net/netfilter/xt_TCPMSS.c

   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>
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 as
   8  * published by the Free Software Foundation.
   9  */
  10
  11 #include <linux/module.h>
  12 #include <linux/skbuff.h>
  13 #include <linux/ip.h>
  14 #include <linux/ipv6.h>
  15 #include <linux/tcp.h>
  16 #include <net/ipv6.h>
  17 #include <net/tcp.h>
  18
  19 #include <linux/netfilter_ipv4/ip_tables.h>
  20 #include <linux/netfilter_ipv6/ip6_tables.h>
  21 #include <linux/netfilter/x_tables.h>
  22 #include <linux/netfilter/xt_tcpudp.h>
  23 #include <linux/netfilter/xt_TCPMSS.h>
  24
  25 MODULE_LICENSE("GPL");
  26 MODULE_AUTHOR("Marc Boucher <marc@mbsi.ca>");
  27 MODULE_DESCRIPTION("Xtables: TCP Maximum Segment Size (MSS) adjustment");
  28 MODULE_ALIAS("ipt_TCPMSS");
  29 MODULE_ALIAS("ip6t_TCPMSS");
  30
  31 static inline unsigned int
  32 optlen(const u_int8_t *opt, unsigned int offset)
  33 {
  34         /* Beware zero-length options: make finite progress */
  35         if (opt[offset] <= TCPOPT_NOP || opt[offset+1] == 0)
  36                 return 1;
  37         else
  38                 return opt[offset+1];
  39 }
  40
  41 static int
  42 tcpmss_mangle_packet(struct sk_buff *skb,
  43                      const struct xt_tcpmss_info *info,
  44                      unsigned int tcphoff,
  45                      unsigned int minlen)
  46 {
  47         struct tcphdr *tcph;
  48         unsigned int tcplen, i;
  49         __be16 oldval;
  50         u16 newmss;
  51         u8 *opt;
  52
  53         if (!skb_make_writable(skb, skb->len))
  54                 return -1;
  55
  56         tcplen = skb->len - tcphoff;
  57         tcph = (struct tcphdr *)(skb_network_header(skb) + tcphoff);
  58
  59         /* Since it passed flags test in tcp match, we know it is is
  60            not a fragment, and has data >= tcp header length.  SYN
  61            packets should not contain data: if they did, then we risk
  62            running over MTU, sending Frag Needed and breaking things
  63            badly. --RR */
  64         if (tcplen != tcph->doff*4) {
  65                 if (net_ratelimit())
  66                         printk(KERN_ERR "xt_TCPMSS: bad length (%u bytes)\n",
  67                                skb->len);
  68                 return -1;
  69         }
  70
  71         if (info->mss == XT_TCPMSS_CLAMP_PMTU) {
  72                 if (dst_mtu(skb->dst) <= minlen) {
  73                         if (net_ratelimit())
  74                                 printk(KERN_ERR "xt_TCPMSS: "
  75                                        "unknown or invalid path-MTU (%u)\n",
  76                                        dst_mtu(skb->dst));
  77                         return -1;
  78                 }
  79                 newmss = dst_mtu(skb->dst) - minlen;
  80         } else
  81                 newmss = info->mss;
  82
  83         opt = (u_int8_t *)tcph;
  84         for (i = sizeof(struct tcphdr); i < tcph->doff*4; i += optlen(opt, i)) {
  85                 if (opt[i] == TCPOPT_MSS && tcph->doff*4 - i >= TCPOLEN_MSS &&
  86                     opt[i+1] == TCPOLEN_MSS) {
  87                         u_int16_t oldmss;
  88
  89                         oldmss = (opt[i+2] << 8) | opt[i+3];
  90
  91                         /* Never increase MSS, even when setting it, as
  92                          * doing so results in problems for hosts that rely
  93                          * on MSS being set correctly.
  94                          */
  95                         if (oldmss <= newmss)
  96                                 return 0;
  97
  98                         opt[i+2] = (newmss & 0xff00) >> 8;
  99                         opt[i+3] = newmss & 0x00ff;
 100
 101                         inet_proto_csum_replace2(&tcph->check, skb,
 102                                                  htons(oldmss), htons(newmss),
 103                                                  0);
 104                         return 0;
 105                 }
 106         }
 107
 108         /*
 109          * MSS Option not found ?! add it..
 110          */
 111         if (skb_tailroom(skb) < TCPOLEN_MSS) {
 112                 if (pskb_expand_head(skb, 0,
 113                                      TCPOLEN_MSS - skb_tailroom(skb),
 114                                      GFP_ATOMIC))
 115                         return -1;
 116                 tcph = (struct tcphdr *)(skb_network_header(skb) + tcphoff);
 117         }
 118
 119         skb_put(skb, TCPOLEN_MSS);
 120
 121         opt = (u_int8_t *)tcph + sizeof(struct tcphdr);
 122         memmove(opt + TCPOLEN_MSS, opt, tcplen - sizeof(struct tcphdr));
 123
 124         inet_proto_csum_replace2(&tcph->check, skb,
 125                                  htons(tcplen), htons(tcplen + TCPOLEN_MSS), 1);
 126         opt[0] = TCPOPT_MSS;
 127         opt[1] = TCPOLEN_MSS;
 128         opt[2] = (newmss & 0xff00) >> 8;
 129         opt[3] = newmss & 0x00ff;
 130
 131         inet_proto_csum_replace4(&tcph->check, skb, 0, *((__be32 *)opt), 0);
 132
 133         oldval = ((__be16 *)tcph)[6];
 134         tcph->doff += TCPOLEN_MSS/4;
 135         inet_proto_csum_replace2(&tcph->check, skb,
 136                                  oldval, ((__be16 *)tcph)[6], 0);
 137         return TCPOLEN_MSS;
 138 }
 139
 140 static unsigned int
 141 tcpmss_tg4(struct sk_buff *skb, const struct net_device *in,
 142            const struct net_device *out, unsigned int hooknum,
 143            const struct xt_target *target, const void *targinfo)
 144 {
 145         struct iphdr *iph = ip_hdr(skb);
 146         __be16 newlen;
 147         int ret;
 148
 149         ret = tcpmss_mangle_packet(skb, targinfo, iph->ihl * 4,
 150                                    sizeof(*iph) + sizeof(struct tcphdr));
 151         if (ret < 0)
 152                 return NF_DROP;
 153         if (ret > 0) {
 154                 iph = ip_hdr(skb);
 155                 newlen = htons(ntohs(iph->tot_len) + ret);
 156                 csum_replace2(&iph->check, iph->tot_len, newlen);
 157                 iph->tot_len = newlen;
 158         }
 159         return XT_CONTINUE;
 160 }
 161
 162 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
 163 static unsigned int
 164 tcpmss_tg6(struct sk_buff *skb, const struct net_device *in,
 165            const struct net_device *out, unsigned int hooknum,
 166            const struct xt_target *target, const void *targinfo)
 167 {
 168         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 169         u8 nexthdr;
 170         int tcphoff;
 171         int ret;
 172
 173         nexthdr = ipv6h->nexthdr;
 174         tcphoff = ipv6_skip_exthdr(skb, sizeof(*ipv6h), &nexthdr);
 175         if (tcphoff < 0)
 176                 return NF_DROP;
 177         ret = tcpmss_mangle_packet(skb, targinfo, tcphoff,
 178                                    sizeof(*ipv6h) + sizeof(struct tcphdr));
 179         if (ret < 0)
 180                 return NF_DROP;
 181         if (ret > 0) {
 182                 ipv6h = ipv6_hdr(skb);
 183                 ipv6h->payload_len = htons(ntohs(ipv6h->payload_len) + ret);
 184         }
 185         return XT_CONTINUE;
 186 }
 187 #endif
 188
 189 #define TH_SYN 0x02
 190
 191 /* Must specify -p tcp --syn */
 192 static inline bool find_syn_match(const struct xt_entry_match *m)
 193 {
 194         const struct xt_tcp *tcpinfo = (const struct xt_tcp *)m->data;
 195
 196         if (strcmp(m->u.kernel.match->name, "tcp") == 0 &&
 197             tcpinfo->flg_cmp & TH_SYN &&
 198             !(tcpinfo->invflags & XT_TCP_INV_FLAGS))
 199                 return true;
 200
 201         return false;
 202 }
 203
 204 static bool
 205 tcpmss_tg4_check(const char *tablename, const void *entry,
 206                  const struct xt_target *target, void *targinfo,
 207                  unsigned int hook_mask)
 208 {
 209         const struct xt_tcpmss_info *info = targinfo;
 210         const struct ipt_entry *e = entry;
 211
 212         if (info->mss == XT_TCPMSS_CLAMP_PMTU &&
 213             (hook_mask & ~((1 << NF_INET_FORWARD) |
 214                            (1 << NF_INET_LOCAL_OUT) |
 215                            (1 << NF_INET_POST_ROUTING))) != 0) {
 216                 printk("xt_TCPMSS: path-MTU clamping only supported in "
 217                        "FORWARD, OUTPUT and POSTROUTING hooks\n");
 218                 return false;
 219         }
 220         if (IPT_MATCH_ITERATE(e, find_syn_match))
 221                 return true;
 222         printk("xt_TCPMSS: Only works on TCP SYN packets\n");
 223         return false;
 224 }
 225
 226 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
 227 static bool
 228 tcpmss_tg6_check(const char *tablename, const void *entry,
 229                  const struct xt_target *target, void *targinfo,
 230                  unsigned int hook_mask)
 231 {
 232         const struct xt_tcpmss_info *info = targinfo;
 233         const struct ip6t_entry *e = entry;
 234
 235         if (info->mss == XT_TCPMSS_CLAMP_PMTU &&
 236             (hook_mask & ~((1 << NF_INET_FORWARD) |
 237                            (1 << NF_INET_LOCAL_OUT) |
 238                            (1 << NF_INET_POST_ROUTING))) != 0) {
 239                 printk("xt_TCPMSS: path-MTU clamping only supported in "
 240                        "FORWARD, OUTPUT and POSTROUTING hooks\n");
 241                 return false;
 242         }
 243         if (IP6T_MATCH_ITERATE(e, find_syn_match))
 244                 return true;
 245         printk("xt_TCPMSS: Only works on TCP SYN packets\n");
 246         return false;
 247 }
 248 #endif
 249
 250 static struct xt_target tcpmss_tg_reg[] __read_mostly = {
 251         {
 252                 .family         = AF_INET,
 253                 .name           = "TCPMSS",
 254                 .checkentry     = tcpmss_tg4_check,
 255                 .target         = tcpmss_tg4,
 256                 .targetsize     = sizeof(struct xt_tcpmss_info),
 257                 .proto          = IPPROTO_TCP,
 258                 .me             = THIS_MODULE,
 259         },
 260 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
 261         {
 262                 .family         = AF_INET6,
 263                 .name           = "TCPMSS",
 264                 .checkentry     = tcpmss_tg6_check,
 265                 .target         = tcpmss_tg6,
 266                 .targetsize     = sizeof(struct xt_tcpmss_info),
 267                 .proto          = IPPROTO_TCP,
 268                 .me             = THIS_MODULE,
 269         },
 270 #endif
 271 };
 272
 273 static int __init tcpmss_tg_init(void)
 274 {
 275         return xt_register_targets(tcpmss_tg_reg, ARRAY_SIZE(tcpmss_tg_reg));
 276 }
 277
 278 static void __exit tcpmss_tg_exit(void)
 279 {
 280         xt_unregister_targets(tcpmss_tg_reg, ARRAY_SIZE(tcpmss_tg_reg));
 281 }
 282
 283 module_init(tcpmss_tg_init);
 284 module_exit(tcpmss_tg_exit);