how to convert a dotted-decimal IP address into host internal IP address

Hi,All,
It is very convenient to use a dotted-decimal ip
address,however,TDISendDataGram requires a ULong-typed IP address.
Hence a convension is necessary.Is there any function in windows kernel
mode which can assume this duty?

 

Write inet_addr() yourself. It's trivial.

 

I think gethostbyname() should do it.

 

In the kernel?
Well, we have the kernel-mode DNS resolver, but it is around 2000 lines of
code or such.

 

RtlIpv4AddressToStringA
RtlIpv4AddressToStringExA
RtlIpv4AddressToStringExW
RtlIpv4AddressToStringW
RtlIpv4StringToAddressA
RtlIpv4StringToAddressExA
RtlIpv4StringToAddressExW
RtlIpv4StringToAddressW
RtlIpv6AddressToStringA
RtlIpv6AddressToStringExA
RtlIpv6AddressToStringExW
RtlIpv6AddressToStringW
RtlIpv6StringToAddressA
RtlIpv6StringToAddressExA
RtlIpv6StringToAddressExW
RtlIpv6StringToAddressW

 

Thank you all. But I think the functions like RtlIPv* don't work.
So I will have to write a special-purpose function like inet_addr.

 

I love responses like this.

What do you mean by "I think the functions like RtlIPv* don't work"? Did
you just have a bad feeling that they were flawed in some way? Did someone
tell you they didn't work? Or have you actually tried one and had it fail?

 

I thought exactly the same thing

Another, beloved, frequently posted reply is like "problem solved,
thanks". Not that we would ever like to know *how* it got solved...

Stephan
---

 

Hi,Stephan and Tim,
Thank you for good advice.
I think the functions like RtlIPv* don't work because I can't find them
in Windows DDK Document.
On the other hand, it is certainly trivial to write an inet_addr
function, but fortunately, the principle of the function is clear to us.
Hence, I would implement it rather than spend time in looking for it.

Steve

fail?

 

Hello Steve,

Trivial? Did you consider

- getting an IP as decimal value (i.e., 127.0.0.1 as "2130706433")?
- getting an IP in octal notation (i.e., 127.0.0.1 as "0177.0.0.1")?

As there are some possibilities which are not frequently used, you might
miss a use.

Kind regards,
Spiro.

 

I just convert a string of "10.0.0.3" into a ULong-typed number of
0X0a000003.
so I can parse the string and calculate its corresponding value.

 

Trivial? Did you consider

2 or 3 screenfuls of C code even with all these things handled.

 

Hello Maxim,

My point was that inet_addr() can process many different types of IPs,
and one does not always consider every single way of it.

If I know which options there are, I don't think it is hard to write
such a function. I only wanted to tell that the "if" part can be quite
hard. ;-)

Kind regards,
Spiro.

 

Hello Steve,

I'm not sure if I understand you correctly:

Does your routine handle

010.0.0.3 as 0x08000003,
0177.0.017.2 as 0x7F000F02,
2130706435 as 0x7F000003,

and so on? If not, you only have a limited inet_addr() function. If you
get external inputs (i.e., from a user), this could really be a
problem.

Regards,
Spiro.

 

If I know which options there are, I don't think it is hard to write

Here is the description of IP address formats understandable by inet_addr()
from StorageCraft KSockets
(http://www.storagecraft.com/products/ksockets.html). It (logic, not code) was
taken from FreeBSD IIRC.

// Possible formats:
// 1 digit group (a) - treat as is
// 2 digit groups (a.b) - treat as ( a << 24 ) | b, a must be < 256, b
must be < 16M
// 3 digit groups (a.b.c) - treat as ( a << 24 ) | ( b << 16 ) | c, a and
b must be < 256,
// c must be < 64K
// 4 digit groups (a.b.c.d) - treat as ( a << 24 ) | ( b << 16 ) | ( c <<
8 ) | d,
// all of them must be < 256
// Any digit group with trailing dot must be < 256, and the last digit
group
// must be < ( 1 << ( 32 - ( NumberOfDigitGroups - 1 ) * 8 ) )

And, surely, any digit group can have a radix of 8 ("0" prefix), 16 ("0x"
prefix) or 10 (no prefix).

 

Hello Maxim,

thanks for the info.

*g* An xBSD would have been the place where I would have looked, too.

Regards,
Spiro.