Documentation/filesystems/proc.txt: add more description for maps/smaps
[cascardo/linux.git] / net / ipv6 / netfilter / ip6t_SYNPROXY.c
1 /*
2  * Copyright (c) 2013 Patrick McHardy <kaber@trash.net>
3  *
4  * This program is free software; you can redistribute it and/or modify
5  * it under the terms of the GNU General Public License version 2 as
6  * published by the Free Software Foundation.
7  */
8
9 #include <linux/module.h>
10 #include <linux/skbuff.h>
11 #include <net/ip6_checksum.h>
12 #include <net/ip6_route.h>
13 #include <net/tcp.h>
14
15 #include <linux/netfilter_ipv6/ip6_tables.h>
16 #include <linux/netfilter/x_tables.h>
17 #include <linux/netfilter/xt_SYNPROXY.h>
18 #include <net/netfilter/nf_conntrack.h>
19 #include <net/netfilter/nf_conntrack_seqadj.h>
20 #include <net/netfilter/nf_conntrack_synproxy.h>
21
22 static struct ipv6hdr *
23 synproxy_build_ip(struct net *net, struct sk_buff *skb,
24                   const struct in6_addr *saddr,
25                   const struct in6_addr *daddr)
26 {
27         struct ipv6hdr *iph;
28
29         skb_reset_network_header(skb);
30         iph = (struct ipv6hdr *)skb_put(skb, sizeof(*iph));
31         ip6_flow_hdr(iph, 0, 0);
32         iph->hop_limit  = net->ipv6.devconf_all->hop_limit;
33         iph->nexthdr    = IPPROTO_TCP;
34         iph->saddr      = *saddr;
35         iph->daddr      = *daddr;
36
37         return iph;
38 }
39
40 static void
41 synproxy_send_tcp(struct net *net,
42                   const struct sk_buff *skb, struct sk_buff *nskb,
43                   struct nf_conntrack *nfct, enum ip_conntrack_info ctinfo,
44                   struct ipv6hdr *niph, struct tcphdr *nth,
45                   unsigned int tcp_hdr_size)
46 {
47         struct dst_entry *dst;
48         struct flowi6 fl6;
49
50         nth->check = ~tcp_v6_check(tcp_hdr_size, &niph->saddr, &niph->daddr, 0);
51         nskb->ip_summed   = CHECKSUM_PARTIAL;
52         nskb->csum_start  = (unsigned char *)nth - nskb->head;
53         nskb->csum_offset = offsetof(struct tcphdr, check);
54
55         memset(&fl6, 0, sizeof(fl6));
56         fl6.flowi6_proto = IPPROTO_TCP;
57         fl6.saddr = niph->saddr;
58         fl6.daddr = niph->daddr;
59         fl6.fl6_sport = nth->source;
60         fl6.fl6_dport = nth->dest;
61         security_skb_classify_flow((struct sk_buff *)skb, flowi6_to_flowi(&fl6));
62         dst = ip6_route_output(net, NULL, &fl6);
63         if (dst->error) {
64                 dst_release(dst);
65                 goto free_nskb;
66         }
67         dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0);
68         if (IS_ERR(dst))
69                 goto free_nskb;
70
71         skb_dst_set(nskb, dst);
72
73         if (nfct) {
74                 nskb->nfct = nfct;
75                 nskb->nfctinfo = ctinfo;
76                 nf_conntrack_get(nfct);
77         }
78
79         ip6_local_out(net, nskb->sk, nskb);
80         return;
81
82 free_nskb:
83         kfree_skb(nskb);
84 }
85
86 static void
87 synproxy_send_client_synack(struct net *net,
88                             const struct sk_buff *skb, const struct tcphdr *th,
89                             const struct synproxy_options *opts)
90 {
91         struct sk_buff *nskb;
92         struct ipv6hdr *iph, *niph;
93         struct tcphdr *nth;
94         unsigned int tcp_hdr_size;
95         u16 mss = opts->mss;
96
97         iph = ipv6_hdr(skb);
98
99         tcp_hdr_size = sizeof(*nth) + synproxy_options_size(opts);
100         nskb = alloc_skb(sizeof(*niph) + tcp_hdr_size + MAX_TCP_HEADER,
101                          GFP_ATOMIC);
102         if (nskb == NULL)
103                 return;
104         skb_reserve(nskb, MAX_TCP_HEADER);
105
106         niph = synproxy_build_ip(net, nskb, &iph->daddr, &iph->saddr);
107
108         skb_reset_transport_header(nskb);
109         nth = (struct tcphdr *)skb_put(nskb, tcp_hdr_size);
110         nth->source     = th->dest;
111         nth->dest       = th->source;
112         nth->seq        = htonl(__cookie_v6_init_sequence(iph, th, &mss));
113         nth->ack_seq    = htonl(ntohl(th->seq) + 1);
114         tcp_flag_word(nth) = TCP_FLAG_SYN | TCP_FLAG_ACK;
115         if (opts->options & XT_SYNPROXY_OPT_ECN)
116                 tcp_flag_word(nth) |= TCP_FLAG_ECE;
117         nth->doff       = tcp_hdr_size / 4;
118         nth->window     = 0;
119         nth->check      = 0;
120         nth->urg_ptr    = 0;
121
122         synproxy_build_options(nth, opts);
123
124         synproxy_send_tcp(net, skb, nskb, skb->nfct, IP_CT_ESTABLISHED_REPLY,
125                           niph, nth, tcp_hdr_size);
126 }
127
128 static void
129 synproxy_send_server_syn(struct net *net,
130                          const struct sk_buff *skb, const struct tcphdr *th,
131                          const struct synproxy_options *opts, u32 recv_seq)
132 {
133         struct synproxy_net *snet = synproxy_pernet(net);
134         struct sk_buff *nskb;
135         struct ipv6hdr *iph, *niph;
136         struct tcphdr *nth;
137         unsigned int tcp_hdr_size;
138
139         iph = ipv6_hdr(skb);
140
141         tcp_hdr_size = sizeof(*nth) + synproxy_options_size(opts);
142         nskb = alloc_skb(sizeof(*niph) + tcp_hdr_size + MAX_TCP_HEADER,
143                          GFP_ATOMIC);
144         if (nskb == NULL)
145                 return;
146         skb_reserve(nskb, MAX_TCP_HEADER);
147
148         niph = synproxy_build_ip(net, nskb, &iph->saddr, &iph->daddr);
149
150         skb_reset_transport_header(nskb);
151         nth = (struct tcphdr *)skb_put(nskb, tcp_hdr_size);
152         nth->source     = th->source;
153         nth->dest       = th->dest;
154         nth->seq        = htonl(recv_seq - 1);
155         /* ack_seq is used to relay our ISN to the synproxy hook to initialize
156          * sequence number translation once a connection tracking entry exists.
157          */
158         nth->ack_seq    = htonl(ntohl(th->ack_seq) - 1);
159         tcp_flag_word(nth) = TCP_FLAG_SYN;
160         if (opts->options & XT_SYNPROXY_OPT_ECN)
161                 tcp_flag_word(nth) |= TCP_FLAG_ECE | TCP_FLAG_CWR;
162         nth->doff       = tcp_hdr_size / 4;
163         nth->window     = th->window;
164         nth->check      = 0;
165         nth->urg_ptr    = 0;
166
167         synproxy_build_options(nth, opts);
168
169         synproxy_send_tcp(net, skb, nskb, &snet->tmpl->ct_general, IP_CT_NEW,
170                           niph, nth, tcp_hdr_size);
171 }
172
173 static void
174 synproxy_send_server_ack(struct net *net,
175                          const struct ip_ct_tcp *state,
176                          const struct sk_buff *skb, const struct tcphdr *th,
177                          const struct synproxy_options *opts)
178 {
179         struct sk_buff *nskb;
180         struct ipv6hdr *iph, *niph;
181         struct tcphdr *nth;
182         unsigned int tcp_hdr_size;
183
184         iph = ipv6_hdr(skb);
185
186         tcp_hdr_size = sizeof(*nth) + synproxy_options_size(opts);
187         nskb = alloc_skb(sizeof(*niph) + tcp_hdr_size + MAX_TCP_HEADER,
188                          GFP_ATOMIC);
189         if (nskb == NULL)
190                 return;
191         skb_reserve(nskb, MAX_TCP_HEADER);
192
193         niph = synproxy_build_ip(net, nskb, &iph->daddr, &iph->saddr);
194
195         skb_reset_transport_header(nskb);
196         nth = (struct tcphdr *)skb_put(nskb, tcp_hdr_size);
197         nth->source     = th->dest;
198         nth->dest       = th->source;
199         nth->seq        = htonl(ntohl(th->ack_seq));
200         nth->ack_seq    = htonl(ntohl(th->seq) + 1);
201         tcp_flag_word(nth) = TCP_FLAG_ACK;
202         nth->doff       = tcp_hdr_size / 4;
203         nth->window     = htons(state->seen[IP_CT_DIR_ORIGINAL].td_maxwin);
204         nth->check      = 0;
205         nth->urg_ptr    = 0;
206
207         synproxy_build_options(nth, opts);
208
209         synproxy_send_tcp(net, skb, nskb, NULL, 0, niph, nth, tcp_hdr_size);
210 }
211
212 static void
213 synproxy_send_client_ack(struct net *net,
214                          const struct sk_buff *skb, const struct tcphdr *th,
215                          const struct synproxy_options *opts)
216 {
217         struct sk_buff *nskb;
218         struct ipv6hdr *iph, *niph;
219         struct tcphdr *nth;
220         unsigned int tcp_hdr_size;
221
222         iph = ipv6_hdr(skb);
223
224         tcp_hdr_size = sizeof(*nth) + synproxy_options_size(opts);
225         nskb = alloc_skb(sizeof(*niph) + tcp_hdr_size + MAX_TCP_HEADER,
226                          GFP_ATOMIC);
227         if (nskb == NULL)
228                 return;
229         skb_reserve(nskb, MAX_TCP_HEADER);
230
231         niph = synproxy_build_ip(net, nskb, &iph->saddr, &iph->daddr);
232
233         skb_reset_transport_header(nskb);
234         nth = (struct tcphdr *)skb_put(nskb, tcp_hdr_size);
235         nth->source     = th->source;
236         nth->dest       = th->dest;
237         nth->seq        = htonl(ntohl(th->seq) + 1);
238         nth->ack_seq    = th->ack_seq;
239         tcp_flag_word(nth) = TCP_FLAG_ACK;
240         nth->doff       = tcp_hdr_size / 4;
241         nth->window     = htons(ntohs(th->window) >> opts->wscale);
242         nth->check      = 0;
243         nth->urg_ptr    = 0;
244
245         synproxy_build_options(nth, opts);
246
247         synproxy_send_tcp(net, skb, nskb, skb->nfct, IP_CT_ESTABLISHED_REPLY,
248                           niph, nth, tcp_hdr_size);
249 }
250
251 static bool
252 synproxy_recv_client_ack(struct net *net,
253                          const struct sk_buff *skb, const struct tcphdr *th,
254                          struct synproxy_options *opts, u32 recv_seq)
255 {
256         struct synproxy_net *snet = synproxy_pernet(net);
257         int mss;
258
259         mss = __cookie_v6_check(ipv6_hdr(skb), th, ntohl(th->ack_seq) - 1);
260         if (mss == 0) {
261                 this_cpu_inc(snet->stats->cookie_invalid);
262                 return false;
263         }
264
265         this_cpu_inc(snet->stats->cookie_valid);
266         opts->mss = mss;
267         opts->options |= XT_SYNPROXY_OPT_MSS;
268
269         if (opts->options & XT_SYNPROXY_OPT_TIMESTAMP)
270                 synproxy_check_timestamp_cookie(opts);
271
272         synproxy_send_server_syn(net, skb, th, opts, recv_seq);
273         return true;
274 }
275
276 static unsigned int
277 synproxy_tg6(struct sk_buff *skb, const struct xt_action_param *par)
278 {
279         const struct xt_synproxy_info *info = par->targinfo;
280         struct net *net = par->net;
281         struct synproxy_net *snet = synproxy_pernet(net);
282         struct synproxy_options opts = {};
283         struct tcphdr *th, _th;
284
285         if (nf_ip6_checksum(skb, par->hooknum, par->thoff, IPPROTO_TCP))
286                 return NF_DROP;
287
288         th = skb_header_pointer(skb, par->thoff, sizeof(_th), &_th);
289         if (th == NULL)
290                 return NF_DROP;
291
292         if (!synproxy_parse_options(skb, par->thoff, th, &opts))
293                 return NF_DROP;
294
295         if (th->syn && !(th->ack || th->fin || th->rst)) {
296                 /* Initial SYN from client */
297                 this_cpu_inc(snet->stats->syn_received);
298
299                 if (th->ece && th->cwr)
300                         opts.options |= XT_SYNPROXY_OPT_ECN;
301
302                 opts.options &= info->options;
303                 if (opts.options & XT_SYNPROXY_OPT_TIMESTAMP)
304                         synproxy_init_timestamp_cookie(info, &opts);
305                 else
306                         opts.options &= ~(XT_SYNPROXY_OPT_WSCALE |
307                                           XT_SYNPROXY_OPT_SACK_PERM |
308                                           XT_SYNPROXY_OPT_ECN);
309
310                 synproxy_send_client_synack(net, skb, th, &opts);
311                 return NF_DROP;
312
313         } else if (th->ack && !(th->fin || th->rst || th->syn)) {
314                 /* ACK from client */
315                 synproxy_recv_client_ack(net, skb, th, &opts, ntohl(th->seq));
316                 return NF_DROP;
317         }
318
319         return XT_CONTINUE;
320 }
321
322 static unsigned int ipv6_synproxy_hook(void *priv,
323                                        struct sk_buff *skb,
324                                        const struct nf_hook_state *nhs)
325 {
326         struct net *net = nhs->net;
327         struct synproxy_net *snet = synproxy_pernet(net);
328         enum ip_conntrack_info ctinfo;
329         struct nf_conn *ct;
330         struct nf_conn_synproxy *synproxy;
331         struct synproxy_options opts = {};
332         const struct ip_ct_tcp *state;
333         struct tcphdr *th, _th;
334         __be16 frag_off;
335         u8 nexthdr;
336         int thoff;
337
338         ct = nf_ct_get(skb, &ctinfo);
339         if (ct == NULL)
340                 return NF_ACCEPT;
341
342         synproxy = nfct_synproxy(ct);
343         if (synproxy == NULL)
344                 return NF_ACCEPT;
345
346         if (nf_is_loopback_packet(skb))
347                 return NF_ACCEPT;
348
349         nexthdr = ipv6_hdr(skb)->nexthdr;
350         thoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
351                                  &frag_off);
352         if (thoff < 0)
353                 return NF_ACCEPT;
354
355         th = skb_header_pointer(skb, thoff, sizeof(_th), &_th);
356         if (th == NULL)
357                 return NF_DROP;
358
359         state = &ct->proto.tcp;
360         switch (state->state) {
361         case TCP_CONNTRACK_CLOSE:
362                 if (th->rst && !test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
363                         nf_ct_seqadj_init(ct, ctinfo, synproxy->isn -
364                                                       ntohl(th->seq) + 1);
365                         break;
366                 }
367
368                 if (!th->syn || th->ack ||
369                     CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL)
370                         break;
371
372                 /* Reopened connection - reset the sequence number and timestamp
373                  * adjustments, they will get initialized once the connection is
374                  * reestablished.
375                  */
376                 nf_ct_seqadj_init(ct, ctinfo, 0);
377                 synproxy->tsoff = 0;
378                 this_cpu_inc(snet->stats->conn_reopened);
379
380                 /* fall through */
381         case TCP_CONNTRACK_SYN_SENT:
382                 if (!synproxy_parse_options(skb, thoff, th, &opts))
383                         return NF_DROP;
384
385                 if (!th->syn && th->ack &&
386                     CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) {
387                         /* Keep-Alives are sent with SEG.SEQ = SND.NXT-1,
388                          * therefore we need to add 1 to make the SYN sequence
389                          * number match the one of first SYN.
390                          */
391                         if (synproxy_recv_client_ack(net, skb, th, &opts,
392                                                      ntohl(th->seq) + 1))
393                                 this_cpu_inc(snet->stats->cookie_retrans);
394
395                         return NF_DROP;
396                 }
397
398                 synproxy->isn = ntohl(th->ack_seq);
399                 if (opts.options & XT_SYNPROXY_OPT_TIMESTAMP)
400                         synproxy->its = opts.tsecr;
401                 break;
402         case TCP_CONNTRACK_SYN_RECV:
403                 if (!th->syn || !th->ack)
404                         break;
405
406                 if (!synproxy_parse_options(skb, thoff, th, &opts))
407                         return NF_DROP;
408
409                 if (opts.options & XT_SYNPROXY_OPT_TIMESTAMP)
410                         synproxy->tsoff = opts.tsval - synproxy->its;
411
412                 opts.options &= ~(XT_SYNPROXY_OPT_MSS |
413                                   XT_SYNPROXY_OPT_WSCALE |
414                                   XT_SYNPROXY_OPT_SACK_PERM);
415
416                 swap(opts.tsval, opts.tsecr);
417                 synproxy_send_server_ack(net, state, skb, th, &opts);
418
419                 nf_ct_seqadj_init(ct, ctinfo, synproxy->isn - ntohl(th->seq));
420
421                 swap(opts.tsval, opts.tsecr);
422                 synproxy_send_client_ack(net, skb, th, &opts);
423
424                 consume_skb(skb);
425                 return NF_STOLEN;
426         default:
427                 break;
428         }
429
430         synproxy_tstamp_adjust(skb, thoff, th, ct, ctinfo, synproxy);
431         return NF_ACCEPT;
432 }
433
434 static int synproxy_tg6_check(const struct xt_tgchk_param *par)
435 {
436         const struct ip6t_entry *e = par->entryinfo;
437
438         if (!(e->ipv6.flags & IP6T_F_PROTO) ||
439             e->ipv6.proto != IPPROTO_TCP ||
440             e->ipv6.invflags & XT_INV_PROTO)
441                 return -EINVAL;
442
443         return nf_ct_l3proto_try_module_get(par->family);
444 }
445
446 static void synproxy_tg6_destroy(const struct xt_tgdtor_param *par)
447 {
448         nf_ct_l3proto_module_put(par->family);
449 }
450
451 static struct xt_target synproxy_tg6_reg __read_mostly = {
452         .name           = "SYNPROXY",
453         .family         = NFPROTO_IPV6,
454         .hooks          = (1 << NF_INET_LOCAL_IN) | (1 << NF_INET_FORWARD),
455         .target         = synproxy_tg6,
456         .targetsize     = sizeof(struct xt_synproxy_info),
457         .checkentry     = synproxy_tg6_check,
458         .destroy        = synproxy_tg6_destroy,
459         .me             = THIS_MODULE,
460 };
461
462 static struct nf_hook_ops ipv6_synproxy_ops[] __read_mostly = {
463         {
464                 .hook           = ipv6_synproxy_hook,
465                 .pf             = NFPROTO_IPV6,
466                 .hooknum        = NF_INET_LOCAL_IN,
467                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM - 1,
468         },
469         {
470                 .hook           = ipv6_synproxy_hook,
471                 .pf             = NFPROTO_IPV6,
472                 .hooknum        = NF_INET_POST_ROUTING,
473                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM - 1,
474         },
475 };
476
477 static int __init synproxy_tg6_init(void)
478 {
479         int err;
480
481         err = nf_register_hooks(ipv6_synproxy_ops,
482                                 ARRAY_SIZE(ipv6_synproxy_ops));
483         if (err < 0)
484                 goto err1;
485
486         err = xt_register_target(&synproxy_tg6_reg);
487         if (err < 0)
488                 goto err2;
489
490         return 0;
491
492 err2:
493         nf_unregister_hooks(ipv6_synproxy_ops, ARRAY_SIZE(ipv6_synproxy_ops));
494 err1:
495         return err;
496 }
497
498 static void __exit synproxy_tg6_exit(void)
499 {
500         xt_unregister_target(&synproxy_tg6_reg);
501         nf_unregister_hooks(ipv6_synproxy_ops, ARRAY_SIZE(ipv6_synproxy_ops));
502 }
503
504 module_init(synproxy_tg6_init);
505 module_exit(synproxy_tg6_exit);
506
507 MODULE_LICENSE("GPL");
508 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");