Magnum One

home *** CD-ROM | disk | FTP | other *** search

/ Magnum One / Magnum One (Mid-American Digital) (Disc Manufacturing).iso / d22 / hdtest.arc / HDTEST.DOC < prev next >

Wrap

Text File | 1989-01-18 | 102.6 KB | 1,964 lines

Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ USER-SUPPORTED SOFTWARE _______________________ Much of the software which is available for your computer seems to have been produced and sold on the assumption that you are a crook, and that you will take any available opportunity to cheat its creators and vendors out of their rightful reward. Such software is copy-protected within an inch of its (and your computer's) life, and is still priced at a level which guarantees its producers a profit even if four or five illegal copies are made for each one sold. User-supported software, increasingly known as "Shareware" is different. You may have got a copy of this package free from a friend or a bulletin board or (for a nominal fee) from a users' group or software library. In any case, you have in your possession a perfectly legal copy of the program. You are further encouraged to make as many copies of the package as you like and to distribute them to all your friends and acquaintances. However, if, after trying the program out, you find ________ that it is useful and want to keep it and continue using it, you must ____ register your copy by sending the registration fee ($35.00,) together with a completed registration form, to the address on the form. If you continue to use the program after an appropriate trial period (and the "Shareware" system lets you determine what is appropriate), you are ___ breaking both the terms of the license under which you have received the software and the law. You may believe that the chances of the law of Copyright being enforced against you are very low (and you may even be right!), and the distributors of "user-supported" software do, in fact, usually simply have to trust you to be honest and willing to pay a reasonable price for a useful program. However, you should bear in mind that unsupported "Shareware" software is likely either to cease to exist altogether or to go commercial. The former eventuality benefits nobody - the latter only the "middlemen". As an additional inducement, Registered users will get a copy of the latest version of HDTEST, which already (at the time of writing) includes a number of significant enhancements over V4.45, and they will be entitled to support and help with HDTEST-related problems (including telephone support) for a year from the date of Registration. Additional features in V4.71 of HDTEST included support for a still wider variety of computer systems and storage media, full DOS 4.0 support, and the removal of the ANSI.SYS requirement for the screen display to work properly. Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ INTRODUCTION ____________ HDTEST was originally written to allow comprehensive Read/Write testing of hard disks on IBM PCs, XTs, ATs, and "true" compatibles. The V4.10 release relaxed many of the hardware restrictions imposed by earlier (V2.xx and V3.xx) versions of the program and made it useful for testing most MS-DOS-compatible storage devices currently in use on IBM-compatible computers, including removable (floppy and hard) disks and logically partitioned hard disks. This release (V4.45) includes BIOS-level support for DOS 3.3 "extended devices" and more options for the handling of totally unreadable sectors, and recognizes (but does not support) large (> 32 Mb) devices under IBM DOS 4.0 and Compaq DOS 3.31. Although many programs, including IBM's diagnostics, allow "surface scanning" and seek testing of hard disks, none that are available to the average user test whether the whole disk can retain the data that is or will be written to it. HDTEST's principal "claim to fame" is the ability thoroughly to test the magnetic media in a device which contains data while preserving the data (although data _________________________ loss is unavoidable if an area which is in use has become unreadable). ____ HDTEST tests the disk much more thoroughly than programs such as DFixer, Mace, and Norton's DT. Mace and DT, like many public domain "disk test" programs, simply read every cluster on the disk and report on any errors encountered. DFixer does better, writing 4 test patterns to every free cluster and checking that they can be read back correctly, but it makes no attempt to test clusters that are already in use - the very ones which contain your precious data! In addition to its testing function, HDTEST has been reported to be useful as an automatic disk repair utility for both hard and floppy disks which have developed (especially) intermittent or persistent read errors, preventing or impeding their normal use. Although HDTEST cannot read the totally unreadable, it tries much harder that DOS does and will almost always, in the course of its operation, allow you to recover the entire readable portion(s) of a file with one or more ________ unreadable clusters in it, optionally either remapping all of what it reads from the disk or replacing bad sectors (the default) or clusters with NULLs. HDTEST is not an alternative to backing up your floppies ___ and/or hard disk - it may sometimes be able to save your neck when you _________ forget or your backup goes bad too! On hard disks which have developed intermittent read errors, running HDTEST can sometimes totally cure ___ the problem, presumably because the "flaky" sectors get rewritten and the minor misalignments and/or weakening of the magnetic "image" which were causing the trouble are corrected when the data is rewritten. In its normal ("slow") mode, HDTEST writes 20 different test patterns to every cluster on the disk and checks that each pattern can be read _____ back correctly, while preserving (and eventually restoring) the contents of the clusters that contain data. HDTEST can even (on many devices) detect hard disk errors which are minor enough to be correctable (and corrected) by the disk controller's own "error- correcting code" (ECC). These errors are normally totally invisible to programs running under DOS and will not usually be picked up by other disk test utilities. They may, however, be early signs of problems that will become significant later, and HDTEST consequently marks these clusters, too, as bad. As long as HDTEST can read the data from a flaky cluster correctly once (and it tries quite hard) it will Page 2 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ be able to save the file of which that cluster is a part intact, while ______ marking the cluster so that DOS does not try to use it again. The program was designed to minimize the likelihood of user error causing data loss on a good (or slightly bad) disk, and to be easy to use - it could not be made very fast, since multiple bit patterns have to be written to and read from each cluster, and this takes a significant time even on a fast hard disk. This version does, however, have a less sensitive "fast" test mode, which writes only two test patterns to each block and is recommended for use when testing/repairing floppy disks and/or to correct intermittent problems or rule out gross problems on a hard disk. The program can be used to test/repair single and double-sided 8- and 9-sector minifloppies (if you have plenty of patience!), as well as ______ AT-format 1.2Mb minifloppies, 8" standard floppies, and 3.5" microfloppies (with the same proviso regarding a need for patience). HDTEST's companion program, HDCHEK, simply examines the logical structure of the system and target device and gives a short report on what it finds. It does not write to the disk, and is provided ___ principally to allow a quick and easy, but fairly reliable, check for compatibility between HDTEST and a given software/hardware configuration. As part of its report, HDCHEK notes at what level (DOS interrupts only, INT 13, or direct entry to the BIOS) it was able to access the target device, which has a bearing on HDTEST's ability to detect certain types of errors. In general, if HDCHEK runs __________ successfully on a system and target device, then HDTEST can be expected to do so, although the latter program does require more memory. The HDTEST package is distributed as "User-supported Software". If you intend to continue using the program, you are required to become a Registered user. Send a completed copy of the Registration Form (which is included in this documentation), together with a check or money order for $35.00 (+ $5.00 for shipping outside the continent of North America) either to me at the address which appears on the form or (for European users) to the address displayed on the European version of the program's copyright screen. Checks and money orders in denominations other than U.S. dollars are welcome, but please add a further $5.00 for bank charges and use the current exchange rate to calclate the fee if you are paying with such a check or money order. Please also add $2.50 (or equivalent) if you want the software on a 1.2 Mb HD floppy or a 3.5" minifloppy, rather than a 360 Kb DSDD floppy, and specify which you want on the Reg. form. I will send you a personalised copy of the latest, considerably enhanced, version of the HDTEST package (currently V4.71), and register you for a year's support. Note that the Registered version of the package will be shipped from the United States (by Airmail, for non-U.S. users), even if you use the European address, but European Registrations sent to that address are forwarded by computer network, and will usually reach me faster than if they were mailed to me in the U.S.A. Registration entitles you to be notified of upgrades, bug fixes, newly discovered incompatibilities, etc., and to ask for (and get) necessary help in adapting the program to work on your hardware. I cannot guarantee to make HDTEST work with every possible combination of hardware and software, but I will promise to try to solve problems as they come up, and I will return your registration fee if I cannot support your Page 3 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ system. Support will be available TO REGISTERED USERS ONLY by mail, ________________________ BIXmail, Compuserve EMail and telephone. Please also bear in mind that my willingness and ability to continue to support and enhance these programs will be directly dependent on users' willingness to pay the registration fee and (to some extent) inversely dependent on the number of questions I get that could be answered by reading the docu- mentation(!). The HD programs were written principally in C, compiled by the Microsoft C compiler (Version 5.10) using its "small" memory model, and linked with code from the distributed Microsoft C object libraries using the Microsoft LINK program (Version 3.65). Additional assembly language routines were assembled by the Microsoft MASM Macro Assembler (Version 5.10). Portions of the distributed program code are consequently Copyright by Microsoft Corp., 1985 - 1988. All rights reserved. These portions are used under the terms of a license from Microsoft Corp. The HD programs were written with care and have been tested on a number of different systems. I cannot, however, give any guarantee as to their performance on a specific system, nor can I accept liability for any actual or consequential loss resulting from their use. The programs may only be used and/or copied in their original form. Unauthorised modifications and/or the use of all or part of the programs in other software, commercial or otherwise, are strictly prohibited. Page 4 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ GETTING STARTED _______________ This section gives step by step instructions for running HDCHEK and HDTEST on the C: drive of a standard IBM PC, XT, or 100% compatible. Users of ATs or AT-compatibles may safely follow them too, but should really read the description of (and use) the /A command line flag. Subsequent sections also describe some of the program's other features and options in more detail, and should help to explain and let you deal with any problems which you may encounter. Note that you should never attempt to run HDTEST under a multi-tasking or multi-user _____ operating system, nor activate any "pop-up" program (e.g. SideKick) while it is running. 1) Use the DISKCOPY program to make a duplicate of the distribution disk, put the original away in a safe place, and use the duplicate as your working disk. 2) Make sure that you have current backups of all the files on drive C:. This is particularly important when running the program for the first time on a new system. 3) Run DOS's CHKDSK program on drive C: and correct any errors reported by it before proceeding further. 4) Make the drive and/or directory which contains HDCHEK.EXE and HDTEST.EXE your current drive and/or directory. 5) Insert a formatted, non-write-protected floppy disk in drive A. HDTEST will write an ASCII log file called HDTEST.LOG to this disk - any other files on it will be undisturbed. 6) Give the command: "HDCHEK C:". After a certain amount of disk activity has taken place, and certainly within a minute or two, your screen should look like one of the examples given in Appendix C. If any errors are reported, DO NOT PROCEED until you have read the rest ______________ of the documentation and/or are sure you know what is going on. 7) If HDCHEK reported no errors, give the command "HDTEST C: A:". ____________________________ After some disk activity, the program will prompt for replies to three or four questions - press <Return> in response to each prompt. The program will report its progress to the screen and will record any problems encountered in the log file on the floppy disk. Depending on the size of your DOS partition and the access time of the disk, a complete run of HDTEST (in its normal, "slow", mode) will take between 2 and 5 hours. Your intervention will only be required if a recovered _____ bad cluster is found in a "hidden", "system", or write-protected file, when you will be asked to approve remapping; otherwise, bad clusters will be automatically remapped. HDTEST can be aborted at any time by pressing <Esc> and responding to the "Are you sure...." prompt by pressing <Return>. 8) If HDTEST finishes testing without finding any bad clusters, it will simply record that fact and the finishing time in the log file and return to the DOS prompt. Otherwise, when read/write testing has Page 5 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ been completed, you will be prompted for permission to complete the remapping process by updating the disk FATs and Directories. Pressing <Return> will allow this to take place, and you will then be returned to the DOS prompt. 9) N.B. If HDTEST reported and remapped any bad clusters and you are ____ ___ using either the DOS (3.3 and above) FASTOPEN utility, or any other resident directory- or file-caching software, you should reboot your system before proceeding with other work on it. 10) If HDTEST reported and remapped some bad clusters, you should also read the section on "Dealing with bad clusters" later in this documentation. Page 6 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ HARDWARE & SOFTWARE REQUIREMENTS ________________________________ HDTEST and HDCHEK make a number of assumptions about the environment in which they are operating. They will, generally speaking, abort with an informative error message if they find one of these assumptions to be invalid! The programs require MS/PC-DOS 2.x or above and (for HDTEST) up to about 256 Kb of free memory (the exact amount required depends on the size and other characteristics of the device being tested) to run - the presence of more memory may (slightly) increase HDTEST's speed when testing devices with large cluster sizes. The device to be tested must be usable as a normal DOS device - commands such as DIR, COPY, and CHKDSK should be valid for it and for files and subdirectories on it. It need not, however, be bootable. Only the DOS partition of a multi-operating-system disk will be tested, but all the partitions of a disk that is divided into multiple DOS partitions can ___ be tested (one by one). This version of HDTEST fully supports DOS 3.3 extended partitions, and will test them as thoroughly as it does normal DOS partitions. HDTEST's support for other disk partitioning software is provided only to the extent that DOS support is provided by the authors of the partitioning software. If you wish to test a new hard disk which is to be configured to ____ support multiple operating systems, you should first configure it as a single DOS partition, then test it in that configuration, and finally reconfigure it to the form in which you are going to use it. Note __ that, under these circumstances, HDTEST can only be used as a test - ____ ____ when you repartition the disk, any remapping and/or marking of bad clusters which HDTEST may have done is lost, unless your partitioning software also allows you to specify bad sectors to be locked out. Ideally, for normal hard disks, the BIOS handler for interrupt 13 should be capable of handling (at least) normal disk reads and writes (sub-functions 2 and 3) in a way that is compatible with that documented for the XT disk adapter in IBM's technical documentation. This should be no problem for most disk systems on IBM PCs and most compatibles. HDTEST may, however, be forced to use the (somewhat more limiting) DOS device interface when used with non-bootable hard disks and/or with hard disks which require special device drivers to be installed, and will be forced to use the DOS device interface when dealing with disks partitioned by means other that the DOS 3.3 utilities. Earlier V4.xx releases of HDCHEK and HDTEST would abort if run on small (<= 10 Mb) hard disks which had been patched to use a cluster size smaller than the DOS standard (and 16-bit FATs). This deficiency has been corrected in the current version, but very small (less than 1024 clusters) partitions may still cause problems if 16- bit FATs are used (there is no good reason why such partitions should use 16-bit FATs). If run under DOS 3.0 or above on an IBM INT 13 compatible hard disk, HDTEST attempts to locate and use the ROM BIOS INT 13 entry point, rather than the normal INT 13 vector (which DOS 3.x intercepts and interferes with). By default, unless the /H option is used to force a different segment address, or the /A, /D or /I flag is set, the program may assume that, if a ROM is present at absolute address ___________________ C8000H, it contains the hard disk BIOS and has initialization code ___ starting at offset 3. This assumption is true for all hard disks which Page 7 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ "plug in" to the IBM PC and XT. It may not be true for all clones which come with hard disks installed, although it is likely to be true for most of them. The AT and most of its clones have no ROM at all at this location, which is OK, too, even if you forget to use the /A flag (but see para 5 in the "Limitations and Restrictions" section). Problems (up to and including system halts) may occur if none of the above flags is set and there is a ROM at C8000H but it does not __ ___ control the hard disk. They are almost guaranteed under these circumstances if the initialization code does not start at offset 3 in the ROM. If you know that your hard disk ROM is located somewhere other than at the default segment address, you can use the /H flag to tell the programs where to look for it. If you have other problems running HDCHEK and HDTEST on non-standard systems, try running the programs with the /I flag (for PCs and XTs) or the /A flag (for ATs), or with the /D flag (for any hardware), and/or with the /C flag. All these command line flags are discussed in more detail below. Note that, since HDTEST remaps bad clusters to free space, there is not much point in running it on a disk which has none (as shown by DIR or CHKDSK), since the program will abort immediately if it finds a bad cluster. Unless you have a really flaky disk, however, you should not need more than a couple of hundred Kb free to remap any reasonably probable number of bad clusters. ANSI.SYS or another ANSI console device driver should be installed on those systems (including the IBM family) which do not support ANSI screen I/O in Video BIOS - nothing terrible will happen if ANSI.SYS is not installed, but the screen display will be very strange (!!). HDTEST has been tested with the Public Domain NANSI driver, as well as with FANSI-CONSOLE (tm). Page 8 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ LIMITATIONS AND RESTRICTIONS ____________________________ The current version of HDTEST has a few limitations, most of which result from the way data is stored internally by the program. None of them are "etched in stone", although some could only be avoided at the cost of slowing down what is already far from a fast process. The known limitations are as follows: 1) Maximum number of clusters per partition = 24,000 (the number is approximate, because it depends on dynamic memory allocation). This limit may be increased to 32,768 in a subsequent version of the program. 2) Maximum cluster size = 65,528 bytes. I do not believe that a DOS disk can ever have clusters larger than 64 Kb, but my limit is a few less than that and may, since sector and cluster sizes tend to run in powers of 2, effectively limit cluster size to 32 Kb - this should be enough for most people!. 3) Because of certain hardware limitations of the chips used to support direct memory access (DMA, the method used to transfer data from disk to memory and vice versa) in the IBM PC and most clones and compatibles, HDTEST may be unable to use available memory efficiently when cluster sizes (and the buffers needed to hold them) are large. This inefficiency can result in "Insufficient Memory" or "Memory Allocation" errors, even on systems with more than the minimum 256 Kb of free memory prescribed above. The only cure for this is to free up (or buy!) more memory (remove RAMdisks, remove resident software, etc.). 4) This version of HDTEST will run on ATs and most AT clones, but may not, even with the /A flag set, be able to detect corrected ECC errors when run on some of the clones - a warning message will be displayed when the program initializes itself if corrected ECC errors are not being detected. The program will also not detect corrected ECC errors when invoked with the /I flag (under DOS 3.0 or above) or the /D flag. 5) The current Shareware version of HDTEST will run under IBM DOS 4.0 and Compaq DOS 3.31 but cannot be used to test the large (> 32Mb) DOS partitions which these DOS versions support. The Registered version of the program fully supports these large partitions. 6) For reasons which are poorly understood (at least by the author!), HDTEST runs very slowly on Epson Equity II systems, particularly when ____ used to test "plug-in" hard drives of the "Hardcard" or "Flashcard" variety. For this hardware, the estimated run times given elsewhere in the documentation need to be multiplied by about 2.5. __________ ___ Page 9 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ RUNNING THE PROGRAMS ____________________ Always run HDCHEK on any new hardware/software configuration before you attempt to run HDTEST (at least the first time). The latter program has been written as conservatively as possible, but it is not possible to predict the behavior of every conceivable combination of hardware and operating system when dealing with a program which bypasses most of the controls which normally protect the user from destroying his world. HDCHEK is invoked with the command line: HDCHEK device [/A] [/Cn] [/D | /I] [/Hnnnn] "device" is the device to be tested (e.g. A:, B:, etc. - the colons are permitted but not required). For versions of HDCHEK from V4.35, ___ the spaces between the device identifier and the flags (and between different flags) are optional - the command lines "HDCHEK C:/C2/D" and "HDCHEK C: /C2 /D" are equivalent. The /A(T) flag is used to tell the HD program that it is being run on an AT or AT-clone (which has its hard disk BIOS in the system ROM at F0000H, rather than in a separate ROM at C8000H). This flag also forces HDTEST to compensate for a known bug in the (IBM) AT's floppy disk BIOS when used on 1.2 Mb floppy disks. The /C(lone BIOS) flag is used in systems which do not support the IBM BIOS's INT 11 "equipment determination" BIOS function (e.g. the Tandy 2000). The parameter ("n") is the number of floppy drives attached to the system, which is normally determined automatically. The /D(OS interrupts) flag forces HDCHEK to use the hardware- independent DOS interrupt 25 for access to the target device. It is usually not required, since the program is usually smart enough to _______ _______ recognize devices which have to be accessed this way (e.g. logical disk partitions, non-bootable hard disks, some removable-media hard disks). If the /D flag is not used, the program will issue a warning ___ message if it finds it has to use the DOS interrupts. N.B. the definition of the /D flag changed at V4.00 - the new /I flag _____________________________________________________________________ has the same effect as the old (pre-V4.00) /D flag. Only one of these _____________________________________________________________________ flags may be present on the command line. _________________________________________ The /I(nterrupt 13) flag has no effect under DOS 2.x. Under DOS 3.0 or above, it forces HDCHEK to use the existing INT 13 vector (which it always does under DOS 2.x), and prevents the program from trying to determine the ROM BIOS entry point; the flag should be used if you know that your system's fixed disk BIOS uses INT 13 but does not follow IBM's "rules" or if the program crashes the system when invoked without it. The /H(ex BIOS address) flag is used to tell the HD programs that the ROM BIOS of the device being tested is located at nnnn, which is interpreted as a Hexadecimal segment address, rather than at C800, which is the default. There must be no space between the H and the number. /HF000 is similar to /A in its effects, but does not correct for the bug in the IBM AT floppy disk BIOS mentioned in the above description of the latter flag. Page 10 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ After a short delay, during most of which the specified device will be active, a description of the device's logical structure will be disp- layed. After a further delay, and more disk activity, a message will be printed giving the number of files and subdirectories found on the disk, and the program will exit back to DOS. Note that the total number of files and subdirectories reported by HDCHEK will be greater than the sum of the counts reported by CHKDSK. This is because HDCHEK counts the "." and ".." entries in subdirectories and CHKDSK doesn't. Errors and/or problems encountered, if any, will be reported on in some detail. A number of example screen displays and log files produced by running HDCHEK (and HDTEST) on different systems will be found in Appendix C at the end of this documentation. Always run the DOS CHKDSK program immediately before running HDTEST on a disk. If CHKDSK reports any problems, its /F option (or other means) should be used to correct them before HDTEST is run. ____________________ HDTEST is invoked with the command line: HDTEST device1 [device2] [optional parameters] "optional parameters" may include one or more of the following: [/A] [/B] [/Cn] [/D | /I] [/F] [/Hnnnn] [/N] [/P] [/Rn] [/Snnnnn] [/Zn] "device1" is the name of the device to be tested, as for HDCHEK. "device2" (which, if present, must be different from "device1") is the name of a device on which a log file (called HDTEST.LOG) is to be written. If "device2" is not specified, errors are reported to the console terminal and the DOS console redirection facility should NOT ___________________________________________________ be used in an attempt to redirect them elsewhere. For versions of ________________________________________________ HDTEST from V4.35, the spaces between the device identifier and the flags (and between different flags) are optional - the command lines "HDTEST C:/A/F" and "HDTEST C: /A /F" are equivalent. The /A, /C, /D, /H, and /I flags have the same meaning and effect that they do for HDCHEK. When used for HDTEST, the /D flag forces the use of INT 25 to read from and INT 26 to write to the target disk. The /B(oot/Partition sector test) flag forces the program to include the boot/partition sector (track 0, head 0, sector 1) in the DOS area for the purposes of read/write testing. The default action is only to perform read and seek tests on this sector, even if read/write testing of the DOS area is specified. Since the boot/partition sector is effectively read-only under normal circumstances (it is only written _______ to during low-level formatting and by FDISK), read/write testing is __ less vital than for the rest of the disk. In addition, at least one hard disk controller (a Western Digital board which allows a single disk to be addressed as two "physical" devices) will not allow test patterns to be written to this sector. Certain devices (chiefly RAMdisks) appear to DOS to be hard disks but do not, in fact, possess boot/partition sectors. On these devices, HDTEST will report check/test results on the boot/partition sector while actually reading/writing the device's boot sector - the latter sector will simply end up being checked/tested twice under these circumstances. Page 11 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ The /F(ast mode) flag causes the program to use only two test patterns (all 1s and all 0s) for read/write testing. It is intended for use on floppy disks and for "quick and dirty" testing of other devices. The /N(o Query) flag disables the individual prompts for confirmation which the program normally issues before remapping clusters which are marked as "Reserved" or which are part of a file which is marked as "System", "Hidden", or "Readonly". The /N flag does not prevent the ___ program from prompting for confirmation before updating the disk. The /P(rint) flag causes the log file to be directed to the PRN device (normally LPT1), and overrides any "device2" specification. The /R(epeat)n flag causes HDTEST to go through n complete testing cycles (omitting n, or setting it to zero, causes HDTEST to run continuously until terminated by pressing <Esc>). Instead of stopping after completing the testing of allocated data clusters, the program immediately starts again with the DOS area. Any bad clusters found and flagged in the first pass should be reported as "already marked as bad" in subsequent passes. The /R flag implicitly sets the /N flag. The /S(tart)nnnnn flag causes HDTEST to start testing occupied ________ clusters at cluster number nnnnn (expressed as a decimal number). There must be no space between the "S" and the number. This option will mainly be useful when a previous run of HDTEST was interrupted (by pressing <Esc>), since it allows you to start where you left off. Note that, in order to be sure that data is not remapped to a bad area, HDTEST always tests all empty clusters before starting to test ______ ___ occupied clusters. For this reason, using the /S flag will not result in a large time saving if your disk is nearly empty, and it will make the most difference on a full, or nearly full, disk. The /Z(ero bad sectors)n flag controls how HDTEST deals with sectors that cannot be read correctly (unrecoverable read errors) when it remaps bad clusters. Versions of HDTEST up to and including V4.10 (somewhat cavalierly) disregarded the remaining contents of the entire cluster if one or more sectors in it were unreadable, and remapped it to a cluster of NULLs. The default action is now to replace only the unreadable sectors with NULLs, transferring the remaining sectors _______ intact to the remapped cluster. There are, however, two other options, which can be chosen by setting n to 0 or 2 (if n is omitted, 1, or any other number, the default action is taken). Setting n to 0 uses the old default (remaps the bad cluster to a cluster of NULLs), while ___ _______ setting n to 2 goes to the other extreme and copies all data read (some of which is almost certainly invalid) to the remapped cluster. It is always safe and often advisable to use the /Z2 option provided ________ that you do not try to use any non-ASCII files that have been remapped ______________________________________________________________________ with unrecovered read errors and this (/Z2) option. Remapped ASCII ___________________________________________________ files can be inspected with an editor and fixed up where necessary - non-ASCII files (e.g. .EXE, .COM, .WKS files) may appear to be undamaged (the program may run, the worksheet be readable, etc.) but can still contain a small change which will cause trouble later, and they cannot easily be checked. The large "holes" which /Z options 0 and 1 leave in such files usually allows it to be much more obvious that they are defective! Page 12 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ After displaying its copyright message, HDTEST reads the FATs and scans the root directory and all subdirectory trees of the specified device before initiating a short dialogue. All questions are of the "Yes/No" type and should be answered by typing "Y" or "N" (case does not matter) and pressing the <Return> key. Pressing the <Return> key alone is equivalent to giving the default response (which appears in square brackets at the end of each prompt). HDTEST then proceeds to test the specified device. By default, the DOS area (boot sector, FATs, and root directory) is not subjected to Read/Write testing, since failures in this area usually make themselves apparent fairly rapidly. The user can, however, choose to run these tests on the DOS area too, or to limit testing to read and seek tests only over the whole partition. In its reports to the screen, HDTEST uses the convention that read- only testing is "Checking" while read/write testing is "Testing" - it always "checks" the boot/partition sector and the rest of the DOS area, but only "tests" them if requested to do so. In "checking" or "testing" a block (sector or cluster), HDTEST first reads it (retrying, if necessary, up to five times) and then does the longest possible seek away from it before attempting to read it again. "Checking" is complete at this point. "Testing" continues with the program sequentially writing 20 (or 2, for "fast mode") different bit patterns to the block and checking that each one can be read correctly, before rewriting the original contents of the block (and checking to ensure that it has been restored correctly). Errors that occur at any point in this process are reported and (if the block is in the data area - errors in the DOS area are always treated as fatal) the block is remapped. If an error is detected in a "Reserved" cluster or in one that is part of a file which is marked as "System", "Hidden", or "Readonly", the user is normally asked for confirmation before the cluster is remapped, otherwise it is assumed that remapping is desired. HDTEST keeps its own copy of the FAT and a map of the device's directory structure in memory. A bad cluster is remapped by transferring its contents to a previously free cluster and (if the bad cluster was in use) changing all necessary FAT and directory entries so that the pointer chain which maintains the structure of the affected file or subdirectory remains valid. The bad cluster's FAT entry is then altered to identify it as bad to DOS. The device's FATs ________ and directories are not altered until the whole device has been scanned, and then only after the user has confirmed that it is OK to proceed. This means that a power failure and/or an inadvertent power- off which occurs while the program is running will cause the loss of, at most, one cluster of data and of no FAT or root directory information (unless you are very unlucky and it happens after you have ____ confirmed that you want a damaged data area to be remapped and while this is being done). If a power failure and/or power-off occurs while a cluster which contains the directory information of a subdirectory is being tested, part or all of that subdirectory will probably become inaccessible to DOS, although the data in it will remain on the disk and be potentially recoverable with the aid of (e.g.) the Norton Utilities (TM). You can safely interrupt the testing process part way through (by Page 13 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ pressing the <Esc> key) with no risk of leaving a partially updated FAT or directory. If you do interrupt the program in this way, you will still be given the option of updating the disk, so that bad clusters which were detected before <Esc> was pressed can be remapped. Errors which occur in the program's initialization phase cause it to abort with an error message and (sometimes) a hexadecimal error code. The meaning of the error code is context-dependent. Errors detected during read-write testing only cause the program to abort if they occur during testing of the DOS area, or if remapping is impossible (only likely to happen if the disk is full). In any event, a description of the error is written to the log file (if any) and/or to the screen. This specifies the cluster or (for the DOS area) sector number and the name of the affected file (if any), and defines the type(s) of error which occurred. Appendix B contains a discussion of some of the more technical aspects of data storage on disks (floppy and hard), and includes more information about some of the possible errors that may be encountered. HDTEST takes a very long time to run to completion because of the ____ large number of reads and writes it does for each cluster. A typical time for full testing ("slow" mode) of an XT disk (10 Mb) would be 2 hrs; full disk testing on a 30 Mb Enhanced AT takes at least 4 hrs. Page 14 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ DEALING WITH BAD CLUSTERS _________________________ It is, unfortunately, not unusual for a new hard disk to have a small, _____ constant, number of bad clusters. IBM's quality control standards ________ allow, for example, 5% of the surface area of a disk to be bad (!). These bad clusters, which correspond to imperfections in the magnetic coating on the disk, will normally be detected and marked by the DOS FORMAT program, and HDTEST will report them as "already marked as bad". This should not be regarded as evidence of any real problem with the disk, although some suppliers (not including IBM) will ___ replace such disks under warranty if pressed to do so. It is also not unusual for a disk to develop a few more defects as it ages. When ___ these do develop, they are also usually constant in size and position, and will be reported as "already marked as bad" by HDTEST on runs subsequent to that during which they are first found. The occasional __________ appearance (and subsequent persistence) of a new bad cluster is, therefore, not necessarily evidence of serious trouble developing. However, if more than one or two new bad clusters are reported by HDTEST, and especially if two consecutive runs (or passes, if the /R flag is set) of HDTEST report different bad clusters, there is a good _________ chance that the hard disk or controller is developing a serious problem - the disk should be backed up immediately and the disk and/or controller sent for repair at the earliest opportunity. New bad clusters in the data area are automatically mapped out by HDTEST, and no further action is necessary. Bad sectors in the DOS area can be handled in two ways: 1) If you are able to do a low-level format on your hard disk (most _________ makers of "add-on" hard disks supply a method of doing this, either in the form of a short program, or by giving instructions for running a program in the ROM BIOS from DEBUG), back up the entire hard disk ______ (including any non-DOS partitions), run the low-level format, use FDISK to repartition the disk, and then run the DOS FORMAT program. If HDTEST now reports no errors in the DOS area you can restore all your files and continue to use the disk. You should, however, use HDTEST to test (at least) the DOS area fairly frequently for a month or so to be sure that the problems are not going to recur. 2) If you do not have the means to do a low-level format, or if you follow the above instructions and HDTEST continues to report errors in the DOS area, your only other choice is to back up the DOS partition, reconfigure the disk with FDISK so that the new DOS partition starts above the bad area, reFORMAT it, and restore all of the old files to the new DOS partition (which will unavoidably be a little smaller than the old). If absolute sector zero (track 0, Head 0, sector 1) becomes bad, the disk is useless and must be replaced. This sector is only subjected to read/write testing if the /B flag is specified on the command line, so __________ any reported errors (in sector -1) during normal testing represent a __ problem reading the sector and are likely to be significant. Problems _______ encountered during read/write testing of this sector do not always ___ indicate a bad disk - certain controller cards may not allow test patterns to be written to it, resulting in a report of multiple errors on read/write testing. Consequently, if multiple errors are reported Page 15 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ on read/write testing of this sector, but read-only testing proceeds without problems and the rest of the disk seems to be OK, there is probably no cause for concern. BUGS AND OTHER UNPLEASANT SUBJECTS __________________________________ Although, like all programmers, I like to think that my product is perfect, it is possible that there are a few bugs lurking in the depths of the program, and likely that some combination of software and hardware exists which HDTEST cannot figure out and/or cope with. If you think you have found a bug and/or if you cannot get either of the programs to work on your system, do the following: 1) Reread this documentation! Make sure that you are doing exactly what the documentation says to do. Make sure (if you are testing a drive with removable media) that the media is not write-protected. 2) Boot your system from a "virgin" DOS disk - preferably a copy of your DOS distribution disk - and try to run the HD programs again. If they run in this environment but not in your normal DOS environment, try reconstructing the system under which they failed by adding things one by one until the problem recurs. If you find a specific piece of software which causes the trouble, let me know about the incompatibility. 3) If the above approaches do not help, or if HDCHEK runs but HDTEST bombs (the reverse should never happen), send me as detailed a description of the problem as possible and I will do my best to diagnose and fix it. A bug report should always include a full ______ description of the system being used, as well as the output from a run of HDCHEK. I can be contacted by U.S. mail at the address on the Registration Form, by BIXMail to pfletcher, or by Compuserve EMail to 72255,305. Be aware that, although I do pay attention to bug and problem reports from people who are not Registered users, the latter will always be given priority. Page 16 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ CHANGES FROM PREVIOUS VERSIONS ______________________________ The first version of the HD programs actually released (not including beta-test versions) was V2.60. The programs have grown and developed gradually, for the most part, but the releases listed below have constituted the most significant "watersheds" - features added and changes made between listed releases are treated here as if first _______ included in the next listed release. Version 4.45 (this version) ____________ 1) Deficiency causing "fallback" to DOS interrupt level under some circumstances when attempting to test "extended devices" on the second physical hard drive fixed (HDCHEK and HDTEST). 2) Limited support for IBM DOS 4.00 and Compaq DOS 3.31 large partitions added - these partitions are now recognized as valid (allowing HDTEST to be used to test "normal" partitions on systems which include large ones) but cannot, themselves, be tested (HDCHEK _______ and HDTEST). Version 4.41 ____________ 1) Deficiency (introduced at V4.00) causing HDCHEK to issue an incorrect error report on an abort resulting from a FAT comparison error corrected. 2) Minor deficiency, causing HDTEST to create an empty log file on the specified device when both a log device and the /P flag were present, corrected. 3) Work-around for hardware multi-sector read ECC problem (see Appendix B) added to HDTEST. 4) Support for small (<= 10 Mb) hard disks patched to have small clusters and 16-bit FATs added. 5) Requirement that option flags be separated from device specifier(s) (or each other) by one or more spaces or tabs, removed. 6) /R option added to HDTEST for multi-pass disk testing. Version 4.30 ____________ 1) Support for BIOS-level access to DOS 3.3 extended partitions added. 2) /Z option added to HDTEST to control remapping of unreadable sectors in bad clusters. Page 17 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ Version 4.10 ____________ 1) Assorted minor changes made to screen display. 2) /C option added to allow the programs to be used on some non-IBM compatible MS-DOS systems. 3) /D option redefined - old /D option became new /I option. 4) /F option for fast read/write testing added to HDTEST. 5) /H option added to allow BIOS-level access to the disk on systems with non-standard hard disk BIOS locations. Version 3.13 ____________ 1) Assorted minor bugs and deficiencies in error reporting and termination routines in HDTEST fixed. 2) Minor bug in HDTEST, causing multiple recovered ECC errors occurring during initial cluster reads to be reported as recoverable errors without further qualification as to their type and origin, fixed. 3) Disk reset added before retry on all errors. 4) Absolute limit on number of files and directories on device removed. 5) Total sector or cluster count shown on screen by HDTEST during testing along with current sector or cluster number. 6) /A option for full functionality on ATs and AT clones added. 7) /B option to include the boot/partition sector in the DOS area for the purposes of read/write testing added to HDTEST. This sector was previously tested with the DOS area. Page 18 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ Version 2.73 ____________ 1) Bug (in V2.60) which could cause bad termination of HDTEST and possible subdirectory inconsistencies if a sector which was not in use was found to be bad, fixed. 2) Bug (in V2.70) which caused HDTEST to enter a terminal input loop if ANSI.SYS was not loaded and <Esc> was entered to abort it, fixed. 3) Minor bug in V2.60 (which resulted in a failed directory scan if HDCHEK was run on a write-protected device) fixed. 4) Internal changes made to avoid interference with expanded memory device handlers. 5) Ability to update disk after HDTEST terminated by user (typing <Esc>) added. 6) /S option to start checking occupied clusters at a given cluster number added to HDTEST. Page 19 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ APPENDIX A - SUMMARY OF FLAG SETTINGS _____________________________________ /A(T) - used to tell the HD program that it is being run on an AT or AT-clone (HDCHEK and HDTEST). /A is not quite equivalent to /HF000. /B(oot/Partition sector test) - used to include the hard disk boot/partition sector in the DOS area for the purposes of read/write testing (HDTEST only). /C(lone BIOS)n - used in systems which do not support the IBM BIOS's INT 11 "equipment determination" BIOS function (e.g. the Tandy 2000). The parameter ("n") is the number of floppy drives attached to the system, which is otherwise determined automatically. /D(OS interrupts) forces the program to use only DOS interrupts (INT 25 and INT 26) for access to the target device. May not be combined with the /I flag (HDCHEK and HDTEST). /F(ast mode) limits read/write testing to the use of two patterns only (HDTEST only). /H(ex BIOS address)nnnn - used to tell the HD programs that the ROM BIOS of the device being tested is located at nnnn, which is interpreted as a Hexadecimal segment address, rather than at C800, which is the default. /I(nterrupt 13) - forces the program to use the existing INT 13 vector. This is the default under DOS 2.x. May not be combined with the /D flag (HDCHEK and HDTEST). /N(o Query) - disables the individual prompts for confirmation which the program normally issues before remapping "special" clusters - does not prevent the program from prompting for confirmation before ___ updating the disk (HDTEST only). /P(rint) - causes the log file to be directed to the PRN device (HDTEST only). /Rn - causes HDTEST to perform n complete testing cycles (or to continue until interrupted for n=0) (HDTEST only). /Snnnnn - causes HDTEST to start testing occupied clusters at cluster ________ number nnnnn (expressed as a decimal number) (HDTEST only). /Zn - controls remapping of unreadable sectors in bad clusters (HDTEST only). n may have one of the following values: 0 - Remap clusters with unreadable sectors to all NULLs 1 (default) - Remap only unreadable sectors to NULLs 2 - Transfer all data read to new cluster Page 20 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ APPENDIX B - DISK ERROR MESSAGES ________________________________ Users of earlier versions of this software have repeatedly requested that I provide a more detailed explanation of the significance of some of the error messages displayed by HDTEST when it encounters a bad sector or cluster on the disk. It is not possible to do this without assuming (or imparting) a more detailed understanding of the "nitty- gritty" details of data storage on hard and floppy disks than most people want or need, let alone possess. This Appendix is not, therefore, "required reading" for the safe or effective use of the HD programs. It is intended as a tutorial for those who would like to understand something more of what is really going on on their hard or ______ floppy disk drive when the little red light comes on. For the purposes of what follows, it is (fortunately) not necessary to know or understand exactly how changes in magnetic flux on the surface of a hard (or floppy) disk are translated into binary data, or vice versa. That could be the subject of another long dissertation. I will simply take for granted that binary data, conceptually organised into 8-bit bytes, can be read and written by the disk drive hardware under the command of its controller (which usually resides on a separate card plugged into one of the computer's expansion slots). I will be describing how blocks of data are physically arranged on a disk and how the disk controller goes about ensuring that the data it reads has not been corrupted since it was written. I propose to start off by running over some of the error-detection and error-correction techniques and procedures that are applied to data stored on magnetic media. These techniques and procedures are necessary because, although magnetic media are very reliable and the technology of recording and playback is well understood, there always remains the possibility that a defect in the medium or some other intercurrent problem may cause the sequence of bytes read back from a disk or tape to be something other than a perfect duplicate of the sequence that was stored. If a single copy of the raw data were all that were recorded, there would be no way of immediately detecting (let alone correcting) such an error. It might be brought to light when the file that was changed caused problems when it is used by or executed as a program, but it might not become apparent until it had propagated itself through and caused problems or errors in multiple other programs or files. This would clearly be an unacceptable state of affairs. In order to prevent this from occurring, then, there should be some means of checking data which is read in from storage for errors, and/or (preferably) of correcting any errors which may have occurred before any use is made of the data. ______ Although error checking and/or correction can (and sometimes does) take place at any level from that of the application program down to that of the device controller, I am only going to describe the means that are used by hard and floppy disk controllers. When a block of data is passed to the controller to be written to the disk, a series of arithmetical operations is performed in sequence on all the bytes in the block, to produce a 2-byte (for diskettes) or multi-byte (for hard drives) code. The diskette controller implements what is known as a cyclic redundancy check (CRC) algorithm, which is optimized to allow detection of all single bit errors and many longer error bursts (it is _________ Page 21 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ reasonable to assume that errors will most probably either affect a single bit at random or, if they result from media defects, a number of adjacent bits). The longer (at least 32-bit) error correcting code (ECC) calculated by a hard disk controller gives enough information to allow detection and correction of, at least, all single bit errors and ______________ detection of most likely longer error bursts. The CRC or ECC is saved on disk with the data and is read back in by the controller when the data is read. The controller then compares the saved CRC or ECC with one newly calculated from the data. If the two are identical, the data is assumed to be valid and a successful read is signalled. If there is a discrepancy, a diskette controller always returns the data with a "CRC error", since the CRC code does not contain enough information to allow the incorrect bit or bits to be identified. On a hard disk, if the ECC comparison suggests a correctable error, it will be corrected, and the read call will return with a "Corrected ECC Error" code but with (presumably) valid data in the transfer buffer. If a more extensive error is detected, on the other hand, the call returns with a "Uncorrectable ECC Error" code (which is the same as the CRC error code). The 32-bit ECC used by IBM and many compatible disk manufacturers allows correction of single-bit errors only. Manufacturers of other, ____ usually more expensive, controllers use longer and more complex ECCs, to allow the correction of some multi-bit errors, but the necessity to minimise the time taken to calculate the ECC (which is part of "access time") sets tight limits on how complex it can be. ECC error correction is unfortunately not infallible. It is possible for a ___ complex, uncorrectable error to result in the same ECC disparity as a __ different, correctable, error, so that invalid "corrected" data can be ______ returned. The longer and more complex the ECC, the less likely this becomes, but there is no way of completely avoiding this ambiguity, short of keeping multiple copies of the data on the disk. In addition, some hard disk controllers* apparently abort with the "corrected ECC error" code immediately on encountering a sector containing such an ___________ error, even if it is not the last sector in the block specified by a ___ (multi-sector) read request. This non-intuitive (to say the least of it) behavior can combine with DOS 3.x's habit of ignoring corrected ________ ECC errors to result in apparently inexplicable disk-related problems, up to and including system crashes (if the file in which the problem occurred contained executable code). This hardware deficiency may be related to the problems which have plagued users of DOS 3.3 on some (previously) apparently IBM-compatible non-IBM hardware, since these also appear to result from deficiencies in the clones' handling of multi-sector disk reads. Prior to V4.36, HDTEST did retry corrected ___ ECC errors which were reported on the first read of a cluster, but would eventually (if the corrected error persisted) accept the data read as valid and copy and remap it to a "good" cluster. This could cause trouble if the controller had the "immediate abort on corrected ECC error" problem described above and the ECC error were not in the ___ last sector of the cluster, since the sectors after the "bad" one would not actually have been read, and would be lost (actually replaced by the corresponding sectors from the previous cluster read) ________ _____________________________________________________________________ * This behavior is known to occur with some early Western Digital controllers (1002A-WX1, 1002S-WX2) but it may also occur with others. Page 22 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ after remapping. On a machine with this problem, DOS would not, of course, have done any better than the earlier versions of HDTEST at reading the bad cluster, and HDTEST at least warned you that there was a problem, but the behavior described was certainly not ideal. From V4.36, HDTEST rereads clusters in which a persistent corrected ECC error is reported on the initial read sector by sector, to ensure that ________________ all the data in the cluster is transferred when it is remapped. In ___ most cases (as long as the ECC error was really correctable), this ______ will save the cluster (and the file of which it is a part) intact. DOS allows the programmer and/or computer user to think of and deal with disk data in terms of files, lines, and/or records, but data is always moved between a disk and memory a sector at a time. Any disk read or write request passed to DOS eventually gets translated into a command to read or write one or more sectors on the disk, commencing at a specified head, track, and sector. A floppy disk may have either one or two usable surfaces, corresponding to the top and bottom of the ________ disk. A hard disk usually has at least two surfaces, and may have more, since multiple physical "platters" are frequently incorporated in a single device. Each surface is normally accessed by and corresponds to a single read/write head, which moves radially across ____ it, although one surface on newer disks is often reserved for "servo" tracks, which are written with a special pattern that allows the controller to ensure that the read/write heads are kept exactly aligned with the data tracks. For a single device, the (data) heads are, by convention, numbered sequentially from zero up to one less than the total number present. Each surface is further divided into a number of circumferential tracks (or cylinders) and each track is ______ _________ divided into a number of sectors. By convention, the tracks on each _______ surface are also numbered sequentially from zero, with track zero being the outermost track. The sectors on each track are normally numbered from one to a number corresponding to the total number of ___ sectors on the track, but adjacent sectors do not normally have ___ consecutive sector numbers. A typical arrangement of the seventeen sectors on a single track of an XT hard disk, for example, would be as follows: 1,8,15,5,12,2,9,16,6,13,3,10,17,7,14,4,11. This "inter- ______ leaving" of sectors helps to make multi-sector disk reads and writes _______ more efficient. The controller requires a finite time to process the data from one sector before it can start reading (or writing) the next. During this time, of course, the disk continues to rotate and one or more of the sectors which physically "follow" the one that was last accessed pass under the read/write head unread. Ideally, the interleave (number of physical sectors separating consecutively numbered sectors) should be such that the next sector is just about to come under the read-write head when the controller is free to deal with it, but, because this time is not exactly predictable, it is better to be ultra-conservative (as IBM is) and make it one or two sectors longer than necessary than to be too optimistic. Too small an interleave value is very undesirable, because it results in the ____ controller having to wait almost a complete disk rotation time for each sector it misses to come round again. At the beginning of each disk sector is an address mark, a short, ____________ recognizable (to the controller) bit sequence containing the track and sector number. The process of writing the address marks for each sector on a disk is known as "low-level" or "sector" formatting. The DOS FORMAT program does this when it is run on floppy disks, but on Page 23 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ hard disks the DOS program merely makes sure that all required address marks are present - they are created by the disk manufacturer or supplier's low-level format program, which may or may not be available to the end user. After each address mark there follows a short gap (to allow the controller time to recognise and respond to the address mark), then the raw data, then the CRC or ECC data. A sector normally contains 512 bytes of raw data, but DOS (and HDTEST) can recognise and deal appropriately with devices with non-standard sector sizes. When a sector read/write command is passed to the disk controller, it (the controller) causes the disk's read/write head to move (seek) to ____ the specified track, waits a few milliseconds for it to settle down, and then starts to "pay attention to" the data passing underneath it. The controller locates and reads the address marks to confirm that the read/write head is positioned at the correct track and to identify each sector. If data is being read from the disk, the raw data and recorded CRC or ECC from the requested sector(s) are transferred into the disk controller's internal buffers and (assuming that no errors are detected), thence to main memory. If data is being written, the controller derives the error-checking data from the raw data and writes both to the disk. Some of this process is under the direct control of the controller hardware, but it is supervised by the device's ROM BIOS, which is responsible for error-reporting. Errors can occur at almost any stage of the read/write process, but (if the disk is working at all) usually result from media defects or problems developing in the mechanical components of the hard disk drive. Media defects, once present, are generally constant in size and position (although jarring a disk while it is being written to may sometimes generate a temporary bad sector). _________ When HDTEST detects an error during read/write testing of a sector or cluster which has not previously been marked as bad, a (normally) four-line error message is printed, either to the screen (by default), to the printer (/P option), or to the logging file (if an output device was specified). The first line of the message simply indicates the sector or cluster number at which the error occurred. The second line indicates what sort of error was detected and during what phase of read/write testing it occurred. The third line indicates whether HDTEST was able to recover and/or remap the data from the bad sector or cluster, and the fourth indicates either the file which a bad cluster formed part of or that the cluster was not in use. If it was not possible to read one or more sectors from a cluster which contained data (unrecoverable read error), a further two-line message is printed which notes how many sectors were unreadable and how the _______ data from the bad cluster has been remapped (which depends on whether a /Zn option was specified on the command line). When the program detects an error during read/write testing of a sector in the DOS area, only the first three lines described above are printed. The remainder of this Appendix contains more details of and further information on the possible contents of lines two and three of HDTEST's disk error message. Page 24 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ The second line of the disk read/write error message will usually take the form: "[Error Type] occurred during [phase of testing]". Possible values for [Error Type] are as follows (in alphabetical order): Address Mark Not Found - this is a fairly common error reported for bad blocks which are present from "scratch" on a new disk, since a significant media defect may well not retain sector formatting data and/or the manufacturer may have made sure that a known bad track on the disk will not be used by deliberately not writing address marks there. If new bad clusters are reported with this error message, it is likely that the hard disk surface has been physically damaged (perhaps by a shock to the drive) but possible that the controller has a problem. Bad Controller - this means what it says, but check that all boards and cables are properly seated before giving up. Bad Sector - the BIOS was instructed to read a sector number which (it believes) does not exist on the device. This indicates either an internal program error or that the program is having a problem determining the physical layout of the disk. Check the values reported by HDCHEK against those in the disk documentation. Bad Track - analogous to "Bad Sector". Buffer Comparison - the data read from a sector differs from that written to it. This should not happen, since it suggests either that the device's error-detection functionality has failed or that it has been overwhelmed. If it does occur, the most likely causes are controller and memory problems (in that order of likelihood). Corrected ECC - as discussed at some length above, the data retrieved from a hard disk was incorrect but the device's error-correction hardware was (probably) able to correct it. DMA Boundary Violation - should not occur (HDTEST positions its buffers to avoid crossing absolute segment boundaries). Suggests either an internal program error or controller/BIOS problem. Hardware Timeout - a seek or other operation was not completed within the maximum time allowed by the BIOS. Check connections, reseat boards, otherwise suspect the mechanics of the disk. Init Fail - should not occur (on a working disk). Internal - internal program error. Make sure that your working copy of HDTEST has not been corrupted. The disk is probably not defective. Invalid Command - should not occur. Treat as Internal Error. Multiple Errors - one or more additional errors occurred while HDTEST was attempting to recover from an initial one. Multiple phase report lines will be printed but the error types cannot be further specified. Recovered - HDTEST recovered from the error and did not report on the error type (should not occur). Page 25 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ Reset Fail - The device could not be reset (after an error). Usually indicates major problems with the disk or controller, but check connections, etc. Sector Not Found - the requested sector's address mark was not found. Likely to result from a media defect. Seek Fail - after a seek to a track, the controller found that the read/write head was not positioned correctly. Usually indicates a significant problem with the (drive) hardware. Sense Fail - the controller cannot find the drive! Check connections, etc. Uncorrected ECC - data error too major to be corrected. Usually a media defect. This message will also be produced if a CRC error is detected on a floppy disk. Undefined - the BIOS cannot figure out what went wrong, but something did! Unknown - likewise! Write Protection - HDTEST is attempting to write to a write-protected device. Possible forms for the rest of the second line are as follows: ... Error (unrecovered) occurred during first read - the cluster could not be read without error in five tries before a seek to the other "end" of the disk. ... Error occurred during first read - the error occurred during the first read but was recovered after one or more retries. ... Error occurred during read after seek - the error occurred when the cluster was read after a seek away. ... Error occurred during pattern read/writes - the error occurred during the pattern read/write phase of testing. ... Error occurred during final write - the error occurred when the saved copy of the cluster was being restored to the disk. ... Error occurred during read after final write - the error occurred when the restored cluster was being read back. ... (Buffer comparison) error occurred after final read - the copy of the cluster read back after restoration to the disk differed from its original memory image. Page 26 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ There are only three basic forms of the third line: This error is (or "These errors are") recoverable (no data loss has occurred). - the bad cluster was read successfully and has been successfully restored to its original form, but it remains suspect and will (with your permission) be remapped. This error is unrecoverable, but remapping is possible. - the bad cluster was read successfully but has not been restored to its ___ original form. Remapping should be successful, but the data will be lost if this is not permitted. This error is unrecoverable, and data may be lost! - the bad cluster was not read successfully and will be remapped as provided by the value of the /Z switch parameter if remapping is permitted. In any event, some data loss has almost certainly occurred. Page 27 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ APPENDIX C - EXAMPLE PROGRAM RUNS _________________________________ Screen display from successful run of HDCHEK on partition occupying the whole of a 20 Mb hard disk: Device C: appears to be the DOS partition of a hard disk It contains 41531 sectors It starts at absolute sector 17 on the disk {THIS IS CORRECT} _______________ The total partition size is approximately 20 Mb There are 4 heads and 17 sectors/track Sector size is 512 bytes Each of the 2 (16 bit) FATs occupies 41 sectors The Root Directory occupies 32 sectors The data area starts at DOS sector 115, and contains 10354 clusters, each consisting of 4 sectors The directory scan was successful; a total of 1260 files and subdirectories was found. Screen display from successful run of HDCHEK on a freshly FORMATted DSDD 9-sector floppy disk: Device A: is not a hard disk - using DOS interrupts Device A: is a floppy disk The disk contains 720 sectors There are 2 heads and 9 sectors/track Sector size is 512 bytes Each of the 2 (12 bit) FATs occupies 2 sectors The Root Directory occupies 7 sectors The data area starts at DOS sector 12, and contains 354 clusters, each consisting of 2 sectors The directory scan was successful; a total of 0 files and subdirectories was found. Screen display from run of HDCHEK on 20 Mb DOS partition - directory structure bad or unrecognizable: Device C: appears to be the DOS partition of a hard disk It contains 41531 sectors It starts at absolute sector 17 on the disk The total partition size is approximately 20 Mb There are 4 heads and 17 sectors/track Sector size is 512 bytes Each of the 2 (16 bit) FATs occupies 41 sectors The Root Directory occupies 32 sectors The data area starts at DOS sector 115, and contains 10354 clusters, each consisting of 4 sectors The preliminary directory scan failed. Page 28 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ Screen display from run of HDCHEK on 20 Mb DOS partition - FATs not identical: FATS are different FAT discrepancy at byte 20F - FAT 1 has FF; FAT 2 has F8 Continue comparison (Y/N)? [Y]: n {user input} _ Device C: appears to be the DOS partition of a hard disk It contains 41531 sectors It starts at absolute sector 17 on the disk The total partition size is approximately 20 Mb There are 4 heads and 17 sectors/track Sector size is 512 bytes Each of the 2 (16 bit) FATs occupies 41 sectors The Root Directory occupies 32 sectors The data area starts at DOS sector 115, and contains 10354 clusters, each consisting of 4 sectors Screen display from run of HDCHEK - program unable to deduce disk structure - disk not recognizable as a DOS device: Unrecognised media descriptor byte! - see error code Error Code = F1 {F1 is not a media descriptor byte used by DOS} Page 29 Documentation for HDTEST V4.45 Copyright 1988, P. R. Fletcher _______________________________ _______________________________ Log file from successful run of HDTEST on 20 Mb DOS partition: HDTEST Log File - Wed Jan 07 11:44:52 1986 Testing Device C: No new bad clusters identified - HDTEST finished. Wed Jan 07 16:10:41 1987 Log file from run of HDTEST on 20 Mb DOS partition - one bad cluster detected and remapped: HDTEST Log File - Mon Jan 05 21:09:32 1987 Testing Device C: Read/Write error at cluster number 7623!! Corrected ECC error occurred during pattern read/writes This error is recoverable (no data loss has occurred). Bad cluster is in file \MYFILES\DATA5\FFT1234.DAT 1 new bad clusters identified, marked, and remapped - HDTEST finished. Tue Jan 06 05:08:04 1987 Page 30 REGISTRATION FORM - HDTEST V4.45 ________________________________ Please complete this form and return it with a check or money order for $35.00 (+ $5.00 for shipping outside North America) or equivalent to: Peter R. Fletcher 1515 West Montgomery Avenue Rosemont PA 19010 U.S.A. European users should send Registrations to the address on the program's copyright screen. None of the information on this form will be passed on to any other individual or commercial organization. -------------------------------------------------------------------------- This information is required to process your Registration: Last Name:____________________________ Other Names:______________________ Address:______________________________ Town/City:________________________ State:_____________ Post Code:____________ Telephone:_________________ (evenings/weekends) I understand that this software is supplied without warranty, express or implied, and agree not to hold its author liable for any direct or consequential loss arising out of my use of it. Signature:______________________________ ------------------------------------------------------------------------- You are requested, but not required, also to answer the following questions: On What type(s) of computer are you using HDTEST? _________________________ CPU (8088, 8086, etc.)? __________ Amount of RAM (kB)? ____________ Type & Size of Hard Disk(s)? ____________________________________ Make and Type of Hard Disk Controller(s)? ______________________________ Do you own a Modem? Y N Are you a member of BIX? Y N Compuserve? Y N How did you obtain/hear about this program? Did you think that the $35.00 registration fee for this program was: Very Good Value? About Right? Excessively High? Are there any other utility programs that you would like to see made available as "User Supported Software"? What sort of programs?