xen-blkfront: don't add indirect pages to list when !feature_persistent
[linux-2.6-block.git] / net / netfilter / xt_TCPMSS.c
CommitLineData
cdd289a2
PM
1/*
2 * This is a module which is used for setting the MSS option in TCP packets.
3 *
4 * Copyright (C) 2000 Marc Boucher <marc@mbsi.ca>
f229f6ce 5 * Copyright (C) 2007 Patrick McHardy <kaber@trash.net>
cdd289a2
PM
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
8bee4bad 11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
cdd289a2
PM
12#include <linux/module.h>
13#include <linux/skbuff.h>
14#include <linux/ip.h>
5a0e3ad6 15#include <linux/gfp.h>
cdd289a2
PM
16#include <linux/ipv6.h>
17#include <linux/tcp.h>
37c08387
JE
18#include <net/dst.h>
19#include <net/flow.h>
cdd289a2 20#include <net/ipv6.h>
37c08387 21#include <net/route.h>
cdd289a2
PM
22#include <net/tcp.h>
23
24#include <linux/netfilter_ipv4/ip_tables.h>
25#include <linux/netfilter_ipv6/ip6_tables.h>
26#include <linux/netfilter/x_tables.h>
27#include <linux/netfilter/xt_tcpudp.h>
28#include <linux/netfilter/xt_TCPMSS.h>
29
30MODULE_LICENSE("GPL");
31MODULE_AUTHOR("Marc Boucher <marc@mbsi.ca>");
2ae15b64 32MODULE_DESCRIPTION("Xtables: TCP Maximum Segment Size (MSS) adjustment");
cdd289a2
PM
33MODULE_ALIAS("ipt_TCPMSS");
34MODULE_ALIAS("ip6t_TCPMSS");
35
36static inline unsigned int
37optlen(const u_int8_t *opt, unsigned int offset)
38{
39 /* Beware zero-length options: make finite progress */
40 if (opt[offset] <= TCPOPT_NOP || opt[offset+1] == 0)
41 return 1;
42 else
43 return opt[offset+1];
44}
45
7722e0d1
G
46static u_int32_t tcpmss_reverse_mtu(struct net *net,
47 const struct sk_buff *skb,
de1389b1
G
48 unsigned int family)
49{
50 struct flowi fl;
51 const struct nf_afinfo *ai;
52 struct rtable *rt = NULL;
53 u_int32_t mtu = ~0U;
54
55 if (family == PF_INET) {
56 struct flowi4 *fl4 = &fl.u.ip4;
57 memset(fl4, 0, sizeof(*fl4));
58 fl4->daddr = ip_hdr(skb)->saddr;
59 } else {
60 struct flowi6 *fl6 = &fl.u.ip6;
61
62 memset(fl6, 0, sizeof(*fl6));
63 fl6->daddr = ipv6_hdr(skb)->saddr;
64 }
65 rcu_read_lock();
66 ai = nf_get_afinfo(family);
67 if (ai != NULL)
7722e0d1 68 ai->route(net, (struct dst_entry **)&rt, &fl, false);
de1389b1
G
69 rcu_read_unlock();
70
71 if (rt != NULL) {
72 mtu = dst_mtu(&rt->dst);
73 dst_release(&rt->dst);
74 }
75 return mtu;
76}
77
cdd289a2 78static int
3db05fea 79tcpmss_mangle_packet(struct sk_buff *skb,
70d19f80 80 const struct xt_action_param *par,
de1389b1 81 unsigned int family,
cdd289a2
PM
82 unsigned int tcphoff,
83 unsigned int minlen)
84{
70d19f80 85 const struct xt_tcpmss_info *info = par->targinfo;
cdd289a2 86 struct tcphdr *tcph;
71ffe9c7
PNA
87 int len, tcp_hdrlen;
88 unsigned int i;
cdd289a2
PM
89 __be16 oldval;
90 u16 newmss;
91 u8 *opt;
92
b396966c
PO
93 /* This is a fragment, no TCP header is available */
94 if (par->fragoff != 0)
1205e1fa 95 return 0;
b396966c 96
3db05fea 97 if (!skb_make_writable(skb, skb->len))
cdd289a2
PM
98 return -1;
99
71ffe9c7
PNA
100 len = skb->len - tcphoff;
101 if (len < (int)sizeof(struct tcphdr))
102 return -1;
103
3db05fea 104 tcph = (struct tcphdr *)(skb_network_header(skb) + tcphoff);
71ffe9c7 105 tcp_hdrlen = tcph->doff * 4;
cdd289a2 106
71ffe9c7 107 if (len < tcp_hdrlen)
cdd289a2 108 return -1;
cdd289a2
PM
109
110 if (info->mss == XT_TCPMSS_CLAMP_PMTU) {
7722e0d1
G
111 struct net *net = dev_net(par->in ? par->in : par->out);
112 unsigned int in_mtu = tcpmss_reverse_mtu(net, skb, family);
de1389b1 113
adf30907 114 if (dst_mtu(skb_dst(skb)) <= minlen) {
e87cc472
JP
115 net_err_ratelimited("unknown or invalid path-MTU (%u)\n",
116 dst_mtu(skb_dst(skb)));
cdd289a2
PM
117 return -1;
118 }
37c08387 119 if (in_mtu <= minlen) {
e87cc472
JP
120 net_err_ratelimited("unknown or invalid path-MTU (%u)\n",
121 in_mtu);
37c08387
JE
122 return -1;
123 }
adf30907 124 newmss = min(dst_mtu(skb_dst(skb)), in_mtu) - minlen;
cdd289a2
PM
125 } else
126 newmss = info->mss;
127
128 opt = (u_int8_t *)tcph;
71ffe9c7
PNA
129 for (i = sizeof(struct tcphdr); i <= tcp_hdrlen - TCPOLEN_MSS; i += optlen(opt, i)) {
130 if (opt[i] == TCPOPT_MSS && opt[i+1] == TCPOLEN_MSS) {
cdd289a2
PM
131 u_int16_t oldmss;
132
133 oldmss = (opt[i+2] << 8) | opt[i+3];
134
17008064
BL
135 /* Never increase MSS, even when setting it, as
136 * doing so results in problems for hosts that rely
137 * on MSS being set correctly.
138 */
139 if (oldmss <= newmss)
cdd289a2
PM
140 return 0;
141
142 opt[i+2] = (newmss & 0xff00) >> 8;
7c4e36bc 143 opt[i+3] = newmss & 0x00ff;
cdd289a2 144
be0ea7d5
PM
145 inet_proto_csum_replace2(&tcph->check, skb,
146 htons(oldmss), htons(newmss),
147 0);
cdd289a2
PM
148 return 0;
149 }
150 }
151
10a19939 152 /* There is data after the header so the option can't be added
71ffe9c7
PNA
153 * without moving it, and doing so may make the SYN packet
154 * itself too large. Accept the packet unmodified instead.
155 */
156 if (len > tcp_hdrlen)
10a19939
SA
157 return 0;
158
cdd289a2
PM
159 /*
160 * MSS Option not found ?! add it..
161 */
3db05fea
HX
162 if (skb_tailroom(skb) < TCPOLEN_MSS) {
163 if (pskb_expand_head(skb, 0,
164 TCPOLEN_MSS - skb_tailroom(skb),
2ca7b0ac 165 GFP_ATOMIC))
cdd289a2 166 return -1;
3db05fea 167 tcph = (struct tcphdr *)(skb_network_header(skb) + tcphoff);
cdd289a2
PM
168 }
169
3db05fea 170 skb_put(skb, TCPOLEN_MSS);
cdd289a2 171
70d19f80
PO
172 /*
173 * IPv4: RFC 1122 states "If an MSS option is not received at
174 * connection setup, TCP MUST assume a default send MSS of 536".
175 * IPv6: RFC 2460 states IPv6 has a minimum MTU of 1280 and a minimum
176 * length IPv6 header of 60, ergo the default MSS value is 1220
177 * Since no MSS was provided, we must use the default values
409b545a 178 */
70d19f80
PO
179 if (par->family == NFPROTO_IPV4)
180 newmss = min(newmss, (u16)536);
181 else
182 newmss = min(newmss, (u16)1220);
409b545a 183
cdd289a2 184 opt = (u_int8_t *)tcph + sizeof(struct tcphdr);
71ffe9c7 185 memmove(opt + TCPOLEN_MSS, opt, len - sizeof(struct tcphdr));
cdd289a2 186
be0ea7d5 187 inet_proto_csum_replace2(&tcph->check, skb,
71ffe9c7 188 htons(len), htons(len + TCPOLEN_MSS), 1);
cdd289a2
PM
189 opt[0] = TCPOPT_MSS;
190 opt[1] = TCPOLEN_MSS;
191 opt[2] = (newmss & 0xff00) >> 8;
7c4e36bc 192 opt[3] = newmss & 0x00ff;
cdd289a2 193
be0ea7d5 194 inet_proto_csum_replace4(&tcph->check, skb, 0, *((__be32 *)opt), 0);
cdd289a2
PM
195
196 oldval = ((__be16 *)tcph)[6];
197 tcph->doff += TCPOLEN_MSS/4;
be0ea7d5
PM
198 inet_proto_csum_replace2(&tcph->check, skb,
199 oldval, ((__be16 *)tcph)[6], 0);
cdd289a2
PM
200 return TCPOLEN_MSS;
201}
202
203static unsigned int
4b560b44 204tcpmss_tg4(struct sk_buff *skb, const struct xt_action_param *par)
cdd289a2 205{
3db05fea 206 struct iphdr *iph = ip_hdr(skb);
cdd289a2
PM
207 __be16 newlen;
208 int ret;
209
70d19f80 210 ret = tcpmss_mangle_packet(skb, par,
de1389b1 211 PF_INET,
37c08387 212 iph->ihl * 4,
cdd289a2
PM
213 sizeof(*iph) + sizeof(struct tcphdr));
214 if (ret < 0)
215 return NF_DROP;
216 if (ret > 0) {
3db05fea 217 iph = ip_hdr(skb);
cdd289a2 218 newlen = htons(ntohs(iph->tot_len) + ret);
be0ea7d5 219 csum_replace2(&iph->check, iph->tot_len, newlen);
cdd289a2
PM
220 iph->tot_len = newlen;
221 }
222 return XT_CONTINUE;
223}
224
c0cd1156 225#if IS_ENABLED(CONFIG_IP6_NF_IPTABLES)
cdd289a2 226static unsigned int
4b560b44 227tcpmss_tg6(struct sk_buff *skb, const struct xt_action_param *par)
cdd289a2 228{
3db05fea 229 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
cdd289a2 230 u8 nexthdr;
75f2811c 231 __be16 frag_off;
cdd289a2
PM
232 int tcphoff;
233 int ret;
234
235 nexthdr = ipv6h->nexthdr;
75f2811c 236 tcphoff = ipv6_skip_exthdr(skb, sizeof(*ipv6h), &nexthdr, &frag_off);
9dc0564e 237 if (tcphoff < 0)
cdd289a2 238 return NF_DROP;
70d19f80 239 ret = tcpmss_mangle_packet(skb, par,
de1389b1 240 PF_INET6,
37c08387 241 tcphoff,
cdd289a2
PM
242 sizeof(*ipv6h) + sizeof(struct tcphdr));
243 if (ret < 0)
244 return NF_DROP;
245 if (ret > 0) {
3db05fea 246 ipv6h = ipv6_hdr(skb);
cdd289a2
PM
247 ipv6h->payload_len = htons(ntohs(ipv6h->payload_len) + ret);
248 }
249 return XT_CONTINUE;
250}
251#endif
252
cdd289a2 253/* Must specify -p tcp --syn */
e1931b78 254static inline bool find_syn_match(const struct xt_entry_match *m)
cdd289a2
PM
255{
256 const struct xt_tcp *tcpinfo = (const struct xt_tcp *)m->data;
257
258 if (strcmp(m->u.kernel.match->name, "tcp") == 0 &&
a3433f35 259 tcpinfo->flg_cmp & TCPHDR_SYN &&
cdd289a2 260 !(tcpinfo->invflags & XT_TCP_INV_FLAGS))
e1931b78 261 return true;
cdd289a2 262
e1931b78 263 return false;
cdd289a2
PM
264}
265
135367b8 266static int tcpmss_tg4_check(const struct xt_tgchk_param *par)
cdd289a2 267{
af5d6dc2
JE
268 const struct xt_tcpmss_info *info = par->targinfo;
269 const struct ipt_entry *e = par->entryinfo;
dcea992a 270 const struct xt_entry_match *ematch;
cdd289a2
PM
271
272 if (info->mss == XT_TCPMSS_CLAMP_PMTU &&
af5d6dc2 273 (par->hook_mask & ~((1 << NF_INET_FORWARD) |
6e23ae2a
PM
274 (1 << NF_INET_LOCAL_OUT) |
275 (1 << NF_INET_POST_ROUTING))) != 0) {
8bee4bad
JE
276 pr_info("path-MTU clamping only supported in "
277 "FORWARD, OUTPUT and POSTROUTING hooks\n");
d6b00a53 278 return -EINVAL;
cdd289a2 279 }
dcea992a
JE
280 xt_ematch_foreach(ematch, e)
281 if (find_syn_match(ematch))
d6b00a53 282 return 0;
8bee4bad 283 pr_info("Only works on TCP SYN packets\n");
d6b00a53 284 return -EINVAL;
cdd289a2
PM
285}
286
c0cd1156 287#if IS_ENABLED(CONFIG_IP6_NF_IPTABLES)
135367b8 288static int tcpmss_tg6_check(const struct xt_tgchk_param *par)
cdd289a2 289{
af5d6dc2
JE
290 const struct xt_tcpmss_info *info = par->targinfo;
291 const struct ip6t_entry *e = par->entryinfo;
dcea992a 292 const struct xt_entry_match *ematch;
cdd289a2
PM
293
294 if (info->mss == XT_TCPMSS_CLAMP_PMTU &&
af5d6dc2 295 (par->hook_mask & ~((1 << NF_INET_FORWARD) |
6e23ae2a
PM
296 (1 << NF_INET_LOCAL_OUT) |
297 (1 << NF_INET_POST_ROUTING))) != 0) {
8bee4bad
JE
298 pr_info("path-MTU clamping only supported in "
299 "FORWARD, OUTPUT and POSTROUTING hooks\n");
d6b00a53 300 return -EINVAL;
cdd289a2 301 }
dcea992a
JE
302 xt_ematch_foreach(ematch, e)
303 if (find_syn_match(ematch))
d6b00a53 304 return 0;
8bee4bad 305 pr_info("Only works on TCP SYN packets\n");
d6b00a53 306 return -EINVAL;
cdd289a2
PM
307}
308#endif
309
d3c5ee6d 310static struct xt_target tcpmss_tg_reg[] __read_mostly = {
cdd289a2 311 {
ee999d8b 312 .family = NFPROTO_IPV4,
cdd289a2 313 .name = "TCPMSS",
d3c5ee6d
JE
314 .checkentry = tcpmss_tg4_check,
315 .target = tcpmss_tg4,
cdd289a2
PM
316 .targetsize = sizeof(struct xt_tcpmss_info),
317 .proto = IPPROTO_TCP,
318 .me = THIS_MODULE,
319 },
c0cd1156 320#if IS_ENABLED(CONFIG_IP6_NF_IPTABLES)
cdd289a2 321 {
ee999d8b 322 .family = NFPROTO_IPV6,
cdd289a2 323 .name = "TCPMSS",
d3c5ee6d
JE
324 .checkentry = tcpmss_tg6_check,
325 .target = tcpmss_tg6,
cdd289a2
PM
326 .targetsize = sizeof(struct xt_tcpmss_info),
327 .proto = IPPROTO_TCP,
328 .me = THIS_MODULE,
329 },
330#endif
331};
332
d3c5ee6d 333static int __init tcpmss_tg_init(void)
cdd289a2 334{
d3c5ee6d 335 return xt_register_targets(tcpmss_tg_reg, ARRAY_SIZE(tcpmss_tg_reg));
cdd289a2
PM
336}
337
d3c5ee6d 338static void __exit tcpmss_tg_exit(void)
cdd289a2 339{
d3c5ee6d 340 xt_unregister_targets(tcpmss_tg_reg, ARRAY_SIZE(tcpmss_tg_reg));
cdd289a2
PM
341}
342
d3c5ee6d
JE
343module_init(tcpmss_tg_init);
344module_exit(tcpmss_tg_exit);