These changes are the raw update to linux-4.4.6-rt14. Kernel sources
[kvmfornfv.git] / kernel / 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 sk_buff *skb, const struct in6_addr *saddr,
24                                        const struct in6_addr *daddr)
25 {
26         struct ipv6hdr *iph;
27
28         skb_reset_network_header(skb);
29         iph = (struct ipv6hdr *)skb_put(skb, sizeof(*iph));
30         ip6_flow_hdr(iph, 0, 0);
31         iph->hop_limit  = 64;   //XXX
32         iph->nexthdr    = IPPROTO_TCP;
33         iph->saddr      = *saddr;
34         iph->daddr      = *daddr;
35
36         return iph;
37 }
38
39 static void
40 synproxy_send_tcp(const struct synproxy_net *snet,
41                   const struct sk_buff *skb, struct sk_buff *nskb,
42                   struct nf_conntrack *nfct, enum ip_conntrack_info ctinfo,
43                   struct ipv6hdr *niph, struct tcphdr *nth,
44                   unsigned int tcp_hdr_size)
45 {
46         struct net *net = nf_ct_net(snet->tmpl);
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 == NULL || 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(const struct synproxy_net *snet,
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(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(snet, skb, nskb, skb->nfct, IP_CT_ESTABLISHED_REPLY,
125                           niph, nth, tcp_hdr_size);
126 }
127
128 static void
129 synproxy_send_server_syn(const struct synproxy_net *snet,
130                          const struct sk_buff *skb, const struct tcphdr *th,
131                          const struct synproxy_options *opts, u32 recv_seq)
132 {
133         struct sk_buff *nskb;
134         struct ipv6hdr *iph, *niph;
135         struct tcphdr *nth;
136         unsigned int tcp_hdr_size;
137
138         iph = ipv6_hdr(skb);
139
140         tcp_hdr_size = sizeof(*nth) + synproxy_options_size(opts);
141         nskb = alloc_skb(sizeof(*niph) + tcp_hdr_size + MAX_TCP_HEADER,
142                          GFP_ATOMIC);
143         if (nskb == NULL)
144                 return;
145         skb_reserve(nskb, MAX_TCP_HEADER);
146
147         niph = synproxy_build_ip(nskb, &iph->saddr, &iph->daddr);
148
149         skb_reset_transport_header(nskb);
150         nth = (struct tcphdr *)skb_put(nskb, tcp_hdr_size);
151         nth->source     = th->source;
152         nth->dest       = th->dest;
153         nth->seq        = htonl(recv_seq - 1);
154         /* ack_seq is used to relay our ISN to the synproxy hook to initialize
155          * sequence number translation once a connection tracking entry exists.
156          */
157         nth->ack_seq    = htonl(ntohl(th->ack_seq) - 1);
158         tcp_flag_word(nth) = TCP_FLAG_SYN;
159         if (opts->options & XT_SYNPROXY_OPT_ECN)
160                 tcp_flag_word(nth) |= TCP_FLAG_ECE | TCP_FLAG_CWR;
161         nth->doff       = tcp_hdr_size / 4;
162         nth->window     = th->window;
163         nth->check      = 0;
164         nth->urg_ptr    = 0;
165
166         synproxy_build_options(nth, opts);
167
168         synproxy_send_tcp(snet, skb, nskb, &snet->tmpl->ct_general, IP_CT_NEW,
169                           niph, nth, tcp_hdr_size);
170 }
171
172 static void
173 synproxy_send_server_ack(const struct synproxy_net *snet,
174                          const struct ip_ct_tcp *state,
175                          const struct sk_buff *skb, const struct tcphdr *th,
176                          const struct synproxy_options *opts)
177 {
178         struct sk_buff *nskb;
179         struct ipv6hdr *iph, *niph;
180         struct tcphdr *nth;
181         unsigned int tcp_hdr_size;
182
183         iph = ipv6_hdr(skb);
184
185         tcp_hdr_size = sizeof(*nth) + synproxy_options_size(opts);
186         nskb = alloc_skb(sizeof(*niph) + tcp_hdr_size + MAX_TCP_HEADER,
187                          GFP_ATOMIC);
188         if (nskb == NULL)
189                 return;
190         skb_reserve(nskb, MAX_TCP_HEADER);
191
192         niph = synproxy_build_ip(nskb, &iph->daddr, &iph->saddr);
193
194         skb_reset_transport_header(nskb);
195         nth = (struct tcphdr *)skb_put(nskb, tcp_hdr_size);
196         nth->source     = th->dest;
197         nth->dest       = th->source;
198         nth->seq        = htonl(ntohl(th->ack_seq));
199         nth->ack_seq    = htonl(ntohl(th->seq) + 1);
200         tcp_flag_word(nth) = TCP_FLAG_ACK;
201         nth->doff       = tcp_hdr_size / 4;
202         nth->window     = htons(state->seen[IP_CT_DIR_ORIGINAL].td_maxwin);
203         nth->check      = 0;
204         nth->urg_ptr    = 0;
205
206         synproxy_build_options(nth, opts);
207
208         synproxy_send_tcp(snet, skb, nskb, NULL, 0, niph, nth, tcp_hdr_size);
209 }
210
211 static void
212 synproxy_send_client_ack(const struct synproxy_net *snet,
213                          const struct sk_buff *skb, const struct tcphdr *th,
214                          const struct synproxy_options *opts)
215 {
216         struct sk_buff *nskb;
217         struct ipv6hdr *iph, *niph;
218         struct tcphdr *nth;
219         unsigned int tcp_hdr_size;
220
221         iph = ipv6_hdr(skb);
222
223         tcp_hdr_size = sizeof(*nth) + synproxy_options_size(opts);
224         nskb = alloc_skb(sizeof(*niph) + tcp_hdr_size + MAX_TCP_HEADER,
225                          GFP_ATOMIC);
226         if (nskb == NULL)
227                 return;
228         skb_reserve(nskb, MAX_TCP_HEADER);
229
230         niph = synproxy_build_ip(nskb, &iph->saddr, &iph->daddr);
231
232         skb_reset_transport_header(nskb);
233         nth = (struct tcphdr *)skb_put(nskb, tcp_hdr_size);
234         nth->source     = th->source;
235         nth->dest       = th->dest;
236         nth->seq        = htonl(ntohl(th->seq) + 1);
237         nth->ack_seq    = th->ack_seq;
238         tcp_flag_word(nth) = TCP_FLAG_ACK;
239         nth->doff       = tcp_hdr_size / 4;
240         nth->window     = htons(ntohs(th->window) >> opts->wscale);
241         nth->check      = 0;
242         nth->urg_ptr    = 0;
243
244         synproxy_build_options(nth, opts);
245
246         synproxy_send_tcp(snet, skb, nskb, skb->nfct, IP_CT_ESTABLISHED_REPLY,
247                           niph, nth, tcp_hdr_size);
248 }
249
250 static bool
251 synproxy_recv_client_ack(const struct synproxy_net *snet,
252                          const struct sk_buff *skb, const struct tcphdr *th,
253                          struct synproxy_options *opts, u32 recv_seq)
254 {
255         int mss;
256
257         mss = __cookie_v6_check(ipv6_hdr(skb), th, ntohl(th->ack_seq) - 1);
258         if (mss == 0) {
259                 this_cpu_inc(snet->stats->cookie_invalid);
260                 return false;
261         }
262
263         this_cpu_inc(snet->stats->cookie_valid);
264         opts->mss = mss;
265         opts->options |= XT_SYNPROXY_OPT_MSS;
266
267         if (opts->options & XT_SYNPROXY_OPT_TIMESTAMP)
268                 synproxy_check_timestamp_cookie(opts);
269
270         synproxy_send_server_syn(snet, skb, th, opts, recv_seq);
271         return true;
272 }
273
274 static unsigned int
275 synproxy_tg6(struct sk_buff *skb, const struct xt_action_param *par)
276 {
277         const struct xt_synproxy_info *info = par->targinfo;
278         struct synproxy_net *snet = synproxy_pernet(par->net);
279         struct synproxy_options opts = {};
280         struct tcphdr *th, _th;
281
282         if (nf_ip6_checksum(skb, par->hooknum, par->thoff, IPPROTO_TCP))
283                 return NF_DROP;
284
285         th = skb_header_pointer(skb, par->thoff, sizeof(_th), &_th);
286         if (th == NULL)
287                 return NF_DROP;
288
289         if (!synproxy_parse_options(skb, par->thoff, th, &opts))
290                 return NF_DROP;
291
292         if (th->syn && !(th->ack || th->fin || th->rst)) {
293                 /* Initial SYN from client */
294                 this_cpu_inc(snet->stats->syn_received);
295
296                 if (th->ece && th->cwr)
297                         opts.options |= XT_SYNPROXY_OPT_ECN;
298
299                 opts.options &= info->options;
300                 if (opts.options & XT_SYNPROXY_OPT_TIMESTAMP)
301                         synproxy_init_timestamp_cookie(info, &opts);
302                 else
303                         opts.options &= ~(XT_SYNPROXY_OPT_WSCALE |
304                                           XT_SYNPROXY_OPT_SACK_PERM |
305                                           XT_SYNPROXY_OPT_ECN);
306
307                 synproxy_send_client_synack(snet, skb, th, &opts);
308                 return NF_DROP;
309
310         } else if (th->ack && !(th->fin || th->rst || th->syn)) {
311                 /* ACK from client */
312                 synproxy_recv_client_ack(snet, skb, th, &opts, ntohl(th->seq));
313                 return NF_DROP;
314         }
315
316         return XT_CONTINUE;
317 }
318
319 static unsigned int ipv6_synproxy_hook(void *priv,
320                                        struct sk_buff *skb,
321                                        const struct nf_hook_state *nhs)
322 {
323         struct synproxy_net *snet = synproxy_pernet(nhs->net);
324         enum ip_conntrack_info ctinfo;
325         struct nf_conn *ct;
326         struct nf_conn_synproxy *synproxy;
327         struct synproxy_options opts = {};
328         const struct ip_ct_tcp *state;
329         struct tcphdr *th, _th;
330         __be16 frag_off;
331         u8 nexthdr;
332         int thoff;
333
334         ct = nf_ct_get(skb, &ctinfo);
335         if (ct == NULL)
336                 return NF_ACCEPT;
337
338         synproxy = nfct_synproxy(ct);
339         if (synproxy == NULL)
340                 return NF_ACCEPT;
341
342         if (nf_is_loopback_packet(skb))
343                 return NF_ACCEPT;
344
345         nexthdr = ipv6_hdr(skb)->nexthdr;
346         thoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
347                                  &frag_off);
348         if (thoff < 0)
349                 return NF_ACCEPT;
350
351         th = skb_header_pointer(skb, thoff, sizeof(_th), &_th);
352         if (th == NULL)
353                 return NF_DROP;
354
355         state = &ct->proto.tcp;
356         switch (state->state) {
357         case TCP_CONNTRACK_CLOSE:
358                 if (th->rst && !test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
359                         nf_ct_seqadj_init(ct, ctinfo, synproxy->isn -
360                                                       ntohl(th->seq) + 1);
361                         break;
362                 }
363
364                 if (!th->syn || th->ack ||
365                     CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL)
366                         break;
367
368                 /* Reopened connection - reset the sequence number and timestamp
369                  * adjustments, they will get initialized once the connection is
370                  * reestablished.
371                  */
372                 nf_ct_seqadj_init(ct, ctinfo, 0);
373                 synproxy->tsoff = 0;
374                 this_cpu_inc(snet->stats->conn_reopened);
375
376                 /* fall through */
377         case TCP_CONNTRACK_SYN_SENT:
378                 if (!synproxy_parse_options(skb, thoff, th, &opts))
379                         return NF_DROP;
380
381                 if (!th->syn && th->ack &&
382                     CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) {
383                         /* Keep-Alives are sent with SEG.SEQ = SND.NXT-1,
384                          * therefore we need to add 1 to make the SYN sequence
385                          * number match the one of first SYN.
386                          */
387                         if (synproxy_recv_client_ack(snet, skb, th, &opts,
388                                                      ntohl(th->seq) + 1))
389                                 this_cpu_inc(snet->stats->cookie_retrans);
390
391                         return NF_DROP;
392                 }
393
394                 synproxy->isn = ntohl(th->ack_seq);
395                 if (opts.options & XT_SYNPROXY_OPT_TIMESTAMP)
396                         synproxy->its = opts.tsecr;
397                 break;
398         case TCP_CONNTRACK_SYN_RECV:
399                 if (!th->syn || !th->ack)
400                         break;
401
402                 if (!synproxy_parse_options(skb, thoff, th, &opts))
403                         return NF_DROP;
404
405                 if (opts.options & XT_SYNPROXY_OPT_TIMESTAMP)
406                         synproxy->tsoff = opts.tsval - synproxy->its;
407
408                 opts.options &= ~(XT_SYNPROXY_OPT_MSS |
409                                   XT_SYNPROXY_OPT_WSCALE |
410                                   XT_SYNPROXY_OPT_SACK_PERM);
411
412                 swap(opts.tsval, opts.tsecr);
413                 synproxy_send_server_ack(snet, state, skb, th, &opts);
414
415                 nf_ct_seqadj_init(ct, ctinfo, synproxy->isn - ntohl(th->seq));
416
417                 swap(opts.tsval, opts.tsecr);
418                 synproxy_send_client_ack(snet, skb, th, &opts);
419
420                 consume_skb(skb);
421                 return NF_STOLEN;
422         default:
423                 break;
424         }
425
426         synproxy_tstamp_adjust(skb, thoff, th, ct, ctinfo, synproxy);
427         return NF_ACCEPT;
428 }
429
430 static int synproxy_tg6_check(const struct xt_tgchk_param *par)
431 {
432         const struct ip6t_entry *e = par->entryinfo;
433
434         if (!(e->ipv6.flags & IP6T_F_PROTO) ||
435             e->ipv6.proto != IPPROTO_TCP ||
436             e->ipv6.invflags & XT_INV_PROTO)
437                 return -EINVAL;
438
439         return nf_ct_l3proto_try_module_get(par->family);
440 }
441
442 static void synproxy_tg6_destroy(const struct xt_tgdtor_param *par)
443 {
444         nf_ct_l3proto_module_put(par->family);
445 }
446
447 static struct xt_target synproxy_tg6_reg __read_mostly = {
448         .name           = "SYNPROXY",
449         .family         = NFPROTO_IPV6,
450         .hooks          = (1 << NF_INET_LOCAL_IN) | (1 << NF_INET_FORWARD),
451         .target         = synproxy_tg6,
452         .targetsize     = sizeof(struct xt_synproxy_info),
453         .checkentry     = synproxy_tg6_check,
454         .destroy        = synproxy_tg6_destroy,
455         .me             = THIS_MODULE,
456 };
457
458 static struct nf_hook_ops ipv6_synproxy_ops[] __read_mostly = {
459         {
460                 .hook           = ipv6_synproxy_hook,
461                 .pf             = NFPROTO_IPV6,
462                 .hooknum        = NF_INET_LOCAL_IN,
463                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM - 1,
464         },
465         {
466                 .hook           = ipv6_synproxy_hook,
467                 .pf             = NFPROTO_IPV6,
468                 .hooknum        = NF_INET_POST_ROUTING,
469                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM - 1,
470         },
471 };
472
473 static int __init synproxy_tg6_init(void)
474 {
475         int err;
476
477         err = nf_register_hooks(ipv6_synproxy_ops,
478                                 ARRAY_SIZE(ipv6_synproxy_ops));
479         if (err < 0)
480                 goto err1;
481
482         err = xt_register_target(&synproxy_tg6_reg);
483         if (err < 0)
484                 goto err2;
485
486         return 0;
487
488 err2:
489         nf_unregister_hooks(ipv6_synproxy_ops, ARRAY_SIZE(ipv6_synproxy_ops));
490 err1:
491         return err;
492 }
493
494 static void __exit synproxy_tg6_exit(void)
495 {
496         xt_unregister_target(&synproxy_tg6_reg);
497         nf_unregister_hooks(ipv6_synproxy_ops, ARRAY_SIZE(ipv6_synproxy_ops));
498 }
499
500 module_init(synproxy_tg6_init);
501 module_exit(synproxy_tg6_exit);
502
503 MODULE_LICENSE("GPL");
504 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");