hosts.allow not always working... misses some IPs

[Uwe Doering]

Kerry B. Rogers wrote:
>>[...]
>>I think the netmask is wrong.  When you apply the third octet of the
>>netmask (251) to the IP address (220) the result will be 216, which is
>>then compared with 220.  Since the numbers differ the rule doesn't
>>apply, which is to be expected.
>>
>>Are you sure that the netmask's third octet shouldn't have been 254, 252
>>or 248 instead for proper masking, depending on the range of addresses
>>you'd like to cover?
> 
> Uwe... how did you come up with netmask 251 applied to 220 equals 216? I'm
> confused about how one
> would determine the proper netmask. I think my formula is wrong and would
> like to get it right. I'm trying to convert the ARIN data line:
> 
> arin|CA|ipv4|199.185.220.0|1280|19940222|assigned
> 
> to a hosts.allow line and come up with:
> 
> smtp : 199.185.220.0/255.255.251.0 : deny
> 
> using the formula:
> 
> MaskFromIPRange = DoubleToIPAddress(IPAddressToDouble("255.255.255.255") -
> (IPAddressToDouble(strLastIP) - IPAddressToDouble(strFirstIP)))
> 
> or, translated symbolically:
> 
> Mask = 255.255.255.255 - 199.185.224.255 - 199.185.220.0
> 
> which (mathematically) is:
> 
> Mask = 4294967295 - 3350847743 - 3350846464
> 
> I guess using 255.255.255.255 and subtracting the difference of the IP range
> is not the proper way to arrive at a netmask. What is? Anyone?

Netmasks are supposed to be calculated bit-wise, not by subtraction, and 
they can cover only address ranges that are a power of two.  So you need 
two ranges in your case: the first 1024 addresses and the remaining 256 
(adds up to 1280).  In C syntax the formular for the netmask would be:

   netmask = ^(number_of_addresses - 1);

This results in

   smtp : 199.185.220.0/255.255.252.0 199.185.224.0/255.255.255.0 : deny

If you don't have a calculator with a binary mode you can easily do this 
bit by bit on a piece of paper.  First write down 1023 (1024 - 1) in 
binary form (all 32 bits representing an IPv4 address), then invert the 
bits, and finally convert them back into a decimal number.  Do the same 
for the second range (256 - 1), and adapt the base address for this 
range accordingly.

Hope this explanation was clear enough.

    Uwe
-- 
Uwe Doering         |  EscapeBox - Managed On-Demand UNIX Servers
gemini at geminix.org  |  http://www.escapebox.net