Skip to content
Pete Batard edited this page Feb 7, 2014 · 30 revisions

Table of Contents

Installing and Compiling libwdi

Where can I get binary snapshots of libwdi?

Because libwdi can be compiled in very different way (as a static or dynamic library, with or without the WinUSB or libusb-win32 drivers, with or without separate 32 or 64 bit binaries, etc.), and these configuration have a major impact on the size of the resulting binaries, producing regular binary snapshots of libwdi to address various user's needs would require a lot of time, as we would need to produce about 18 different versions to satisfy everyone. We might produce a limited subset of binary snapshots in the future, but for the time being, we require the library to be configured and recompiled according to your needs.

How can I specify a custom signed inf/driver files for use with libwdi?

See the paragraph 'Using a custom driver file with libwdi' in the installation & compilation guide.

Is it possible to cross-compile libwdi?

Absolutely. This can be very convenient if you have a multilib MinGW-w64 cross compiler installed on a Linux host for instance. You will of course need to have the WinUSB (from the MS WDK), libusb-win32, libusbK or your own user driver files available on the build platform.
An example of configuring libwdi for cross-compilation using MinGW-w64, with the libusb-win32 driver, would go something like this:

./configure --host=x86_64-w64-mingw32 --with-libusb0="/usr/src/libusb-win32"

The embedder fails to run during cross-compilation

By default, the configure script expects your default build platform's compiler to be able to produce executables that run on that platform. This is necessary so that the build process can compile and run the custom embedder, to add the driver files to the library. If however your default compiler does not produce executables that run on your build platform, the process will fail. This can occur for instance, if you are using a multilib MinGW-w64 on a Windows 32 platform, but the default for gcc is to produce 64 bit executables. If this is the case, you should define a CC_FOR_BUILD environment variable with the relevant options. For instance, in the scenario above, you would issue the following, to ensure that MinGW-w64 produces a 32 bit embedder executable that can be run during the build process: export CC_FOR_BUILD="x86_64-w64-mingw32-gcc -m32

How can I create a 32+64 bit compatible version of libwdi?

  1. Use Visual Studio (>= 2005) and select Win32 for the target. Note that Visual Studio Express cannot be used, as it doesn't include a 64 bit compiler.
  2. Use the Windows Driver Kit or DDK (v 7.x or later), and run ddk_build.cmd from one of the Windows XP 32 bit command prompts
  3. Use a multilib version of MinGW-w64. Multilib means a compiler that supports both the -m32 and -m64 options. Currently, the binaries provided on the project page are not multilib, so you need to use TDM64 which only runs under Windows or recompile MinGW-w64 yourself.

How can I create Windows 2000 compatible version of libwdi?

Currently, the only officially supported way to do so is to use MinGW32. You may try WDK 6001 version as well but it is unsupported. Take note that MinGW-w64 only officially supports XP onwards (although the libwdi binaries produced appear to work fine on Windows 2000).

How can I create a DLL with the DDK environment?

Run ddk_build DLL. Make sure the DLL parameter is specified as uppercase.

I get "System error: -2147154677" when converting to Visual C++ Express 2010

Unfortunately, Microsoft's Visual C++ Express 2010 is a crippled version that cannot import project files that contain 64 bit builds. Currently, the only solution is to either not use Visual Studio Express 2010 or re-create the project files manually.

I get "fatal error RC1050: Cannot open include file 'afxres.h'" during compilation

This can occur when compiling Zadig with Visual Studio Express, and is due to Visual Studio Express not officially supporting MFC applications. You can however get zadig compiled if you edit the 'zadig.rc' file and replace the line

#include "afxres.h"
with
#include <windows.h>
#define IDC_STATIC -1

I get "Warning: linker path does not have real file for library -lsetupapi/-lole32" with MinGW-w64

If this happens with a recent version of libtool (2.2.8 or later), this can be due to a MinGW-w64 packaging problem with the 64 bit libraries in the x86_64-w64-mingw32/lib directory. One way to address that issue is to move all the libraries from lib/ into the lib64/ directory. If you fixed the MinGW-w64 packaging issue yet still have similar issues, you may need to update libtool to version 2.2.8 or later.

I get "gcc.exe: !CreateProcess: No such file or directory" in MinGW

You are probably attempting to cross compile on an environment where the default gcc cannot produce executables. Because libwdi requires the compilation and execution of an embedder.exe native to the host, the host's default gcc must be able to produce executables that run on the compilation platform.

Running libwdi

What are these "USB\VID_####&PID_####[&MI_##] (Autogenerated)" certificates that libwdi installs in the Trusted certificate stores?

From version 1.1.0, if you are installing a driver on Vista or later versions of Windows, libwdi will attempt to generate and self-sign driver package, and then mark the certificate used for the signing process as trusted, by installing it in the system stores, in order to avoid further security prompts during the driver installation.
During that process, libwdi will copy two autogenerated certificates to the Root (Trusted Root Certification Authorities) and Trusted Publisher system certificate stores.

While the installation of certificates into these stores, without asking the end user, may sound like questionable practice, please understand that none of these certificate actually introduce a security liability as nobody, not even the authors of libwdi, have the ability to reuse these certificates in order to mark malicious software as trusted. This is because:

  1. The private key for the certificate is generated at runtime, on the end user machine, and is not accessible to anybody else but the libwdi application that does the signing process for a limited time only
  2. after the driver package has been signed, the private key is destroyed. This effectively means that the certificate is a one-time use only, and cannot be used for anything but to validate driver package that was signed by libwdi, that one time. As a matter of fact, even if you re-install the exact same device using the same driver, libwdi has to generate a brand new certificate (and a new temporary private key), and replace the old one, because it is impossible to reuse it.
We do take security matters very seriously (which is the reason why we implemented the no-reuse policy above, with the obvious drawback of multiplying certificates), and since we are 100% open source, we invite anyone who wishes to do so to look at our PKI implementation, to want to confirm for themselves that, depite the additional certificates being added as trusted, the system's security is not being diminished in any way.

For more information, see our Certification Practice Statement page

How can I use libwdi to install a specific USB device with an installer?

When run from within an installer script, the wdi-simple sample from examples should be able to achieve just that. An example of Inno Setup installer script is also provided with the libwdi examples. See the wdi-simple.iss sample.

Does libwdi support internationalization?

Absolutely. The inf file created by libwdi is written in Unicode (UTF-16) format, to preserve any international characters, and the libwdi API calls and string parameters fully support UTF-8, meaning that you can name your devices with non English characters, or use directories in a different locale than English, and libwdi will happily process that data.

Can two libwdi applications concurrently access libwdi at the same time?

No. Just like Windows provides mechanisms to prevent simultaneous driver installations, most of the libwdi API calls will prevent concurrent access to libwdi to ensure that only one application at a time has the ability to install a system driver.

Why is my driver installation so slow?

The default on Windows Vista and later is to create restore a point before installing any driver. Depending on the system, this can be a fairly lengthy operation (20 seconds or, in extreme cases, up to 5 minutes). Because we are installing non bootable drivers, for devices (USB) that can be easily removed in case of problems, from v1.1.0, libwdi does its best to disable the creation of restore points. However, if the Local Group Policies are overridden or cannot be modified, you may have to contend with a restore point and a lengthy driver installation.

Are the extracted driver files removed after installation?

They aren't. The reason behind this is twofold. First, it is a bad idea to have an application running in elevated mode to delete files, as, if the extraction directory was user selected, we could end up deleting important data. Secondly, it is possible to use libwdi to setup the driver files for the next time the device is plugged in, in which case we must ensure that the files are still available. Deletion of the file is therefore left either to the end user or your application.

Why doesn't libwdi use dpinst.exe instead of a custom installer?

The two main reasons here are the size factor (each dpinst is about 1 MB in size or 300 KB compressed, whereas our installer is a lot smaller) as well as the ability to have greater control over the installation process. We might provide an option to embed and use dpinst in future versions of the library.

Why doesn't libwdi offer the option to set additional inf parameters?

If you ask this, you probably would like to have libwdi/Zadig set the class GUID, name or provider in the installed driver package. We have taken the deliberate decision not to add this feature, because:

  1. With the introduction of WCID for WinUSB devices in Windows 8 and then in Windows 7, Microsoft has effectively defined a unique class name, Universal Serial Bus devices as well as a unique class GUID, {88bae032-5a81-49f0-bc3d-a4ff138216d6}, for all USB devices that rely on the WinUSB generic driver. Therefore, if you were to set your own class properties for utilization in your application, not only will you not be able to use WCID as a result, but you will also prevent compatible WCID devices to work with your application, which we see as very detrimental for end users. This also applies even if you don't plan to use WinUSB, as WCID is not limited to WinUSB and the same restrictions will occur with other drivers.
  2. The main reason people want to set class parameters in the inf is either to place their own branding in the device manager (which makes it more confusing for end-users to locate their devices) or to define a class GUID that restricts the generic access that libwdi provides to hardware and software ("walled garden"). We don't think restrictions that impair the ability for end users to use the device of their choosing, including a cloned version, should be part of the libwdi generic driver installation process.
This being said, if you really want to have your own class properties defined in the inf, and you know the VID:PIDs of your devices, which should always be the case if you want to restrict the class from the default generic one, there's nothing preventing you from reusing WinUSB, libusb-win32 or libusbK as a custom driver with a customized inf that you control.

Zadig

I'm using Zadig on XP, and it doesn't seem to work

This is a known issue for v2.0.1.156 and later due to switching to Microsoft KMDF 1.11, as per the latest WinUSB drivers, which is incompatible with XP. Newer versions of Zadig see 2 releases, one with KDMF 1.09 for XP users, and one with 1.11 for Vista and later. Please make sure you download the version that is suitable for your platform.

Help, Zadig replaced the driver for the wrong device! How can I delete it?

First of all, I have to stress out that Zadig does exactly what you tell it to do, so if you select the wrong device, and Zadig happen to replace it's driver, the responsibility is yours. Zadig will try to warn about some of the common system drivers, such as HID or Mass Storage, to try to prevent you from replacing a driver you really don't want to replace, but there's only so many drivers Zadig can guess you probably don't want to replace. So, please try to pay attention to the device you select next time, and don't just close your eyes and press the install button on the first device you see, OK?

Now, restoring an old system driver, if Zadig replaced it, is pretty simple (as long as you didn't replace the HID driver for your mouse or keyboard). Let's say I inadvertently let Zadig install a driver for an USB Flash Drive, and replaced the Windows default Mass Storage device, and now I find that I want to access my device back as a flash drive:

  1. Open Device Manager
  2. Under Universal Serial Bus devices (or libusbk devices/libusb-win32 devices if you didn't choose WinUSB as your driver), right click on your device. In this example, ADATA UFD is our device.
  3. Select Uninstall and on the dialog box that appears, check the box that says Delete the driver software for this device then click OK.
  4. Once the uninstallation is complete, unplug and replug your device. The previous driver should be reinstalled (in our case, the Flash Drive reappears under the Disk drives category).

Clone this wiki locally