Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / net / ipv4 / netfilter / nf_conntrack_proto_icmp.c
1 /* (C) 1999-2001 Paul `Rusty' Russell
2  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
3  * (C) 2006-2010 Patrick McHardy <kaber@trash.net>
4  *
5  * This program is free software; you can redistribute it and/or modify
6  * it under the terms of the GNU General Public License version 2 as
7  * published by the Free Software Foundation.
8  */
9
10 #include <linux/types.h>
11 #include <linux/timer.h>
12 #include <linux/netfilter.h>
13 #include <linux/in.h>
14 #include <linux/icmp.h>
15 #include <linux/seq_file.h>
16 #include <net/ip.h>
17 #include <net/checksum.h>
18 #include <linux/netfilter_ipv4.h>
19 #include <net/netfilter/nf_conntrack_tuple.h>
20 #include <net/netfilter/nf_conntrack_l4proto.h>
21 #include <net/netfilter/nf_conntrack_core.h>
22 #include <net/netfilter/nf_conntrack_zones.h>
23 #include <net/netfilter/nf_log.h>
24
25 static unsigned int nf_ct_icmp_timeout __read_mostly = 30*HZ;
26
27 static inline struct nf_icmp_net *icmp_pernet(struct net *net)
28 {
29         return &net->ct.nf_ct_proto.icmp;
30 }
31
32 static bool icmp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff,
33                               struct nf_conntrack_tuple *tuple)
34 {
35         const struct icmphdr *hp;
36         struct icmphdr _hdr;
37
38         hp = skb_header_pointer(skb, dataoff, sizeof(_hdr), &_hdr);
39         if (hp == NULL)
40                 return false;
41
42         tuple->dst.u.icmp.type = hp->type;
43         tuple->src.u.icmp.id = hp->un.echo.id;
44         tuple->dst.u.icmp.code = hp->code;
45
46         return true;
47 }
48
49 /* Add 1; spaces filled with 0. */
50 static const u_int8_t invmap[] = {
51         [ICMP_ECHO] = ICMP_ECHOREPLY + 1,
52         [ICMP_ECHOREPLY] = ICMP_ECHO + 1,
53         [ICMP_TIMESTAMP] = ICMP_TIMESTAMPREPLY + 1,
54         [ICMP_TIMESTAMPREPLY] = ICMP_TIMESTAMP + 1,
55         [ICMP_INFO_REQUEST] = ICMP_INFO_REPLY + 1,
56         [ICMP_INFO_REPLY] = ICMP_INFO_REQUEST + 1,
57         [ICMP_ADDRESS] = ICMP_ADDRESSREPLY + 1,
58         [ICMP_ADDRESSREPLY] = ICMP_ADDRESS + 1
59 };
60
61 static bool icmp_invert_tuple(struct nf_conntrack_tuple *tuple,
62                               const struct nf_conntrack_tuple *orig)
63 {
64         if (orig->dst.u.icmp.type >= sizeof(invmap) ||
65             !invmap[orig->dst.u.icmp.type])
66                 return false;
67
68         tuple->src.u.icmp.id = orig->src.u.icmp.id;
69         tuple->dst.u.icmp.type = invmap[orig->dst.u.icmp.type] - 1;
70         tuple->dst.u.icmp.code = orig->dst.u.icmp.code;
71         return true;
72 }
73
74 /* Print out the per-protocol part of the tuple. */
75 static void icmp_print_tuple(struct seq_file *s,
76                             const struct nf_conntrack_tuple *tuple)
77 {
78         seq_printf(s, "type=%u code=%u id=%u ",
79                    tuple->dst.u.icmp.type,
80                    tuple->dst.u.icmp.code,
81                    ntohs(tuple->src.u.icmp.id));
82 }
83
84 static unsigned int *icmp_get_timeouts(struct net *net)
85 {
86         return &icmp_pernet(net)->timeout;
87 }
88
89 /* Returns verdict for packet, or -1 for invalid. */
90 static int icmp_packet(struct nf_conn *ct,
91                        const struct sk_buff *skb,
92                        unsigned int dataoff,
93                        enum ip_conntrack_info ctinfo,
94                        u_int8_t pf,
95                        unsigned int hooknum,
96                        unsigned int *timeout)
97 {
98         /* Do not immediately delete the connection after the first
99            successful reply to avoid excessive conntrackd traffic
100            and also to handle correctly ICMP echo reply duplicates. */
101         nf_ct_refresh_acct(ct, ctinfo, skb, *timeout);
102
103         return NF_ACCEPT;
104 }
105
106 /* Called when a new connection for this protocol found. */
107 static bool icmp_new(struct nf_conn *ct, const struct sk_buff *skb,
108                      unsigned int dataoff, unsigned int *timeouts)
109 {
110         static const u_int8_t valid_new[] = {
111                 [ICMP_ECHO] = 1,
112                 [ICMP_TIMESTAMP] = 1,
113                 [ICMP_INFO_REQUEST] = 1,
114                 [ICMP_ADDRESS] = 1
115         };
116
117         if (ct->tuplehash[0].tuple.dst.u.icmp.type >= sizeof(valid_new) ||
118             !valid_new[ct->tuplehash[0].tuple.dst.u.icmp.type]) {
119                 /* Can't create a new ICMP `conn' with this. */
120                 pr_debug("icmp: can't create new conn with type %u\n",
121                          ct->tuplehash[0].tuple.dst.u.icmp.type);
122                 nf_ct_dump_tuple_ip(&ct->tuplehash[0].tuple);
123                 return false;
124         }
125         return true;
126 }
127
128 /* Returns conntrack if it dealt with ICMP, and filled in skb fields */
129 static int
130 icmp_error_message(struct net *net, struct nf_conn *tmpl, struct sk_buff *skb,
131                  enum ip_conntrack_info *ctinfo,
132                  unsigned int hooknum)
133 {
134         struct nf_conntrack_tuple innertuple, origtuple;
135         const struct nf_conntrack_l4proto *innerproto;
136         const struct nf_conntrack_tuple_hash *h;
137         u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE;
138
139         NF_CT_ASSERT(skb->nfct == NULL);
140
141         /* Are they talking about one of our connections? */
142         if (!nf_ct_get_tuplepr(skb,
143                                skb_network_offset(skb) + ip_hdrlen(skb)
144                                                        + sizeof(struct icmphdr),
145                                PF_INET, &origtuple)) {
146                 pr_debug("icmp_error_message: failed to get tuple\n");
147                 return -NF_ACCEPT;
148         }
149
150         /* rcu_read_lock()ed by nf_hook_slow */
151         innerproto = __nf_ct_l4proto_find(PF_INET, origtuple.dst.protonum);
152
153         /* Ordinarily, we'd expect the inverted tupleproto, but it's
154            been preserved inside the ICMP. */
155         if (!nf_ct_invert_tuple(&innertuple, &origtuple,
156                                 &nf_conntrack_l3proto_ipv4, innerproto)) {
157                 pr_debug("icmp_error_message: no match\n");
158                 return -NF_ACCEPT;
159         }
160
161         *ctinfo = IP_CT_RELATED;
162
163         h = nf_conntrack_find_get(net, zone, &innertuple);
164         if (!h) {
165                 pr_debug("icmp_error_message: no match\n");
166                 return -NF_ACCEPT;
167         }
168
169         if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
170                 *ctinfo += IP_CT_IS_REPLY;
171
172         /* Update skb to refer to this connection */
173         skb->nfct = &nf_ct_tuplehash_to_ctrack(h)->ct_general;
174         skb->nfctinfo = *ctinfo;
175         return NF_ACCEPT;
176 }
177
178 /* Small and modified version of icmp_rcv */
179 static int
180 icmp_error(struct net *net, struct nf_conn *tmpl,
181            struct sk_buff *skb, unsigned int dataoff,
182            enum ip_conntrack_info *ctinfo, u_int8_t pf, unsigned int hooknum)
183 {
184         const struct icmphdr *icmph;
185         struct icmphdr _ih;
186
187         /* Not enough header? */
188         icmph = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_ih), &_ih);
189         if (icmph == NULL) {
190                 if (LOG_INVALID(net, IPPROTO_ICMP))
191                         nf_log_packet(net, PF_INET, 0, skb, NULL, NULL,
192                                       NULL, "nf_ct_icmp: short packet ");
193                 return -NF_ACCEPT;
194         }
195
196         /* See ip_conntrack_proto_tcp.c */
197         if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING &&
198             nf_ip_checksum(skb, hooknum, dataoff, 0)) {
199                 if (LOG_INVALID(net, IPPROTO_ICMP))
200                         nf_log_packet(net, PF_INET, 0, skb, NULL, NULL, NULL,
201                                       "nf_ct_icmp: bad HW ICMP checksum ");
202                 return -NF_ACCEPT;
203         }
204
205         /*
206          *      18 is the highest 'known' ICMP type. Anything else is a mystery
207          *
208          *      RFC 1122: 3.2.2  Unknown ICMP messages types MUST be silently
209          *                discarded.
210          */
211         if (icmph->type > NR_ICMP_TYPES) {
212                 if (LOG_INVALID(net, IPPROTO_ICMP))
213                         nf_log_packet(net, PF_INET, 0, skb, NULL, NULL, NULL,
214                                       "nf_ct_icmp: invalid ICMP type ");
215                 return -NF_ACCEPT;
216         }
217
218         /* Need to track icmp error message? */
219         if (icmph->type != ICMP_DEST_UNREACH &&
220             icmph->type != ICMP_SOURCE_QUENCH &&
221             icmph->type != ICMP_TIME_EXCEEDED &&
222             icmph->type != ICMP_PARAMETERPROB &&
223             icmph->type != ICMP_REDIRECT)
224                 return NF_ACCEPT;
225
226         return icmp_error_message(net, tmpl, skb, ctinfo, hooknum);
227 }
228
229 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
230
231 #include <linux/netfilter/nfnetlink.h>
232 #include <linux/netfilter/nfnetlink_conntrack.h>
233
234 static int icmp_tuple_to_nlattr(struct sk_buff *skb,
235                                 const struct nf_conntrack_tuple *t)
236 {
237         if (nla_put_be16(skb, CTA_PROTO_ICMP_ID, t->src.u.icmp.id) ||
238             nla_put_u8(skb, CTA_PROTO_ICMP_TYPE, t->dst.u.icmp.type) ||
239             nla_put_u8(skb, CTA_PROTO_ICMP_CODE, t->dst.u.icmp.code))
240                 goto nla_put_failure;
241         return 0;
242
243 nla_put_failure:
244         return -1;
245 }
246
247 static const struct nla_policy icmp_nla_policy[CTA_PROTO_MAX+1] = {
248         [CTA_PROTO_ICMP_TYPE]   = { .type = NLA_U8 },
249         [CTA_PROTO_ICMP_CODE]   = { .type = NLA_U8 },
250         [CTA_PROTO_ICMP_ID]     = { .type = NLA_U16 },
251 };
252
253 static int icmp_nlattr_to_tuple(struct nlattr *tb[],
254                                 struct nf_conntrack_tuple *tuple)
255 {
256         if (!tb[CTA_PROTO_ICMP_TYPE] ||
257             !tb[CTA_PROTO_ICMP_CODE] ||
258             !tb[CTA_PROTO_ICMP_ID])
259                 return -EINVAL;
260
261         tuple->dst.u.icmp.type = nla_get_u8(tb[CTA_PROTO_ICMP_TYPE]);
262         tuple->dst.u.icmp.code = nla_get_u8(tb[CTA_PROTO_ICMP_CODE]);
263         tuple->src.u.icmp.id = nla_get_be16(tb[CTA_PROTO_ICMP_ID]);
264
265         if (tuple->dst.u.icmp.type >= sizeof(invmap) ||
266             !invmap[tuple->dst.u.icmp.type])
267                 return -EINVAL;
268
269         return 0;
270 }
271
272 static int icmp_nlattr_tuple_size(void)
273 {
274         return nla_policy_len(icmp_nla_policy, CTA_PROTO_MAX + 1);
275 }
276 #endif
277
278 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
279
280 #include <linux/netfilter/nfnetlink.h>
281 #include <linux/netfilter/nfnetlink_cttimeout.h>
282
283 static int icmp_timeout_nlattr_to_obj(struct nlattr *tb[],
284                                       struct net *net, void *data)
285 {
286         unsigned int *timeout = data;
287         struct nf_icmp_net *in = icmp_pernet(net);
288
289         if (tb[CTA_TIMEOUT_ICMP_TIMEOUT]) {
290                 *timeout =
291                         ntohl(nla_get_be32(tb[CTA_TIMEOUT_ICMP_TIMEOUT])) * HZ;
292         } else {
293                 /* Set default ICMP timeout. */
294                 *timeout = in->timeout;
295         }
296         return 0;
297 }
298
299 static int
300 icmp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data)
301 {
302         const unsigned int *timeout = data;
303
304         if (nla_put_be32(skb, CTA_TIMEOUT_ICMP_TIMEOUT, htonl(*timeout / HZ)))
305                 goto nla_put_failure;
306         return 0;
307
308 nla_put_failure:
309         return -ENOSPC;
310 }
311
312 static const struct nla_policy
313 icmp_timeout_nla_policy[CTA_TIMEOUT_ICMP_MAX+1] = {
314         [CTA_TIMEOUT_ICMP_TIMEOUT]      = { .type = NLA_U32 },
315 };
316 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
317
318 #ifdef CONFIG_SYSCTL
319 static struct ctl_table icmp_sysctl_table[] = {
320         {
321                 .procname       = "nf_conntrack_icmp_timeout",
322                 .maxlen         = sizeof(unsigned int),
323                 .mode           = 0644,
324                 .proc_handler   = proc_dointvec_jiffies,
325         },
326         { }
327 };
328 #ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
329 static struct ctl_table icmp_compat_sysctl_table[] = {
330         {
331                 .procname       = "ip_conntrack_icmp_timeout",
332                 .maxlen         = sizeof(unsigned int),
333                 .mode           = 0644,
334                 .proc_handler   = proc_dointvec_jiffies,
335         },
336         { }
337 };
338 #endif /* CONFIG_NF_CONNTRACK_PROC_COMPAT */
339 #endif /* CONFIG_SYSCTL */
340
341 static int icmp_kmemdup_sysctl_table(struct nf_proto_net *pn,
342                                      struct nf_icmp_net *in)
343 {
344 #ifdef CONFIG_SYSCTL
345         pn->ctl_table = kmemdup(icmp_sysctl_table,
346                                 sizeof(icmp_sysctl_table),
347                                 GFP_KERNEL);
348         if (!pn->ctl_table)
349                 return -ENOMEM;
350
351         pn->ctl_table[0].data = &in->timeout;
352 #endif
353         return 0;
354 }
355
356 static int icmp_kmemdup_compat_sysctl_table(struct nf_proto_net *pn,
357                                             struct nf_icmp_net *in)
358 {
359 #ifdef CONFIG_SYSCTL
360 #ifdef CONFIG_NF_CONNTRACK_PROC_COMPAT
361         pn->ctl_compat_table = kmemdup(icmp_compat_sysctl_table,
362                                        sizeof(icmp_compat_sysctl_table),
363                                        GFP_KERNEL);
364         if (!pn->ctl_compat_table)
365                 return -ENOMEM;
366
367         pn->ctl_compat_table[0].data = &in->timeout;
368 #endif
369 #endif
370         return 0;
371 }
372
373 static int icmp_init_net(struct net *net, u_int16_t proto)
374 {
375         int ret;
376         struct nf_icmp_net *in = icmp_pernet(net);
377         struct nf_proto_net *pn = &in->pn;
378
379         in->timeout = nf_ct_icmp_timeout;
380
381         ret = icmp_kmemdup_compat_sysctl_table(pn, in);
382         if (ret < 0)
383                 return ret;
384
385         ret = icmp_kmemdup_sysctl_table(pn, in);
386         if (ret < 0)
387                 nf_ct_kfree_compat_sysctl_table(pn);
388
389         return ret;
390 }
391
392 static struct nf_proto_net *icmp_get_net_proto(struct net *net)
393 {
394         return &net->ct.nf_ct_proto.icmp.pn;
395 }
396
397 struct nf_conntrack_l4proto nf_conntrack_l4proto_icmp __read_mostly =
398 {
399         .l3proto                = PF_INET,
400         .l4proto                = IPPROTO_ICMP,
401         .name                   = "icmp",
402         .pkt_to_tuple           = icmp_pkt_to_tuple,
403         .invert_tuple           = icmp_invert_tuple,
404         .print_tuple            = icmp_print_tuple,
405         .packet                 = icmp_packet,
406         .get_timeouts           = icmp_get_timeouts,
407         .new                    = icmp_new,
408         .error                  = icmp_error,
409         .destroy                = NULL,
410         .me                     = NULL,
411 #if IS_ENABLED(CONFIG_NF_CT_NETLINK)
412         .tuple_to_nlattr        = icmp_tuple_to_nlattr,
413         .nlattr_tuple_size      = icmp_nlattr_tuple_size,
414         .nlattr_to_tuple        = icmp_nlattr_to_tuple,
415         .nla_policy             = icmp_nla_policy,
416 #endif
417 #if IS_ENABLED(CONFIG_NF_CT_NETLINK_TIMEOUT)
418         .ctnl_timeout           = {
419                 .nlattr_to_obj  = icmp_timeout_nlattr_to_obj,
420                 .obj_to_nlattr  = icmp_timeout_obj_to_nlattr,
421                 .nlattr_max     = CTA_TIMEOUT_ICMP_MAX,
422                 .obj_size       = sizeof(unsigned int),
423                 .nla_policy     = icmp_timeout_nla_policy,
424         },
425 #endif /* CONFIG_NF_CT_NETLINK_TIMEOUT */
426         .init_net               = icmp_init_net,
427         .get_net_proto          = icmp_get_net_proto,
428 };