1 |
1275 |
phoenix |
/*
|
2 |
|
|
* NET3: Implementation of the ICMP protocol layer.
|
3 |
|
|
*
|
4 |
|
|
* Alan Cox, <alan@redhat.com>
|
5 |
|
|
*
|
6 |
|
|
* Version: $Id: icmp.c,v 1.1.1.1 2004-04-15 01:13:46 phoenix Exp $
|
7 |
|
|
*
|
8 |
|
|
* This program is free software; you can redistribute it and/or
|
9 |
|
|
* modify it under the terms of the GNU General Public License
|
10 |
|
|
* as published by the Free Software Foundation; either version
|
11 |
|
|
* 2 of the License, or (at your option) any later version.
|
12 |
|
|
*
|
13 |
|
|
* Some of the function names and the icmp unreach table for this
|
14 |
|
|
* module were derived from [icmp.c 1.0.11 06/02/93] by
|
15 |
|
|
* Ross Biro, Fred N. van Kempen, Mark Evans, Alan Cox, Gerhard Koerting.
|
16 |
|
|
* Other than that this module is a complete rewrite.
|
17 |
|
|
*
|
18 |
|
|
* Fixes:
|
19 |
|
|
* Clemens Fruhwirth : introduce global icmp rate limiting
|
20 |
|
|
* with icmp type masking ability instead
|
21 |
|
|
* of broken per type icmp timeouts.
|
22 |
|
|
* Mike Shaver : RFC1122 checks.
|
23 |
|
|
* Alan Cox : Multicast ping reply as self.
|
24 |
|
|
* Alan Cox : Fix atomicity lockup in ip_build_xmit
|
25 |
|
|
* call.
|
26 |
|
|
* Alan Cox : Added 216,128 byte paths to the MTU
|
27 |
|
|
* code.
|
28 |
|
|
* Martin Mares : RFC1812 checks.
|
29 |
|
|
* Martin Mares : Can be configured to follow redirects
|
30 |
|
|
* if acting as a router _without_ a
|
31 |
|
|
* routing protocol (RFC 1812).
|
32 |
|
|
* Martin Mares : Echo requests may be configured to
|
33 |
|
|
* be ignored (RFC 1812).
|
34 |
|
|
* Martin Mares : Limitation of ICMP error message
|
35 |
|
|
* transmit rate (RFC 1812).
|
36 |
|
|
* Martin Mares : TOS and Precedence set correctly
|
37 |
|
|
* (RFC 1812).
|
38 |
|
|
* Martin Mares : Now copying as much data from the
|
39 |
|
|
* original packet as we can without
|
40 |
|
|
* exceeding 576 bytes (RFC 1812).
|
41 |
|
|
* Willy Konynenberg : Transparent proxying support.
|
42 |
|
|
* Keith Owens : RFC1191 correction for 4.2BSD based
|
43 |
|
|
* path MTU bug.
|
44 |
|
|
* Thomas Quinot : ICMP Dest Unreach codes up to 15 are
|
45 |
|
|
* valid (RFC 1812).
|
46 |
|
|
* Andi Kleen : Check all packet lengths properly
|
47 |
|
|
* and moved all kfree_skb() up to
|
48 |
|
|
* icmp_rcv.
|
49 |
|
|
* Andi Kleen : Move the rate limit bookkeeping
|
50 |
|
|
* into the dest entry and use a token
|
51 |
|
|
* bucket filter (thanks to ANK). Make
|
52 |
|
|
* the rates sysctl configurable.
|
53 |
|
|
* Yu Tianli : Fixed two ugly bugs in icmp_send
|
54 |
|
|
* - IP option length was accounted wrongly
|
55 |
|
|
* - ICMP header length was not accounted at all.
|
56 |
|
|
* Tristan Greaves : Added sysctl option to ignore bogus broadcast
|
57 |
|
|
* responses from broken routers.
|
58 |
|
|
*
|
59 |
|
|
* To Fix:
|
60 |
|
|
*
|
61 |
|
|
* - Should use skb_pull() instead of all the manual checking.
|
62 |
|
|
* This would also greatly simply some upper layer error handlers. --AK
|
63 |
|
|
*
|
64 |
|
|
*/
|
65 |
|
|
|
66 |
|
|
#include <linux/config.h>
|
67 |
|
|
#include <linux/types.h>
|
68 |
|
|
#include <linux/sched.h>
|
69 |
|
|
#include <linux/kernel.h>
|
70 |
|
|
#include <linux/fcntl.h>
|
71 |
|
|
#include <linux/socket.h>
|
72 |
|
|
#include <linux/in.h>
|
73 |
|
|
#include <linux/inet.h>
|
74 |
|
|
#include <linux/netdevice.h>
|
75 |
|
|
#include <linux/string.h>
|
76 |
|
|
#include <linux/netfilter_ipv4.h>
|
77 |
|
|
#include <net/snmp.h>
|
78 |
|
|
#include <net/ip.h>
|
79 |
|
|
#include <net/route.h>
|
80 |
|
|
#include <net/protocol.h>
|
81 |
|
|
#include <net/icmp.h>
|
82 |
|
|
#include <net/tcp.h>
|
83 |
|
|
#include <net/udp.h>
|
84 |
|
|
#include <net/raw.h>
|
85 |
|
|
#include <linux/skbuff.h>
|
86 |
|
|
#include <net/sock.h>
|
87 |
|
|
#include <linux/errno.h>
|
88 |
|
|
#include <linux/timer.h>
|
89 |
|
|
#include <linux/init.h>
|
90 |
|
|
#include <asm/system.h>
|
91 |
|
|
#include <asm/uaccess.h>
|
92 |
|
|
#include <net/checksum.h>
|
93 |
|
|
|
94 |
|
|
/*
|
95 |
|
|
* Build xmit assembly blocks
|
96 |
|
|
*/
|
97 |
|
|
|
98 |
|
|
struct icmp_bxm
|
99 |
|
|
{
|
100 |
|
|
struct sk_buff *skb;
|
101 |
|
|
int offset;
|
102 |
|
|
int data_len;
|
103 |
|
|
|
104 |
|
|
unsigned int csum;
|
105 |
|
|
struct {
|
106 |
|
|
struct icmphdr icmph;
|
107 |
|
|
__u32 times[3];
|
108 |
|
|
} data;
|
109 |
|
|
int head_len;
|
110 |
|
|
struct ip_options replyopts;
|
111 |
|
|
unsigned char optbuf[40];
|
112 |
|
|
};
|
113 |
|
|
|
114 |
|
|
/*
|
115 |
|
|
* Statistics
|
116 |
|
|
*/
|
117 |
|
|
|
118 |
|
|
struct icmp_mib icmp_statistics[NR_CPUS*2];
|
119 |
|
|
|
120 |
|
|
/* An array of errno for error messages from dest unreach. */
|
121 |
|
|
/* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOS_UNREACH and SR_FAIELD MUST be considered 'transient errs'. */
|
122 |
|
|
|
123 |
|
|
struct icmp_err icmp_err_convert[] = {
|
124 |
|
|
{ ENETUNREACH, 0 }, /* ICMP_NET_UNREACH */
|
125 |
|
|
{ EHOSTUNREACH, 0 }, /* ICMP_HOST_UNREACH */
|
126 |
|
|
{ ENOPROTOOPT, 1 }, /* ICMP_PROT_UNREACH */
|
127 |
|
|
{ ECONNREFUSED, 1 }, /* ICMP_PORT_UNREACH */
|
128 |
|
|
{ EMSGSIZE, 0 }, /* ICMP_FRAG_NEEDED */
|
129 |
|
|
{ EOPNOTSUPP, 0 }, /* ICMP_SR_FAILED */
|
130 |
|
|
{ ENETUNREACH, 1 }, /* ICMP_NET_UNKNOWN */
|
131 |
|
|
{ EHOSTDOWN, 1 }, /* ICMP_HOST_UNKNOWN */
|
132 |
|
|
{ ENONET, 1 }, /* ICMP_HOST_ISOLATED */
|
133 |
|
|
{ ENETUNREACH, 1 }, /* ICMP_NET_ANO */
|
134 |
|
|
{ EHOSTUNREACH, 1 }, /* ICMP_HOST_ANO */
|
135 |
|
|
{ ENETUNREACH, 0 }, /* ICMP_NET_UNR_TOS */
|
136 |
|
|
{ EHOSTUNREACH, 0 }, /* ICMP_HOST_UNR_TOS */
|
137 |
|
|
{ EHOSTUNREACH, 1 }, /* ICMP_PKT_FILTERED */
|
138 |
|
|
{ EHOSTUNREACH, 1 }, /* ICMP_PREC_VIOLATION */
|
139 |
|
|
{ EHOSTUNREACH, 1 } /* ICMP_PREC_CUTOFF */
|
140 |
|
|
};
|
141 |
|
|
|
142 |
|
|
extern int sysctl_ip_default_ttl;
|
143 |
|
|
|
144 |
|
|
/* Control parameters for ECHO replies. */
|
145 |
|
|
int sysctl_icmp_echo_ignore_all;
|
146 |
|
|
int sysctl_icmp_echo_ignore_broadcasts;
|
147 |
|
|
|
148 |
|
|
/* Control parameter - ignore bogus broadcast responses? */
|
149 |
|
|
int sysctl_icmp_ignore_bogus_error_responses;
|
150 |
|
|
|
151 |
|
|
/*
|
152 |
|
|
* Configurable global rate limit.
|
153 |
|
|
*
|
154 |
|
|
* ratelimit defines tokens/packet consumed for dst->rate_token bucket
|
155 |
|
|
* ratemask defines which icmp types are ratelimited by setting
|
156 |
|
|
* it's bit position.
|
157 |
|
|
*
|
158 |
|
|
* default:
|
159 |
|
|
* dest unreachable (3), source quench (4),
|
160 |
|
|
* time exceeded (11), parameter problem (12)
|
161 |
|
|
*/
|
162 |
|
|
|
163 |
|
|
int sysctl_icmp_ratelimit = 1*HZ;
|
164 |
|
|
int sysctl_icmp_ratemask = 0x1818;
|
165 |
|
|
|
166 |
|
|
/*
|
167 |
|
|
* ICMP control array. This specifies what to do with each ICMP.
|
168 |
|
|
*/
|
169 |
|
|
|
170 |
|
|
struct icmp_control
|
171 |
|
|
{
|
172 |
|
|
unsigned long *output; /* Address to increment on output */
|
173 |
|
|
unsigned long *input; /* Address to increment on input */
|
174 |
|
|
void (*handler)(struct sk_buff *skb);
|
175 |
|
|
short error; /* This ICMP is classed as an error message */
|
176 |
|
|
};
|
177 |
|
|
|
178 |
|
|
static struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];
|
179 |
|
|
|
180 |
|
|
/*
|
181 |
|
|
* The ICMP socket(s). This is the most convenient way to flow control
|
182 |
|
|
* our ICMP output as well as maintain a clean interface throughout
|
183 |
|
|
* all layers. All Socketless IP sends will soon be gone.
|
184 |
|
|
*/
|
185 |
|
|
|
186 |
|
|
static struct inode __icmp_inode[NR_CPUS];
|
187 |
|
|
#define icmp_socket (&__icmp_inode[smp_processor_id()].u.socket_i)
|
188 |
|
|
#define icmp_socket_cpu(X) (&__icmp_inode[(X)].u.socket_i)
|
189 |
|
|
|
190 |
|
|
static int icmp_xmit_lock(void)
|
191 |
|
|
{
|
192 |
|
|
local_bh_disable();
|
193 |
|
|
if (unlikely(!spin_trylock(&icmp_socket->sk->lock.slock))) {
|
194 |
|
|
/* This can happen if the output path signals a
|
195 |
|
|
* dst_link_failure() for an outgoing ICMP packet.
|
196 |
|
|
*/
|
197 |
|
|
local_bh_enable();
|
198 |
|
|
return 1;
|
199 |
|
|
}
|
200 |
|
|
return 0;
|
201 |
|
|
}
|
202 |
|
|
|
203 |
|
|
static void icmp_xmit_unlock(void)
|
204 |
|
|
{
|
205 |
|
|
spin_unlock_bh(&icmp_socket->sk->lock.slock);
|
206 |
|
|
}
|
207 |
|
|
|
208 |
|
|
/*
|
209 |
|
|
* Send an ICMP frame.
|
210 |
|
|
*/
|
211 |
|
|
|
212 |
|
|
/*
|
213 |
|
|
* Check transmit rate limitation for given message.
|
214 |
|
|
* The rate information is held in the destination cache now.
|
215 |
|
|
* This function is generic and could be used for other purposes
|
216 |
|
|
* too. It uses a Token bucket filter as suggested by Alexey Kuznetsov.
|
217 |
|
|
*
|
218 |
|
|
* Note that the same dst_entry fields are modified by functions in
|
219 |
|
|
* route.c too, but these work for packet destinations while xrlim_allow
|
220 |
|
|
* works for icmp destinations. This means the rate limiting information
|
221 |
|
|
* for one "ip object" is shared - and these ICMPs are twice limited:
|
222 |
|
|
* by source and by destination.
|
223 |
|
|
*
|
224 |
|
|
* RFC 1812: 4.3.2.8 SHOULD be able to limit error message rate
|
225 |
|
|
* SHOULD allow setting of rate limits
|
226 |
|
|
*
|
227 |
|
|
* Shared between ICMPv4 and ICMPv6.
|
228 |
|
|
*/
|
229 |
|
|
#define XRLIM_BURST_FACTOR 6
|
230 |
|
|
int xrlim_allow(struct dst_entry *dst, int timeout)
|
231 |
|
|
{
|
232 |
|
|
unsigned long now;
|
233 |
|
|
|
234 |
|
|
now = jiffies;
|
235 |
|
|
dst->rate_tokens += now - dst->rate_last;
|
236 |
|
|
dst->rate_last = now;
|
237 |
|
|
if (dst->rate_tokens > XRLIM_BURST_FACTOR*timeout)
|
238 |
|
|
dst->rate_tokens = XRLIM_BURST_FACTOR*timeout;
|
239 |
|
|
if (dst->rate_tokens >= timeout) {
|
240 |
|
|
dst->rate_tokens -= timeout;
|
241 |
|
|
return 1;
|
242 |
|
|
}
|
243 |
|
|
return 0;
|
244 |
|
|
}
|
245 |
|
|
|
246 |
|
|
static inline int icmpv4_xrlim_allow(struct rtable *rt, int type, int code)
|
247 |
|
|
{
|
248 |
|
|
struct dst_entry *dst = &rt->u.dst;
|
249 |
|
|
|
250 |
|
|
if (type > NR_ICMP_TYPES)
|
251 |
|
|
return 1;
|
252 |
|
|
|
253 |
|
|
/* Don't limit PMTU discovery. */
|
254 |
|
|
if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
|
255 |
|
|
return 1;
|
256 |
|
|
|
257 |
|
|
/* No rate limit on loopback */
|
258 |
|
|
if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
|
259 |
|
|
return 1;
|
260 |
|
|
|
261 |
|
|
/* Limit if icmp type is enabled in ratemask. */
|
262 |
|
|
if((1 << type) & sysctl_icmp_ratemask)
|
263 |
|
|
return xrlim_allow(dst, sysctl_icmp_ratelimit);
|
264 |
|
|
else
|
265 |
|
|
return 1;
|
266 |
|
|
}
|
267 |
|
|
|
268 |
|
|
/*
|
269 |
|
|
* Maintain the counters used in the SNMP statistics for outgoing ICMP
|
270 |
|
|
*/
|
271 |
|
|
|
272 |
|
|
static void icmp_out_count(int type)
|
273 |
|
|
{
|
274 |
|
|
if (type>NR_ICMP_TYPES)
|
275 |
|
|
return;
|
276 |
|
|
(icmp_pointers[type].output)[(smp_processor_id()*2+!in_softirq())*sizeof(struct icmp_mib)/sizeof(unsigned long)]++;
|
277 |
|
|
ICMP_INC_STATS(IcmpOutMsgs);
|
278 |
|
|
}
|
279 |
|
|
|
280 |
|
|
/*
|
281 |
|
|
* Checksum each fragment, and on the first include the headers and final checksum.
|
282 |
|
|
*/
|
283 |
|
|
|
284 |
|
|
static int icmp_glue_bits(const void *p, char *to, unsigned int offset, unsigned int fraglen)
|
285 |
|
|
{
|
286 |
|
|
struct icmp_bxm *icmp_param = (struct icmp_bxm *)p;
|
287 |
|
|
struct icmphdr *icmph;
|
288 |
|
|
unsigned int csum;
|
289 |
|
|
|
290 |
|
|
if (offset) {
|
291 |
|
|
icmp_param->csum=skb_copy_and_csum_bits(icmp_param->skb,
|
292 |
|
|
icmp_param->offset+(offset-icmp_param->head_len),
|
293 |
|
|
to, fraglen,icmp_param->csum);
|
294 |
|
|
return 0;
|
295 |
|
|
}
|
296 |
|
|
|
297 |
|
|
/*
|
298 |
|
|
* First fragment includes header. Note that we've done
|
299 |
|
|
* the other fragments first, so that we get the checksum
|
300 |
|
|
* for the whole packet here.
|
301 |
|
|
*/
|
302 |
|
|
csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
|
303 |
|
|
to, icmp_param->head_len,
|
304 |
|
|
icmp_param->csum);
|
305 |
|
|
csum=skb_copy_and_csum_bits(icmp_param->skb,
|
306 |
|
|
icmp_param->offset,
|
307 |
|
|
to+icmp_param->head_len,
|
308 |
|
|
fraglen-icmp_param->head_len,
|
309 |
|
|
csum);
|
310 |
|
|
icmph=(struct icmphdr *)to;
|
311 |
|
|
icmph->checksum = csum_fold(csum);
|
312 |
|
|
return 0;
|
313 |
|
|
}
|
314 |
|
|
|
315 |
|
|
/*
|
316 |
|
|
* Driving logic for building and sending ICMP messages.
|
317 |
|
|
*/
|
318 |
|
|
|
319 |
|
|
static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
|
320 |
|
|
{
|
321 |
|
|
struct sock *sk=icmp_socket->sk;
|
322 |
|
|
struct ipcm_cookie ipc;
|
323 |
|
|
struct rtable *rt = (struct rtable*)skb->dst;
|
324 |
|
|
u32 daddr;
|
325 |
|
|
|
326 |
|
|
if (ip_options_echo(&icmp_param->replyopts, skb))
|
327 |
|
|
return;
|
328 |
|
|
|
329 |
|
|
if (icmp_xmit_lock())
|
330 |
|
|
return;
|
331 |
|
|
|
332 |
|
|
icmp_param->data.icmph.checksum=0;
|
333 |
|
|
icmp_param->csum=0;
|
334 |
|
|
icmp_out_count(icmp_param->data.icmph.type);
|
335 |
|
|
|
336 |
|
|
sk->protinfo.af_inet.tos = skb->nh.iph->tos;
|
337 |
|
|
sk->protinfo.af_inet.ttl = sysctl_ip_default_ttl;
|
338 |
|
|
daddr = ipc.addr = rt->rt_src;
|
339 |
|
|
ipc.opt = NULL;
|
340 |
|
|
if (icmp_param->replyopts.optlen) {
|
341 |
|
|
ipc.opt = &icmp_param->replyopts;
|
342 |
|
|
if (ipc.opt->srr)
|
343 |
|
|
daddr = icmp_param->replyopts.faddr;
|
344 |
|
|
}
|
345 |
|
|
if (ip_route_output(&rt, daddr, rt->rt_spec_dst, RT_TOS(skb->nh.iph->tos), 0))
|
346 |
|
|
goto out;
|
347 |
|
|
if (icmpv4_xrlim_allow(rt, icmp_param->data.icmph.type,
|
348 |
|
|
icmp_param->data.icmph.code)) {
|
349 |
|
|
ip_build_xmit(sk, icmp_glue_bits, icmp_param,
|
350 |
|
|
icmp_param->data_len+icmp_param->head_len,
|
351 |
|
|
&ipc, rt, MSG_DONTWAIT);
|
352 |
|
|
}
|
353 |
|
|
ip_rt_put(rt);
|
354 |
|
|
out:
|
355 |
|
|
icmp_xmit_unlock();
|
356 |
|
|
}
|
357 |
|
|
|
358 |
|
|
|
359 |
|
|
/*
|
360 |
|
|
* Send an ICMP message in response to a situation
|
361 |
|
|
*
|
362 |
|
|
* RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header. MAY send more (we do).
|
363 |
|
|
* MUST NOT change this header information.
|
364 |
|
|
* MUST NOT reply to a multicast/broadcast IP address.
|
365 |
|
|
* MUST NOT reply to a multicast/broadcast MAC address.
|
366 |
|
|
* MUST reply to only the first fragment.
|
367 |
|
|
*/
|
368 |
|
|
|
369 |
|
|
void icmp_send(struct sk_buff *skb_in, int type, int code, u32 info)
|
370 |
|
|
{
|
371 |
|
|
struct iphdr *iph;
|
372 |
|
|
int room;
|
373 |
|
|
struct icmp_bxm icmp_param;
|
374 |
|
|
struct rtable *rt = (struct rtable*)skb_in->dst;
|
375 |
|
|
struct ipcm_cookie ipc;
|
376 |
|
|
u32 saddr;
|
377 |
|
|
u8 tos;
|
378 |
|
|
|
379 |
|
|
if (!rt)
|
380 |
|
|
return;
|
381 |
|
|
|
382 |
|
|
/*
|
383 |
|
|
* Find the original header. It is expected to be valid, of course.
|
384 |
|
|
* Check this, icmp_send is called from the most obscure devices
|
385 |
|
|
* sometimes.
|
386 |
|
|
*/
|
387 |
|
|
iph = skb_in->nh.iph;
|
388 |
|
|
|
389 |
|
|
if ((u8*)iph < skb_in->head || (u8*)(iph+1) > skb_in->tail)
|
390 |
|
|
return;
|
391 |
|
|
|
392 |
|
|
/*
|
393 |
|
|
* No replies to physical multicast/broadcast
|
394 |
|
|
*/
|
395 |
|
|
if (skb_in->pkt_type!=PACKET_HOST)
|
396 |
|
|
return;
|
397 |
|
|
|
398 |
|
|
/*
|
399 |
|
|
* Now check at the protocol level
|
400 |
|
|
*/
|
401 |
|
|
if (rt->rt_flags&(RTCF_BROADCAST|RTCF_MULTICAST))
|
402 |
|
|
return;
|
403 |
|
|
|
404 |
|
|
/*
|
405 |
|
|
* Only reply to fragment 0. We byte re-order the constant
|
406 |
|
|
* mask for efficiency.
|
407 |
|
|
*/
|
408 |
|
|
if (iph->frag_off&htons(IP_OFFSET))
|
409 |
|
|
return;
|
410 |
|
|
|
411 |
|
|
/*
|
412 |
|
|
* If we send an ICMP error to an ICMP error a mess would result..
|
413 |
|
|
*/
|
414 |
|
|
if (icmp_pointers[type].error) {
|
415 |
|
|
/*
|
416 |
|
|
* We are an error, check if we are replying to an ICMP error
|
417 |
|
|
*/
|
418 |
|
|
if (iph->protocol==IPPROTO_ICMP) {
|
419 |
|
|
u8 inner_type;
|
420 |
|
|
|
421 |
|
|
if (skb_copy_bits(skb_in,
|
422 |
|
|
skb_in->nh.raw + (iph->ihl<<2)
|
423 |
|
|
+ offsetof(struct icmphdr, type)
|
424 |
|
|
- skb_in->data,
|
425 |
|
|
&inner_type, 1))
|
426 |
|
|
return;
|
427 |
|
|
|
428 |
|
|
/*
|
429 |
|
|
* Assume any unknown ICMP type is an error. This isn't
|
430 |
|
|
* specified by the RFC, but think about it..
|
431 |
|
|
*/
|
432 |
|
|
if (inner_type>NR_ICMP_TYPES || icmp_pointers[inner_type].error)
|
433 |
|
|
return;
|
434 |
|
|
}
|
435 |
|
|
}
|
436 |
|
|
|
437 |
|
|
if (icmp_xmit_lock())
|
438 |
|
|
return;
|
439 |
|
|
|
440 |
|
|
/*
|
441 |
|
|
* Construct source address and options.
|
442 |
|
|
*/
|
443 |
|
|
|
444 |
|
|
#ifdef CONFIG_IP_ROUTE_NAT
|
445 |
|
|
/*
|
446 |
|
|
* Restore original addresses if packet has been translated.
|
447 |
|
|
*/
|
448 |
|
|
if (rt->rt_flags&RTCF_NAT && IPCB(skb_in)->flags&IPSKB_TRANSLATED) {
|
449 |
|
|
iph->daddr = rt->key.dst;
|
450 |
|
|
iph->saddr = rt->key.src;
|
451 |
|
|
}
|
452 |
|
|
#endif
|
453 |
|
|
|
454 |
|
|
saddr = iph->daddr;
|
455 |
|
|
if (!(rt->rt_flags & RTCF_LOCAL))
|
456 |
|
|
saddr = 0;
|
457 |
|
|
|
458 |
|
|
tos = icmp_pointers[type].error ?
|
459 |
|
|
((iph->tos & IPTOS_TOS_MASK) | IPTOS_PREC_INTERNETCONTROL) :
|
460 |
|
|
iph->tos;
|
461 |
|
|
|
462 |
|
|
if (ip_route_output(&rt, iph->saddr, saddr, RT_TOS(tos), 0))
|
463 |
|
|
goto out;
|
464 |
|
|
|
465 |
|
|
if (ip_options_echo(&icmp_param.replyopts, skb_in))
|
466 |
|
|
goto ende;
|
467 |
|
|
|
468 |
|
|
|
469 |
|
|
/*
|
470 |
|
|
* Prepare data for ICMP header.
|
471 |
|
|
*/
|
472 |
|
|
|
473 |
|
|
icmp_param.data.icmph.type=type;
|
474 |
|
|
icmp_param.data.icmph.code=code;
|
475 |
|
|
icmp_param.data.icmph.un.gateway = info;
|
476 |
|
|
icmp_param.data.icmph.checksum=0;
|
477 |
|
|
icmp_param.csum=0;
|
478 |
|
|
icmp_param.skb=skb_in;
|
479 |
|
|
icmp_param.offset=skb_in->nh.raw - skb_in->data;
|
480 |
|
|
icmp_out_count(icmp_param.data.icmph.type);
|
481 |
|
|
icmp_socket->sk->protinfo.af_inet.tos = tos;
|
482 |
|
|
icmp_socket->sk->protinfo.af_inet.ttl = sysctl_ip_default_ttl;
|
483 |
|
|
ipc.addr = iph->saddr;
|
484 |
|
|
ipc.opt = &icmp_param.replyopts;
|
485 |
|
|
if (icmp_param.replyopts.srr) {
|
486 |
|
|
ip_rt_put(rt);
|
487 |
|
|
if (ip_route_output(&rt, icmp_param.replyopts.faddr, saddr, RT_TOS(tos), 0))
|
488 |
|
|
goto out;
|
489 |
|
|
}
|
490 |
|
|
|
491 |
|
|
if (!icmpv4_xrlim_allow(rt, type, code))
|
492 |
|
|
goto ende;
|
493 |
|
|
|
494 |
|
|
/* RFC says return as much as we can without exceeding 576 bytes. */
|
495 |
|
|
|
496 |
|
|
room = rt->u.dst.pmtu;
|
497 |
|
|
if (room > 576)
|
498 |
|
|
room = 576;
|
499 |
|
|
room -= sizeof(struct iphdr) + icmp_param.replyopts.optlen;
|
500 |
|
|
room -= sizeof(struct icmphdr);
|
501 |
|
|
|
502 |
|
|
icmp_param.data_len=skb_in->len-icmp_param.offset;
|
503 |
|
|
if (icmp_param.data_len > room)
|
504 |
|
|
icmp_param.data_len = room;
|
505 |
|
|
icmp_param.head_len = sizeof(struct icmphdr);
|
506 |
|
|
|
507 |
|
|
ip_build_xmit(icmp_socket->sk, icmp_glue_bits, &icmp_param,
|
508 |
|
|
icmp_param.data_len+sizeof(struct icmphdr),
|
509 |
|
|
&ipc, rt, MSG_DONTWAIT);
|
510 |
|
|
|
511 |
|
|
ende:
|
512 |
|
|
ip_rt_put(rt);
|
513 |
|
|
out:
|
514 |
|
|
icmp_xmit_unlock();
|
515 |
|
|
}
|
516 |
|
|
|
517 |
|
|
|
518 |
|
|
/*
|
519 |
|
|
* Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH.
|
520 |
|
|
*/
|
521 |
|
|
|
522 |
|
|
static void icmp_unreach(struct sk_buff *skb)
|
523 |
|
|
{
|
524 |
|
|
struct iphdr *iph;
|
525 |
|
|
struct icmphdr *icmph;
|
526 |
|
|
int hash, protocol;
|
527 |
|
|
struct inet_protocol *ipprot;
|
528 |
|
|
struct sock *raw_sk;
|
529 |
|
|
u32 info = 0;
|
530 |
|
|
|
531 |
|
|
/*
|
532 |
|
|
* Incomplete header ?
|
533 |
|
|
* Only checks for the IP header, there should be an
|
534 |
|
|
* additional check for longer headers in upper levels.
|
535 |
|
|
*/
|
536 |
|
|
|
537 |
|
|
if (!pskb_may_pull(skb, sizeof(struct iphdr))) {
|
538 |
|
|
ICMP_INC_STATS_BH(IcmpInErrors);
|
539 |
|
|
return;
|
540 |
|
|
}
|
541 |
|
|
|
542 |
|
|
icmph = skb->h.icmph;
|
543 |
|
|
iph = (struct iphdr *) skb->data;
|
544 |
|
|
|
545 |
|
|
if (iph->ihl<5) {
|
546 |
|
|
/* Mangled header, drop. */
|
547 |
|
|
ICMP_INC_STATS_BH(IcmpInErrors);
|
548 |
|
|
return;
|
549 |
|
|
}
|
550 |
|
|
|
551 |
|
|
if(icmph->type==ICMP_DEST_UNREACH) {
|
552 |
|
|
switch(icmph->code & 15) {
|
553 |
|
|
case ICMP_NET_UNREACH:
|
554 |
|
|
break;
|
555 |
|
|
case ICMP_HOST_UNREACH:
|
556 |
|
|
break;
|
557 |
|
|
case ICMP_PROT_UNREACH:
|
558 |
|
|
break;
|
559 |
|
|
case ICMP_PORT_UNREACH:
|
560 |
|
|
break;
|
561 |
|
|
case ICMP_FRAG_NEEDED:
|
562 |
|
|
if (ipv4_config.no_pmtu_disc) {
|
563 |
|
|
if (net_ratelimit())
|
564 |
|
|
printk(KERN_INFO "ICMP: %u.%u.%u.%u: fragmentation needed and DF set.\n",
|
565 |
|
|
NIPQUAD(iph->daddr));
|
566 |
|
|
} else {
|
567 |
|
|
info = ip_rt_frag_needed(iph, ntohs(icmph->un.frag.mtu));
|
568 |
|
|
if (!info)
|
569 |
|
|
goto out;
|
570 |
|
|
}
|
571 |
|
|
break;
|
572 |
|
|
case ICMP_SR_FAILED:
|
573 |
|
|
if (net_ratelimit())
|
574 |
|
|
printk(KERN_INFO "ICMP: %u.%u.%u.%u: Source Route Failed.\n", NIPQUAD(iph->daddr));
|
575 |
|
|
break;
|
576 |
|
|
default:
|
577 |
|
|
break;
|
578 |
|
|
}
|
579 |
|
|
if (icmph->code>NR_ICMP_UNREACH)
|
580 |
|
|
goto out;
|
581 |
|
|
} else if (icmph->type == ICMP_PARAMETERPROB) {
|
582 |
|
|
info = ntohl(icmph->un.gateway)>>24;
|
583 |
|
|
}
|
584 |
|
|
|
585 |
|
|
/*
|
586 |
|
|
* Throw it at our lower layers
|
587 |
|
|
*
|
588 |
|
|
* RFC 1122: 3.2.2 MUST extract the protocol ID from the passed header.
|
589 |
|
|
* RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the transport layer.
|
590 |
|
|
* RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to transport layer.
|
591 |
|
|
*/
|
592 |
|
|
|
593 |
|
|
/*
|
594 |
|
|
* Check the other end isnt violating RFC 1122. Some routers send
|
595 |
|
|
* bogus responses to broadcast frames. If you see this message
|
596 |
|
|
* first check your netmask matches at both ends, if it does then
|
597 |
|
|
* get the other vendor to fix their kit.
|
598 |
|
|
*/
|
599 |
|
|
|
600 |
|
|
if (!sysctl_icmp_ignore_bogus_error_responses)
|
601 |
|
|
{
|
602 |
|
|
|
603 |
|
|
if (inet_addr_type(iph->daddr) == RTN_BROADCAST)
|
604 |
|
|
{
|
605 |
|
|
if (net_ratelimit())
|
606 |
|
|
printk(KERN_WARNING "%u.%u.%u.%u sent an invalid ICMP type %u, code %u error to a broadcast: %u.%u.%u.%u on %s\n",
|
607 |
|
|
NIPQUAD(skb->nh.iph->saddr),
|
608 |
|
|
icmph->type, icmph->code,
|
609 |
|
|
NIPQUAD(iph->daddr),
|
610 |
|
|
skb->dev->name);
|
611 |
|
|
goto out;
|
612 |
|
|
}
|
613 |
|
|
}
|
614 |
|
|
|
615 |
|
|
/* Checkin full IP header plus 8 bytes of protocol to
|
616 |
|
|
* avoid additional coding at protocol handlers.
|
617 |
|
|
*/
|
618 |
|
|
if (!pskb_may_pull(skb, iph->ihl*4+8))
|
619 |
|
|
goto out;
|
620 |
|
|
|
621 |
|
|
iph = (struct iphdr *) skb->data;
|
622 |
|
|
protocol = iph->protocol;
|
623 |
|
|
|
624 |
|
|
/*
|
625 |
|
|
* Deliver ICMP message to raw sockets. Pretty useless feature?
|
626 |
|
|
*/
|
627 |
|
|
|
628 |
|
|
/* Note: See raw.c and net/raw.h, RAWV4_HTABLE_SIZE==MAX_INET_PROTOS */
|
629 |
|
|
hash = protocol & (MAX_INET_PROTOS - 1);
|
630 |
|
|
read_lock(&raw_v4_lock);
|
631 |
|
|
if ((raw_sk = raw_v4_htable[hash]) != NULL)
|
632 |
|
|
{
|
633 |
|
|
while ((raw_sk = __raw_v4_lookup(raw_sk, protocol, iph->daddr,
|
634 |
|
|
iph->saddr, skb->dev->ifindex)) != NULL) {
|
635 |
|
|
raw_err(raw_sk, skb, info);
|
636 |
|
|
raw_sk = raw_sk->next;
|
637 |
|
|
iph = (struct iphdr *)skb->data;
|
638 |
|
|
}
|
639 |
|
|
}
|
640 |
|
|
read_unlock(&raw_v4_lock);
|
641 |
|
|
|
642 |
|
|
/*
|
643 |
|
|
* This can't change while we are doing it.
|
644 |
|
|
* Callers have obtained BR_NETPROTO_LOCK so
|
645 |
|
|
* we are OK.
|
646 |
|
|
*/
|
647 |
|
|
|
648 |
|
|
ipprot = (struct inet_protocol *) inet_protos[hash];
|
649 |
|
|
while (ipprot) {
|
650 |
|
|
struct inet_protocol *nextip;
|
651 |
|
|
|
652 |
|
|
nextip = (struct inet_protocol *) ipprot->next;
|
653 |
|
|
|
654 |
|
|
/*
|
655 |
|
|
* Pass it off to everyone who wants it.
|
656 |
|
|
*/
|
657 |
|
|
|
658 |
|
|
/* RFC1122: OK. Passes appropriate ICMP errors to the */
|
659 |
|
|
/* appropriate protocol layer (MUST), as per 3.2.2. */
|
660 |
|
|
|
661 |
|
|
if (protocol == ipprot->protocol && ipprot->err_handler)
|
662 |
|
|
ipprot->err_handler(skb, info);
|
663 |
|
|
|
664 |
|
|
ipprot = nextip;
|
665 |
|
|
}
|
666 |
|
|
out:;
|
667 |
|
|
}
|
668 |
|
|
|
669 |
|
|
|
670 |
|
|
/*
|
671 |
|
|
* Handle ICMP_REDIRECT.
|
672 |
|
|
*/
|
673 |
|
|
|
674 |
|
|
static void icmp_redirect(struct sk_buff *skb)
|
675 |
|
|
{
|
676 |
|
|
struct iphdr *iph;
|
677 |
|
|
unsigned long ip;
|
678 |
|
|
|
679 |
|
|
if (skb->len < sizeof(struct iphdr)) {
|
680 |
|
|
ICMP_INC_STATS_BH(IcmpInErrors);
|
681 |
|
|
return;
|
682 |
|
|
}
|
683 |
|
|
|
684 |
|
|
/*
|
685 |
|
|
* Get the copied header of the packet that caused the redirect
|
686 |
|
|
*/
|
687 |
|
|
if (!pskb_may_pull(skb, sizeof(struct iphdr)))
|
688 |
|
|
return;
|
689 |
|
|
|
690 |
|
|
iph = (struct iphdr *) skb->data;
|
691 |
|
|
ip = iph->daddr;
|
692 |
|
|
|
693 |
|
|
switch (skb->h.icmph->code & 7) {
|
694 |
|
|
case ICMP_REDIR_NET:
|
695 |
|
|
case ICMP_REDIR_NETTOS:
|
696 |
|
|
/*
|
697 |
|
|
* As per RFC recommendations now handle it as
|
698 |
|
|
* a host redirect.
|
699 |
|
|
*/
|
700 |
|
|
|
701 |
|
|
case ICMP_REDIR_HOST:
|
702 |
|
|
case ICMP_REDIR_HOSTTOS:
|
703 |
|
|
ip_rt_redirect(skb->nh.iph->saddr, ip, skb->h.icmph->un.gateway, iph->saddr, iph->tos, skb->dev);
|
704 |
|
|
break;
|
705 |
|
|
default:
|
706 |
|
|
break;
|
707 |
|
|
}
|
708 |
|
|
}
|
709 |
|
|
|
710 |
|
|
/*
|
711 |
|
|
* Handle ICMP_ECHO ("ping") requests.
|
712 |
|
|
*
|
713 |
|
|
* RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo requests.
|
714 |
|
|
* RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be included in the reply.
|
715 |
|
|
* RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring echo requests, MUST have default=NOT.
|
716 |
|
|
* See also WRT handling of options once they are done and working.
|
717 |
|
|
*/
|
718 |
|
|
|
719 |
|
|
static void icmp_echo(struct sk_buff *skb)
|
720 |
|
|
{
|
721 |
|
|
if (!sysctl_icmp_echo_ignore_all) {
|
722 |
|
|
struct icmp_bxm icmp_param;
|
723 |
|
|
|
724 |
|
|
icmp_param.data.icmph=*skb->h.icmph;
|
725 |
|
|
icmp_param.data.icmph.type=ICMP_ECHOREPLY;
|
726 |
|
|
icmp_param.skb=skb;
|
727 |
|
|
icmp_param.offset=0;
|
728 |
|
|
icmp_param.data_len=skb->len;
|
729 |
|
|
icmp_param.head_len=sizeof(struct icmphdr);
|
730 |
|
|
icmp_reply(&icmp_param, skb);
|
731 |
|
|
}
|
732 |
|
|
}
|
733 |
|
|
|
734 |
|
|
/*
|
735 |
|
|
* Handle ICMP Timestamp requests.
|
736 |
|
|
* RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
|
737 |
|
|
* SHOULD be in the kernel for minimum random latency.
|
738 |
|
|
* MUST be accurate to a few minutes.
|
739 |
|
|
* MUST be updated at least at 15Hz.
|
740 |
|
|
*/
|
741 |
|
|
|
742 |
|
|
static void icmp_timestamp(struct sk_buff *skb)
|
743 |
|
|
{
|
744 |
|
|
struct timeval tv;
|
745 |
|
|
struct icmp_bxm icmp_param;
|
746 |
|
|
|
747 |
|
|
/*
|
748 |
|
|
* Too short.
|
749 |
|
|
*/
|
750 |
|
|
|
751 |
|
|
if (skb->len < 4) {
|
752 |
|
|
ICMP_INC_STATS_BH(IcmpInErrors);
|
753 |
|
|
return;
|
754 |
|
|
}
|
755 |
|
|
|
756 |
|
|
/*
|
757 |
|
|
* Fill in the current time as ms since midnight UT:
|
758 |
|
|
*/
|
759 |
|
|
do_gettimeofday(&tv);
|
760 |
|
|
icmp_param.data.times[1] = htonl((tv.tv_sec % 86400) * 1000 + tv.tv_usec / 1000);
|
761 |
|
|
icmp_param.data.times[2] = icmp_param.data.times[1];
|
762 |
|
|
if (skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4))
|
763 |
|
|
BUG();
|
764 |
|
|
icmp_param.data.icmph=*skb->h.icmph;
|
765 |
|
|
icmp_param.data.icmph.type=ICMP_TIMESTAMPREPLY;
|
766 |
|
|
icmp_param.data.icmph.code=0;
|
767 |
|
|
icmp_param.skb=skb;
|
768 |
|
|
icmp_param.offset=0;
|
769 |
|
|
icmp_param.data_len=0;
|
770 |
|
|
icmp_param.head_len=sizeof(struct icmphdr)+12;
|
771 |
|
|
icmp_reply(&icmp_param, skb);
|
772 |
|
|
}
|
773 |
|
|
|
774 |
|
|
|
775 |
|
|
/*
|
776 |
|
|
* Handle ICMP_ADDRESS_MASK requests. (RFC950)
|
777 |
|
|
*
|
778 |
|
|
* RFC1122 (3.2.2.9). A host MUST only send replies to
|
779 |
|
|
* ADDRESS_MASK requests if it's been configured as an address mask
|
780 |
|
|
* agent. Receiving a request doesn't constitute implicit permission to
|
781 |
|
|
* act as one. Of course, implementing this correctly requires (SHOULD)
|
782 |
|
|
* a way to turn the functionality on and off. Another one for sysctl(),
|
783 |
|
|
* I guess. -- MS
|
784 |
|
|
*
|
785 |
|
|
* RFC1812 (4.3.3.9). A router MUST implement it.
|
786 |
|
|
* A router SHOULD have switch turning it on/off.
|
787 |
|
|
* This switch MUST be ON by default.
|
788 |
|
|
*
|
789 |
|
|
* Gratuitous replies, zero-source replies are not implemented,
|
790 |
|
|
* that complies with RFC. DO NOT implement them!!! All the idea
|
791 |
|
|
* of broadcast addrmask replies as specified in RFC950 is broken.
|
792 |
|
|
* The problem is that it is not uncommon to have several prefixes
|
793 |
|
|
* on one physical interface. Moreover, addrmask agent can even be
|
794 |
|
|
* not aware of existing another prefixes.
|
795 |
|
|
* If source is zero, addrmask agent cannot choose correct prefix.
|
796 |
|
|
* Gratuitous mask announcements suffer from the same problem.
|
797 |
|
|
* RFC1812 explains it, but still allows to use ADDRMASK,
|
798 |
|
|
* that is pretty silly. --ANK
|
799 |
|
|
*
|
800 |
|
|
* All these rules are so bizarre, that I removed kernel addrmask
|
801 |
|
|
* support at all. It is wrong, it is obsolete, nobody uses it in
|
802 |
|
|
* any case. --ANK
|
803 |
|
|
*
|
804 |
|
|
* Furthermore you can do it with a usermode address agent program
|
805 |
|
|
* anyway...
|
806 |
|
|
*/
|
807 |
|
|
|
808 |
|
|
static void icmp_address(struct sk_buff *skb)
|
809 |
|
|
{
|
810 |
|
|
#if 0
|
811 |
|
|
if (net_ratelimit())
|
812 |
|
|
printk(KERN_DEBUG "a guy asks for address mask. Who is it?\n");
|
813 |
|
|
#endif
|
814 |
|
|
}
|
815 |
|
|
|
816 |
|
|
/*
|
817 |
|
|
* RFC1812 (4.3.3.9). A router SHOULD listen all replies, and complain
|
818 |
|
|
* loudly if an inconsistency is found.
|
819 |
|
|
*/
|
820 |
|
|
|
821 |
|
|
static void icmp_address_reply(struct sk_buff *skb)
|
822 |
|
|
{
|
823 |
|
|
struct rtable *rt = (struct rtable*)skb->dst;
|
824 |
|
|
struct net_device *dev = skb->dev;
|
825 |
|
|
struct in_device *in_dev;
|
826 |
|
|
struct in_ifaddr *ifa;
|
827 |
|
|
u32 mask;
|
828 |
|
|
|
829 |
|
|
if (skb->len < 4 || !(rt->rt_flags&RTCF_DIRECTSRC))
|
830 |
|
|
return;
|
831 |
|
|
|
832 |
|
|
in_dev = in_dev_get(dev);
|
833 |
|
|
if (!in_dev)
|
834 |
|
|
return;
|
835 |
|
|
read_lock(&in_dev->lock);
|
836 |
|
|
if (in_dev->ifa_list &&
|
837 |
|
|
IN_DEV_LOG_MARTIANS(in_dev) &&
|
838 |
|
|
IN_DEV_FORWARD(in_dev)) {
|
839 |
|
|
if (skb_copy_bits(skb, 0, &mask, 4))
|
840 |
|
|
BUG();
|
841 |
|
|
for (ifa=in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
|
842 |
|
|
if (mask == ifa->ifa_mask && inet_ifa_match(rt->rt_src, ifa))
|
843 |
|
|
break;
|
844 |
|
|
}
|
845 |
|
|
if (!ifa && net_ratelimit()) {
|
846 |
|
|
printk(KERN_INFO "Wrong address mask %u.%u.%u.%u from %s/%u.%u.%u.%u\n",
|
847 |
|
|
NIPQUAD(mask), dev->name, NIPQUAD(rt->rt_src));
|
848 |
|
|
}
|
849 |
|
|
}
|
850 |
|
|
read_unlock(&in_dev->lock);
|
851 |
|
|
in_dev_put(in_dev);
|
852 |
|
|
}
|
853 |
|
|
|
854 |
|
|
static void icmp_discard(struct sk_buff *skb)
|
855 |
|
|
{
|
856 |
|
|
}
|
857 |
|
|
|
858 |
|
|
/*
|
859 |
|
|
* Deal with incoming ICMP packets.
|
860 |
|
|
*/
|
861 |
|
|
|
862 |
|
|
int icmp_rcv(struct sk_buff *skb)
|
863 |
|
|
{
|
864 |
|
|
struct icmphdr *icmph;
|
865 |
|
|
struct rtable *rt = (struct rtable*)skb->dst;
|
866 |
|
|
|
867 |
|
|
ICMP_INC_STATS_BH(IcmpInMsgs);
|
868 |
|
|
|
869 |
|
|
switch (skb->ip_summed) {
|
870 |
|
|
case CHECKSUM_HW:
|
871 |
|
|
if ((u16)csum_fold(skb->csum) == 0)
|
872 |
|
|
break;
|
873 |
|
|
NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG "icmp v4 hw csum failure\n"));
|
874 |
|
|
case CHECKSUM_NONE:
|
875 |
|
|
if ((u16)csum_fold(skb_checksum(skb, 0, skb->len, 0)))
|
876 |
|
|
goto error;
|
877 |
|
|
default:;
|
878 |
|
|
}
|
879 |
|
|
|
880 |
|
|
if (!pskb_pull(skb, sizeof(struct icmphdr)))
|
881 |
|
|
goto error;
|
882 |
|
|
|
883 |
|
|
icmph = skb->h.icmph;
|
884 |
|
|
|
885 |
|
|
/*
|
886 |
|
|
* 18 is the highest 'known' ICMP type. Anything else is a mystery
|
887 |
|
|
*
|
888 |
|
|
* RFC 1122: 3.2.2 Unknown ICMP messages types MUST be silently discarded.
|
889 |
|
|
*/
|
890 |
|
|
if (icmph->type > NR_ICMP_TYPES)
|
891 |
|
|
goto error;
|
892 |
|
|
|
893 |
|
|
|
894 |
|
|
/*
|
895 |
|
|
* Parse the ICMP message
|
896 |
|
|
*/
|
897 |
|
|
|
898 |
|
|
if (rt->rt_flags&(RTCF_BROADCAST|RTCF_MULTICAST)) {
|
899 |
|
|
/*
|
900 |
|
|
* RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
|
901 |
|
|
* silently ignored (we let user decide with a sysctl).
|
902 |
|
|
* RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
|
903 |
|
|
* discarded if to broadcast/multicast.
|
904 |
|
|
*/
|
905 |
|
|
if (icmph->type == ICMP_ECHO &&
|
906 |
|
|
sysctl_icmp_echo_ignore_broadcasts) {
|
907 |
|
|
goto error;
|
908 |
|
|
}
|
909 |
|
|
if (icmph->type != ICMP_ECHO &&
|
910 |
|
|
icmph->type != ICMP_TIMESTAMP &&
|
911 |
|
|
icmph->type != ICMP_ADDRESS &&
|
912 |
|
|
icmph->type != ICMP_ADDRESSREPLY) {
|
913 |
|
|
goto error;
|
914 |
|
|
}
|
915 |
|
|
}
|
916 |
|
|
|
917 |
|
|
icmp_pointers[icmph->type].input[smp_processor_id()*2*sizeof(struct icmp_mib)/sizeof(unsigned long)]++;
|
918 |
|
|
(icmp_pointers[icmph->type].handler)(skb);
|
919 |
|
|
|
920 |
|
|
drop:
|
921 |
|
|
kfree_skb(skb);
|
922 |
|
|
return 0;
|
923 |
|
|
error:
|
924 |
|
|
ICMP_INC_STATS_BH(IcmpInErrors);
|
925 |
|
|
goto drop;
|
926 |
|
|
}
|
927 |
|
|
|
928 |
|
|
/*
|
929 |
|
|
* This table is the definition of how we handle ICMP.
|
930 |
|
|
*/
|
931 |
|
|
|
932 |
|
|
static struct icmp_control icmp_pointers[NR_ICMP_TYPES+1] = {
|
933 |
|
|
/* ECHO REPLY (0) */
|
934 |
|
|
{ &icmp_statistics[0].IcmpOutEchoReps, &icmp_statistics[0].IcmpInEchoReps, icmp_discard, 0 },
|
935 |
|
|
{ &icmp_statistics[0].dummy, &icmp_statistics[0].IcmpInErrors, icmp_discard, 1 },
|
936 |
|
|
{ &icmp_statistics[0].dummy, &icmp_statistics[0].IcmpInErrors, icmp_discard, 1 },
|
937 |
|
|
/* DEST UNREACH (3) */
|
938 |
|
|
{ &icmp_statistics[0].IcmpOutDestUnreachs, &icmp_statistics[0].IcmpInDestUnreachs, icmp_unreach, 1 },
|
939 |
|
|
/* SOURCE QUENCH (4) */
|
940 |
|
|
{ &icmp_statistics[0].IcmpOutSrcQuenchs, &icmp_statistics[0].IcmpInSrcQuenchs, icmp_unreach, 1 },
|
941 |
|
|
/* REDIRECT (5) */
|
942 |
|
|
{ &icmp_statistics[0].IcmpOutRedirects, &icmp_statistics[0].IcmpInRedirects, icmp_redirect, 1 },
|
943 |
|
|
{ &icmp_statistics[0].dummy, &icmp_statistics[0].IcmpInErrors, icmp_discard, 1 },
|
944 |
|
|
{ &icmp_statistics[0].dummy, &icmp_statistics[0].IcmpInErrors, icmp_discard, 1 },
|
945 |
|
|
/* ECHO (8) */
|
946 |
|
|
{ &icmp_statistics[0].IcmpOutEchos, &icmp_statistics[0].IcmpInEchos, icmp_echo, 0 },
|
947 |
|
|
{ &icmp_statistics[0].dummy, &icmp_statistics[0].IcmpInErrors, icmp_discard, 1 },
|
948 |
|
|
{ &icmp_statistics[0].dummy, &icmp_statistics[0].IcmpInErrors, icmp_discard, 1 },
|
949 |
|
|
/* TIME EXCEEDED (11) */
|
950 |
|
|
{ &icmp_statistics[0].IcmpOutTimeExcds, &icmp_statistics[0].IcmpInTimeExcds, icmp_unreach, 1 },
|
951 |
|
|
/* PARAMETER PROBLEM (12) */
|
952 |
|
|
{ &icmp_statistics[0].IcmpOutParmProbs, &icmp_statistics[0].IcmpInParmProbs, icmp_unreach, 1 },
|
953 |
|
|
/* TIMESTAMP (13) */
|
954 |
|
|
{ &icmp_statistics[0].IcmpOutTimestamps, &icmp_statistics[0].IcmpInTimestamps, icmp_timestamp, 0 },
|
955 |
|
|
/* TIMESTAMP REPLY (14) */
|
956 |
|
|
{ &icmp_statistics[0].IcmpOutTimestampReps, &icmp_statistics[0].IcmpInTimestampReps, icmp_discard, 0 },
|
957 |
|
|
/* INFO (15) */
|
958 |
|
|
{ &icmp_statistics[0].dummy, &icmp_statistics[0].dummy, icmp_discard, 0 },
|
959 |
|
|
/* INFO REPLY (16) */
|
960 |
|
|
{ &icmp_statistics[0].dummy, &icmp_statistics[0].dummy, icmp_discard, 0 },
|
961 |
|
|
/* ADDR MASK (17) */
|
962 |
|
|
{ &icmp_statistics[0].IcmpOutAddrMasks, &icmp_statistics[0].IcmpInAddrMasks, icmp_address, 0 },
|
963 |
|
|
/* ADDR MASK REPLY (18) */
|
964 |
|
|
{ &icmp_statistics[0].IcmpOutAddrMaskReps, &icmp_statistics[0].IcmpInAddrMaskReps, icmp_address_reply, 0 }
|
965 |
|
|
};
|
966 |
|
|
|
967 |
|
|
void __init icmp_init(struct net_proto_family *ops)
|
968 |
|
|
{
|
969 |
|
|
int err, i;
|
970 |
|
|
|
971 |
|
|
for (i = 0; i < NR_CPUS; i++) {
|
972 |
|
|
__icmp_inode[i].i_mode = S_IFSOCK;
|
973 |
|
|
__icmp_inode[i].i_sock = 1;
|
974 |
|
|
__icmp_inode[i].i_uid = 0;
|
975 |
|
|
__icmp_inode[i].i_gid = 0;
|
976 |
|
|
init_waitqueue_head(&__icmp_inode[i].i_wait);
|
977 |
|
|
init_waitqueue_head(&__icmp_inode[i].u.socket_i.wait);
|
978 |
|
|
|
979 |
|
|
icmp_socket_cpu(i)->inode = &__icmp_inode[i];
|
980 |
|
|
icmp_socket_cpu(i)->state = SS_UNCONNECTED;
|
981 |
|
|
icmp_socket_cpu(i)->type = SOCK_RAW;
|
982 |
|
|
|
983 |
|
|
if ((err=ops->create(icmp_socket_cpu(i), IPPROTO_ICMP)) < 0)
|
984 |
|
|
panic("Failed to create the ICMP control socket.\n");
|
985 |
|
|
|
986 |
|
|
icmp_socket_cpu(i)->sk->allocation=GFP_ATOMIC;
|
987 |
|
|
|
988 |
|
|
/* Enough space for 2 64K ICMP packets, including
|
989 |
|
|
* sk_buff struct overhead.
|
990 |
|
|
*/
|
991 |
|
|
icmp_socket_cpu(i)->sk->sndbuf =
|
992 |
|
|
(2 * ((64 * 1024) + sizeof(struct sk_buff)));
|
993 |
|
|
|
994 |
|
|
icmp_socket_cpu(i)->sk->protinfo.af_inet.ttl = MAXTTL;
|
995 |
|
|
icmp_socket_cpu(i)->sk->protinfo.af_inet.pmtudisc = IP_PMTUDISC_DONT;
|
996 |
|
|
|
997 |
|
|
/* Unhash it so that IP input processing does not even
|
998 |
|
|
* see it, we do not wish this socket to see incoming
|
999 |
|
|
* packets.
|
1000 |
|
|
*/
|
1001 |
|
|
icmp_socket_cpu(i)->sk->prot->unhash(icmp_socket_cpu(i)->sk);
|
1002 |
|
|
}
|
1003 |
|
|
}
|