]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - net/netfilter/xt_TCPMSS.c
netfilter: nf_defrag_ipv6.o included twice
[mirror_ubuntu-bionic-kernel.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
46static int
3db05fea 47tcpmss_mangle_packet(struct sk_buff *skb,
70d19f80 48 const struct xt_action_param *par,
37c08387 49 unsigned int in_mtu,
cdd289a2
PM
50 unsigned int tcphoff,
51 unsigned int minlen)
52{
70d19f80 53 const struct xt_tcpmss_info *info = par->targinfo;
cdd289a2 54 struct tcphdr *tcph;
71ffe9c7
PNA
55 int len, tcp_hdrlen;
56 unsigned int i;
cdd289a2
PM
57 __be16 oldval;
58 u16 newmss;
59 u8 *opt;
60
b396966c
PO
61 /* This is a fragment, no TCP header is available */
62 if (par->fragoff != 0)
63 return XT_CONTINUE;
64
3db05fea 65 if (!skb_make_writable(skb, skb->len))
cdd289a2
PM
66 return -1;
67
71ffe9c7
PNA
68 len = skb->len - tcphoff;
69 if (len < (int)sizeof(struct tcphdr))
70 return -1;
71
3db05fea 72 tcph = (struct tcphdr *)(skb_network_header(skb) + tcphoff);
71ffe9c7 73 tcp_hdrlen = tcph->doff * 4;
cdd289a2 74
71ffe9c7 75 if (len < tcp_hdrlen)
cdd289a2 76 return -1;
cdd289a2
PM
77
78 if (info->mss == XT_TCPMSS_CLAMP_PMTU) {
adf30907 79 if (dst_mtu(skb_dst(skb)) <= minlen) {
e87cc472
JP
80 net_err_ratelimited("unknown or invalid path-MTU (%u)\n",
81 dst_mtu(skb_dst(skb)));
cdd289a2
PM
82 return -1;
83 }
37c08387 84 if (in_mtu <= minlen) {
e87cc472
JP
85 net_err_ratelimited("unknown or invalid path-MTU (%u)\n",
86 in_mtu);
37c08387
JE
87 return -1;
88 }
adf30907 89 newmss = min(dst_mtu(skb_dst(skb)), in_mtu) - minlen;
cdd289a2
PM
90 } else
91 newmss = info->mss;
92
93 opt = (u_int8_t *)tcph;
71ffe9c7
PNA
94 for (i = sizeof(struct tcphdr); i <= tcp_hdrlen - TCPOLEN_MSS; i += optlen(opt, i)) {
95 if (opt[i] == TCPOPT_MSS && opt[i+1] == TCPOLEN_MSS) {
cdd289a2
PM
96 u_int16_t oldmss;
97
98 oldmss = (opt[i+2] << 8) | opt[i+3];
99
17008064
BL
100 /* Never increase MSS, even when setting it, as
101 * doing so results in problems for hosts that rely
102 * on MSS being set correctly.
103 */
104 if (oldmss <= newmss)
cdd289a2
PM
105 return 0;
106
107 opt[i+2] = (newmss & 0xff00) >> 8;
7c4e36bc 108 opt[i+3] = newmss & 0x00ff;
cdd289a2 109
be0ea7d5
PM
110 inet_proto_csum_replace2(&tcph->check, skb,
111 htons(oldmss), htons(newmss),
112 0);
cdd289a2
PM
113 return 0;
114 }
115 }
116
10a19939 117 /* There is data after the header so the option can't be added
71ffe9c7
PNA
118 * without moving it, and doing so may make the SYN packet
119 * itself too large. Accept the packet unmodified instead.
120 */
121 if (len > tcp_hdrlen)
10a19939
SA
122 return 0;
123
cdd289a2
PM
124 /*
125 * MSS Option not found ?! add it..
126 */
3db05fea
HX
127 if (skb_tailroom(skb) < TCPOLEN_MSS) {
128 if (pskb_expand_head(skb, 0,
129 TCPOLEN_MSS - skb_tailroom(skb),
2ca7b0ac 130 GFP_ATOMIC))
cdd289a2 131 return -1;
3db05fea 132 tcph = (struct tcphdr *)(skb_network_header(skb) + tcphoff);
cdd289a2
PM
133 }
134
3db05fea 135 skb_put(skb, TCPOLEN_MSS);
cdd289a2 136
70d19f80
PO
137 /*
138 * IPv4: RFC 1122 states "If an MSS option is not received at
139 * connection setup, TCP MUST assume a default send MSS of 536".
140 * IPv6: RFC 2460 states IPv6 has a minimum MTU of 1280 and a minimum
141 * length IPv6 header of 60, ergo the default MSS value is 1220
142 * Since no MSS was provided, we must use the default values
409b545a 143 */
70d19f80
PO
144 if (par->family == NFPROTO_IPV4)
145 newmss = min(newmss, (u16)536);
146 else
147 newmss = min(newmss, (u16)1220);
409b545a 148
cdd289a2 149 opt = (u_int8_t *)tcph + sizeof(struct tcphdr);
71ffe9c7 150 memmove(opt + TCPOLEN_MSS, opt, len - sizeof(struct tcphdr));
cdd289a2 151
be0ea7d5 152 inet_proto_csum_replace2(&tcph->check, skb,
71ffe9c7 153 htons(len), htons(len + TCPOLEN_MSS), 1);
cdd289a2
PM
154 opt[0] = TCPOPT_MSS;
155 opt[1] = TCPOLEN_MSS;
156 opt[2] = (newmss & 0xff00) >> 8;
7c4e36bc 157 opt[3] = newmss & 0x00ff;
cdd289a2 158
be0ea7d5 159 inet_proto_csum_replace4(&tcph->check, skb, 0, *((__be32 *)opt), 0);
cdd289a2
PM
160
161 oldval = ((__be16 *)tcph)[6];
162 tcph->doff += TCPOLEN_MSS/4;
be0ea7d5
PM
163 inet_proto_csum_replace2(&tcph->check, skb,
164 oldval, ((__be16 *)tcph)[6], 0);
cdd289a2
PM
165 return TCPOLEN_MSS;
166}
167
db1a75bd
JE
168static u_int32_t tcpmss_reverse_mtu(const struct sk_buff *skb,
169 unsigned int family)
37c08387 170{
a1bbb0e6 171 struct flowi fl;
37c08387
JE
172 const struct nf_afinfo *ai;
173 struct rtable *rt = NULL;
174 u_int32_t mtu = ~0U;
175
a1bbb0e6
DM
176 if (family == PF_INET) {
177 struct flowi4 *fl4 = &fl.u.ip4;
178 memset(fl4, 0, sizeof(*fl4));
179 fl4->daddr = ip_hdr(skb)->saddr;
180 } else {
181 struct flowi6 *fl6 = &fl.u.ip6;
db1a75bd 182
a1bbb0e6 183 memset(fl6, 0, sizeof(*fl6));
4e3fd7a0 184 fl6->daddr = ipv6_hdr(skb)->saddr;
a1bbb0e6 185 }
37c08387 186 rcu_read_lock();
db1a75bd 187 ai = nf_get_afinfo(family);
37c08387 188 if (ai != NULL)
0fae2e77 189 ai->route(&init_net, (struct dst_entry **)&rt, &fl, false);
37c08387
JE
190 rcu_read_unlock();
191
192 if (rt != NULL) {
d8d1f30b
CG
193 mtu = dst_mtu(&rt->dst);
194 dst_release(&rt->dst);
37c08387
JE
195 }
196 return mtu;
197}
198
cdd289a2 199static unsigned int
4b560b44 200tcpmss_tg4(struct sk_buff *skb, const struct xt_action_param *par)
cdd289a2 201{
3db05fea 202 struct iphdr *iph = ip_hdr(skb);
cdd289a2
PM
203 __be16 newlen;
204 int ret;
205
70d19f80 206 ret = tcpmss_mangle_packet(skb, par,
db1a75bd 207 tcpmss_reverse_mtu(skb, PF_INET),
37c08387 208 iph->ihl * 4,
cdd289a2
PM
209 sizeof(*iph) + sizeof(struct tcphdr));
210 if (ret < 0)
211 return NF_DROP;
212 if (ret > 0) {
3db05fea 213 iph = ip_hdr(skb);
cdd289a2 214 newlen = htons(ntohs(iph->tot_len) + ret);
be0ea7d5 215 csum_replace2(&iph->check, iph->tot_len, newlen);
cdd289a2
PM
216 iph->tot_len = newlen;
217 }
218 return XT_CONTINUE;
219}
220
c0cd1156 221#if IS_ENABLED(CONFIG_IP6_NF_IPTABLES)
cdd289a2 222static unsigned int
4b560b44 223tcpmss_tg6(struct sk_buff *skb, const struct xt_action_param *par)
cdd289a2 224{
3db05fea 225 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
cdd289a2 226 u8 nexthdr;
75f2811c 227 __be16 frag_off;
cdd289a2
PM
228 int tcphoff;
229 int ret;
230
231 nexthdr = ipv6h->nexthdr;
75f2811c 232 tcphoff = ipv6_skip_exthdr(skb, sizeof(*ipv6h), &nexthdr, &frag_off);
9dc0564e 233 if (tcphoff < 0)
cdd289a2 234 return NF_DROP;
70d19f80 235 ret = tcpmss_mangle_packet(skb, par,
db1a75bd 236 tcpmss_reverse_mtu(skb, PF_INET6),
37c08387 237 tcphoff,
cdd289a2
PM
238 sizeof(*ipv6h) + sizeof(struct tcphdr));
239 if (ret < 0)
240 return NF_DROP;
241 if (ret > 0) {
3db05fea 242 ipv6h = ipv6_hdr(skb);
cdd289a2
PM
243 ipv6h->payload_len = htons(ntohs(ipv6h->payload_len) + ret);
244 }
245 return XT_CONTINUE;
246}
247#endif
248
cdd289a2 249/* Must specify -p tcp --syn */
e1931b78 250static inline bool find_syn_match(const struct xt_entry_match *m)
cdd289a2
PM
251{
252 const struct xt_tcp *tcpinfo = (const struct xt_tcp *)m->data;
253
254 if (strcmp(m->u.kernel.match->name, "tcp") == 0 &&
a3433f35 255 tcpinfo->flg_cmp & TCPHDR_SYN &&
cdd289a2 256 !(tcpinfo->invflags & XT_TCP_INV_FLAGS))
e1931b78 257 return true;
cdd289a2 258
e1931b78 259 return false;
cdd289a2
PM
260}
261
135367b8 262static int tcpmss_tg4_check(const struct xt_tgchk_param *par)
cdd289a2 263{
af5d6dc2
JE
264 const struct xt_tcpmss_info *info = par->targinfo;
265 const struct ipt_entry *e = par->entryinfo;
dcea992a 266 const struct xt_entry_match *ematch;
cdd289a2
PM
267
268 if (info->mss == XT_TCPMSS_CLAMP_PMTU &&
af5d6dc2 269 (par->hook_mask & ~((1 << NF_INET_FORWARD) |
6e23ae2a
PM
270 (1 << NF_INET_LOCAL_OUT) |
271 (1 << NF_INET_POST_ROUTING))) != 0) {
8bee4bad
JE
272 pr_info("path-MTU clamping only supported in "
273 "FORWARD, OUTPUT and POSTROUTING hooks\n");
d6b00a53 274 return -EINVAL;
cdd289a2 275 }
dcea992a
JE
276 xt_ematch_foreach(ematch, e)
277 if (find_syn_match(ematch))
d6b00a53 278 return 0;
8bee4bad 279 pr_info("Only works on TCP SYN packets\n");
d6b00a53 280 return -EINVAL;
cdd289a2
PM
281}
282
c0cd1156 283#if IS_ENABLED(CONFIG_IP6_NF_IPTABLES)
135367b8 284static int tcpmss_tg6_check(const struct xt_tgchk_param *par)
cdd289a2 285{
af5d6dc2
JE
286 const struct xt_tcpmss_info *info = par->targinfo;
287 const struct ip6t_entry *e = par->entryinfo;
dcea992a 288 const struct xt_entry_match *ematch;
cdd289a2
PM
289
290 if (info->mss == XT_TCPMSS_CLAMP_PMTU &&
af5d6dc2 291 (par->hook_mask & ~((1 << NF_INET_FORWARD) |
6e23ae2a
PM
292 (1 << NF_INET_LOCAL_OUT) |
293 (1 << NF_INET_POST_ROUTING))) != 0) {
8bee4bad
JE
294 pr_info("path-MTU clamping only supported in "
295 "FORWARD, OUTPUT and POSTROUTING hooks\n");
d6b00a53 296 return -EINVAL;
cdd289a2 297 }
dcea992a
JE
298 xt_ematch_foreach(ematch, e)
299 if (find_syn_match(ematch))
d6b00a53 300 return 0;
8bee4bad 301 pr_info("Only works on TCP SYN packets\n");
d6b00a53 302 return -EINVAL;
cdd289a2
PM
303}
304#endif
305
d3c5ee6d 306static struct xt_target tcpmss_tg_reg[] __read_mostly = {
cdd289a2 307 {
ee999d8b 308 .family = NFPROTO_IPV4,
cdd289a2 309 .name = "TCPMSS",
d3c5ee6d
JE
310 .checkentry = tcpmss_tg4_check,
311 .target = tcpmss_tg4,
cdd289a2
PM
312 .targetsize = sizeof(struct xt_tcpmss_info),
313 .proto = IPPROTO_TCP,
314 .me = THIS_MODULE,
315 },
c0cd1156 316#if IS_ENABLED(CONFIG_IP6_NF_IPTABLES)
cdd289a2 317 {
ee999d8b 318 .family = NFPROTO_IPV6,
cdd289a2 319 .name = "TCPMSS",
d3c5ee6d
JE
320 .checkentry = tcpmss_tg6_check,
321 .target = tcpmss_tg6,
cdd289a2
PM
322 .targetsize = sizeof(struct xt_tcpmss_info),
323 .proto = IPPROTO_TCP,
324 .me = THIS_MODULE,
325 },
326#endif
327};
328
d3c5ee6d 329static int __init tcpmss_tg_init(void)
cdd289a2 330{
d3c5ee6d 331 return xt_register_targets(tcpmss_tg_reg, ARRAY_SIZE(tcpmss_tg_reg));
cdd289a2
PM
332}
333
d3c5ee6d 334static void __exit tcpmss_tg_exit(void)
cdd289a2 335{
d3c5ee6d 336 xt_unregister_targets(tcpmss_tg_reg, ARRAY_SIZE(tcpmss_tg_reg));
cdd289a2
PM
337}
338
d3c5ee6d
JE
339module_init(tcpmss_tg_init);
340module_exit(tcpmss_tg_exit);