |
2381 | fl4->daddr = fl4->saddr; | |
2382 | if (!fl4->daddr) | |
2383 | fl4->daddr = fl4->saddr = htonl(INADDR_LOOPBACK); | |
b40afd0e | 2384 | dev_out = net->loopback_dev; |
1fb9489b | 2385 | fl4->flowi4_oif = LOOPBACK_IFINDEX; |
3abd1ade | 2386 | res->type = RTN_LOCAL; |
1da177e4 LT |
2387 | flags |= RTCF_LOCAL; |
2388 | goto make_route; | |
2389 | } | |
2390 | ||
3abd1ade | 2391 | err = fib_lookup(net, fl4, res, 0); |
0315e382 | 2392 | if (err) { |
3abd1ade DA |
2393 | res->fi = NULL; |
2394 | res->table = NULL; | |
6104e112 | 2395 | if (fl4->flowi4_oif && |
e58e4159 DA |
2396 | (ipv4_is_multicast(fl4->daddr) || |
2397 | !netif_index_is_l3_master(net, fl4->flowi4_oif))) { | |
1da177e4 LT |
2398 | /* Apparently, routing tables are wrong. Assume, |
2399 | that the destination is on link. | |
2400 | ||
2401 | WHY? DW. | |
2402 | Because we are allowed to send to iface | |
2403 | even if it has NO routes and NO assigned | |
2404 | addresses. When oif is specified, routing | |
2405 | tables are looked up with only one purpose: | |
2406 | to catch if destination is gatewayed, rather than | |
2407 | direct. Moreover, if MSG_DONTROUTE is set, | |
2408 | we send packet, ignoring both routing tables | |
2409 | and ifaddr state. --ANK | |
2410 | ||
2411 | ||
2412 | We could make it even if oif is unknown, | |
2413 | likely IPv6, but we do not. | |
2414 | */ | |
2415 | ||
813b3b5d DM |
2416 | if (fl4->saddr == 0) |
2417 | fl4->saddr = inet_select_addr(dev_out, 0, | |
2418 | RT_SCOPE_LINK); | |
3abd1ade | 2419 | res->type = RTN_UNICAST; |
1da177e4 LT |
2420 | goto make_route; |
2421 | } | |
0315e382 | 2422 | rth = ERR_PTR(err); |
1da177e4 LT |
2423 | goto out; |
2424 | } | |
1da177e4 | 2425 | |
3abd1ade | 2426 | if (res->type == RTN_LOCAL) { |
813b3b5d | 2427 | if (!fl4->saddr) { |
3abd1ade DA |
2428 | if (res->fi->fib_prefsrc) |
2429 | fl4->saddr = res->fi->fib_prefsrc; | |
9fc3bbb4 | 2430 | else |
813b3b5d | 2431 | fl4->saddr = fl4->daddr; |
9fc3bbb4 | 2432 | } |
5f02ce24 DA |
2433 | |
2434 | /* L3 master device is the loopback for that domain */ | |
3abd1ade | 2435 | dev_out = l3mdev_master_dev_rcu(FIB_RES_DEV(*res)) ? : |
b7c8487c | 2436 | net->loopback_dev; |
813b3b5d | 2437 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2438 | flags |= RTCF_LOCAL; |
2439 | goto make_route; | |
2440 | } | |
2441 | ||
3abd1ade | 2442 | fib_select_path(net, res, fl4, skb); |
1da177e4 | 2443 | |
3abd1ade | 2444 | dev_out = FIB_RES_DEV(*res); |
813b3b5d | 2445 | fl4->flowi4_oif = dev_out->ifindex; |
1da177e4 LT |
2446 | |
2447 | ||
2448 | make_route: | |
3abd1ade | 2449 | rth = __mkroute_output(res, fl4, orig_oif, dev_out, flags); |
1da177e4 | 2450 | |
010c2708 | 2451 | out: |
b23dd4fe | 2452 | return rth; |
1da177e4 | 2453 | } |
d8c97a94 | 2454 | |
ae2688d5 JW |
2455 | static struct dst_entry *ipv4_blackhole_dst_check(struct dst_entry *dst, u32 cookie) |
2456 | { | |
2457 | return NULL; | |
2458 | } | |
2459 | ||
ebb762f2 | 2460 | static unsigned int ipv4_blackhole_mtu(const struct dst_entry *dst) |
ec831ea7 | 2461 | { |
618f9bc7 SK |
2462 | unsigned int mtu = dst_metric_raw(dst, RTAX_MTU); |
2463 | ||
2464 | return mtu ? : dst->dev->mtu; | |
ec831ea7 RD |
2465 | } |
2466 | ||
6700c270 DM |
2467 | static void ipv4_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk, |
2468 | struct sk_buff *skb, u32 mtu) | |
14e50e57 DM |
2469 | { |
2470 | } | |
2471 | ||
6700c270 DM |
2472 | static void ipv4_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk, |
2473 | struct sk_buff *skb) | |
b587ee3b DM |
2474 | { |
2475 | } | |
2476 | ||
0972ddb2 HB |
2477 | static u32 *ipv4_rt_blackhole_cow_metrics(struct dst_entry *dst, |
2478 | unsigned long old) | |
2479 | { | |
2480 | return NULL; | |
2481 | } | |
2482 | ||
14e50e57 DM |
2483 | static struct dst_ops ipv4_dst_blackhole_ops = { |
2484 | .family = AF_INET, | |
ae2688d5 | 2485 | .check = ipv4_blackhole_dst_check, |
ebb762f2 | 2486 | .mtu = ipv4_blackhole_mtu, |
214f45c9 | 2487 | .default_advmss = ipv4_default_advmss, |
14e50e57 | 2488 | .update_pmtu = ipv4_rt_blackhole_update_pmtu, |
b587ee3b | 2489 | .redirect = ipv4_rt_blackhole_redirect, |
0972ddb2 | 2490 | .cow_metrics = ipv4_rt_blackhole_cow_metrics, |
d3aaeb38 | 2491 | .neigh_lookup = ipv4_neigh_lookup, |
14e50e57 DM |
2492 | }; |
2493 | ||
2774c131 | 2494 | struct dst_entry *ipv4_blackhole_route(struct net *net, struct dst_entry *dst_orig) |
14e50e57 | 2495 | { |
2774c131 | 2496 | struct rtable *ort = (struct rtable *) dst_orig; |
f5b0a874 | 2497 | struct rtable *rt; |
14e50e57 | 2498 | |
f5b0a874 | 2499 | rt = dst_alloc(&ipv4_dst_blackhole_ops, NULL, 1, DST_OBSOLETE_NONE, 0); |
14e50e57 | 2500 | if (rt) { |
d8d1f30b | 2501 | struct dst_entry *new = &rt->dst; |
14e50e57 | 2502 | |
14e50e57 | 2503 | new->__use = 1; |
352e512c | 2504 | new->input = dst_discard; |
ede2059d | 2505 | new->output = dst_discard_out; |
14e50e57 | 2506 | |
d8d1f30b | 2507 | new->dev = ort->dst.dev; |
14e50e57 DM |
2508 | if (new->dev) |
2509 | dev_hold(new->dev); | |
2510 | ||
9917e1e8 | 2511 | rt->rt_is_input = ort->rt_is_input; |
5e2b61f7 | 2512 | rt->rt_iif = ort->rt_iif; |
5943634f | 2513 | rt->rt_pmtu = ort->rt_pmtu; |
14e50e57 | 2514 | |
ca4c3fc2 | 2515 | rt->rt_genid = rt_genid_ipv4(net); |
14e50e57 DM |
2516 | rt->rt_flags = ort->rt_flags; |
2517 | rt->rt_type = ort->rt_type; | |
14e50e57 | 2518 | rt->rt_gateway = ort->rt_gateway; |
155e8336 | 2519 | rt->rt_uses_gateway = ort->rt_uses_gateway; |
14e50e57 | 2520 | |
caacf05e | 2521 | INIT_LIST_HEAD(&rt->rt_uncached); |
14e50e57 DM |
2522 | dst_free(new); |
2523 | } | |
2524 | ||
2774c131 DM |
2525 | dst_release(dst_orig); |
2526 | ||
2527 | return rt ? &rt->dst : ERR_PTR(-ENOMEM); | |
14e50e57 DM |
2528 | } |
2529 | ||
9d6ec938 | 2530 | struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4, |
6f9c9615 | 2531 | const struct sock *sk) |
1da177e4 | 2532 | { |
9d6ec938 | 2533 | struct rtable *rt = __ip_route_output_key(net, flp4); |
1da177e4 | 2534 | |
b23dd4fe DM |
2535 | if (IS_ERR(rt)) |
2536 | return rt; | |
1da177e4 | 2537 | |
56157872 | 2538 | if (flp4->flowi4_proto) |
f92ee619 SK |
2539 | rt = (struct rtable *)xfrm_lookup_route(net, &rt->dst, |
2540 | flowi4_to_flowi(flp4), | |
2541 | sk, 0); | |
1da177e4 | 2542 | |
b23dd4fe | 2543 | return rt; |
1da177e4 | 2544 | } |
d8c97a94 ACM |
2545 | EXPORT_SYMBOL_GPL(ip_route_output_flow); |
2546 | ||
3765d35e | 2547 | /* called with rcu_read_lock held */ |
c36ba660 | 2548 | static int rt_fill_info(struct net *net, __be32 dst, __be32 src, u32 table_id, |
15e47304 | 2549 | struct flowi4 *fl4, struct sk_buff *skb, u32 portid, |
3765d35e | 2550 | u32 seq, struct rtable *rt) |
1da177e4 | 2551 | { |
1da177e4 | 2552 | struct rtmsg *r; |
be403ea1 | 2553 | struct nlmsghdr *nlh; |
2bc8ca40 | 2554 | unsigned long expires = 0; |
f185071d | 2555 | u32 error; |
521f5490 | 2556 | u32 metrics[RTAX_MAX]; |
be403ea1 | 2557 | |
d3166e0c | 2558 | nlh = nlmsg_put(skb, portid, seq, RTM_NEWROUTE, sizeof(*r), 0); |
51456b29 | 2559 | if (!nlh) |
26932566 | 2560 | return -EMSGSIZE; |
be403ea1 TG |
2561 | |
2562 | r = nlmsg_data(nlh); | |
1da177e4 LT |
2563 | r->rtm_family = AF_INET; |
2564 | r->rtm_dst_len = 32; | |
2565 | r->rtm_src_len = 0; | |
d6c0a4f6 | 2566 | r->rtm_tos = fl4->flowi4_tos; |
8a430ed5 | 2567 | r->rtm_table = table_id < 256 ? table_id : RT_TABLE_COMPAT; |
c36ba660 | 2568 | if (nla_put_u32(skb, RTA_TABLE, table_id)) |
f3756b79 | 2569 | goto nla_put_failure; |
1da177e4 LT |
2570 | r->rtm_type = rt->rt_type; |
2571 | r->rtm_scope = RT_SCOPE_UNIVERSE; | |
2572 | r->rtm_protocol = RTPROT_UNSPEC; | |
2573 | r->rtm_flags = (rt->rt_flags & ~0xFFFF) | RTM_F_CLONED; | |
2574 | if (rt->rt_flags & RTCF_NOTIFY) | |
2575 | r->rtm_flags |= RTM_F_NOTIFY; | |
df4d9254 HFS |
2576 | if (IPCB(skb)->flags & IPSKB_DOREDIRECT) |
2577 | r->rtm_flags |= RTCF_DOREDIRECT; | |
be403ea1 | 2578 | |
930345ea | 2579 | if (nla_put_in_addr(skb, RTA_DST, dst)) |
f3756b79 | 2580 | goto nla_put_failure; |
1a00fee4 | 2581 | if (src) { |
1da177e4 | 2582 | r->rtm_src_len = 32; |
930345ea | 2583 | if (nla_put_in_addr(skb, RTA_SRC, src)) |
f3756b79 | 2584 | goto nla_put_failure; |
1da177e4 | 2585 | } |
f3756b79 DM |
2586 | if (rt->dst.dev && |
2587 | nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex)) | |
2588 | goto nla_put_failure; | |
c7066f70 | 2589 | #ifdef CONFIG_IP_ROUTE_CLASSID |
f3756b79 DM |
2590 | if (rt->dst.tclassid && |
2591 | nla_put_u32(skb, RTA_FLOW, rt->dst.tclassid)) | |
2592 | goto nla_put_failure; | |
1da177e4 | 2593 | #endif |
41347dcd | 2594 | if (!rt_is_input_route(rt) && |
d6c0a4f6 | 2595 | fl4->saddr != src) { |
930345ea | 2596 | if (nla_put_in_addr(skb, RTA_PREFSRC, fl4->saddr)) |
f3756b79 DM |
2597 | goto nla_put_failure; |
2598 | } | |
155e8336 | 2599 | if (rt->rt_uses_gateway && |
930345ea | 2600 | nla_put_in_addr(skb, RTA_GATEWAY, rt->rt_gateway)) |
f3756b79 | 2601 | goto nla_put_failure; |
be403ea1 | 2602 | |
ee9a8f7a SK |
2603 | expires = rt->dst.expires; |
2604 | if (expires) { | |
2605 | unsigned long now = jiffies; | |
2606 | ||
2607 | if (time_before(now, expires)) | |
2608 | expires -= now; | |
2609 | else | |
2610 | expires = 0; | |
2611 | } | |
2612 | ||
521f5490 | 2613 | memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics)); |
ee9a8f7a | 2614 | if (rt->rt_pmtu && expires) |
521f5490 JA |
2615 | metrics[RTAX_MTU - 1] = rt->rt_pmtu; |
2616 | if (rtnetlink_put_metrics(skb, metrics) < 0) | |
be403ea1 TG |
2617 | goto nla_put_failure; |
2618 | ||
b4869889 | 2619 | if (fl4->flowi4_mark && |
68aaed54 | 2620 | nla_put_u32(skb, RTA_MARK, fl4->flowi4_mark)) |
f3756b79 | 2621 | goto nla_put_failure; |
963bfeee | 2622 | |
622ec2c9 LC |
2623 | if (!uid_eq(fl4->flowi4_uid, INVALID_UID) && |
2624 | nla_put_u32(skb, RTA_UID, | |
2625 | from_kuid_munged(current_user_ns(), fl4->flowi4_uid))) | |
2626 | goto nla_put_failure; | |
2627 | ||
d8d1f30b | 2628 | error = rt->dst.error; |
be403ea1 | 2629 | |
c7537967 | 2630 | if (rt_is_input_route(rt)) { |
8caaf7b6 ND |
2631 | #ifdef CONFIG_IP_MROUTE |
2632 | if (ipv4_is_multicast(dst) && !ipv4_is_local_multicast(dst) && | |
2633 | IPV4_DEVCONF_ALL(net, MC_FORWARDING)) { | |
2634 | int err = ipmr_get_route(net, skb, | |
2635 | fl4->saddr, fl4->daddr, | |
9f09eaea | 2636 | r, portid); |
2cf75070 | 2637 | |
8caaf7b6 | 2638 | if (err <= 0) { |
0c8d803f DA |
2639 | if (err == 0) |
2640 | return 0; | |
2641 | goto nla_put_failure; | |
8caaf7b6 ND |
2642 | } |
2643 | } else | |
2644 | #endif | |
91146153 | 2645 | if (nla_put_u32(skb, RTA_IIF, skb->dev->ifindex)) |
8caaf7b6 | 2646 | goto nla_put_failure; |
1da177e4 LT |
2647 | } |
2648 | ||
f185071d | 2649 | if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, error) < 0) |
e3703b3d | 2650 | goto nla_put_failure; |
be403ea1 | 2651 | |
053c095a JB |
2652 | nlmsg_end(skb, nlh); |
2653 | return 0; | |
1da177e4 | 2654 | |
be403ea1 | 2655 | nla_put_failure: |
26932566 PM |
2656 | nlmsg_cancel(skb, nlh); |
2657 | return -EMSGSIZE; | |
1da177e4 LT |
2658 | } |
2659 | ||
c21ef3e3 DA |
2660 | static int inet_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, |
2661 | struct netlink_ext_ack *extack) | |
1da177e4 | 2662 | { |
3b1e0a65 | 2663 | struct net *net = sock_net(in_skb->sk); |
d889ce3b TG |
2664 | struct rtmsg *rtm; |
2665 | struct nlattr *tb[RTA_MAX+1]; | |
3765d35e | 2666 | struct fib_result res = {}; |
1da177e4 | 2667 | struct rtable *rt = NULL; |
d6c0a4f6 | 2668 | struct flowi4 fl4; |
9e12bb22 AV |
2669 | __be32 dst = 0; |
2670 | __be32 src = 0; | |
2671 | u32 iif; | |
d889ce3b | 2672 | int err; |
963bfeee | 2673 | int mark; |
1da177e4 | 2674 | struct sk_buff *skb; |
c36ba660 | 2675 | u32 table_id = RT_TABLE_MAIN; |
622ec2c9 | 2676 | kuid_t uid; |
1da177e4 | 2677 | |
fceb6435 | 2678 | err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv4_policy, |
c21ef3e3 | 2679 | extack); |
d889ce3b TG |
2680 | if (err < 0) |
2681 | goto errout; | |
2682 | ||
2683 | rtm = nlmsg_data(nlh); | |
2684 | ||
1da177e4 | 2685 | skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
51456b29 | 2686 | if (!skb) { |
d889ce3b TG |
2687 | err = -ENOBUFS; |
2688 | goto errout; | |
2689 | } | |
1da177e4 LT |
2690 | |
2691 | /* Reserve room for dummy headers, this skb can pass | |
2692 | through good chunk of routing engine. | |
2693 | */ | |
459a98ed | 2694 | skb_reset_mac_header(skb); |
c1d2bbe1 | 2695 | skb_reset_network_header(skb); |
d2c962b8 | 2696 | |
67b61f6c JB |
2697 | src = tb[RTA_SRC] ? nla_get_in_addr(tb[RTA_SRC]) : 0; |
2698 | dst = tb[RTA_DST] ? nla_get_in_addr(tb[RTA_DST]) : 0; | |
d889ce3b | 2699 | iif = tb[RTA_IIF] ? nla_get_u32(tb[RTA_IIF]) : 0; |
963bfeee | 2700 | mark = tb[RTA_MARK] ? nla_get_u32(tb[RTA_MARK]) : 0; |
622ec2c9 LC |
2701 | if (tb[RTA_UID]) |
2702 | uid = make_kuid(current_user_ns(), nla_get_u32(tb[RTA_UID])); | |
2703 | else | |
2704 | uid = (iif ? INVALID_UID : current_uid()); | |
1da177e4 | 2705 | |
bbadb9a2 FL |
2706 | /* Bugfix: need to give ip_route_input enough of an IP header to |
2707 | * not gag. | |
2708 | */ | |
2709 | ip_hdr(skb)->protocol = IPPROTO_UDP; | |
2710 | ip_hdr(skb)->saddr = src; | |
2711 | ip_hdr(skb)->daddr = dst; | |
2712 | ||
2713 | skb_reserve(skb, MAX_HEADER + sizeof(struct iphdr)); | |
2714 | ||
d6c0a4f6 DM |
2715 | memset(&fl4, 0, sizeof(fl4)); |
2716 | fl4.daddr = dst; | |
2717 | fl4.saddr = src; | |
2718 | fl4.flowi4_tos = rtm->rtm_tos; | |
2719 | fl4.flowi4_oif = tb[RTA_OIF] ? nla_get_u32(tb[RTA_OIF]) : 0; | |
2720 | fl4.flowi4_mark = mark; | |
622ec2c9 | 2721 | fl4.flowi4_uid = uid; |
d6c0a4f6 | 2722 | |
3765d35e DA |
2723 | rcu_read_lock(); |
2724 | ||
1da177e4 | 2725 | if (iif) { |
d889ce3b TG |
2726 | struct net_device *dev; |
2727 | ||
3765d35e | 2728 | dev = dev_get_by_index_rcu(net, iif); |
51456b29 | 2729 | if (!dev) { |
d889ce3b TG |
2730 | err = -ENODEV; |
2731 | goto errout_free; | |
2732 | } | |
2733 | ||
1da177e4 LT |
2734 | skb->protocol = htons(ETH_P_IP); |
2735 | skb->dev = dev; | |
963bfeee | 2736 | skb->mark = mark; |
3765d35e DA |
2737 | err = ip_route_input_rcu(skb, dst, src, rtm->rtm_tos, |
2738 | dev, &res); | |
d889ce3b | 2739 | |
511c3f92 | 2740 | rt = skb_rtable(skb); |
d8d1f30b CG |
2741 | if (err == 0 && rt->dst.error) |
2742 | err = -rt->dst.error; | |
1da177e4 | 2743 | } else { |
3765d35e | 2744 | rt = ip_route_output_key_hash_rcu(net, &fl4, &res, skb); |
b23dd4fe DM |
2745 | err = 0; |
2746 | if (IS_ERR(rt)) | |
2747 | err = PTR_ERR(rt); | |
1da177e4 | 2748 | } |
d889ce3b | 2749 | |
1da177e4 | 2750 | if (err) |
d889ce3b | 2751 | goto errout_free; |
1da177e4 | 2752 | |
1da177e4 LT |
2753 | if (rtm->rtm_flags & RTM_F_NOTIFY) |
2754 | rt->rt_flags |= RTCF_NOTIFY; | |
2755 | ||
c36ba660 DA |
2756 | if (rtm->rtm_flags & RTM_F_LOOKUP_TABLE) |
2757 | table_id = rt->rt_table_id; | |
2758 | ||
b6179813 RP |
2759 | if (rtm->rtm_flags & RTM_F_FIB_MATCH) |
2760 | err = fib_dump_info(skb, NETLINK_CB(in_skb).portid, | |
2761 | nlh->nlmsg_seq, RTM_NEWROUTE, table_id, | |
2762 | rt->rt_type, res.prefix, res.prefixlen, | |
2763 | fl4.flowi4_tos, res.fi, 0); | |
2764 | else | |
2765 | err = rt_fill_info(net, dst, src, table_id, &fl4, skb, | |
2766 | NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, | |
2767 | rt); | |
7b46a644 | 2768 | if (err < 0) |
d889ce3b | 2769 | goto errout_free; |
1da177e4 | 2770 | |
3765d35e DA |
2771 | rcu_read_unlock(); |
2772 | ||
15e47304 | 2773 | err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); |
d889ce3b | 2774 | errout: |
2942e900 | 2775 | return err; |
1da177e4 | 2776 | |
d889ce3b | 2777 | errout_free: |
3765d35e | 2778 | rcu_read_unlock(); |
1da177e4 | 2779 | kfree_skb(skb); |
d889ce3b | 2780 | goto errout; |
1da177e4 LT |
2781 | } |
2782 | ||
1da177e4 LT |
2783 | void ip_rt_multicast_event(struct in_device *in_dev) |
2784 | { | |
4ccfe6d4 | 2785 | rt_cache_flush(dev_net(in_dev->dev)); |
1da177e4 LT |
2786 | } |
2787 | ||
2788 | #ifdef CONFIG_SYSCTL | |
082c7ca4 G |
2789 | static int ip_rt_gc_interval __read_mostly = 60 * HZ; |
2790 | static int ip_rt_gc_min_interval __read_mostly = HZ / 2; | |
2791 | static int ip_rt_gc_elasticity __read_mostly = 8; | |
2792 | ||
fe2c6338 | 2793 | static int ipv4_sysctl_rtcache_flush(struct ctl_table *__ctl, int write, |
8d65af78 | 2794 | void __user *buffer, |
1da177e4 LT |
2795 | size_t *lenp, loff_t *ppos) |
2796 | { | |
5aad1de5 TT |
2797 | struct net *net = (struct net *)__ctl->extra1; |
2798 | ||
1da177e4 | 2799 | if (write) { |
5aad1de5 TT |
2800 | rt_cache_flush(net); |
2801 | fnhe_genid_bump(net); | |
1da177e4 | 2802 | return 0; |
e905a9ed | 2803 | } |
1da177e4 LT |
2804 | |
2805 | return -EINVAL; | |
2806 | } | |
2807 | ||
fe2c6338 | 2808 | static struct ctl_table ipv4_route_table[] = { |
1da177e4 | 2809 | { |
1da177e4 LT |
2810 | .procname = "gc_thresh", |
2811 | .data = &ipv4_dst_ops.gc_thresh, | |
2812 | .maxlen = sizeof(int), | |
2813 | .mode = 0644, | |
6d9f239a | 2814 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2815 | }, |
2816 | { | |
1da177e4 LT |
2817 | .procname = "max_size", |
2818 | .data = &ip_rt_max_size, | |
2819 | .maxlen = sizeof(int), | |
2820 | .mode = 0644, | |
6d9f239a | 2821 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2822 | }, |
2823 | { | |
2824 | /* Deprecated. Use gc_min_interval_ms */ | |
e905a9ed | 2825 | |
1da177e4 LT |
2826 | .procname = "gc_min_interval", |
2827 | .data = &ip_rt_gc_min_interval, | |
2828 | .maxlen = sizeof(int), | |
2829 | .mode = 0644, | |
6d9f239a | 2830 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
2831 | }, |
2832 | { | |
1da177e4 LT |
2833 | .procname = "gc_min_interval_ms", |
2834 | .data = &ip_rt_gc_min_interval, | |
2835 | .maxlen = sizeof(int), | |
2836 | .mode = 0644, | |
6d9f239a | 2837 | .proc_handler = proc_dointvec_ms_jiffies, |
1da177e4 LT |
2838 | }, |
2839 | { | |
1da177e4 LT |
2840 | .procname = "gc_timeout", |
2841 | .data = &ip_rt_gc_timeout, | |
2842 | .maxlen = sizeof(int), | |
2843 | .mode = 0644, | |
6d9f239a | 2844 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 | 2845 | }, |
9f28a2fc ED |
2846 | { |
2847 | .procname = "gc_interval", | |
2848 | .data = &ip_rt_gc_interval, | |
2849 | .maxlen = sizeof(int), | |
2850 | .mode = 0644, | |
2851 | .proc_handler = proc_dointvec_jiffies, | |
2852 | }, | |
1da177e4 | 2853 | { |
1da177e4 LT |
2854 | .procname = "redirect_load", |
2855 | .data = &ip_rt_redirect_load, | |
2856 | .maxlen = sizeof(int), | |
2857 | .mode = 0644, | |
6d9f239a | 2858 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2859 | }, |
2860 | { | |
1da177e4 LT |
2861 | .procname = "redirect_number", |
2862 | .data = &ip_rt_redirect_number, | |
2863 | .maxlen = sizeof(int), | |
2864 | .mode = 0644, | |
6d9f239a | 2865 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2866 | }, |
2867 | { | |
1da177e4 LT |
2868 | .procname = "redirect_silence", |
2869 | .data = &ip_rt_redirect_silence, | |
2870 | .maxlen = sizeof(int), | |
2871 | .mode = 0644, | |
6d9f239a | 2872 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2873 | }, |
2874 | { | |
1da177e4 LT |
2875 | .procname = "error_cost", |
2876 | .data = &ip_rt_error_cost, | |
2877 | .maxlen = sizeof(int), | |
2878 | .mode = 0644, | |
6d9f239a | 2879 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2880 | }, |
2881 | { | |
1da177e4 LT |
2882 | .procname = "error_burst", |
2883 | .data = &ip_rt_error_burst, | |
2884 | .maxlen = sizeof(int), | |
2885 | .mode = 0644, | |
6d9f239a | 2886 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2887 | }, |
2888 | { | |
1da177e4 LT |
2889 | .procname = "gc_elasticity", |
2890 | .data = &ip_rt_gc_elasticity, | |
2891 | .maxlen = sizeof(int), | |
2892 | .mode = 0644, | |
6d9f239a | 2893 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2894 | }, |
2895 | { | |
1da177e4 LT |
2896 | .procname = "mtu_expires", |
2897 | .data = &ip_rt_mtu_expires, | |
2898 | .maxlen = sizeof(int), | |
2899 | .mode = 0644, | |
6d9f239a | 2900 | .proc_handler = proc_dointvec_jiffies, |
1da177e4 LT |
2901 | }, |
2902 | { | |
1da177e4 LT |
2903 | .procname = "min_pmtu", |
2904 | .data = &ip_rt_min_pmtu, | |
2905 | .maxlen = sizeof(int), | |
2906 | .mode = 0644, | |
6d9f239a | 2907 | .proc_handler = proc_dointvec, |
1da177e4 LT |
2908 | }, |
2909 | { | |
1da177e4 LT |
2910 | .procname = "min_adv_mss", |
2911 | .data = &ip_rt_min_advmss, | |
2912 | .maxlen = sizeof(int), | |
2913 | .mode = 0644, | |
6d9f239a | 2914 | .proc_handler = proc_dointvec, |
1da177e4 | 2915 | }, |
f8572d8f | 2916 | { } |
1da177e4 | 2917 | }; |
39a23e75 | 2918 | |
39a23e75 DL |
2919 | static struct ctl_table ipv4_route_flush_table[] = { |
2920 | { | |
39a23e75 DL |
2921 | .procname = "flush", |
2922 | .maxlen = sizeof(int), | |
2923 | .mode = 0200, | |
6d9f239a | 2924 | .proc_handler = ipv4_sysctl_rtcache_flush, |
39a23e75 | 2925 | }, |
f8572d8f | 2926 | { }, |
39a23e75 DL |
2927 | }; |
2928 | ||
2929 | static __net_init int sysctl_route_net_init(struct net *net) | |
2930 | { | |
2931 | struct ctl_table *tbl; | |
2932 | ||
2933 | tbl = ipv4_route_flush_table; | |
09ad9bc7 | 2934 | if (!net_eq(net, &init_net)) { |
39a23e75 | 2935 | tbl = kmemdup(tbl, sizeof(ipv4_route_flush_table), GFP_KERNEL); |
51456b29 | 2936 | if (!tbl) |
39a23e75 | 2937 | goto err_dup; |
464dc801 EB |
2938 | |
2939 | /* Don't export sysctls to unprivileged users */ | |
2940 | if (net->user_ns != &init_user_ns) | |
2941 | tbl[0].procname = NULL; | |
39a23e75 DL |
2942 | } |
2943 | tbl[0].extra1 = net; | |
2944 | ||
ec8f23ce | 2945 | net->ipv4.route_hdr = register_net_sysctl(net, "net/ipv4/route", tbl); |
51456b29 | 2946 | if (!net->ipv4.route_hdr) |
39a23e75 DL |
2947 | goto err_reg; |
2948 | return 0; | |
2949 | ||
2950 | err_reg: | |
2951 | if (tbl != ipv4_route_flush_table) | |
2952 | kfree(tbl); | |
2953 | err_dup: | |
2954 | return -ENOMEM; | |
2955 | } | |
2956 | ||
2957 | static __net_exit void sysctl_route_net_exit(struct net *net) | |
2958 | { | |
2959 | struct ctl_table *tbl; | |
2960 | ||
2961 | tbl = net->ipv4.route_hdr->ctl_table_arg; | |
2962 | unregister_net_sysctl_table(net->ipv4.route_hdr); | |
2963 | BUG_ON(tbl == ipv4_route_flush_table); | |
2964 | kfree(tbl); | |
2965 | } | |
2966 | ||
2967 | static __net_initdata struct pernet_operations sysctl_route_ops = { | |
2968 | .init = sysctl_route_net_init, | |
2969 | .exit = sysctl_route_net_exit, | |
2970 | }; | |
1da177e4 LT |
2971 | #endif |
2972 | ||
3ee94372 | 2973 | static __net_init int rt_genid_init(struct net *net) |
9f5e97e5 | 2974 | { |
ca4c3fc2 | 2975 | atomic_set(&net->ipv4.rt_genid, 0); |
5aad1de5 | 2976 | atomic_set(&net->fnhe_genid, 0); |
436c3b66 DM |
2977 | get_random_bytes(&net->ipv4.dev_addr_genid, |
2978 | sizeof(net->ipv4.dev_addr_genid)); | |
9f5e97e5 DL |
2979 | return 0; |
2980 | } | |
2981 | ||
3ee94372 NH |
2982 | static __net_initdata struct pernet_operations rt_genid_ops = { |
2983 | .init = rt_genid_init, | |
9f5e97e5 DL |
2984 | }; |
2985 | ||
c3426b47 DM |
2986 | static int __net_init ipv4_inetpeer_init(struct net *net) |
2987 | { | |
2988 | struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL); | |
2989 | ||
2990 | if (!bp) | |
2991 | return -ENOMEM; | |
2992 | inet_peer_base_init(bp); | |
2993 | net->ipv4.peers = bp; | |
2994 | return 0; | |
2995 | } | |
2996 | ||
2997 | static void __net_exit ipv4_inetpeer_exit(struct net *net) | |
2998 | { | |
2999 | struct inet_peer_base *bp = net->ipv4.peers; | |
3000 | ||
3001 | net->ipv4.peers = NULL; | |
56a6b248 | 3002 | inetpeer_invalidate_tree(bp); |
c3426b47 DM |
3003 | kfree(bp); |
3004 | } | |
3005 | ||
3006 | static __net_initdata struct pernet_operations ipv4_inetpeer_ops = { | |
3007 | .init = ipv4_inetpeer_init, | |
3008 | .exit = ipv4_inetpeer_exit, | |
3009 | }; | |
9f5e97e5 | 3010 | |
c7066f70 | 3011 | #ifdef CONFIG_IP_ROUTE_CLASSID |
7d720c3e | 3012 | struct ip_rt_acct __percpu *ip_rt_acct __read_mostly; |
c7066f70 | 3013 | #endif /* CONFIG_IP_ROUTE_CLASSID */ |
1da177e4 | 3014 | |
1da177e4 LT |
3015 | int __init ip_rt_init(void) |
3016 | { | |
424c4b70 | 3017 | int rc = 0; |
5055c371 | 3018 | int cpu; |
1da177e4 | 3019 | |
73f156a6 ED |
3020 | ip_idents = kmalloc(IP_IDENTS_SZ * sizeof(*ip_idents), GFP_KERNEL); |
3021 | if (!ip_idents) | |
3022 | panic("IP: failed to allocate ip_idents\n"); | |
3023 | ||
3024 | prandom_bytes(ip_idents, IP_IDENTS_SZ * sizeof(*ip_idents)); | |
3025 | ||
355b590c ED |
3026 | ip_tstamps = kcalloc(IP_IDENTS_SZ, sizeof(*ip_tstamps), GFP_KERNEL); |
3027 | if (!ip_tstamps) | |
3028 | panic("IP: failed to allocate ip_tstamps\n"); | |
3029 | ||
5055c371 ED |
3030 | for_each_possible_cpu(cpu) { |
3031 | struct uncached_list *ul = &per_cpu(rt_uncached_list, cpu); | |
3032 | ||
3033 | INIT_LIST_HEAD(&ul->head); | |
3034 | spin_lock_init(&ul->lock); | |
3035 | } | |
c7066f70 | 3036 | #ifdef CONFIG_IP_ROUTE_CLASSID |
0dcec8c2 | 3037 | ip_rt_acct = __alloc_percpu(256 * sizeof(struct ip_rt_acct), __alignof__(struct ip_rt_acct)); |
1da177e4 LT |
3038 | if (!ip_rt_acct) |
3039 | panic("IP: failed to allocate ip_rt_acct\n"); | |
1da177e4 LT |
3040 | #endif |
3041 | ||
e5d679f3 AD |
3042 | ipv4_dst_ops.kmem_cachep = |
3043 | kmem_cache_create("ip_dst_cache", sizeof(struct rtable), 0, | |
20c2df83 | 3044 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 3045 | |
14e50e57 DM |
3046 | ipv4_dst_blackhole_ops.kmem_cachep = ipv4_dst_ops.kmem_cachep; |
3047 | ||
fc66f95c ED |
3048 | if (dst_entries_init(&ipv4_dst_ops) < 0) |
3049 | panic("IP: failed to allocate ipv4_dst_ops counter\n"); | |
3050 | ||
3051 | if (dst_entries_init(&ipv4_dst_blackhole_ops) < 0) | |
3052 | panic("IP: failed to allocate ipv4_dst_blackhole_ops counter\n"); | |
3053 | ||
89aef892 DM |
3054 | ipv4_dst_ops.gc_thresh = ~0; |
3055 | ip_rt_max_size = INT_MAX; | |
1da177e4 | 3056 | |
1da177e4 LT |
3057 | devinet_init(); |
3058 | ip_fib_init(); | |
3059 | ||
73b38711 | 3060 | if (ip_rt_proc_init()) |
058bd4d2 | 3061 | pr_err("Unable to create route proc files\n"); |
1da177e4 LT |
3062 | #ifdef CONFIG_XFRM |
3063 | xfrm_init(); | |
703fb94e | 3064 | xfrm4_init(); |
1da177e4 | 3065 | #endif |
c7ac8679 | 3066 | rtnl_register(PF_INET, RTM_GETROUTE, inet_rtm_getroute, NULL, NULL); |
63f3444f | 3067 | |
39a23e75 DL |
3068 | #ifdef CONFIG_SYSCTL |
3069 | register_pernet_subsys(&sysctl_route_ops); | |
3070 | #endif | |
3ee94372 | 3071 | register_pernet_subsys(&rt_genid_ops); |
c3426b47 | 3072 | register_pernet_subsys(&ipv4_inetpeer_ops); |
1da177e4 LT |
3073 | return rc; |
3074 | } | |
3075 | ||
a1bc6eb4 | 3076 | #ifdef CONFIG_SYSCTL |
eeb61f71 AV |
3077 | /* |
3078 | * We really need to sanitize the damn ipv4 init order, then all | |
3079 | * this nonsense will go away. | |
3080 | */ | |
3081 | void __init ip_static_sysctl_init(void) | |
3082 | { | |
4e5ca785 | 3083 | register_net_sysctl(&init_net, "net/ipv4/route", ipv4_route_table); |
eeb61f71 | 3084 | } |
a1bc6eb4 | 3085 | #endif |