** Programmer's Technical Reference for MSDOS and the IBM PC ** 旼컴컴컴컴컴컴컴컴컴컴컴컴컴컴� � Shareware Version, 03/21/90 � � Please Register Your Copy � 읕컴컴컴컴컴컴컴컴컴컴컴컴컴컴� Copyright (c) 1987, 1990 Dave Williams C H A P T E R O N E DOS TECHNICAL INFORMATION SOME HISTORY쳐컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴� Development of MSDOS/PCDOS began in October 1980, when IBM began searching the market for an operating system for the yet-to-be-introduced IBM PC. Microsoft had no real operating system to sell, but after some research licensed Seattle Computer Products' 86-DOS operating system, which had been written by a man named Tim Paterson earlier in 1980 for use on that company's line of 8086, S100 bus micros. 86-DOS (also called QDOS, for Quick and Dirty Operating System) had been written as more or less a 16-bit version of CP/M, since Digital Research was showing no hurry in introducing CP/M-86. This code was hurriedly polished up and presented to IBM for evaluation. IBM had originally intended to use Digital Research's CP/M operating system, which was the industry standard at the time. Folklore reports everything from obscure legal entanglements to outright snubbing of the IBM representatives by Digital, irregardless, IBM found itself left with Microsoft's offering of "Microsoft Disk Operating System 1.0". An agreement was reached between the two, and IBM agreed to accept 86-DOS as the main operating system for thir new PC. Microsoft purchased all rights to 86-DOS in July 1981, and "IBM PC-DOS 1.0" was ready for the introduction of the IBM PC in October 1981. IBM subjected the operating system to an extensive quality-assurance program, found well over 300 bugs, and decided to rewrite the programs. This is why PC-DOS is copyrighted by both IBM and Microsoft. It is sometimes amusing to reflect on the fact that the IBM PC was not originally intended to run MSDOS. The target operating system at the end of the development was for a (not yet in existence) 8086 version of CP/M. On the other hand, when DOS was originally written the IBM PC did not yet exist! Although PC-DOS was bundled with the computer, Digital Research's CP/M-86 would probably have been the main operating system for the PC except for two things - Digital Research wanted $495 for CP/M-86 (considering PC-DOS was essentially free) and many software developers found it easier to port existing CP/M software to DOS than to the new version of CP/M. After eight years the wheel has turned a full circle. Digital Research aided IBM in writing DOS 4.0, which was subsequently licensed back to Microsoft, which has dropped further development of the operating system to tilt at the windmills of OS/2. MSDOS and PC-DOS have been run on more than just the IBM-PC and clones. There was an expansion board for the Apple ][ that allowed one to run (some) well - behaved DOS programs. There are expansion boards for the Commodore Amiga 2000, the Apple MacIntosh II, and the IBM RT PC allowing them to run DOS, the Atari STs can run an emulator program called PC-Ditto and boot any version of MSDOS with full functionality. Microsoft's OS/2 operating system incorporates the so-called "DOS Compatibility Box" to run DOS software; QNX will run DOS as a task, and a few Unix systems offer "DOS Merge" or control programs that allow DOS to run as a task. WHAT IS DOS?쳐컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴� DOS exists as a high-level interface between an application program and the computer. DOS stands for "Disk Operating System", which reflects the fact that its main original purpose was to provide an interface between the computer and its disk drives. DOS now lets your programs do simple memory management, I/O from the system console, and assorted system tasks (time and date, etc) as well as managing disk operations. Versions 3.1 and up also incorporate basic networking functions. With the introduction of installable device drivers and TSR (terminate but stay resident) programs in DOS 2.0, the basic DOS functions may be expanded to cover virtually any scale of operations required. OTHER OPERATING SYSTEMS쳐컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴 Various other operating systems are availible for the IBM PC. These include: Digital Research CP/M-86 Digital Research Concurrent CP/M-86 (multitasking) Digital Research Concurrent DOS (PC-DOS compatible, multitasking) Digital Research Concurrent DOS XM (PC-DOS compatible, multitasking, multiuser) Digital Research Concurrent DOS 386 (same, for 80386 computers) Digital Research DR-DOS 3.31 and 4.0 (PC-DOS clones) Wendin-DOS (PC-DOS compatible, multitasking, multiuser) QNX (multitasking, multiuser) UNIX (various systems from IBM itself, Microsoft-SCO, Bell, and various UNIX clones, single and multi user) MINIX (multitasking UNIX clone) PC-MOS/386 (PC-DOS clone, multiuser, multitasking) "Shell" programs exist which use DOS only for disk management while they more or less comprise a new operating system. These include: DesQview Windows OmniView GEM TopView TaskView NEC V-series CPUs can execute Intel 8080/8085 8-bit instructions as well as the 16-bit 8088-up instructions. They can run standard Digital Research 8-bit CP/M and MP/M directly, as well as other operating systems developed for that processor. SPECIFIC VERSIONS OF MS/PC-DOS쳐컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴� DOS 1.x is essentially 86-DOS. DOS 2.x kept the multiple file layout (the two hidden files and COMMAND.COM) but for all practical purposes is an entirely different operating system with backwards compatibility with 1.x. I seriously doubt there has been much code from 1.x retained in 2.x. DOS 3.x is merely an enhancement of 2.x; there seems little justification for jumping a whole version number. DOS 4.0, originating as it did from outside Microsoft, can justify a version jump. Unfortunately, 4.x seems to have very little to reason justify its existence - virtually all of its core features can be found in one version or another of DOS 3.x. DOS version nomenclature: major.minor.minor. The digit to the left of the decimal point indicates a major DOS version change. 1.0 was the first version. 2.0 added support for subdirectorites, 3.0 added support for networking, 4.0 added support for Lotus-Intel-Microsoft EMS. The first minor version indicates customization for a major application. For example, 2.1 for the PCjr, 3.3 for the PS/2s. The second minor version does not seem to have any particular meaning. The main versions of DOS are: PC-DOS 1.0 October 1981 original release PC-DOS 1.1 June 1982 bugfix, double sided drive support MS-DOS 1.25 June 1982 for early compatibles PC-DOS 2.0 March 1983 for PC/XT, many UNIX-like functions PC-DOS 2.1 October 1983 for PCjr, bugfixes for 2.0 MS-DOS 2.11 October 1983 compatible equivalent to 2.1 PC-DOS 3.0 August 1984 1.2 meg drive for PC/AT PC-DOS 3.1 November 1984 bugfix for 3.0, implemented network support MS-DOS 2.25 October 1985 compatible; extended foreign language support PC-DOS 3.2 July 1986 720k 3.5 inch drive support for Convertible PC-DOS 3.3 April 1987 for PS/2 series, 1.44 meg, multiple DOS partitions PC-DOS 4.0 August 1988 over-32 meg DOS partitions, EMS support IBM's PC-DOS is considered to be the "standard" version of DOS; Microsoft has sold MS-DOS over the counter only since version 3.2 (previously, Microsoft sold its versions only to OEMs). Most versions of DOS functionally duplicate the external DOS commands such as DISKCOPY, etc. Microsoft has announced that they will sell MS-DOS 4.0 only to OEMs, so it looks like the "generic" MS-DOS was short-lived. Some versions of MS-DOS varied from PC-DOS in the availible external commands. Some OEMs only licensed the basic operating system code (the xxxDOS and xxxBIO programs, and COMMAND.COM) from Microsoft, and either wrote the rest themselves or contracted them from outside software houses like Phoenix. Most of the external programs for DOS 3.x and 4.x are written in "C" while the 1.x and 2.x utilities were written in assembly language. Other OEMs required customized versions of DOS for their specific hardware configurations, such as Sanyo 55x and early Tandy computers, which were unable to exchange their DOS with the IBM version. At least two versions of DOS have been modified to be run entirely out of ROM. The Sharp PC5000 had MSDOS 1.25 in ROM, and the Toshiba 1000 and some Tandy models have MSDOS 2.11 in ROM. PC-DOS 3.0 was extremely buggy on release. It does not handle the DOS environment correctly and there are numerous documented problems with the batch file parser. The network support code is also nonfunctional in this DOS version. It is recommended that users upgrade to at least version 3.1. DEC MSDOS versions 2.11 for the Rainbow had the ANSI.SYS device driver built into the main code. The Rainbow also supported a unique quad density, single sided floppy drive. IBM had a version 1.85 of PC-DOS in April 1983, after the introduction of DOS 2.0. It was evidently for internal use only, supported multiple drive file searches (a primitive form of PATH), builtin MODE sommands for screen support, a /P parameter for TYPE for paused screens, an editable command stack like the public domain DOSEDIT.COM utility, and could be set up to remain completely resident in RAM instead of a resident/transient part like normal DOS. It is a pity some of the neat enhancements didn't make it into DOS 2.0. IBM also had an internal use only version 3.4, evidently used while developing DOS 4.0. Some versions of DOS used in compatibles do not maintain the 1.x, 2.x, ..... numbering system. Columbia Data Products computers labeled DOS 1.25 as DOS 2.0. Early Compaqs labeled DOS 2.0 as DOS 1.x. Other versions incorporated special features - Compaq DOS 3.31 and Wyse DOS 3.21 both support 32-bit File Allocation Tables. AT&T DOS 3.1 differs from generic MSDOS 3.10 in its use of cluster-size and file allocation table structures. AT&T DOS appears to use rules not from version 3, but rather those from version 2. Epson Equity III and ComputerLand 3.10 DOS's appear to use cluster techniques that are a cross between versions 2 and 3. On type DOS partitions, these DOS's use 3.x rules if the partition is larger than 32,680 sectors in total size. This implies 16 bit FAT entries as well. On partitions below this size, they will use 2.x rules, including the 12 bit FAT entries. Zenith DOS 3.x and Wyse DOS 3.2 have a builtin internal device driver to handle up to 4 32Mb DOS partitions on a single hard disk. Wyse DOS 3.31 will handle single partitions up to 512Mb with a 32-bit FAT. According to PC Week Magazine, July 4, 1988, Arabic versions of MSDOS are shipping with a hardware copy-protection system from Rainbow Technologies. This is similar to the short-lived system used by AutoCAD 2.52 and a very few other MSDOS programs, where an adapter block is plugged into the parallel port and software makes use of coded bytes within the block. This type of copy protection has been common on Commodore products for several years, where it is called a "dongle." The AutoCAD dongle was defeated by a small program written within weeks of version 2.52's debut. Version 2.62 was released 3 months later, without the dongle. The DOS dongle will, however, prevent the system from booting at all unless it is found. This makes the Arabic version of MSDOS the first copy-protected operating system, a dubious distinction at best. The modifications to the operating system to support the dongle are not known at this time. Versions of DOS sold in Great Britain are either newer than those sold in the US or use a different numbering system. DOS 3.4, 4.0, 4.1, 4.2, and 4.3 had been released there between the US releases of 3.3 and 4.0. MSDOS 4.0 was introduced in mid-1987 in Europe (at SICOB in Paris and sometime earlier by Apricot Computer in the UK). It apparently offers multitasking PROVIDED applications are specially written for it. David Fraser (UK MD of Microsoft) is on record saying that DOS 4 is unlikely to set the world alight and is of interest only to specific OEM's who want its features for networking/comms etc. Standard DOS applications will run under DOS 4.x as a foreground task according to uncertain information. It differs from earlier versions only in allowing background tasks to run. No info about how many tasks/memory requirements etc. at this time. It will run an MS-DOS 3.2 or earlier type task in a foreground partition, and a specially written task in the background. Microsoft changed their OEM licensing agreements between DOS versions 2.x and 3.x. OEM versions of DOS 3.x must maintain certain data areas and undocumented functions in order to provide compatibility with the networking features of the operating system. For this reason, resident programs will be much more reliable when operating under DOS 3.x. IBM's release of DOS 4.0 (and the immediate subsequent release of a bugfix) is a dubious step "forward." DOS 4.0 is the first version of DOS to come with a warranty; the catch is that IBM warrants it only for a very slim list of IBM-packaged software. 4.0 has some minor EMS support, support for large hard disks, and not much else. With its voracious RAM requirements and lack of compatibility with previous versions of DOS (many major software packages crash under DOS 4.0), plus the increase in price to a cool $150, there has been no great rush to go to the newest DOS. THE OPERATING SYSTEM HIERARCHY쳐컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴� The Disk Operating System (DOS) and the ROM BIOS serve as an insulating layer between the application program and the machine, and as a source of services to the application program. The system hierarchy may be thought of as a tree, with the lowest level being the actual hardware. The 8088 or V20 processor sees the computer's address space as a ladder two bytes wide and one million bytes long. Parts of this ladder are in ROM, parts in RAM, and parts are not assigned. There are also 256 "ports" that the processor can use to control devices. The hardware is normally addressed by the ROM BIOS, which will always know where everything is in its particular system. The chips may usually also be written to directly, by telling the processor to write to a specific address or port. This sometimes does not work as the chips may not always be at the same addresses or have the same functions from machine to machine. DOS STRUCTURE쳐컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴 DOS consists of four components: * The boot record * The ROM BIOS interface (IBMBIO.COM or IO.SYS) * The DOS program file (IBMDOS.COM or MSDOS.SYS) * The command processor (COMMAND.COM or aftermarket replacement) * The Boot Record The boot record begins on track 0, sector 1, side 0 of every diskette formatted by the DOS FORMAT command. The boot record is placed on diskettes to produce an error message if you try to start up the system with a nonsystem diskette in drive A. For hard disks, the boot record resides on the first sector of the DOS partition. All media supported by DOS use one sector for the boot record. * Read Only Memory (ROM) BIOS Interface and Extensions The file IBMBIO.COM or IO.SYS is the interface module to the ROM BIOS. This file provides a low-level interface to the ROM BIOS device routines and may contain extensions or changes to the system board ROMs. Some compatibles do not have a ROM BIOS to extend, and load the entire BIOS from disk. (Sanyo 55x, Viasyn machines). Some versions of MSDOS, such as those supplied to Tandy, are named IBMBIO.COM but are not IBM files. * The DOS Program The actual DOS program is the file IBMDOS.COM or MSDOS.SYS. It provides a high- level interface for user (application) programs. This program consists of file management routines, data blocking/deblocking for the disk routines, and a variety of built-in functions easily accessible by user programs. When a user program calls these function routines, they accept high-level information by way of register and control block contents. When a user program calls DOS to perform an operation, these functions translate the requirement into one or more calls to IBMBIO.COM, MSDOS.SYS or system hardware to complete the request. * The Command Interpreter The command interpreter, COMMAND.COM, is the part you interact with on the command line. It consists of these parts: Resident Portion: The resident portion resides in memory immediately following IBMDOS.COM and its data area. This portion contains routines to process interrupts 22h (Terminate Address), 23h (Ctrl-Break Handler), and 24h (Critical Error Handler), as well as a routine to reload the transient portion if needed. For DOS 3.x, this portion also contains a routine to load and execute external commands, such as files with exensions of COM or EXE. When a program terminates, a checksum is used to determine if the application program overlaid the transient portion of COMMAND.COM. If so, the resident portion will reload the transient portion from the area designated by COMSPEC= in the DOS environment. If COMMAND.COM cannot be found, the system will halt. NOTE: DOS 3.3+ checks for the presence of a hard disk, and will default to COMSPEC=C:\. Previous versions default to COMSPEC=A:\. Under some DOS versions, if COMMAND.COM is not immediately availible for reloading (i.e., swapping to a floppy with COMMAND.COM on it) DOS may crash. All standard DOS error handling is done within the resident portion of COMMAND.COM. This includes displaying error messages and interpreting the replies to the "Abort, Retry, Ignore, Fail?" message. DOS INITIALIZATION쳐컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴� The system is initialized by a software reset (Ctrl-Alt-Del), a hardware reset (reset button), or by turning the computer on. The Intel 80x8x series processors always look for their first instruction at the end of their address space (0FFFF0h) when powered up or reset. This address contains a jump to the first instruction for the ROM BIOS. Built-in ROM programs (Power-On Self-Test, or POST, in the IBM) check machine status and run inspection programs of various sorts. Some machines set up a reserved RAM area with bytes indicating installed equipment (AT and PCjr). The ROM routine looks for a disk drive at A: or an option ROM (usually a hard disk) at absolute address C:800h. If no floppy drive or option ROM is found, the BIOS calls int 19h (ROM BASIC if it is an IBM) or displays error message. If a bootable disk is found, the ROM BIOS loads the first sector of information from the disk and then jumps into the RAM location holding that code. This code normally is a routine to load the rest of the code off the disk, or to "boot" the system. The following actions occur after a system initialization: 1. The boot record is read into memory and given control. 2. The boot record then checks the root directory to assure that the first two files are IBMBIO.COM and IBMDOS.COM. These two files must be the first two files, and they must be in that order (IBMBIO.COM first, with its sectors in contiguous order). NOTE: IBMDOS.COM need not be contiguous in version 3.x+. 3. The boot record loads IBMBIO.COM into memory. 4. The initialization code in IBMBIO.COM loads IBMDOS.COM, determines equipment status, resets the disk system, initializes the attached devices, sets the system parameters and loads any installable device drivers according to the CONFIG.SYS file in the root directory (if present), sets the low-numbered interrupt vectors, relocates IBMDOS.COM downward, and calls the first byte of DOS. NOTE: CONFIG.SYS may be a hidden file. 5. DOS initializes its internal working tables, initializes the interrupt vectors for interrupts 20h through 27h, and builds a Program Segment Prefix for COMMAND.COM at the lowest available segment. For DOS versions 3.10 up, DOS also initializes the vectors for interrupts 0Fh through 3Fh. An initialization routine is included in the resident portion and assumes control during startup. This routine contains the AUTOEXEC.BAT file handler and determines the segment address where user application programs may be loaded. The initialization routine is then no longer needed and is overlaid by the first program COMMAND.COM loads. NOTE: AUTOEXEC.BAT may be a hidden file. 6. IBMBIO.COM uses the EXEC function call to load and start the top-level command processor. The default command processor is COMMAND.COM in the root directory of the boot drive. If COMMAND.COM is in a subdirectory or another command processor is to be used, it must be specified by a SHELL= statement in the CONFIG.SYS file. A transient portion is loaded at the high end of memory. This is the command processor itself, containing all of the internal command processors and the batch file processor. For DOS 2.x, this portion also contains a routine to load and execute external commands, such as files with extensions of COM or EXE. This portion of COMMAND.COM also produces the DOS prompt (such as "A>"), reads the command from the standard input device (usually the keyboard or a batch file), and executes the command. For external commands, it builds a command line and issues an EXEC function call to load and transfer control to the program. NOTE: COMMAND.COM may be a hidden file. NOTE: For IBM DOS 2.x, the transient portion of the command processor contains the EXEC routine that loads and executes external commands. For MSDOS 2.x+ and IBM DOS 3.x+, the resident portion of the command processor contains the EXEC routine. NOTE: IBMBIO only checks for a file *named* "COMMAND.COM". It will load any file of that name if no SHELL= command is used. IBM's original documentation spoke of installing alternate command interpreters (programs other than COMMAND.COM) with the SHELL= statement in CONFIG.SYS. Unfortunately, IBM chose not to document much of the interaction between IBMDOS.COM and IBMBIO.COM. By the time much of the interaction was widely understood, too many commercial software programs had been written to use peculiarities of COMMAND.COM itself. There are only two well-known replacements for COMMAND.COM - the commercial Command Plus and the shareware 4DOS.