]> nv-tegra.nvidia Code Review - linux-2.6.git/blobdiff - net/netfilter/xt_TCPMSS.c
x25: use limited socket backlog
[linux-2.6.git] / net / netfilter / xt_TCPMSS.c
index 8e76d1f52fbe3079cf574e41b9750d2b77d2f264..0e357ac9a2a831b85fee911847b4c2427d8255d6 100644 (file)
 #include <linux/ip.h>
 #include <linux/ipv6.h>
 #include <linux/tcp.h>
+#include <net/dst.h>
+#include <net/flow.h>
 #include <net/ipv6.h>
+#include <net/route.h>
 #include <net/tcp.h>
 
 #include <linux/netfilter_ipv4/ip_tables.h>
@@ -24,7 +27,7 @@
 
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("Marc Boucher <marc@mbsi.ca>");
-MODULE_DESCRIPTION("x_tables TCP MSS modification module");
+MODULE_DESCRIPTION("Xtables: TCP Maximum Segment Size (MSS) adjustment");
 MODULE_ALIAS("ipt_TCPMSS");
 MODULE_ALIAS("ip6t_TCPMSS");
 
@@ -41,6 +44,7 @@ optlen(const u_int8_t *opt, unsigned int offset)
 static int
 tcpmss_mangle_packet(struct sk_buff *skb,
                     const struct xt_tcpmss_info *info,
+                    unsigned int in_mtu,
                     unsigned int tcphoff,
                     unsigned int minlen)
 {
@@ -56,27 +60,25 @@ tcpmss_mangle_packet(struct sk_buff *skb,
        tcplen = skb->len - tcphoff;
        tcph = (struct tcphdr *)(skb_network_header(skb) + tcphoff);
 
-       /* Since it passed flags test in tcp match, we know it is is
-          not a fragment, and has data >= tcp header length.  SYN
-          packets should not contain data: if they did, then we risk
-          running over MTU, sending Frag Needed and breaking things
-          badly. --RR */
-       if (tcplen != tcph->doff*4) {
-               if (net_ratelimit())
-                       printk(KERN_ERR "xt_TCPMSS: bad length (%u bytes)\n",
-                              skb->len);
+       /* Header cannot be larger than the packet */
+       if (tcplen < tcph->doff*4)
                return -1;
-       }
 
        if (info->mss == XT_TCPMSS_CLAMP_PMTU) {
-               if (dst_mtu(skb->dst) <= minlen) {
+               if (dst_mtu(skb_dst(skb)) <= minlen) {
                        if (net_ratelimit())
                                printk(KERN_ERR "xt_TCPMSS: "
                                       "unknown or invalid path-MTU (%u)\n",
-                                      dst_mtu(skb->dst));
+                                      dst_mtu(skb_dst(skb)));
+                       return -1;
+               }
+               if (in_mtu <= minlen) {
+                       if (net_ratelimit())
+                               printk(KERN_ERR "xt_TCPMSS: unknown or "
+                                      "invalid path-MTU (%u)\n", in_mtu);
                        return -1;
                }
-               newmss = dst_mtu(skb->dst) - minlen;
+               newmss = min(dst_mtu(skb_dst(skb)), in_mtu) - minlen;
        } else
                newmss = info->mss;
 
@@ -88,19 +90,29 @@ tcpmss_mangle_packet(struct sk_buff *skb,
 
                        oldmss = (opt[i+2] << 8) | opt[i+3];
 
-                       if (info->mss == XT_TCPMSS_CLAMP_PMTU &&
-                           oldmss <= newmss)
+                       /* Never increase MSS, even when setting it, as
+                        * doing so results in problems for hosts that rely
+                        * on MSS being set correctly.
+                        */
+                       if (oldmss <= newmss)
                                return 0;
 
                        opt[i+2] = (newmss & 0xff00) >> 8;
                        opt[i+3] = newmss & 0x00ff;
 
-                       nf_proto_csum_replace2(&tcph->check, skb,
-                                              htons(oldmss), htons(newmss), 0);
+                       inet_proto_csum_replace2(&tcph->check, skb,
+                                                htons(oldmss), htons(newmss),
+                                                0);
                        return 0;
                }
        }
 
+       /* There is data after the header so the option can't be added
+          without moving it, and doing so may make the SYN packet
+          itself too large. Accept the packet unmodified instead. */
+       if (tcplen > tcph->doff*4)
+               return 0;
+
        /*
         * MSS Option not found ?! add it..
         */
@@ -117,42 +129,65 @@ tcpmss_mangle_packet(struct sk_buff *skb,
        opt = (u_int8_t *)tcph + sizeof(struct tcphdr);
        memmove(opt + TCPOLEN_MSS, opt, tcplen - sizeof(struct tcphdr));
 
-       nf_proto_csum_replace2(&tcph->check, skb,
-                              htons(tcplen), htons(tcplen + TCPOLEN_MSS), 1);
+       inet_proto_csum_replace2(&tcph->check, skb,
+                                htons(tcplen), htons(tcplen + TCPOLEN_MSS), 1);
        opt[0] = TCPOPT_MSS;
        opt[1] = TCPOLEN_MSS;
        opt[2] = (newmss & 0xff00) >> 8;
        opt[3] = newmss & 0x00ff;
 
-       nf_proto_csum_replace4(&tcph->check, skb, 0, *((__be32 *)opt), 0);
+       inet_proto_csum_replace4(&tcph->check, skb, 0, *((__be32 *)opt), 0);
 
        oldval = ((__be16 *)tcph)[6];
        tcph->doff += TCPOLEN_MSS/4;
-       nf_proto_csum_replace2(&tcph->check, skb,
-                               oldval, ((__be16 *)tcph)[6], 0);
+       inet_proto_csum_replace2(&tcph->check, skb,
+                                oldval, ((__be16 *)tcph)[6], 0);
        return TCPOLEN_MSS;
 }
 
+static u_int32_t tcpmss_reverse_mtu(const struct sk_buff *skb,
+                                   unsigned int family)
+{
+       struct flowi fl = {};
+       const struct nf_afinfo *ai;
+       struct rtable *rt = NULL;
+       u_int32_t mtu     = ~0U;
+
+       if (family == PF_INET)
+               fl.fl4_dst = ip_hdr(skb)->saddr;
+       else
+               fl.fl6_dst = ipv6_hdr(skb)->saddr;
+
+       rcu_read_lock();
+       ai = nf_get_afinfo(family);
+       if (ai != NULL)
+               ai->route((struct dst_entry **)&rt, &fl);
+       rcu_read_unlock();
+
+       if (rt != NULL) {
+               mtu = dst_mtu(&rt->u.dst);
+               dst_release(&rt->u.dst);
+       }
+       return mtu;
+}
+
 static unsigned int
-xt_tcpmss_target4(struct sk_buff *skb,
-                 const struct net_device *in,
-                 const struct net_device *out,
-                 unsigned int hooknum,
-                 const struct xt_target *target,
-                 const void *targinfo)
+tcpmss_tg4(struct sk_buff *skb, const struct xt_target_param *par)
 {
        struct iphdr *iph = ip_hdr(skb);
        __be16 newlen;
        int ret;
 
-       ret = tcpmss_mangle_packet(skb, targinfo, iph->ihl * 4,
+       ret = tcpmss_mangle_packet(skb, par->targinfo,
+                                  tcpmss_reverse_mtu(skb, PF_INET),
+                                  iph->ihl * 4,
                                   sizeof(*iph) + sizeof(struct tcphdr));
        if (ret < 0)
                return NF_DROP;
        if (ret > 0) {
                iph = ip_hdr(skb);
                newlen = htons(ntohs(iph->tot_len) + ret);
-               nf_csum_replace2(&iph->check, iph->tot_len, newlen);
+               csum_replace2(&iph->check, iph->tot_len, newlen);
                iph->tot_len = newlen;
        }
        return XT_CONTINUE;
@@ -160,12 +195,7 @@ xt_tcpmss_target4(struct sk_buff *skb,
 
 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
 static unsigned int
-xt_tcpmss_target6(struct sk_buff *skb,
-                 const struct net_device *in,
-                 const struct net_device *out,
-                 unsigned int hooknum,
-                 const struct xt_target *target,
-                 const void *targinfo)
+tcpmss_tg6(struct sk_buff *skb, const struct xt_target_param *par)
 {
        struct ipv6hdr *ipv6h = ipv6_hdr(skb);
        u8 nexthdr;
@@ -176,7 +206,9 @@ xt_tcpmss_target6(struct sk_buff *skb,
        tcphoff = ipv6_skip_exthdr(skb, sizeof(*ipv6h), &nexthdr);
        if (tcphoff < 0)
                return NF_DROP;
-       ret = tcpmss_mangle_packet(skb, targinfo, tcphoff,
+       ret = tcpmss_mangle_packet(skb, par->targinfo,
+                                  tcpmss_reverse_mtu(skb, PF_INET6),
+                                  tcphoff,
                                   sizeof(*ipv6h) + sizeof(struct tcphdr));
        if (ret < 0)
                return NF_DROP;
@@ -203,72 +235,66 @@ static inline bool find_syn_match(const struct xt_entry_match *m)
        return false;
 }
 
-static bool
-xt_tcpmss_checkentry4(const char *tablename,
-                     const void *entry,
-                     const struct xt_target *target,
-                     void *targinfo,
-                     unsigned int hook_mask)
+static bool tcpmss_tg4_check(const struct xt_tgchk_param *par)
 {
-       const struct xt_tcpmss_info *info = targinfo;
-       const struct ipt_entry *e = entry;
+       const struct xt_tcpmss_info *info = par->targinfo;
+       const struct ipt_entry *e = par->entryinfo;
+       const struct xt_entry_match *ematch;
 
        if (info->mss == XT_TCPMSS_CLAMP_PMTU &&
-           (hook_mask & ~((1 << NF_IP_FORWARD) |
-                          (1 << NF_IP_LOCAL_OUT) |
-                          (1 << NF_IP_POST_ROUTING))) != 0) {
+           (par->hook_mask & ~((1 << NF_INET_FORWARD) |
+                          (1 << NF_INET_LOCAL_OUT) |
+                          (1 << NF_INET_POST_ROUTING))) != 0) {
                printk("xt_TCPMSS: path-MTU clamping only supported in "
                       "FORWARD, OUTPUT and POSTROUTING hooks\n");
                return false;
        }
-       if (IPT_MATCH_ITERATE(e, find_syn_match))
-               return true;
+       xt_ematch_foreach(ematch, e)
+               if (find_syn_match(ematch))
+                       return true;
        printk("xt_TCPMSS: Only works on TCP SYN packets\n");
        return false;
 }
 
 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
-static bool
-xt_tcpmss_checkentry6(const char *tablename,
-                     const void *entry,
-                     const struct xt_target *target,
-                     void *targinfo,
-                     unsigned int hook_mask)
+static bool tcpmss_tg6_check(const struct xt_tgchk_param *par)
 {
-       const struct xt_tcpmss_info *info = targinfo;
-       const struct ip6t_entry *e = entry;
+       const struct xt_tcpmss_info *info = par->targinfo;
+       const struct ip6t_entry *e = par->entryinfo;
+       const struct xt_entry_match *ematch;
 
        if (info->mss == XT_TCPMSS_CLAMP_PMTU &&
-           (hook_mask & ~((1 << NF_IP6_FORWARD) |
-                          (1 << NF_IP6_LOCAL_OUT) |
-                          (1 << NF_IP6_POST_ROUTING))) != 0) {
+           (par->hook_mask & ~((1 << NF_INET_FORWARD) |
+                          (1 << NF_INET_LOCAL_OUT) |
+                          (1 << NF_INET_POST_ROUTING))) != 0) {
                printk("xt_TCPMSS: path-MTU clamping only supported in "
                       "FORWARD, OUTPUT and POSTROUTING hooks\n");
                return false;
        }
-       if (IP6T_MATCH_ITERATE(e, find_syn_match))
-               return true;
+       xt_ematch_foreach(ematch, e)
+               if (find_syn_match(ematch))
+                       return true;
        printk("xt_TCPMSS: Only works on TCP SYN packets\n");
        return false;
 }
 #endif
 
-static struct xt_target xt_tcpmss_reg[] __read_mostly = {
+static struct xt_target tcpmss_tg_reg[] __read_mostly = {
        {
-               .family         = AF_INET,
+               .family         = NFPROTO_IPV4,
                .name           = "TCPMSS",
-               .checkentry     = xt_tcpmss_checkentry4,
-               .target         = xt_tcpmss_target4,
+               .checkentry     = tcpmss_tg4_check,
+               .target         = tcpmss_tg4,
                .targetsize     = sizeof(struct xt_tcpmss_info),
                .proto          = IPPROTO_TCP,
                .me             = THIS_MODULE,
        },
 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
        {
-               .family         = AF_INET6,
+               .family         = NFPROTO_IPV6,
                .name           = "TCPMSS",
-               .checkentry     = xt_tcpmss_checkentry6,
-               .target         = xt_tcpmss_target6,
+               .checkentry     = tcpmss_tg6_check,
+               .target         = tcpmss_tg6,
                .targetsize     = sizeof(struct xt_tcpmss_info),
                .proto          = IPPROTO_TCP,
                .me             = THIS_MODULE,
@@ -276,15 +302,15 @@ static struct xt_target xt_tcpmss_reg[] __read_mostly = {
 #endif
 };
 
-static int __init xt_tcpmss_init(void)
+static int __init tcpmss_tg_init(void)
 {
-       return xt_register_targets(xt_tcpmss_reg, ARRAY_SIZE(xt_tcpmss_reg));
+       return xt_register_targets(tcpmss_tg_reg, ARRAY_SIZE(tcpmss_tg_reg));
 }
 
-static void __exit xt_tcpmss_fini(void)
+static void __exit tcpmss_tg_exit(void)
 {
-       xt_unregister_targets(xt_tcpmss_reg, ARRAY_SIZE(xt_tcpmss_reg));
+       xt_unregister_targets(tcpmss_tg_reg, ARRAY_SIZE(tcpmss_tg_reg));
 }
 
-module_init(xt_tcpmss_init);
-module_exit(xt_tcpmss_fini);
+module_init(tcpmss_tg_init);
+module_exit(tcpmss_tg_exit);