Table of Contents
6.1 Introduction
6.2 First-level System Regeneration
6.3 Second-level System Regeneration
6.4 Sample GETSYS and PUTSYS Program
6.5 Disk Organization
6.6 The BIOS Entry Points
6.7 A Sample BIOS
6.8 A Sample Cold Start Loader
6.9 Reserved Locations in Page Zero
6.10 Disk Parameter Tables
6.11 The DISKDEF Macro Library
6.12 Sector Blocking and Deblocking
Tables
6-1 Standard Memory Size Values
6-2 Common Values for CP/M Systems
6-3 CP/M Disk Sector Allocation
6-4 IOBYTE Field Values
6-5 BIOS Entry Points
6-6 Reserved Locations in Page Zero
6-7 Disk Parameter Headers
6-8 BSH and BLM Values
6-9 EXM Values
6-10 BLS Tabulation
Figures
6-1 IOBYTE Fields
6-2 Disk Parameter Header Format
6-3 Disk Parameter Header Table
6-4 Disk Parameter Block Format
6-5 AL0 and AL1
Listings
6-1 GETSYS Program
6-2 BIOS Entry Points
The standard CP/M system assumes operation on an Intel MDS-800 microcomputer development system, but is designed so you can alter a specific set of subroutines that define the hardware operating environment.
Although standard CP/M 2 is configured for single-density floppy disks, field alteration features allow adaptation to a wide variety of disk subsystems from single drive minidisks to high- capacity, hard disk systems. To simplify the following adaptation process, it is assumed that CP/M 2 is first configured for single-density floppy disks where minimal editing and debugging tools are available. If an earlier version of CP/M is available, the customizing process is eased considerably. In this latter case, you might want to review the system generation process and skip to later sections that discuss system alteration for nonstandard disk systems.
To achieve device independence, CP/M is separated into three distinct modules:
Of these modules, only the BIOS is dependent upon the particular hardware. You can patch the distribution version of CP/M to provide a new BIOS that provides a customized interface between the remaining CP/M modules and the hardware system. This document provides a step-by-step procedure for patching a new BIOS into CP/M.
All disk-dependent portions of CP/M 2 are placed into a BIOS, a resident disk parameter block, which is either hand coded or produced automatically using the disk definition macro library provided with CP/M 2. The end user need only specify the maximum number of active disks, the starting and ending sector numbers, the data allocation size, the maximum extent of the logical disk, directory size information, and reserved track values. The macros use this information to generate the appropriate tables and table references for use during CP/M 2 operation. Deblocking information is provided, which aids in assembly or disassembly of sector sizes that are multiples of the fundamental 128-byte data unit, and the system alteration manual includes general purpose subroutines that use the deblocking information to take advantage of larger sector sizes. Use of these subroutines, together with the table-drive data access algorithms, makes CP/M 2 a universal data management system.
File expansion is achieved by providing up to 512 logical file extents, where each logical extent contains 16K bytes of data. CP/M 2 is structured, however, so that as much as 128K bytes of data are addressed by a single physical extent, corresponding to a single directory entry, iuaintaining compatibility with previous versions while taking advantage of directory space. If CP/M is being tailored to a computer system for the first time, the new BIOS requires some simple software development and testing. The standard BIOS is listed in Appendix A and can be used as a model for the customized package. A skeletal version of the BIOS given in Appendix B can serve as the basis for a modified BIOS.
In addition to the BIOS, you must write a simple memory loader, called GETSYS, which brings the operating system into memory. To patch the new BIOS into CP/M, you must write the reverse of GETSYS, called PUTSYS, which places an altered version of CP/M back onto the disk. PUTSYS can be derived from GETSYS by changing the disk read commands into disk write commands. Sample skeletal GETSYS and PUTSYS programs are described in Section 6.4 and listed in Appendix C.
To make the CP/M system load automatically, you must also supply a cold start loader, similar to the one provided with CP/M, listed in Appendices A and D. A skeletal form of a cold start loader is given in Appendix E, which servies as a model for the loader.
6.2 First-Level System Regeneration
The procedure to patch the CP/M system is given below. Address references in each step are shown with H denoting the hexadecimal radix, and are given for a 20K CP/M system. For larger CP/M systems, a bias is added to each address that is shown with a +b following it, where b is equal to the memory size-20K. Values for b in various standard memory sizes are listed in Table 6-1.
Note that the standard distribution version of CP/M is set for operation within a 20K CP/M system. Therefore, you must first bring up the 20K CP/M system, then configure it for actual memory size (see Section 6.3).
Follow these steps to patch your CP/M system:
If difficulties are encountered, use whatever debug facilities are available to trace and breakpoint the CBIOS.
SAVE 1 X.COM
All commands must be followed by a carriage return. CP/M responds with another prompt after several disk accesses:
A>
If it does not, debug the disk write functions and retry.
DIR
CP/M responds with
A:X COM
ERA X.COM
CP/M responds with the A prompt. This is now an operational system that only requires a bootstrap loader to function completely.
DIR
CP/M responds with a list of files that are provided on the initialized disk. The file DDT.COM is the memory image for the debugger. Note that from now on, you must always reboot the CP/M system (CTRL-C is sufficient) when the disk is removed and replaced by another disk, unless the new disk is to be Read-Only.
DDT
See Section 4 for operating procedures.
Copyright (c), 1983
Digital Research
on each copy that is made with the COPY program.
You should now have a good copy of the customized CP/M system. Although the CBIOS portion of CP/M belongs to the user, the modified version cannot be legally copied.
It should be noted that the system remains file-compatible with all other CP/M systems (assuming media compatibility) which allows transfer of nonproprietary software between CP/M users.
6.3 Second-Level System Generation
Once the system is running, the next step is to configure CP/M for the desired memory size. Usually, a memory image is first produced with the MOVCPM program (system relocator) and then placed into a named disk file. The disk file can then be loaded, examined, patched, and replaced using the debugger and the system generation program (refer to Section 1).
The CBIOS and BOOT are modified using ED and assembled using ASM, producing files called CBIOS.HEX and BOOT.HEX, which contain the code for CBIOS and BOOT in Intel hex format.
To get the memory image of CP/M into the TPA configured for the desired memorv size, type the command:
MOVCPM xx*
where xx is the memory size in decimal K bytes, for example, 32 for 32K. The response is as follows:
CONSTRUCTING xxK CP/M VERS 2.0
READY FOR "SYSGEN" OR
"SAVE 34 CPMxx.COM"
An image of CP/M in the TPA is configured for the requested memory size. The memory image is at location 0900H through 227FH, that is, the BOOT is at 0900H, the CCP is at 980H, the BDOS starts at 1180H, and the BIOS is at 1F80H. Note that the memory image has the standard MDS-800 BIOS and BOOT on it. It is now necessary to save the memory image in a file so that you can patch the CBIOS and CBOOT into it:
SAVE 34 CPMxx.COM
The memory image created by the MOVCPM program is offset by a negative bias so that it loads into the free area of the TPA, and does not interfere with the operation of CP/M in higher memory. This memory image can be subsequently loaded under DDT and examined or changed in preparation for a new generation of the system. DDT is loaded with the memory image by typing:
DDT CPMxx.COM Loads DDT, then reads the CP/M image.
DDT should respond with the following:
NEXT PC
2300 0100
- (The DDT prompt)
You can then give the display and disassembly commands to examine portions of the memory image between 900H and 227FH. Note, however, that to find any particular address within the memory image, you must apply the negative bias to the CP/M address to find the actual address. Track 00, sector 01, is loaded to location 900H (the user should find the cold start loader at 900H to 97FH); track 00, sector 02, is loaded into 980H (this is the base of the CCP); and so on through the entire CP/M system load. In a 20K system, for example, the CCP resides at the CP/M address 3400H, but is placed into memory at 980H by the SYSGEN program. Thus, the negative bias, denoted by n, satisfies
3400H + n = 980H, or n = 980H - 3400H
Assuming two's complement arithmetic, n = D580H, which can be checked by
3400H+D580H = 10980H = 0980H (ignoring high-order overflow).
Note that for larger systems, n satisfies
(3400H + b) + n = 980H, or
n = 980H - (3400H + b), or
n = D580H - b
The value of n for common CP/M systems is given below.
If you want to locate the address x within the memory image loaded under DDT in a 20K system, first type
Hx,n Hexadecimal sum and difference
and DDT responds with the value of x + n (sum) and x - n (difference). The first number printed by DDT is the actual memory address in the image where the data or code is located. For example, the following DDT command:
H3400,D580
produces 980H as the sum, which is where the CCP is located in the memory image under DDT.
Type the L command to disassemble portions of the BIOS located at (4A00H + b) - n, which, when one uses the H command, produces an actual address of 1F80H. The disassembly command would thus be as follows:
L1F80
It is now necessary to patch in the CBOOT and CBIOS routines. The BOOT resides at location 0900H in the memory image. If the actual load address is n, then to calculate the bias (m), type the command:
H900,n Subtract load address from target address.
The second number typed by DDT in response to the command is the desired bias (in). For example, if the BOOT executes at 0080H, the command
H900,80
produces
0980 0880 Sum and difference in hex.
Therefore, the bias in would be 0880H. To read-in the BOOT, give the command:
ICBOOT.HEX Input file CBOOT.HEX
Then
Rm Read CBOOT with a bias of in (= 900H - n).
Examine the CBOOT with
L900
You are now ready to replace the CBIOS by examining the area at 1F80H, where the original version of the CBIOS resides, and then typing
ICBIOS.HEX Ready the hex file for loading.
Assume that the CBIOS is being integrated into a 20K CP/M system and thus originates at location 4A00H. To locate the CBIOS properly in the memory image under DDT, you must apply the negative bias n for a 20K system when loading the hex file. This is accomplished by typing
RD580 Read the file with bias D580H.
Upon completion of the read, reexamine the area where the CBIOS has been loaded (use an L1F80 command) to ensure that it is properly loaded. When you are satisfied that the change has been made, return from DDT using a CTRL-C or G0 command.
SYSGEN is used to replace the patched memory image back onto a disk (you use a test disk until sure of the patch) as shown in the following interaction:
SYSGEN | Start the SYSGEN program. |
SYSGEN VERSION 2.0 | Sign-on message from SYSGEN. |
SOURCE DRIVE NAME (OR RETURN TO SKIP) | Respond with a carriage return to skip the CP/M read operation because the system is already in memory. |
DESTINATION DRIVE NAME (OR RETURN TO REBOOT) | Respond with B to write the new system to the disk in drive B. |
DESTINATION ON B THEN TYPE RETURN | Place a scratch disk in drive B, then press RETURN. |
FUNCTION COMPLETE DESTINATION DRIVE NAME (OR RETURN TO REBOOT) |
Place the scratch disk in drive A, then perform a cold start to bring up the newly configured CP/M system.
The new CP/M system is then tested and the Digital Research copyright notice is placed on the disk, as specified in the Licensing Agreement:
Copyright (c), 1979
Digital Research
6.4 Sample GETSYS and PUTSYS Programs
The following program provides a framework for the GETSYS and
PUTSYS programs referenced in Sections 6.1
and 6.2. To read and
write the specific sectors, you must insert the READSEC and
WRITESEC subroutines.
Listing 6-1. GETSYS Program
; GETSYS PROGRAM -- READ TRACKS 0 AND 1 TO MEMORY AT 3380H
; REGISTER USE
; A (SCRATCH REGISTER)
; B TRACK COUNT (O, 1)
; C SECTOR COUNT (1,2,...,26)
; DE (SCRATCH REGISTER PAIR)
; HL LOAD ADDRESS
; SP SET TO STACK ADDRESS
START: LXI SP,3380H ; SET STACK POINTER TO SCRATCH
; AREA
LXI H,3380H ; SET BASE LOAD ADDRESS
MVI B,0 ; START WITH TRACK 0
RDTRK: ; READ NEXT TRACK (INITIALLY 0)
MVI C,1 ; READ STARTING WITH SECTOR 1
DRSEC: ; READ NEXT SECTOR
CALL RDSEC ; USER SUPPLIED SUBROUTINE
LXI D,128 ; MOVE LOAD ADDRESS TO NEXT 1/2
; PAGE
DAD D ; HL = HL + 128
INR C ; SECTOR = SECTOR + 1
MOV A,C ; CHECK FOR END OF TRACK
CPI 27
JC RDSEC ; CARRY GENERATED IF SECTOR <27
;
;
; ARRIVE HERE AT END OF TRACK, MOVE TO NEXT TRACK
INR B
MOV A,B ; TEST FOR LAST TRACK
CPI 2
JC RDTRK ; CARRY GENERATED IF TRACK <2
;
;
; USER SUPPLIED SUBROUTINE TO READ THE DISK
READSEC:
; ENTER WITH TRACK NUMBER IN REGISTER B,
; SECTOR NUMBER IN REGISTER C,
; AND ADDRESS TO FILL IN HL
;
PUSH B ; SAVE B AND C REGISTERS
PUSH H ; SAVE HL REGISTERS
;
***********************************************
PERFORM DISK READ AT THIS POINT, BRANCH TO
LABEL "START" IF AN ERROR OCCURS
***********************************************
;
POP H ; RECOVER HL
POP B ; RECOVER B AND C REGISTERS
RET ; BACK TO MAIN PROGRAM
;
END START
This program is assembled and listed in Appendix B for reference purposes, with an assumed origin of 100H. The hexadecimal operation codes that are listed on the left might be useful if the program has to be entered through the panel switches.
The PUTSYS program can be constructed from GETSYS by changing only a few operations in the GETSYS program given above, as shown in Appendix C. The register pair HL becomes the dump address, next address to write, and operations on these registers do not change within the program. The READSEC subroutine is replaced by a WRITESEC subroutine, which performs the opposite function; data from address HL is written to the track given by register B and sector given by register C. It is often useful to combine GETSYS and PUTSYS into a single program during the test and development phase, as shown in Appendix C.
The sector allocation for the standard distribution version of CP/M is given here for reference purposes. The first sector contains an optional software boot section (see the table on the following page). Disk controllers are often set up to bring track 0, sector 1, into memory at a specific location, often location 0000H. The program in this sector, called BOOT, has the responsibility of bringing the remaining sectors into memory starting at location 3400H + b. If the controller does not have a built-in sector load, the program in track 0, sector 1 can be ignored. In this case, load the program from track 0, sector 2, to location 3400H + b.
As an example, the Intel MDS-800 hardware cold start loader brings track 0, sector 1, into absolute address 3000H. Upon loading this sector, control transfers to location 3000H, where the bootstrap operation commences by loading the remainder of track 0 and all of track 1 into memory, starting at 3400H + b. Note that this bootstrap loader is of little use in a non-MDS environment, although it is useful to examine it because some of the boot actions will have to be duplicated in the user's cold start loader.
The entry points into the BIOS from the cold start loader and BDOS are detailed below. Entry to the BIOS is through a jump vector located at 4A00H + b, as shown below. See Appendices A and B. The jump vector is a sequence of 17 jump instructions that send program control to the individual BIOS subroutines. The BIOS subroutines might be empty for certain functions (they might contain a single RET operation) during reconfiguration of CP/M, but the entries must be present in the jump vector.
The jump vector at 4A00H + b takes the form shown below, where
the individual jump addresses are given to the left:
Listing 6-2. BIOS Entry Points
4A00H+b JMP BOOT ;ARRIVE HERE FROM COLD START LOAD
4A03H+b JMP WBOOT ;ARRIVE HERE FOR WARM START
4A06H+b JMP CONST ;CHECK FOR CONSOLE CHAR READY
4A09H+b JMP CONIN ;READ CONSOLE CHARACTER IN
4A0CH+b JMP CONOUT ;WRITE CONSOLE CHARACTER OUT
4A0FH+b JMP LIST ;WRITE LISTING CHARACTER OUT
4A12H+b JMP PUNCH ;WRITE CHARACTER TO PUNCH DEVICE
4A15H+b JMP READER ;READ READER DEVICE
4A18H+b JMP HOME ;MOVE TO TRACK 00 ON SELECTED DISK
4A1BH+b JMP SELDSK ;SELECT DISK DRIVE
4A1EH+b JMP SETTRK ;SET TRACK NUMBER
4A21H+b JMP SETSEC ;SET SECTOR NUMBER
4A24H+b JMP SETDMA ;SET DMA ADDRESS
4A27H+b JMP READ ;READ SELECTED SECTOR
4A2AH+b JMP WRITE ;WRITE SELECTED SECTOR
4A2DH+b JMP LISTST ;RETURN LIST STATUS
4A30H+b JMP SECTRAN ;SECTOR TRANSLATE SUBROUTINE
Each jump address corresponds to a particular subroutine that performs the specific function, as outlined below. There are three major divisions in the jump table: the system reinitialization, which results from calls on BOOT and WBOOT; simple character I/O, performed by calls on CONST, CONIN, CONOUT, LIST, PUNCH, READER, and LISTST; and disk I/O, performed by calls on HOME, SELDSK, SETTRK, SETSEC, SETDMA, READ, WRITE, and SECTRAN.
All simple character I/O operations are assumed to be performed in ASCII, upper- and lower-case, with high-order (parity bit) set to zero. An end-of-file condition for an input device is given by an ASCII CTRL-Z (1AH). Peripheral devices are seen by CP/M as logical devices and are assigned to physical devices within the BIOS.
To operate, the BDOS needs only the CONST, CONIN, and CONOUT subroutines. LIST, PUNCH, and READER can be used by PIP, but not the BDOS. Further, the LISTST entry is currently used only by DESPOOL, the print spooling utility. Thus, the initial version of CBIOS can have empty subroutines for the remaining ASCII devices.
The following list describes the characteristics of each device.
A single peripheral can be assigned as the LIST, PUNCH, and READER device simultaneously. If no peripheral device is assigned as the LIST, PUNCH, or READER device, the CBIOS gives an appropriate error message so that the system does not hang if the device is accessed by PIP or some other user program. Alternately, the PUNCH and LIST routines can just simply return, and the READER routine can return with a 1AH (CTRL-Z) in register A to indicate immediate end-of-file.
For added flexibility, you can optionally implement the IOBYTE function, which allows reassignment of physical devices. The IOBYTE function creates a mapping of logical-to-physical devices that can be altered during CP/M processing, see the STAT command in Section 1.6.1.
The definition of the IOBYTE function corresponds to the Intel standard as follows: a single location in memory, currently location 0003H, is maintained, called IOBYTE, which defines the logical-to-physical device mapping that is in effect at a particular time. The mapping is performed by splitting the IOBYTE into four distinct fields of two bits each, called the CONSOLE, READER, PUNCH, and LIST fields, as shown in the following figure.
The value in each field can be in the range 0-3, defining the assigned source or destination of each logical device. Table 6-4 gives the values that can be assigned to each field.
The implementation of the IOBYTE is optional and effects only the organization of the CBIOS. No CP/M system programs use the IOBYTE (although they tolerate the existence of the IOBYTE at location 0003H) except for PIP, which allows access to the physical devices, and STAT, which allows logical-physical assignments to be made or displayed. For more information see Section 1. In any case the IOBYTE implementation should be omitted until the basic CBIOS is fully implemented and tested; then you should add the IOBYTE to increase the facilities.
Disk I/O is always performed through a sequence of calls on the various disk access subroutines that set up the disk number to access, the track and sector on a particular disk, and the Direct Memory Access (DMA) address involved in the I/O operation. After all these parameters have been set up, a call is made to the READ or WRITE function to perform the actual I/O operation.
There is often a single call to SELDSK to select a disk drive, followed by a number of read or write operations to the selected disk before selecting another drive for subsequent operations. Similarly, there might be a single call to set the DMA address, followed by several calls that read or write from the selected DMA address before the DMA address is changed. The track and sector subroutines are always called before the READ or WRITE operations are performed.
The READ and WRITE routines should perform several retries (10 is standard) before reporting the error condition to the BDOS. If the error condition is returned to the BDOS, it reports the error to the user. The HOME subroutine might or might not actually perform the track 00 seek, depending upon controller characteristics; the important point is that track 00 has been selected for the next operation and is often treated in exactly the same manner as SETTRK with a parameter of 00.
The following table describes the exact responsibilities of each BIOS entry point subroutine.
Refer to
Section 6.9 for complete details of page
zero use. Upon completion of the initialization,
the WBOOT program must branch to the CCP at 3400H
+ b to restart the system. Upon entry to the CCP,
register C is set to the drive to select after
system initialization. The WBOOT routine should
read location 4 in memory, verify that is a legal
drive, and pass it to the CCP in register C.
If there is an attempt to select a nonexistent
drive, SELDSK returns HL = 0000H as an error
indicator. Although SELDSK must return the header
address on each call, it is advisable to postpone
the physical disk select operation until an I/O
function (seek, read, or write) is actually
performed, because disk selects often occur
without ultimately performing any disk I/O, , and
many controllers unload the head of the current
disk before selecting the new drive. This causes
an excessive amount of noise and disk wear. The
least significant bit of register E is zero if
this is the first occurrence of the drive select
since the last cold or warm start.
0 - no errors occurred
1 - nonrecoverable error condition occurred
Currently, CP/M responds only to a zero or nonzero
value as the return code. That is, if the value in
register A is 0, CP/M assumes that the disk
operation was completed properly. If an error
occurs the CBIOS should attempt at least 10
retries to see if the error is recoverable. When
an error is reported the BDOS prints the message
BDOS ERR ON x: BAD SECTOR. The operator then has
the option of pressing a carriage return to ignore
the error, or CTRL-C to abort.
In general, SECTRAN receives a logical sector
number relative to zero in BC and a translate
table address in DE. The sector number is used as
an index into the translate table, with the
resulting physical sector number in HL. For
standard systems, the table and indexing code is
provided in the CBIOS and need not be changed.
Entry Point Function
BOOT The BOOT entry point gets control from the cold
start loader and is responsible for basic system
initialization, including sending a sign-on
message, which can be omitted in the first
version. If the IOBYTE function is implemented, it
must be set at this point. The various system
parameters that are set by the WBOOT entry point
must be initialized, and control is transferred to
the CCP at 3400 + b for further processing. Note
that register C must be set to zero to select
drive A.
WBOOT The WBOOT entry point gets control when a warm
start occurs. A warm start is performed whenever a
user program branches to location 0000H, or when
the CPU is reset from the front panel. The CP/M
system must be loaded from the first two tracks of
drive A up to, but not including, the BIOS, or
CBIOS, if the user has completed the patch. System
parameters must be initialized as follows:
CONST You should sample the status of the currently
assigned console device and return 0FFH in
register A if a character is ready to read and 00H
in register A if no console characters are ready.
CONIN The next console character is read into register
A, and the parity bit is set, high-order bit, to
zero. If no console character is ready, wait
until a character is typed before returning.
CONOUT The character is sent from register C to the
console output device. The character is in ASCII,
with high-order parity bit set to zero. You might
want to include a time-out on a line-feed or
carriage return, if the console device requires
some time interval at the end of the line (such as
a TI Silent 700 terminal). You can filter out
control characters that cause the console device
to react in a strange way (CTRL-Z causes the Lear-
Siegler terminal to clear the screen, for
example).
LIST The character is sent from register C to the
currently assigned listing device. The character
is in ASCII with zero parity bit.
PUNCH The character is sent from register C to the
currently assigned punch device. The character is
in ASCII with zero parity.
READER The next character is read from the currently
assigned reader device into register A with zero
parity (high-order bit must be zero); an end-of-
file condition is reported by returning an ASCII
CTRL-Z(1AH).
HOME The disk head of the currently selected disk
(initially disk A) is moved to the track 00
position. If the controller allows access to the
track 0 flag from the drive, the head is stepped
until the track 0 flag is detected. If the
controller does not support this feature, the HOME
call is translated into a call to SETTRK with a
parameter of 0.
SELDSK The disk drive given by register C is selected for
further operations, where register C contains 0
for drive A, 1 for drive B, and so on up to 15 for
drive P (the standard CP/M distribution version
supports four drives). On each disk select, SELDSK
must return in HL the base address of a 16-byte
area, called the Disk Parameter Header, described
in Section 6.10. For standard floppy disk drives,
the contents of the header and associated tables
do not change; thus, the program segment included
in the sample CBIOS performs this operation
automatically.
SETTRK Register BC contains the track number for
subsequent disk accesses on the currently selected
drive. The sector number in BC is the same as the
number returned from the SECTRAN entry point. You
can choose to seek the selected track at this time
or delay the seek until the next read or write
actually occurs. Register BC can take on values
in the range 0-76 corresponding to valid track
numbers for standard floppy disk drives and 0-
65535 for nonstandard disk subsystems.
SETSEC Register BC contains the sector number, 1 through
26, for subsequent disk accesses on the currently
selected drive. The sector number in BC is the
same as the number returned from the SECTRAN entry
point. You can choose to send this information to
the controller at this point or delay sector
selection until a read or write operation occurs.
SETDMA Register BC contains the DMA (Disk Memory Access)
address for subsequent read or write operations.
For example, if B = 00H and C = 80H when SETDMA is
called, all subsequent read operations read their
data into 80H through 0FFH and all subsequent
write operations get their data from 80H through
0FFH, until the next call to SETDMA occurs. The
initial DMA address is assumed to be 80H. The
controller need not actually support Direct Memory
Access. If, for example, all data transfers are
through I/O ports, the CBIOS that is constructed
uses the 128 byte area starting at the selected DMA
address for the memory buffer during the
subsequent read or write operations.
READ Assuming the drive has been selected, the track
has been set, and the DMA address has been
specified, the READ subroutine attempts to read
eone sector based upon these parameters and
returns the following error codes in register A:
WRITE Data is written from the currently selected DMA
address to the currently selected drive, track,
and sector. For floppy disks, the data should be
marked as nondeleted data to maintain
compatibility with other CP/M systems. The error
codes given in the READ command are returned in
register A, with error recovery attempts as
described above.
LISTST You return the ready status of the list device
used by the DESPOOL program to improve console
response during its operation. The value 00 is
returned in A if the list device is not ready to
accept a character and 0FFH if a character can be
sent to the printer. A 00 value should be
returned if LIST status is not implemented.
SECTRAN Logical-to-physical sector translation is
performed to improve the overall response of CP/M.
Standard CP/M systems are shipped with a skew
factor of 6, where six physical sectors are
skipped between each logical read operation. This
skew factor allows enough time between sectors for
most programs to load their buffers without
missing the next sector. In particular computer
systems that use fast processors, memory, and disk
subsystems, the skew factor might be changed to
improve overall response. However, the user should
maintain a single-density IBM-compatible version
of CP/M for information transfer into and out of
the computer system, using a skew factor of 6.
The program shown in Appendix B can serve as a basis for your first BIOS. The simplest functions are assumed in this BIOS, so that you can enter it through a front panel, if absolutely necessary. You must alter and insert code into the subroutines for CONST, CONIN, CONOUT, READ, WRITE, and WAITIO subroutines. Storage is reserved for user-supplied code in these regions. The scratch area reserved in page zero (see Section 6.9) for the BIOS is used in this program, so that it could be implemented in ROM, if desired.
Once operational, this skeletal version can be enhanced to print the initial sign-on message and perform better error recovery. The subroutines for LIST, PUNCH, and READER can be filled out and the IOBYTE function can be implemented.
6.8 A Sample Cold Start Loader
The program shown in Appendix E can serve as a basis for a cold start loader. The disk read function must be supplied by the user, and the program must be loaded somehow starting at location 0000. Space is reserved for the patch code so that the total amount of storage required for the cold start loader is 128 bytes.
Eventually, you might want to get this loader onto the first disk sector (track 0, sector 1) and cause the controller to load it into memory automatically upon system start up. Alternatively, the cold start loader can be placed into ROM, and above the CP/M system. In this case, it is necessary to originate the program at a higher address and key in a jump instruction at system start up that branches to the loader. Subsequent warm starts do not require this key-in operation, because the entry point WBOOT gets control, thus bringing the system in from disk automatically. The skeletal cold start loader has minimal error recovery, which might be enhanced in later versions.
6.9 Reserved Locations in Page Zero
Main memory page zero, between locations 0H and 0FFH, contains several segments of code and data that are used during CP/M processing. The code and data areas are given in the following table.
This information is set up for normal operation under the CP/M system, but can be overwritten by a transient program if the BDOS facilities are not required by the transient.
If, for example, a particular program performs only simple I/O and must begin execution at location 0, it can first be loaded into the TPA, using normal CP/M facilities, with a small memory move program that gets control when loaded. The memory move program must get control from location 0100H, which is the assumed beginning of all transient programs. The move program can then proceed to the entire memory image down to location 0 and pass control to the starting address of the memory load.
If the BIOS is overwritten or if location 0, containing the warm start entry point, is overwritten, the operator must bring the CP/M system back into memory with a cold start sequence.
Tables are included in the BIOS that describe the particular characteristics of the disk subsystem used with CP/M. These tables can be either hand-coded, as shown in the sample CBIOS in Appendix B, or automatically generated using the DISKDEF macro library, as shown in Appendix F. The purpose here is to describe the elements of these tables.
In general, each disk drive has an associated (16-byte) disk parameter header that contains information about the disk drive and provides a scratch pad area for certain BDOS operations. The format of the disk parameter header for each drive is shown in Figure 6-2, where each element is a word (16-bit) value.
The meaning of each Disk Parameter Header (DPH) element is detailed in Table 6-7.
Given n disk drives, the DPHs are arranged in a table whose first
row of 16 bytes corresponds to drive 0, with the last row
corresponding to drive n-1. In the following figure the label
DPBASE defines the base address of the DPH table.
Figure 6-3. Disk Parameter Header Table
DPBASE:
00 XLT00 0000 0000 0000 DIRBUF DBP00 CSV00 ALV00
01 XLT01 0000 0000 0000 DIRBUF DBP01 CSV01 ALV01
(AND SO ON THROUGH)
n-1 XLTn-1 0000 0000 0000 DIRBUF DBPn-1 CSVn-1 ALVn-1
A responsibility of the SELDSK subroutine is to return the base address of the DPH for the selected drive. The following sequence of operations returns the table address, with a 0000H returned if the selected drive does not exist.
NDISKS EQU 4 ;NUMBER OF DISK DRIVES
.......
SELDSK: ;SELECT DISK GIYEN BY BC
LSI H,0000H ;ERROR CODE
MOV A,C ;DRIVE OK?
CPI NDISKS ;CY IF SO
RNC ;RET IF ERROR
;NO ERROR, CONTINUE
MOV L,C ;LOW(DISK)
MOV H,B ;HIGH(DISK)
DAD H
DAD H ;*4
DAD H ;*B
DAD H ;*IC
LXI D,DPBASE ;FIRST DP
DAD D ;DPH(DISK)
RET
The translation vectors, XLT00 through XLTn-1, are located elsewhere in the BIOS, and simply correspond one-for-one with the logical sector numbers zero through the sector count 1. The Disk Parameter Block (DPB) for each drive is more complex. As shown in Figure 6-4, each particular DPB, that is addressed by one or more DPHS, takes the general form:
where each is a byte or word value, as shown by the 8b or 16b indicator below the field.
The following field abbreviations are used in Figure 6-4:
BLS BSH BLM
1,024 3 7
2,048 4 15
4,096 5 31
8,192 6 63
16,384 7 127
where all values are in decimal. The value of EXM depends upon both the BLS and whether the DSM value is less than 256 or greater than 255, as shown in Table 6-9.
BLS EXM values
DSM<256 DSM>255
1,024 0 N/A
2,048 1 0
4,096 3 1
8,192 7 3
16,384 15 7
The value of DSM is the maximum data block number supported by this particular drive, measured in BLS units. The product (DSM + 1) is the total number of bytes held by the drive and must be within the capacity of the physical disk, not counting the reserved operating system tracks.
The DRM entry is the one less than the total number of directory entries that can take on a 16-bit value. The values of AL0 and AL1, however, are determined by DRM. The values AL0 and AL1 can together be considered a string of 16-bits, as shown in Figure 6-5.
AL0 AL1
00 01 02 03 04 05 06 07
08 09 10 11 12 13 14 15
Position 00 corresponds to the high-order bit of the byte labeled AL0 and 15 corresponds to the low-order bit of the byte labeled AL1. Each bit position reserves a data block for number of directory entries, thus allowing a total of 16 data blocks to be assigned for directory entries (bits are assigned starting at 00 and filled to the right until position 15). Each directory entry occupies 32 bytes, resulting in the following tabulation:
Thus, if DRM = 127 (128 directory entries) and BLS = 1024, there are 32 directory entries per block, requiring 4 reserved blocks. In this case, the 4 high-order bits of AL0 are set, resulting in the values AL0 = 0F0H and AL1 = 00H.
The CKS value is determined as follows: if the disk drive media is removable, then CKS = (DRM + 1)/4, where DRM is the last directory entry number. If the media are fixed, then set CKS = 0 (no directory records are checked in this case).
Finally, the OFF field determines the number of tracks that are skipped at the beginning of the physical disk. This value is automatically added whenever SETTRK is called and can be used as a mechanism for skipping reserved operating system tracks or for partitioning a large disk into smaller segmented sections.
To complete the discussion of the DPB, several DPHs can address the same DPB if their drive characteristics are identical. Further, the DPB can be dynamically changed when a new drive is addressed by simply changing the pointer in the DPH; because the BDOS copies the DPB values to a local area whenever the SELDSK function is invoked.
Returning back to DPH for a particular drive, the two address values CSV and ALV remain. Both addresses reference an area of uninitialized memory following the BIOS. The areas must be unique for each drive, and the size of each area is determined by the values in the DPB.
The size of the area addressed by CSV is CKS bytes, which is sufficient to hold the directory check information for this particular drive. If CKS = (DRM + 1)/4, you must reserve (DRM + 1)/4 bytes for directory check use. If CKS = 0, no storage is reserved.
The size of the area addressed by ALV is determined by the maximum number of data blocks allowed for this particular disk and is computed as (DSM/8) + 1.
The CBIOS shown in Appendix B demonstrates an instance of these tables for standard 8-inch, single-density drives. It might be useful to examine this program and compare the tabular values with the definitions given above.
A macro library called DISKDEF (shown in
Appendix F), greatly
simplifies the table construction process. You must have access
to the MAC macro assembler, of course, to use the DISKDEF
facility, while the macro library is included with all CP/M 2
distribution disks.
A BIOS disk definition consists of the following sequence of
macro statements:
where the MACLIB statement loads the DISKDEF.LIB file, on the
same disk as the BIOS, into MAC's internal tables. The DISKS
macro call follows, which specifies the number of drives to be
configured with the user's system, where n is an integer in the
range 1 to 16. A series of DISKDEF macro calls then follow that
define the characteristics of each logical disk, 0 through n - 1,
corresponding to logical drives A through P. The DISKS and
DISKDEF macros generate the in-line fixed data tables described
in the previous section and thus must be placed in a
nonexecutable portion of the BIOS, typically directly following
the BIOS jump vector.
The remaining portion of the BIOS is defined following the
DISKDEF macros, with the ENDEF macro call immediately preceding
the END statement. The ENDEF (End of Diskdef) macro generates the
necessary uninitialized RAM areas that are located in memory
above the BIOS.
The DISKDEF macro call takes the form:
where
The value dn is the drive number being defined with this DISKDEF
macro invocation. The fsc parameter accounts for differing sector
numbering systems and is usually 0 to 1. The lsc is the last
numbered sector on a track. When present, the skf parameter
defines the sector skew factor, which is used to create a sector
translation table according to the skew.
If the number of sectors is less than 256, a single-byte table is
created, otherwise each translation table element occupies two
bytes. No translation table is created if the skf parameter is
omitted, or equal to 0.
The bls parameter specifies the number of bytes allocated to each
data block, and takes on the values 1024, 2048, 4096, 8192, or
16384. Generally, performance increases with larger data block
sizes because there are fewer directory references, and logically
connected data records are physically close on the disk. Further,
each directory entry addresses more data and the BIOS-resident
RAM space is reduced.
The dks parameter specifies the total disk size in bls units.
That is, if the bls = 2048 and dks = 1000, the total disk
capacity is 2,048,000 bytes. If dks is greater than 255, the
block size parameter bls must be greater than 1024. The value of
dir is the total number of directory entries that might exceed
255, if desired.
The cks parameter determines the number of directory items to
check on each directory scan and is used internally to detect
changed disks during system operation, where an intervening cold
or warm start has not occurred. When this situation is detected,
CP/M automatically marks the disk Read-Only so that data is not
subsequently destroyed.
As stated in the previous section, the value of cks = dir when
the medium is easily changed, as is the case with a floppy disk
subsystem. If the disk is permanently mounted, the value of cks
is typically 0, because the probability of changing disks without
a restart is low.
The ofs value determines the number of tracks to skip when this
particular drive is addressed, which can be used to reserve
additional operating system space or to simulate several logical
drives on a single large capacity physical drive. Finally, the
[0] parameter is included when file compatibility is required
with versions of 1.4 that have been modified for higher density
disks. This parameter ensures that only 16K is allocated for each
directory record, as was the case for previous versions.
Normally, this parameter is not included.
For convenience and economy of table space, the special form:
gives disk i the same characteristics as a previously defined
drive j. A standard fourdrive, single-density system, which is
compatible with version 1.4, is defined using the following macro
invocations:
with all disks having the same parameter values of 26 sectors per
track, numbered 1 through 26, with 6 sectors skipped between each
access, 1024 bytes per data block, 243 data blocks for a total of
243K-byte disk capacity, 64 checked directory entries, and two
operating system tracks.
The DISKS macro generates n DPHS, starting at the DPH table
address DPBASE generated by the macro. Each disk header block
contains sixteen bytes, as described above, and correspond one-
for-one to each of the defined drives. In the four-drive standard
system, for example, the DISKS macro generates a table of the
form:
where the DPH labels are included for reference purposes to show
the beginning table addresses for each drive 0 through 3. The
values contained within the DPH are described in detail in the
previous section. The check and allocation vector addresses are
generated by the ENDEF macro in the RAM area following the BIOS
code and tables.
Note that if the skf (skew factor) parameter is omitted, or equal
to 0, the translation table is omitted and a 0000H value is
inserted in the XLT position of the DPH for the disk. In a
subsequent call to perform the logical-to-physical translation,
SECTRAN receives a translation table address of DE = 0000H and
simply returns the original logical sector from BC in the HL
register pair.
A translate table is constructed when the skf parameter is
present, and the (nonzero) table address is placed into the
corresponding DPHS. The following, for example, is constructed
when the standard skew factor skf = 6 is specified in the DISKDEF
macro call:
Following the ENDEF macro call, a number of uninitialized data
areas are defined. These data areas need not be a part of the
BIOS that is loaded upon cold start, but must be available
between the BIOS and the end of memory. The size of the
uninitialized RAM area is determined by EQU statements generated
by the ENDEF macro. For a standard four-drive system, the ENDEF
macro might produce the following EQU statement:
which indicates that uninitialized RAM begins at location 4C72H,
ends at 4DB0H-1, and occupies 013CH bytes. You must ensure that
these addresses are free for use after the system is loaded.
After modification, you can use the STAT program to check drive
characteristics, because STAT uses the disk parameter block to
decode the drive information. A STAT command of the form:
decodes the disk parameter block for drive d (d = A,...,P) and
displays the following values:
Three examples of DISKDEF macro invocations are shown below with
corresponding STAT parameter values. The last example produces a
full 8-megabyte system.
Upon each call to BIOS WRITE entry point, the CP/M BDOS includes
information that allows effective sector blocking and deblocking
where the host disk subsystem has a sector size that is a
multiple of the basic 128-byte unit. The purpose here is to
present a general-purpose algorithm that can be included within
the BIOS and that uses the BDOS information to perform the
operations automatically.
On each call to WRITE, the BDOS provides the following
information in register C:
Condition 0 occurs whenever the next write operation is into a
previously written area, such as a random mode record update;
when the write is to other than the first sector of an
unallocated block; or when the write is not into the directory
area. Condition 1 occurs when a write into the directory area is
performed. Condition 2 occurs when the first record (only) of a
newly allocated data block is written. In most cases, application
programs read or write multiple 128-byte sectors in sequence;
thus, there is little overhead involved in either operation when
blocking and deblocking records, because preread operations can
be avoided when writing records.
Appendix G
lists the blocking and deblocking algorithms in
skeletal form; this file is included on your CP/M disk.
Generally, the algorithms map all CP/M sector read operations
onto the host disk through an intermediate buffer that is the
size of the host disk sector. Throughout the program, values and
variables that relate to the CP/M sector involved in a seek
are prefixed by sek, while those related to the host
disk system are prefixed by hst. The equate statements beginning
on
line 29 of
Appendix G define the mapping between CP/M and the
host system, and must be changed if other than the sample host
system is involved.
The entry points BOOT and WBOOT must contain the initialization
code starting on
line 57, while the SELDSK entry point must be
augmented by the code starting on
line 65. Note that although the
SELDSK entry point computes and returns the Disk Parameter Header
address, it does not physically select the host disk at this
point (it is selected later at READHST or WRITEHST). Further,
SETTRK and SETMA simply store the values, but do not take any
other action at this point. SECTRAN performs a trivial function
of returning the physical sector number.
The principal entry points are READ and WRITE, starting on
lines
110 and
125,
respectively. These subroutines take the place of
your previous READ and WRITE operations.
The actual physical read or write takes place at either WRITEHST
or READHST, where all values have been prepared: hstdsk is the
host disk number, hsttrk is the host track number, and hstsec is
the host sector number, which may require translation to physical
sector number. You must insert code at this point that performs
the full sector read or write into or out of the buffer at hstbuf
of length hstsiz. All other mapping functions are performed by
the algorithms.
This particular algorithm was tested using an 80-megabyte hard
disk unit that was originally configured for 128-byte sectors,
producing approximately 35 megabytes of formatted storage. When
configured for 512-byte host sectors, usable storage increased to
57 megabytes, with a corresponding 400% improvement in overall
response. In this situation, there is no apparent overhead
involved in deblocking sectors, with the advantage that user
programs still maintain 128-byte sectors. This is primarily
because of the information provided by the BDOS, which eliminates
the necessity for preread operations.
6.11 The DISKDEF Macro Library
MACLIB DISKDEF
DISKS n
DISKDEF 0,. . .
DISKDEF 1,. . .
.....
DISKDEF n - 1
ENDEF
DISKDEF dn,fsc,lsc,[skf],bls dks,dir,cks,ofs,[0]
DISKDEF i,j
DISKS 4
DISKDEF 0,1,26,6,1024,243,64,2
DISKDEF 1,0
DISKDEF 2,0
DISKDEF 3,0
......
ENDEF
DPBASE EQU $
DPE0: DW XLT0,0000H,0000H,0000H,DIRBUF,DPB0,CSV0,ALV0
DPE1: DW XLT0,0000H,0000H,0000H,DIRBUF,DPB0,CSV1,ALV1
DPE2: DW XLT0,0000H,0000H,0000H,DIRBUF,DPB0,CSV2,ALV2
DPE3: DW XLT0,0000H,0000H,0000H,DIRBUF,DPB0,CSV3,ALV3
XLT0: DB 1,7,13,19,25,5,11,17,23,3,9,15,21
DB 2,8,14,20,26,6,12,18,24,4,10,16,22
4C72 = BEGDAT EQU $ (data areas)
4DB0 = ENDDAT EQU $
013C = DATSIZ EQU $-BEGDAT
STAT D:DSK:
r: 128-byte record capacity
k: kilobyte drive capacity
d: 32-byte directory entries
c: checked directory entries
e: records/extent
b: records/block
s: sectors/track
t: reserved tracks
DISKDEF 0,1,58,,2048,256,128,128,2
r=4096, k=512, d=128, c=128, e=256, b=16, s=58, t=2
DISKDEF 0,1,58,,2048,1024,300,0,2
r=16348, k=2048, d=300, c=0, e=128, b=16, s=58, t=2
DISKDEF 0,1,58,,16348,512,128,128,2
r=65536, k=8192, d=128, c=128, e=1024, b=128, s=58, t=2
6.12 Sector Blocking and Deblocking
0 (normal sector write)
1 (write to directory sector)
2 (write to the first sector of a new data block)