Limitations on Writing Device Drivers in C (10826)


The information in this article applies to:

  • Microsoft C for MS-DOS 6.0
  • Microsoft C for MS-DOS 6.0a
  • Microsoft C for MS-DOS 6.0ax
  • Microsoft C/C++ for MS-DOS 7.0
  • Microsoft Visual C++ 1.0
  • Microsoft Visual C++ 1.5
This article was previously published under Q10826

SUMMARY

You cannot develop a complete device driver only in C or C++. You must write some assembly code and you cannot use the majority of functions in the C run-time library.

The three major problems that occur when you develop a device driver in C are as follows:

  • The .EXE file loader is in COMMAND.COM which is not available when MS-DOS loads a device driver. Therefore, a device driver must be the binary image of a single-segment program. Microsoft C/C++ version 7.0 introduced the ability to generate code in the tiny memory model in which the application stores its code and data in one memory segment.
  • The C run-time initialization code requires an appropriate PSP (program segment prefix) and requires MS-DOS to be available. Even though the source code to the startup code is provided with the C compiler, most of the startup code is in assembler and would require major modifications to be useful in a driver.
  • The entry-exit sequences used by C functions are inappropriate for use in an MS-DOS driver.
You can work around some of these limitations by using the C compiler to generate assembly code. The -Fa switch directs the compiler to produce an assembly code listing that you can edit and use as part of the driver source code. Depending on the complexity of the MS-DOS device driver, you could write most of the driver in C and modify and maintain the code relatively easily.
Modification Type:MajorLast Reviewed:12/1/2003
Keywords:kb16bitonly KB10826
©2003 Microsoft Corporation. All rights reserved.