Files
basil00_WinDivert/doc/divert.html
T
basil00 ee386df032 Figured out a way to split the "address" part of the buffer from the packet
data itself.  This makes for a much cleaner interface.

sys/divert.c
dll/divert.c
include/*.h
	DivertRecv and DivertSend now use IOCTLs instead of reads/writes.
	The 'address' parameter is passed by pointer to the driver, which
	writes directly to it (after sanity checks).
	This means that the data buffer now only contains the packet, which
	help to avoid some messy code.

examples/*/*.c
	Update the examples to reflect the new API.

doc/divert.html
	Update the documentation to reflect the new API.
2011-08-21 17:59:13 +08:00

956 lines
31 KiB
HTML

<!doctype html>
<!--
divert.html
(C) 2011, all rights reserved,
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
-->
<html>
<head>
<title>Divert Documentation</title>
</head>
<body>
<h1>Divert: Windows Packet Divert</h1>
<h2>Table of Contents</h2>
<ul>
<li><a href="#introduction">1. Introduction</a></li>
<li><a href="#building">2. Building</a></li>
<ul>
<li><a href="#driver_signing">2.2 Driver Signing</a></li>
</ul>
<li><a href="#installing">3. Installing</a></li>
<li><a href="#uninstalling">4. Uninstalling</a></li>
<li><a href="#programming_api">5. Programming API</a></li>
<ul>
<li><a href="#divert_address">5.1 DIVERT_ADDRESS</a></li>
<li><a href="#divert_open">5.2 DivertOpen</a></li>
<li><a href="#divert_recv">5.3 DivertRecv</a></li>
<li><a href="#divert_send">5.4 DivertSend</a></li>
<li><a href="#divert_close">5.5 DivertClose</a></li>
</ul>
<li><a href="#helper_programming_api">6. Helper Programming API</a></li>
<ul>
<li><a href="#divert_iphdr">6.1 DIVERT_IPHDR</a></li>
<li><a href="#divert_ipv6hdr">6.2 DIVERT_IPV6HDR</a></li>
<li><a href="#divert_icmphdr">6.3 DIVERT_ICMPHDR</a></li>
<li><a href="#divert_icmpv6hdr">6.4 DIVERT_ICMPV6HDR</a></li>
<li><a href="#divert_tcphdr">6.5 DIVERT_TCPHDR</a></li>
<li><a href="#divert_udphdr">6.6 DIVERT_UDPHDR</a></li>
<li><a href="#divert_helper_parse">6.7 DivertHelperParse</a></li>
<li><a href="#divert_helper_calc_checksums">6.8 DivertHelperCalcChecksums</a></li>
</ul>
<li><a href="#filter_language">7. Filter Language</a></li>
<ul>
<li><a href="#filter_examples">7.2 Filter Examples</a></li>
</ul>
<li><a href="#samples">8. Samples</a></li>
<li><a href="#known_issues">9. Known Issues</a></li>
<li><a href="#license">10. License</a></li>
</ul>
<hr>
<a name="introduction"><h2>1. Introduction</h2></a>
<p>
This package (<tt>divert</tt>) provides user-mode packet
capture/modification/blocking/re-injection for Windows Vista and later.
</p>
<p>
The main features of the <tt>divert</tt> package are:
<ul>
<li> User-mode packet capture, dropping, modification.
<li> Simple, high-level, programming API.
<li> Fully documented with sample programs.
<li> Full IPv6 support.
<li> Modern WDF/WFP driver implementation, as opposed to an older
WDM/NDIS intermediate driver.
<li> Open source; Licensed under GNU General Public License version 3.
See the <a href="#license">License</a> for more information.
</ul>
</p><p>
The functionality provided by <tt>divert</tt> is very similar to
<tt>DIVERT</tt> sockets in FreeBSD/MacOS and <tt>NETLINK</tt> sockets in Linux,
as well as some commercial packages, e.g. <tt>WinPktFilter</tt>, for Windows.
</p>
<hr>
<a name="building"><h2>2. Building</h2></a>
<p>
To build the <tt>divert</tt> package from source, you must follow these
steps:
<ol>
<li> Download and install the latest <a href="http://www.microsoft.com/whdc/devtools/wdk/default.mspx">Windows Driver Kit</a>.
<li> Open a <i>Free Build Environment</i> console (or
<i>Checked Build Environment</i> for debugging).
<li> In the <tt>divert</tt> root directory, run the command:
<pre>
build -cZg
</pre>
This will build the following files and place them in the
<tt>divert\install</tt> subdirectory:
<ul>
<li> <tt>divert.dll</tt>: User-mode library.
<li> <tt>divert.sys</tt>: Kernel-mode WDF/WFP callout driver.
<li> <tt>divert.inf</tt>: INF file for <tt>divert.sys</tt>.
<li> <tt>WdfCoInstaller*.dll</tt>: Microsoft WDF co-installer for
<tt>divert.sys</tt>.
<li> <tt>*.exe</tt>: Sample applications from the <tt>divert\examples</tt>
directory.
</ul>
</ol>
</p>
<a name="driver_signing"><h3>2.2 Driver Signing</h3></a>
<p>
Before the <tt>divert</tt> package can be used, the <tt>divert.sys</tt> driver
must be digitally signed.
See <a href="http://msdn.microsoft.com/en-us/windows/hardware/gg487317.aspx">Driver Signing Requirements for Windows</a>
for more information.
</p>
<p>
Driver signing is <b>not</b> provided by this package, although this is
something we wish to change in the future.
If you wish to use this package, you must sign the driver yourself.
</p>
<p>
If you wish to test this package, you can set up a <i>test certificate</i>.
See <a href="http://msdn.microsoft.com/en-us/library/ff554705.aspx">
Test-Signing Driver Packages</a> for more information.
</p>
<hr>
<a name="installing"><h2>3. Installing</h2></a>
<p>
The <tt>divert</tt> package does not require any special installation.
Simply ensure that the <tt>divert.dll</tt>, <tt>divert.sys</tt>,
<tt>divert.inf</tt>, and <tt>WdfCoInstaller*.dll</tt> files are in your
application's home directory.
</p>
<p>
The <tt>divert</tt> driver is installed on demand, i.e., when your application
makes a call to <tt>DivertOpen()</tt> from <tt>divert.dll</tt>
(see programming API below).
</p>
<p>
The <tt>WdfCoInstaller*.dll</tt> file is relatively bloated compared to the
other files.
Blame Microsoft.
</p>
<hr>
<a name="uninstalling"><h2>4. Uninstalling</h2></a>
<p>
To uninstall, simply delete the <tt>divert.dll</tt>, <tt>divert.sys</tt>,
<tt>divert.inf</tt>, and <tt>WdfCoInstaller*.dll</tt> files.
If the <tt>divert</tt> driver was already demand-started, it will be removed
automatically after the next reboot.
To immediately remove it, your uninstaller can issue the following commands:
<pre>
sc stop divert
sc delete divert
</pre>
Note however this is not recommended, as this may interfere with other
applications using the <tt>divert</tt> package.
</p>
<hr>
<a name="programming_api"><h2>5. Programming API</h2></a>
<p>
To use the <tt>divert</tt> package, a program/application must:
<ol>
<li> Include the <tt>divert.h</tt> header file
<pre>
#include "divert.h"
</pre>
<li> Link or dynamically load the <tt>divert.dll</tt> dynamic link library.
</ol>
<a name="divert_address"><h3>5.1 DIVERT_ADDRESS</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
typedef struct
{
UINT32 IfIdx;
UINT32 SubIfIdx;
UINT8 Direction;
} <b>DIVERT_ADDRESS</b>, *<b>PDIVERT_ADDRESS</b>;
</pre>
</td></tr></table>
<dl><dd>
<p>
<b>Fields</b>
<ul>
<li> <tt>IfIdx</tt>: The interface index on which the packet arrived
(for inbound packets), or is to be sent (for outbound packets).</li>
<li> <tt>SubIfIdx</tt>: The sub-interface index for <tt>IfIdx</tt>.</li>
<li> <tt>Direction</tt>: The packet's direction.
The possible values are
<ul>
<li> <tt>DIVERT_PACKET_DIRECTION_OUTBOUND</tt> with value 0 for outbound
packets.</li>
<li> <tt>DIVERT_PACKET_DIRECTION_INBOUND</tt> with value 1 for inbound
packets.</li>
</ul></li>
</ul>
</p><p>
<b>Remarks</b><br>
The <tt>DIVERT_ADDRESS</tt> structure represents where a captured or injected
packet is being sent.
This includes the packets network interfaces, and the packet's direction.
</p>
</dd></dl>
<a name="divert_open"><h3>5.2 DivertOpen</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
HANDLE <b>DivertOpen</b>(
__in const char *filter
);
</pre>
</td></tr></table>
<dl><dd>
<p>
<b>Parameters</b><br>
<ul>
<li> <tt>filter</tt>: A packet filter string specified in the <tt>divert</tt>
<a href="#filter_language">filter language</a>.</li>
</ul>
</p><p>
<b>Return Value</b><br>
A valid <tt>divert</tt> <tt>HANDLE</tt> on success, or
<tt>INVALID_HANDLE_VALUE</tt> if an error occurred.
Use <tt>GetLastError()</tt> to get the reason for the error.
</p><p>
<b>Remarks</b><br>
Opens a <tt>divert</tt> packet capture handle for the given filter.
Any packet that matches the filter will be diverted to the handle, and
can be read by calling <tt>DivertRecv()</tt>.
</p>
<p>
A typical application is only interested in a subset of all traffic.
In this case the filter should <i>match this subset as closely as
possible</i>.
This avoids unnecessary overheads introduced by diverting packets to an
application, only to have the application re-inject them.
</p>
<p>
Calling <tt>DivertOpen()</tt> for the first time will automatically load
the <tt>divert.sys</tt> driver.
This driver will remain installed until the next reboot, or if the driver
is explicitly removed, e.g. by issuing the following commands:
<pre>
sc stop divert
sc delete divert
</pre>
This model helps ensure the driver is not loaded unless it is required to be.
</p>
</dd></dl>
<a name="divert_recv"><h3>5.3 DivertRecv</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
BOOL <b>DivertRecv</b>(
__in HANDLE handle,
__out PVOID pPacket,
__in UINT packetLen,
__out PDIVERT_ADDRESS pAddr,
__out_opt UINT *recvLen
);
</pre>
</td></tr></table>
<dl><dd>
<p>
<b>Parameters</b><br>
<ul>
<li> <tt>handle</tt>: A valid <tt>divert</tt> handle created by
<tt>DivertOpen()</tt>.</li>
<li> <tt>pPacket</tt>: A buffer for the captured packet.</li>
<li> <tt>packetLen</tt>: The length of the buffer <tt>pPacket</tt>.</li>
<li> <tt>pAddr</tt>: The <tt>DIVERT_ADDRESS</tt> of the captured packet.</li>
<li> <tt>recvLen</tt>: The total number of bytes written to <tt>pPacket</tt>.
Can be <tt>NULL</tt> if this information is not required.</li>
</ul>
</p><p>
<b>Return Value</b><br>
<tt>TRUE</tt> if a packet was successfully received, or <tt>FALSE</tt> if
an error occurred.
Use <tt>GetLastError()</tt> to get the reason for the error.
</p><p>
<b>Remarks</b><br>
Receives a diverted packet that matches the filter passed to
<tt>DivertOpen()</tt>.
The received packet is guaranteed to match the filter.
</p><p>
The contents of the captured packet are written to <tt>pPacket</tt>.
If the captured packet is larger than the <tt>pPacket</tt> buffer length,
then the packet will be truncated.
If <tt>recvLen</tt> is non-<tt>NULL</tt>, then the total number of bytes
written to <tt>pPacket</tt> is placed there.
The address of the captured packet is written to <tt>pAddr</tt>.
</p><p>
An application should call <tt>DivertRecv()</tt> <i>as soon as possible</i>
after a successful call to <tt>DivertOpen()</tt>.
When a <tt>divert</tt> handle is open, any packet that matches the filter will
be captured and queued until handled by <tt>DivertRecv()</tt>.
The packet queue never exceeds a fixed length (currently 1024 packets), after
which packets are dropped.
Furthermore packets are not queued indefinitely.
A packet that has been queued longer than a specific time (currently 100ms)
will be dropped.
To avoid packets being dropped that application must process packets as fast
as possible.
</p>
</dd></dl>
<a name="divert_send"><h3>5.4 DivertSend</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
BOOL <b>DivertSend</b>(
__in HANDLE handle,
__in PVOID pPacket,
__in UINT packetLen,
__in PDIVERT_ADDRESS pAddr,
__out_opt UINT *sendLen
);
</pre>
</td></tr></table>
<dl><dd>
<p>
<b>Parameters</b><br>
<ul>
<li> <tt>handle</tt>: A valid <tt>divert</tt> handle created by
<tt>DivertOpen()</tt>.</li>
<li> <tt>pPacket</tt>: A buffer containing the packet to be injected.</li>
<li> <tt>packetLen</tt>: The total length of the buffer <tt>pPacket</tt>.</li>
<li> <tt>pAddr</tt>: The <tt>DIVERT_ADDRESS</tt> for the injected packet.</li>
<li> <tt>sendLen</tt>: The total number of bytes injected.
Can be <tt>NULL</tt> if this information is not required.</li>
</ul>
</p><p>
<b>Return Value</b><br>
<tt>TRUE</tt> if a packet was successfully injected, or <tt>FALSE</tt> if
an error occurred.
Use <tt>GetLastError()</tt> to get the reason for the error.
</p><p>
<b>Remarks</b><br>
Injects a packet into the network stack.
The injected packet may be one received from <tt>DivertRecv()</tt>, or a
modified version, or a completely new packet.
Injected packets cannot be read again by <tt>DivertRecv()</tt>.
</p><p>
The <tt>pAddr</tt> parameter determines how the packet is injected.
If the <tt>Direction</tt> field is <tt>DIVERT_PACKET_DIRECTION_OUTBOUND</tt>,
the packet is injected into the <i>outbound</i> path (i.e. a packet leaving
this computer).
Else, if <tt>Direction</tt> is <tt>DIVERT_PACKET_DIRECTION_INBOUND</tt>,
the packet is injected into the <i>inbound</i> path (i.e. a packet arriving at
this computer).
Note that the <tt>Direction</tt> field, and <i>not</i> the IP addresses in
the injected packet, is used to determine the packet's direction.
</p><p>
For packets injected into the <i>inbound</i> path, the <tt>IfIdx</tt> and
<tt>SubIfIdx</tt> fields are assumed to contain valid interface numbers.
These may be retrieved from <tt>DivertRecv()</tt> (for packet modification),
or from the <a href="http://msdn.microsoft.com/en-us/library/aa366073%28v=VS.85%29.aspx">IP Helper API</a>.
</p><p>
For <i>outbound</i> injected packets, the <tt>IfIdx</tt> and <tt>SubIfIdx</tt>
fields are ignored and may be arbitrary values.
Injecting an inbound packet on the outbound path <i>may</i> work (for some
types of packets), however this should be considered "undocumented" behavior,
and may change in future.
</p>
</dd></dl>
<a name="divert_close"><h3>5.5 DivertClose</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
BOOL <b>DivertClose</b>(
__in HANDLE handle
);
</pre>
</td></tr></table>
<dl><dd>
<p>
<b>Parameters</b><br>
<ul>
<li> <tt>handle</tt>: A valid <tt>divert</tt> handle created by
<tt>DivertOpen()</tt>.</li>
</ul>
</p><p>
<b>Return Value</b><br>
<tt>TRUE</tt> if successful, <tt>FALSE</tt> if an error occurred.
Use <tt>GetLastError()</tt> to get the reason for the error.
</p><p>
<b>Remarks</b><br>
Closes a handle created by <tt>DivertOpen()</tt>.
</p>
<dd></dl>
<hr>
<a name="helper_programming_api"><h2>6. Helper Programming API</h2></a>
The <tt>divert</tt> helper programming API is a collection of definitions
and functions designed to make writing <tt>divert</tt> applications easier.
The use of the helper API is completely optional.
<a name="divert_iphdr"><h3>6.1 DIVERT_IPHDR</h3>
<table border="1" cellpadding="5"><tr><td>
<pre>
typedef struct
{
UINT8 HdrLength:4;
UINT8 Version:4;
UINT8 TOS;
UINT16 Length;
UINT16 Id;
UINT16 <i>...</i>;
UINT8 TTL;
UINT8 Protocol;
UINT16 Checksum;
UINT32 SrcAddr;
UINT32 DstAddr;
} <b>DIVERT_IPHDR</b>, *<b>PDIVERT_IPHDR</b>;
</pre>
</td></tr></table>
<dl><dd>
<p>
<b>Fields</b><br>
See <a href="http://en.wikipedia.org/wiki/IPv4#Packet_structure">here</a>
for more information.
</p><p>
<b>Remarks</b><br>
IPv4 header definition.
</p><p>
The following fields can only be get/set using the following macro
definitions:
<ul>
<li><i>FragOff</i> with <tt>DIVERT_IPHDR_GET_FRAGOFF(<i>hdr</i>)</tt> and
<tt>DIVERT_IPHDR_SET_FRAGOFF(<i>hdr</i>, <i>val</i>)</tt></li>
<li><i>MF</i> with <tt>DIVERT_IPHDR_GET_MF(<i>hdr</i>)</tt> and
<tt>DIVERT_IPHDR_SET_MF(<i>hdr</i>, <i>val</i>)</tt></li>
<li><i>DF</i> with <tt>DIVERT_IPHDR_GET_DF(<i>hdr</i>)</tt> and
<tt>DIVERT_IPHDR_SET_DF(<i>hdr</i>, <i>val</i>)</tt></li>
<li><i>Reserved</i> with <tt>DIVERT_IPHDR_GET_RESERVED(<i>hdr</i>)</tt> and
<tt>DIVERT_IPHDR_SET_RESERVED(<i>hdr</i>, <i>val</i>)</tt></li>
</ul>
</p>
</dl></dd>
<a name="divert_ipv6hdr"><h3>6.2 DIVERT_IPV6HDR</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
typedef struct
{
UINT32 Version:4;
UINT32 ...:28;
UINT16 Length;
UINT8 NextHdr;
UINT8 HopLimit;
UINT32 SrcAddr[4];
UINT32 DstAddr[4];
} <b>DIVERT_IPV6HDR</b>, *<b>PDIVERT_IPV6HDR</b>;
</pre>
</td></tr></table>
<dl><dd>
<b>Fields</b><br>
See <a href="http://en.wikipedia.org/wiki/IPv6_packet#Fixed_header">here</a>
for more information.
</p><p>
<b>Remarks</b><br>
IPv6 header definition.
</p><p>
The following fields can only be get/set using the following macro
definitions:
<ul>
<li><i>TrafficClass</i> with
<tt>DIVERT_IPV6HDR_GET_TRAFFICCLASS(<i>hdr</i>)</tt> and
<tt>DIVERT_IPV6HDR_SET_TRAFFICCLASS(<i>hdr</i>, <i>val</i>)</tt></li>
<li><i>FlowLabel</i> with <tt>DIVERT_IPV6HDR_GET_FLOWLABEL(<i>hdr</i>)</tt> and
<tt>DIVERT_IPV6HDR_SET_FLOWLABEL(<i>hdr</i>, <i>val</i>)</tt></li>
</ul>
</p>
</dl></dd>
<a name="divert_icmphdr"><h3>6.3 DIVERT_ICMPHDR</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
typedef struct
{
UINT8 Type;
UINT8 Code;
UINT16 Checksum;
UINT32 Body;
} <b>DIVERT_ICMPHDR</b>, *<b>PDIVERT_ICMPHDR</b>;
</pre>
</td></tr></table>
<dl><dd>
<b>Fields</b><br>
See <a href="http://en.wikipedia.org/wiki/Internet_Control_Message_Protocol#ICMP_segment_structure">here</a>
for more information.
</p><p>
<b>Remarks</b><br>
ICMP header definition.
</p>
</dl></dd>
<a name="divert_icmpv6hdr"><h3>6.4 DIVERT_ICMPV6HDR</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
typedef struct
{
UINT8 Type;
UINT8 Code;
UINT16 Checksum;
UINT32 Body;
} <b>DIVERT_ICMPV6HDR</b>, *<b>PDIVERT_ICMPV6HDR</b>;
</pre>
</td></tr></table>
<dl><dd>
<b>Fields</b><br>
See <a href="http://en.wikipedia.org/wiki/ICMPv6#Packet_format">here</a> for
more information.
</p><p>
<b>Remarks</b><br>
ICMPv6 header definition.
</p>
</dl></dd>
<a name="divert_tcphdr"><h3>6.5 DIVERT_TCPHDR</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
typedef struct
{
UINT16 SrcPort;
UINT16 DstPort;
UINT32 SeqNum;
UINT32 AckNum;
UINT16 Reserved1:4;
UINT16 HdrLength:4;
UINT16 Fin:1;
UINT16 Syn:1;
UINT16 Rst:1;
UINT16 Psh:1;
UINT16 Ack:1;
UINT16 Urg:1;
UINT16 Reserved2:2;
UINT16 Window;
UINT16 Checksum;
UINT16 UrgPtr;
} <b>DIVERT_TCPHDR</b>, *<b>PDIVERT_TCPHDR</b>;
</pre>
</td></tr></table>
<dl><dd>
<b>Fields</b><br>
See <a href="http://en.wikipedia.org/wiki/Transmission_Control_Protocol#TCP_segment_structure">here</a>
for more information.
</p><p>
<b>Remarks</b><br>
TCP header definition.
</p>
</dl></dd>
<a name="divert_udphdr"><h3>6.6 DIVERT_UDPHDR</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
typedef struct
{
UINT16 SrcPort;
UINT16 DstPort;
UINT16 Length;
UINT16 Checksum;
} <b>DIVERT_UDPHDR</b>, *<b>PDIVERT_UDPHDR</b>;
</pre>
</td></tr></table>
<dl><dd>
<b>Fields</b><br>
See <a href="http://en.wikipedia.org/wiki/User_Datagram_Protocol#Packet_structure">here</a>
for more information.
</p><p>
<b>Remarks</b><br>
UDP header definition.
</p>
</dl></dd>
<a name="divert_helper_parse"><h3>6.7 DivertHelperParse</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
BOOL <b>DivertHelperParse</b>(
__in PVOID pPacket,
__in UINT packetLen,
__out_opt PDIVERT_IPHDR *ppIpHdr,
__out_opt PDIVERT_IPV6HDR *ppIpv6Hdr,
__out_opt PDIVERT_ICMPHDR *ppIcmpHdr,
__out_opt PDIVERT_ICMPV6HDR *ppIcmpv6Hdr,
__out_opt PDIVERT_TCPHDR *ppTcpHdr,
__out_opt PDIVERT_UDPHDR *ppUdpHdr,
__out_opt PVOID *ppData,
__out_opt UINT *pDataLen
);
</pre>
</td></tr></table>
<dl><dd>
<p>
<b>Parameters</b><br>
<ul>
<li> <tt>pPacket</tt>: The packet to be parsed.</li>
<li> <tt>packetLen</tt>: The total length of the packet <tt>pPacket</tt>.</li>
<li> <tt>ppIpHdr</tt>: Output pointer to a <tt>DIVERT_IPHDR</tt>.</li>
<li> <tt>ppIpv6Hdr</tt>: Output pointer to a <tt>DIVERT_IPV6HDR</tt>.</li>
<li> <tt>ppIcmpHdr</tt>: Output pointer to a <tt>DIVERT_ICMPHDR</tt>.</li>
<li> <tt>ppIcmpv6Hdr</tt>: Output pointer to a <tt>DIVERT_ICMPV6HDR</tt>.</li>
<li> <tt>ppTcpHdr</tt>: Output pointer to a <tt>DIVERT_TCPHDR</tt>.</li>
<li> <tt>ppUdpHdr</tt>: Output pointer to a <tt>DIVERT_UDPHDR</tt>.</li>
<li> <tt>ppData</tt>: Output pointer to the packet's data/payload.</li>
<li> <tt>pDataLen</tt> Output data/payload length.</li>
</ul>
</p><p>
<b>Return Value</b><br>
<tt>TRUE</tt> if all expected (non-<tt>NULL</tt>) outputs were present,
<tt>FALSE</tt> otherwise.
Note that <tt>FALSE</tt> may sometimes be a legitimate return value, e.g.,
when both <tt>ppIpHdr</tt> and <tt>ppIpv6Hdr</tt> are non-<tt>NULL</tt>.
</p><p>
<b>Remarks</b><br>
Parses a raw packet (e.g. one captured using <tt>DivertRecv</tt>) into the
various packet headers and/or payloads that may or may not be present.
</p><p>
Each output parameter may be <tt>NULL</tt> or non-<tt>NULL</tt>.
For non-<tt>NULL</tt> parameters, this function will write the pointer to
the corresponding header/payload if it exists, or will write <tt>NULL</tt>
otherwise.
Any non-<tt>NULL</tt> pointer that is returned
<ol>
<li> Is a pointer into the original <tt>pPacket</tt> packet; and</li>
<li> There is enough space in <tt>pPacket</tt> to fit the header.</li>
</ol>
</p><p>
This function does not do any verification of the header/payload contents,
other length and the minimal information required to parse the headers
themselves.
<p>
</dd></dl>
<a name="divert_helper_calc_checksums"><h3>6.8 DivertHelperCalcChecksums</h3></a>
<table border="1" cellpadding="5"><tr><td>
<pre>
UINT <b>DivertHelperCalcChecksums</b>(
__inout PVOID pPacket,
__in UINT packetLen,
__in UINT64 flags
);
</pre>
</td></tr></table>
<dl><dd>
<p>
<b>Parameters</b><br>
<ul>
<li> <tt>pPacket</tt>: The packet to be modified.</li>
<li> <tt>packetLen</tt>: The total length of the packet <tt>pPacket</tt>.</li>
<li> <tt>flags</tt>: One or more of the following flags:
<ul>
<li> <tt>DIVERT_HELPER_NO_IP_CHECKSUM</tt>: Do not calculate the IPv4
checksum.</li>
<li> <tt>DIVERT_HELPER_NO_ICMP_CHECKSUM</tt>: Do not calculate the ICMP
checksum.</li>
<li> <tt>DIVERT_HELPER_NO_ICMPV6_CHECKSUM</tt>: Do not calculate the ICMPv6
checksum.</li>
<li> <tt>DIVERT_HELPER_NO_TCP_CHECKSUM</tt>: Do not calculate the TCP
checksum.</li>
<li> <tt>DIVERT_HELPER_NO_UDP_CHECKSUM</tt>: Do not calculate the UDP
checksum.</li>
</ul></li>
</ul>
</p><p>
<b>Return Value</b><br>
The number of checksums calculated.
</p><p>
<b>Remarks</b><br>
(Re)calculates the checksum for any IPv4/ICMP/ICMPv6/TCP/UDP checksum present
in the given packet.
Individual checksum calculations may be disabled via the appropriate flag.
Typically this function should be used before a packet is injected.
</p><p>
This function will calculate each checksum from scratch, even if the existing
checksum is correct.
This may be inefficient for some applications.
For better performance, incremental checksum calculations should be used
instead (not provided by this API).
<p>
</dd></dl>
<hr>
<a name="filter_language"><h2>7. Filter Language</h2></a>
<p>
The <tt>DivertOpen()</tt> function accepts a string containing a
<i>filter expression</i>.
Only packets that match the filter expression are diverted.
Any other packet is allowed to continue as per normal.
</p><p>
Filter allows an application to select only the subset of traffic that is of
interest.
For example, a URL blacklist filter would only be interested in packets that
contain URLs.
This could be achieved via the following filter.
<pre>
HANDLE handle = DivertOpen(
"outbound and "
"tcp.PayloadLength &gt; 0 and "
"tcp.DstPort == 80");
</pre>
This filter specifies that we should only divert traffic that is
<ol>
<li>outbound;</li>
<li>contains a non-empty payload; and</li>
<li>has TCP destination port 80 (i.e. HTTP web traffic).
</ol>
</p><p>
A <i>filter</i> is a Boolean expression of the form:
<pre>
<i>FILTER</i> := true | false | <i>FILTER</i> and <i>FILTER</i> | <i>FILTER</i> or <i>FILTER</i> | (<i>FILTER</i>) | <i>TEST</i>
</pre>
C-style syntax <tt>&amp;&amp;</tt>, <tt>||</tt>, and <tt>!</tt> may also
be used instead of <tt>and</tt>, <tt>or</tt>, and <tt>not</tt>, respectively.
A <i>test</i> is of the following form:
<pre>
<i>TEST</i> := <i>TEST0</i> | not <i>TEST0</i>
<i>TEST0</i> := <i>FIELD</i> | <i>FIELD</i> op <i>VAL</i>
</pre>
where <tt>op</tt> is one of the following:
</p><p>
<center>
<table border="1" cellpadding="5">
<tr><td><tt>==</tt> or <tt>=</tt></td><td>Equal</td></tr>
<tr><td><tt>!=</tt></td><td>Not equal</td></tr>
<tr><td><tt>&lt;</tt></td><td>Less-than</td></tr>
<tr><td><tt>&gt;</tt></td><td>Greater-than</td></tr>
<tr><td><tt>&lt;=</tt></td><td>Less-than-or-equal</td></tr>
<tr><td><tt>&gt;=</tt></td><td>Greater-than-or-equal</td></tr>
</table>
</center>
</p><p>
and <tt><i>VAL</i></tt> is a decimal number, hexadecimal number, or IP
address.
If the "<tt>op <i>VAL</i></tt>" is missing, the test is implicitly
"<tt><i>FIELD</i> != 0</tt>".
</p><p>
Finally a <i>field</i> is some property about the packet.
The possible fields are:
</p><p>
<center>
<table border="1" cellpadding="5">
<tr><td><tt>outbound</tt></td><td>Is outbound?</td></tr>
<tr><td><tt>inbound</tt></td><td>Is inbound?</td></tr>
<tr><td><tt>ifIdx</tt></td><td>Interface index</td></tr>
<tr><td><tt>subIfIdx</tt></td><td>Sub-interface index</td></tr>
<tr><td><tt>ip</tt></td><td>Is IPv4?</td></tr>
<tr><td><tt>ipv6</tt></td><td>Is IPv6?</td></tr>
<tr><td><tt>icmp</tt></td><td>Is ICMP?</td></tr>
<tr><td><tt>icmpv6</tt></td><td>Is ICMPv6?</td></tr>
<tr><td><tt>tcp</tt></td><td>Is TCP?</td></tr>
<tr><td><tt>udp</tt></td><td>Is UDP?</td></tr>
<tr><td><tt>ip.*</tt></td><td>IPv4 fields (see <tt>DIVERT_IPHDR</tt>)</td></tr>
<tr><td><tt>ipv6.*</tt></td><td>IPv6 fields (see <tt>DIVERT_IPV6HDR</tt>)</td></tr>
<tr><td><tt>icmp.*</tt></td><td>ICMP fields (see <tt>DIVERT_ICMPHDR</tt>)</td></tr>
<tr><td><tt>icmpv6.*</tt></td><td>ICMPV6 fields (see <tt>DIVERT_ICMPV6HDR</tt>)</td></tr>
<tr><td><tt>tcp.*</tt></td><td>TCP fields (see <tt>DIVERT_TCPHDR</tt>)</td></tr>
<tr><td><tt>tcp.PayloadLength</tt></td><td>The TCP payload length</td></tr>
<tr><td><tt>udp.*</tt></td><td>UDP fields (see <tt>DIVERT_UDPHDR</tt>)</td></tr>
<tr><td><tt>udp.PayloadLength</tt></td><td>The UDP payload length</td></tr>
</table>
</center>
</p><p>
A <i>test</i> also fails if the field is missing.
E.g. the test "<tt>tcp.DstPort == 80</tt>" will fail if the packet does not
contain a TCP header.
</p>
<a name="filter_examples"><h3>7.1 Filter Examples</h3></a>
<p>
<ol>
<li>
Divert all outbound web traffic:
<pre>
HANDLE handle = DivertOpen(
"outbound and "
"(tcp.DstPort == 80 or udp.DstPort == 53)"
);
</pre>
</li>
<li>
Divert all inbound TCP SYNs:
<pre>
HANDLE handle = DivertOpen(
"inbound and "
"tcp.Syn"
);
</pre>
</li>
<li>
Divert only (inbound) local traffic:
<pre>
HANDLE handle = DivertOpen(
"inbound and ("
"(ip.DstAddr &gt;= 127.0.0.1 and ip.DstAddr &lt;= 127.255.255.255) or"
"ipv6.DstAddr == ::1)"
);
</pre>
</li>
<li>
Divert all traffic:
<pre>
HANDLE handle = DivertOpen("true");
</pre>
</li>
<li>
Divert no traffic:
<pre>
HANDLE handle = DivertOpen("false");
</pre>
(This is not very useful).
</ol>
<hr>
<a name="samples"><h2>8. Samples</h2></a>
<p>
Some samples have been provided to demonstrate the <tt>divert</tt> API.
The sample programs are:
<ul>
<li><tt>webfilter.exe</tt>: A simple URL blacklist filter.
This program monitors outbound HTTP traffic.
If it finds a URL request that matches the blacklist, it hijacks the
TCP connection, reseting the connection at the server's end, and
sending a simple block-page to the browser.
The blacklist(s) are specified at the command-line.</li>
<li><tt>netdump.exe</tt>: A simple packet sniffer based on the
<tt>divert</tt> filter language.
This program takes a filter specified at the command line, and prints
information about any packet that matches the filter.
Unlike the other samples, this program does not modify or block packets,
it simply re-injects them.</li>
<li><tt>netfilter.exe</tt>: A simple firewall based on the <tt>divert</tt>
filter language.
This program takes a filter specified at the command line, and blocks
any packet that matches the filter.
It blocks TCP by sending a TCP reset, UDP by an ICMP message, and all
other traffic it simply drops.
This is similar to the Linux <tt>iptables</tt> command with the
<tt>-j REJECT</tt> option.</li>
</ul>
</p><p>
The samples are intended for educational purposes only, and are not
fully-featured applications.
</p>
<hr>
<a name="known_issues"><h2>9. Known Issues</h2></a>
<p>
There are some limitations to the <tt>divert</tt> package.
They are
<ul>
<li><i>Injecting inbound ICMP/ICMPv6 messages</i>:
For some ICMP/ICMPv6 messages, inbound injection does not work.
An error will be returned and the packet will be lost.
It is suspected that this is an issue with the WFP framework on which
<tt>divert</tt> is built.
The work-around is to inject inbound ICMP messages as <tt>outbound</tt>.
</li>
<li><i>No IPv6 extension header support</i>:
Currently there is no filter support for IPv6 packets with extension
headers.
The work around is to capture all IPv6 traffic.
<li><i>Injected packets are never re-captured</i>:
An injected packet will <i>never</i> be captured again by any
<tt>divert</tt> handle.
This is necessary to prevent packet loops and deadlocks.
In the future we intend to implement priorities for <tt>divert</tt>
handles to allow packets to be seen by multiple <tt>divert</tt> handles.
<li><i>Speed</i>:
The <tt>divert</tt> driver is not re-entrant, and thus is not as
efficient as it could be.
In the future we plan to rectify this.
</ul>
</p><p>
All of the samples use some variant of the following basic template for
<tt>divert</tt> applications.
The basic idea is to open a <tt>divert</tt> handle, then enter a
capture-modify-reinject loop:
<pre>
HANDLE handle; // Divert handle
DIVERT_ADDRESS addr; // Packet address
char packet[MAXBUF]; // Packet buffer
UINT packetLen;
handle = DivertOpen("..."); // Open some filter
if (handle == INVALID_HANDLE_VALUE);
{
// Handle error
exit(1);
}
// Main capture-modify-inject loop:
while (TRUE)
{
if (!DivertRecv(handle, packet, sizeof(packet), &amp;addr, &amp;packetLen))
{
// Handle recv error
continue;
}
// Modify packet.
if (!DivertSend(handle, packet, packetLen, &amp;addr, NULL))
{
// Handle send error
continue;
}
}
</pre>
</p>
<hr>
<a name="license"><h2>10. License</h2></a>
<p>
This package is distributed <i>strictly</i> under the
<a href="http://www.gnu.org/licenses/gpl-3.0.txt">GNU Public License (GPL) Version 3</a>.
Please note the following:
<pre>
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see &lt;http://www.gnu.org/licenses/&gt;.
</pre>
</p>
<p>
Other licenses (including commercial licenses) <i>may</i> be available on
request.
For more information please contact:<br>
<tt>basil</tt> AT <tt>reqrypt</tt> DOT <tt>org</tt>
</p>
<br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br>
<br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br>
<br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br>
</body>
</html>