fix for /usr/src/linux/net/ipv4/ip_default_ttl usage

Hi All,

i've been playing with different IP setting available in the linux kernel ( 2.4.18 ) and found a stange behaviour with /usr/src/linux/net/ipv4/ip_default_ttl :
the value that you set here is not taken into account in the following cases :

- ICMP reply ( of any kind )
- TCP RST .

since the sockets used in this cases are the one created at the init of the given protocols, and therefore the ttl value used are the one that are setup at boot time and not the one that you will set later when using the sysctl or the proc interface.

I'm not sure that this has been left out on purpose since it will defeat the goal of changing the default ttl to confuse OS fingerprinting softwares. Therefore, i've made a small patch ( against kernel 2.4.18 ) that will change the behaviour and make the default ttl be used in these cases.

I hope it will be useful to some of you ....
Regards

PS : i'm not on the list so please CC me if you reply to this email.

Charles-Edouard Ruault
PGP Key ID 4370AF2D

@@ -139,6 +139,8 @@
   { EHOSTUNREACH,	1 }	/*	ICMP_PREC_CUTOFF	*/
 };
 
+extern int sysctl_ip_default_ttl;
+
 /* Control parameters for ECHO replies. */
 int sysctl_icmp_echo_ignore_all;
 int sysctl_icmp_echo_ignore_broadcasts;

@@ -354,6 +356,7 @@
 	icmp_out_count(icmp_param->data.icmph.type);
 
 	sk->protinfo.af_inet.tos = skb->nh.iph->tos;
+	sk->protinfo.af_inet.ttl = sysctl_ip_default_ttl;
 	daddr = ipc.addr = rt->rt_src;
 	ipc.opt = NULL;
 	if (icmp_param->replyopts.optlen) {

@@ -64,7 +64,7 @@
 #include <linux/ipsec.h>
 
 extern int sysctl_ip_dynaddr;
-
+extern int sysctl_ip_default_ttl;
 /* Check TCP sequence numbers in ICMP packets. */
 #define ICMP_MIN_LENGTH 8
 

@@ -1072,6 +1072,7 @@
 	arg.n_iov = 1;
 	arg.csumoffset = offsetof(struct tcphdr, check) / 2; 
 
+	tcp_socket->sk->protinfo.af_inet.ttl = sysctl_ip_default_ttl;
 	ip_send_reply(tcp_socket->sk, skb, &arg, sizeof rth);
 
 	TCP_INC_STATS_BH(TcpOutSegs);