Forum Replies Created
-
Posts
-
I only have one adaptor in the PC. So everything must go through that network card.
If you have single ethernet interface then it does not mean that you have the only adapter seen by WinpkFilter, because you always have the dial-up(NDISWANIP) interface.
If i put the adaptor in to tunnel mode and dont have SendPacketToAdapter or SendPacketToMSTCP then all packets should be dropped and no networking should function.
You are right, if you filter in tunnel mode and don’t reinject packets to stack then any network activity is disabled. I suppose something is wrong with your code, definitions or the network interface you filter.
I have the only idea, probably you filter the adapter different from the one used for the e-mail and network browsing.
There may be 2 application which are accessing same IP/ port/ protocol like Netscape and Internet explorer accessing yahoo.com same time.
Right, destination IP/protocol/port can be the same, but source ports will be different for the case described.
IMHO, in theory (hardly likely that it will ever happen in real life) it is possible that application running on NAT system access the same IP/port as another application running behind the NAT and the same source port used in NAT table and for local application running on NAT system. In this case it is not possible to distinguish if packet should be NAT’ed or not. But even in this rare case you still can use sequence and aknowledgement fields in TCP header to determine wheather the NAT should be applied (it will work for TCP connection only).
You can use IP/port/protocol information gathered on the TDI level and match it against information extracted from IP packet. If it is the same then the packet is probably originated from the application you intercepted on the TDI level.
So what about if some application use this port number at the NAT computer? Is there any conflicts if use same port number ( with some application at NAT computer )?
The conflict is possible, you will just bypass those applications (they may be not working properly with your NAT application, but it won’t make any problems to NAT). However, you can just reserve required port range by your own application.
Another question: how to get binding IP address for each interface adapter, and how to get windows route table?
IP Helper API
When I replace the Client(IP, PORT) of the inner packet with the NAT External(ip, port), what port number should I used for the NAT external? The same with the Client Port or a new one? Is that must be a free number that no one uses it? How do I generate this port number.
When you substitute inner source IP with external one but don’t substitute inner port this is called static NAT. When you substitute inner source port in addition to IP this is called dynamic NAT. You can use any free port number (an example just take some rarely used port range like 10000-64000 and allocate ports from there).
PS: where can I get the code of Checksum Calculation?
That depends from the progarmming environment you use, an example some C and Delhi code is available on this board.
If you want to realize your own packet forwarding you can do it with API’s you have mentioned. If you filter two adapters and you can read packet from one and send them to another, here is no crime. Just don’t forget that MTU for the interfaces can be different.
One of the approaches (the easiest from my point of view) is filtering Server(WAN) interface substituting the original IP (if it is from LAN) in the outgoing packet with Server WAN card IP (port substitution also should be done in order to perform backward mapping). For incoming packets on the Server (WAN) interface you should match against NAT table and perform the reversed operation. Please note that packet forwarding should be enabled on the Server. You don’t need to route packets youself, TCP/IP will do it for you.
What environment do you use for the development? There is a C header file iphlp.h which has some sample protocol header definitions (IP header, TCP header, UDP header and etc…). If you are using Delphi then you can use the ones below:
TIPHeaderPtr = ^TIPHeader;
TIPHeader = packed record
VerLen: Byte; //HL
TOS: Byte;
TotalLen: Word;
Identifer: Word;
FragOffsets: Word;
TTL: Byte;
Protocol: Byte;
CheckSum: Word;
SourceIp: DWORD;
DestIp: DWORD;
// Options: DWORD;
end;
TTCPHeaderPtr = ^TTCPHeader;
TTCPHeader = packed record
SourcePort:Word;
DestPort:Word;
SequenceNumber:DWord;
AcknowledgementNumber:DWord;
Offset:Byte; //only left 4 bits. Header length in 32-bit segments
Flags:Byte;
Window:Word;
Checksum:Word; //includes speudo header instead of TCP header.
UrgentPointer:Word;
end;
pTCPData: PChar;And probably it makes sense for you to look at this topic:
http://ntkernel.com/forum/viewtopic.php?t=114We are trying to provide as good support as we can 🙄
Is it possible to use both packet level and application level filtering with in the same program ie. inspect packet using winPKfilter then a LSP/TDI filter?
Yes, sure.
Does your company provide a application level filter?
We provide localhost monitor API, which is TDI filter for Windows NT/2000/XP/2003. LSP sample is avalaible as a part of MSDN. Simple TDI filter for Windows 9x can be found in the VTOOLSD samples (from Compuware).
WinpkFilter allows you implementing packet filtering technologies in both user or kernel mode (the one you choose depends from your expirience and requirements). But this is packet level filtering (it is not possible to determine application context on the NDIS level). For application level filtering you would need LSP or TDI filter.
