Difference between revisions of "OSWORD &7F"
(→Implementation: cize+count vs bytes transferred in DDOS, EDOS) |
m (→Opus DDOS: ~) |
||
Line 344: | Line 344: | ||
The count of bytes to be transferred is computed from the ''size+count'' | The count of bytes to be transferred is computed from the ''size+count'' | ||
− | parameter, but the sectors' N records are not matched | + | parameter, but the sectors' N records are not matched with ''size''. |
There is one density flag and one double-stepping flag shared by all drives | There is one density flag and one double-stepping flag shared by all drives |
Revision as of 19:23, 15 April 2021
Contents
Specification
On entry: | On exit: | |
Control block | ||
XY?&00 | Drive | |
XY!&01 | Data address | |
XY?&05 | Number of parameters (n) | |
XY?&06 | Command | |
XY+&07... | Parameters | |
XY?(7+n) | Result |
No contents of the control block should be relied upon on exit other than the result byte. Some systems update the control block, some systems do not.
Command | Parameters | Meaning |
&40 | <track> <sector> <size+count> <size> <field> | Scan data ** |
&44 | <track> <sector> <size+count> <size> <field> | Scan data and deleted data multi-sector ** |
&4A | <track> <sector> | Write data 128 bytes |
&4B | <track> <sector> <size+count> | Write data |
&4E | <track> <sector> | Write deleted data 128 bytes |
&4F | <track> <sector> <size+count> | Write deleted data |
&52 | <track> <sector> | Read data 128 bytes |
&53 | <track> <sector> <size+count> | Read data |
&56 | <track> <sector> | Read data and deleted data 128 bytes |
&57 | <track> <sector> <size+count> | Read data and deleted data |
&5B | <track> <&00> <count> | Read IDs |
&5E | <track> <sector> | Verify data and deleted data 128 bytes |
&5F | <track> <sector> <size+count> | Verify data and deleted data |
&63 | <track> <gap3> <size+count> <gap5> <gap1> | Format track |
&64 | <track> | (EDOS) Write track |
&65 | <track> | (EDOS) Read track |
&69 | <track> | Seek |
&6C | – | Read drive status and reset 'not ready' |
&75 | <&0D> <step> <settle> <load> | Initialise |
&75 | <&1n> <bad 1> <bad 2> <current track> | Specify bad tracks |
&76 | – | (EDOS) Force interrupt |
&7A | <register> <value> | Write special register |
&7D | <register> | Read special register |
&E0 | <track> <nine more bytes> | (1770) Read track |
&F0 | <track> <nine more bytes> | (1770) Write track |
** Not possible on a BBC as needs DMA hardware.
Parameters
Drive
The drive number specifies the drive to access. If bit 7 is set, then the previously-used drive is used, and the drive status check is skipped.
The drive number can be expressed as a binary number %pxDDDdsd.
- b7 : p - If 1, selects the previously-used drive and density
- b5-b3 : DDD - select the drive density:
- %xx0=Acorn single density
- %xx1=Acorn double density
- %0xx=Watford current density
- %10x=Watford single density
- %11x=Watford double density
- b2,b0 : dd - physical drive number
- b1 : s - side
Single density DFSs do not have any ability to select double density disk access. Acorn DFSs select the density with b3, Watford DFSs select the density with b4 and b5. To specify a density in a manner compatible with both Acorn and Watford DFSs specify the density with both methods:
%001000sd - to specify single density %001110sd - to specify double density
Most DFSs do not implement physical drives numbers larger than 1 (logical drives greater than 3). Selecting the previously selected drive can only be relied upon if DFS is the currently selected filing system. A drive number of &FF should be passed to select the previously used drive.
DFS 2.xx in the Master MOS 3.20 ROM incorrectly passes the density bit in b3 to the drive control register b3 instead of b5, so OSWORD &7F cannot access double density disks without patching the DFS[1].
Data Address
This is the address to transfer data to or from.
Track
This is the logical track to transfer data to or from. This may be different from the physical track number if the drive has previously been stepped to the correct physical track.
Sector
This is the logical sector number to start data transfer from. Sectors can be numbered anything from &00 to &FF, but single-density DFS-formatted disks use &00-&09 and double-density DFS and ADFS disks use &00-&0F.
Size + Count
Bits 0 to 4 specify the number of sectors to be read or written. Specifying zero sectors is undefined and may result in unpredictable behaviour.
Bits 5 to 7 specify the sector size as 128*2^size, as follows:
b7 | b6 | b5 | Sector size |
0 | 0 | 0 | 128 bytes |
0 | 0 | 1 | 256 bytes |
0 | 1 | 0 | 512 bytes |
0 | 1 | 1 | 1024 bytes |
1 | 0 | 0 | 2048 bytes (8271 only) |
1 | 0 | 1 | 4096 bytes (8271 only) |
1 | 1 | 0 | 8192 bytes (8271 only) |
1 | 1 | 1 | 16384 bytes (8271 only) |
Special Registers
Command &7A and &7D write and read internal special registers. These registers are:
- &00 40/80 flag (Opus) : reserved:dstep:0:0:0:0:0:0 0=single, 1=double
- &03 Density (Opus) : reserved:densy:0:0:0:0:0:0 0=single, 1=double
- &04 SROM (Opus) : &00-&0F
- &06 Scan Sector (sector any error occured at)
- &10 Bad Track Register 1 (drive 0/2)
- &11 Bad Track Register 2 (drive 0/2)
- &12 Track
- &13 Scan Count LSB
- &14 Scan Count MSB
- &17 DMA Mode Register 1:1:0:0:0:0:<single head>:<no DMA>
- &18 Bad Track 1 (drive 1/3)
- &19 Bad Track 2 (drive 1/3)
- &1A Track (drive 1/3)
- &22 Drive Control Input
- &23 Drive Control Output SEL1:SEL0:FAULT:LOW:LOAD:DIR:SEEK:WREN
What Special Registers are supported depends on the specific hardware and DFS firmware.
Result
The result is returned to the byte after the last parameter. The location is addressed by XY?(7+XY?5). For command &6C this location contains the drive status; for command &7D it is the requested special register; and for command &7A the result is undefined. Values returned by the other commands are as made up as follows:
- b7-b6: always zero (1770 DFS returns 11 if command unrecognised)
- b5: deleted data found
- b4-b3: completion type: 00=no error, 01=recoverable error, 10=unrecoverable error, 11=failure
- b2-b1: completion code
- b0: always zero
This gives the following result values:
type | code | Result | Meaning |
00 | 00 | &00 | Successful completion |
00 | 01 | &02 | Success, scan met equal ** |
00 | 10 | &04 | Success, scan met not equal ** |
00 | 11 | &06 | unused |
01 | 00 | &08 | Clock error |
01 | 01 | &0A | Late DMA ** / Late ISR |
01 | 10 | &0C | ID CRC error |
01 | 11 | &0E | Data CRC error |
10 | 00 | &10 | Drive not ready |
10 | 01 | &12 | Disk write protected |
10 | 10 | &14 | Track 0 not found |
10 | 11 | &16 | Write error |
11 | 00 | &18 | Sector not found |
11 | 01 | &1A | unused |
11 | 10 | &1C | unused |
11 | 11 | &1E | Drive not present/drive empty |
** Not possible on a BBC as needs DMA hardware.
The result byte will have &20 added to it if deleted data has been read.
Intel documentation states that result type 01 should be retried up to a total of ten times before abandoning. In practice:
- results types 01 and 10, other than &10 - Drive not ready - and &12 - Disk write protected, should be retried up to ten times
- result &10 should be retried indefinitely until some other result occurs
- result &18 - Sector not found - could be followed by a seek to track zero then one additional attempt.
- While result &1E - Drive not present/empty - is not mentioned in official documentation, if left trying to access an empty drive this result will eventually occur. As such, emulators such as BeebEm will give Disk fault 1E if trying to access an empty drive. 1770 DFSs usually return Disk error 18 (Sector not found) instead.
The result byte can be preloaded with a value to return if no OSWORD &7F routine exists. For instance, preloading the result with &00 makes all such failed calls appear to succeed, preloading with &18 makes all calls appear to return 'Sector not found'. HADFS preloads the result byte with &1E to give a default result of 'Drive not present'.
Read Drive Status (&6C) resets any 'not ready' error and returns the drive status as the result byte:
- b7 : unused
- b6 : READY1
- b5 : FAULT
- b4 : INDEX
- b3 : WR PROTECT
- b2 : READY0
- b1 : TRACK0
- b0 : COUNT
8271 Commands
The 8271 command passed in the control block is actually a bitmap made up as follows:
+---+---+---+---+---+---+---+---+ | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 | +---+---+---+---+---+---+---+---+ | | | || | | | | | | +---+---+ | +---+----Size: 00 multiple scan data | | || | 01 no transfer | | || | 10 128 bytes | | || | 11 multiple sectors | | || | | | || +------------Data: 0 ignore deleted data | | || 1 accept deleted data | | || | | |+--------------Direction: 0 Write to FDC | | | 1 Read from FDC | | | | | +-----------------Command: 000 Scan | | 001 Write | | 010 Read | | 011 Verify | | 100 Format | | 101 Seek | | 110 Initialise | | 111 Special Registers | | | +---------------------------------- Drive select 0 +-------------------------------------- Drive select 1
The drive select bits must be set to %01 in all commands that operate the drive, to ensure that exactly one drive responds. The command byte value lies in the range &40 to &7F.[2]
Acorn DFS toggles both bits when the drive parameter selects drive 1 or 3.[3] Then it writes the command to the 8271 command register where bits 6 and 7 directly control the select lines to drive 0 and 1, respectively. Each attached drive responds only when its select line is set to logical 1. The 8271 uses bit 7 internally to choose which track register to update, and resets most drive control outputs when bits 6 or 7 change. Bit 5 of special register &23 controls the output on pin 1 and selects which side of the disc to use.[4]
Emulating implementations of OSWORD &7F typically mask off bits 6 and 7 of the command byte, and select the drive and side using their private control register. They seldom recognise the 128-byte versions of commands.
Bit 4 is used to select the appropriate NMI and Tube code to transfer to or from the floppy disk controller.
Obviously, the bitmap above is a simplification: the real effect of each command bit is more complex than the labels imply. None of the undocumented commands are defined. They are not available in implementations of OSWORD &7F that emulate an 8271 controller, though some undocumented commands on a real 8271 exhibit repeatable and exploitable behaviour.
Implementation
Some DFSs, such as DFS 2.45 supplied in the Master MOS 3.50 ROM, only respond to OSWORD &7F if DFS is the current filing system and fail silently if any other filing system is selected.
Acorn DFS
This is the original implementation of OSWORD &7F. It is a fairly simple interface to the Intel 8271 floppy disc controller; it passes the command and parameters to the controller, supplies the appropriate NMI service routine to fetch or store the disc data, and returns the result.
Acorn 1770 DFS
This emulates key commands of the 8271. Software double-stepping is available. Double density access can be made by setting bit 3 of the drive number. Unfortunately, the programmers didn't realise that the double density bit was moved in the BBC Master hardware, so double density access is not possible on the Master without a patched filing system[5].
Watford DDFS
Density selectable with b4-b5 of the drive number.
Opus DDOS
Command &5B auto-detects the disc density by trying each density twice until an ID is read. It does not seek the specified track and reads one ID only, from the track currently under the read head.
Special register &12 is write-only and sets the current track number of the physical drive selected by XY?&00.
The count of bytes to be transferred is computed from the size+count parameter, but the sectors' N records are not matched with size.
There is one density flag and one double-stepping flag shared by all drives and exposed in the Read/write Special Registers commands.
The result of command &6C (read drive status) contains only the write-protect state in bit 3, ORed with &44. Otherwise the returned status byte is the WD controller status byte masked according to the command (e.g. bit 5 is cleared if the command accepts deleted data). &00 indicates success; any other value indicates an error. No translation to an equivalent 8271 result is carried out.
Slogger Challenger
As Opus DDOS except special registers &12 and &1A are read/write and emulate those in the 8271. XY?&00 b3=1 forces double density operation.
Opus EDOS
This emulates key commands of the 8271, and maintains one density flag and
one double-stepping flag for each drive (two pairs of flags per disc.) Only
a catalogue read operation updates the flags, one drive at a time, though a
manual *OPT 6
or *OPT 8
setting during the
operation will force the flags to a desired value. Commands that read the
catalogue include *CAT
, *CATGEN
, OSFIND
&40, OSGBPB 5 or 8, and (with an EDOSPAT patched ROM) OSWORD &71,
&7D or &7E. All raise an error if a valid catalogue is not found.
The Format Track command formats 256-byte sectors only, in double density if 16 or more sectors are specified, otherwise single. The density of operation cannot be directly specified to other commands.
The sector size parameter is ignored in favour of the sectors' N records, which alone determine the number of bytes transferred.
Seven WD controller status values are translated to Intel equivalents (&00, &0A, &0C, &0E, &12, &18, &20); a WD status with an unexpected combination of errors is returned with bit 0 set.
Coding
The following routine can be used to perform an OSWORD &7F call that takes
three parameters, such as Write (&4B) and Read (&53), or Track Read (&E0)
and Track Write (&F0) which take ten parameters. It requires X%
=>15-byte
control block, Y%=X%DIV256
. den%=1
for single
density and, if supported by the hardware, den%=2
for double density.
It returns a result of &1E (not present) if no OSWORD &7F call is available.
DEFFNdisk(addr%,cmd%,drv%,trk%,sec%,num%,den%):LOCAL fs% fs%=FNfs:IF fs%<>4:*FX143,18,4 REPEAT X%?0=drv%+den%*24+8:X%!1=addr%:X%?5=3-7*(cmd%>127) X%?6=cmd%:X%?7=trk%:X%?8=sec%:X%!9=num%OR&1E20 A%=127:CALL&FFF1:A%=X%?(X%?5+7) UNTIL A%<>&10:IF fs%<>4:OSCLI"FX143,18,"+STR$ fs% =A% DEFFNfs:LOCAL A%,E%,Y%:=(USR&FFDA)AND&FF
Tube note
The Tube MOS copies a 16-byte control block over the Tube, so on a read operation, any data read to immediately after the control block will be overwitten by the returned control block if it is within 16 bytes of the start of the control block.
If code is to run properly, the destination address of any reads must be at least 16 bytes away from the start of the control block, for example:
DIM X% 15, T% 255 Y% = X% DIV 256 X%!1 = T% etc...
If blocks are declared with variables in the heap, the variable definitions count towards the 16-byte distance, so the following is ok:
DIM ctrl% 12, data% 255 X%=ctrl%:Y%=X% DIV 256 X%!1 = data% etc...
As the variable data%
is assigned space between the two
buffers, the value of data% - ctrl%
exceeds 15.
See Also
- http://mdfs.net/Docs/Comp/BBC/Osword/Osword7F
- http://mdfs.net/Docs/Comp/BBC/Osword/DFSOsword
- OSWORD &72
- OSWORD &62
References
- ↑ http://mdfs.net/ROMs/Filing/Disk
- ↑ The 8271 data sheet specifies command &35 to Initialise, i.e. both drive select lines zero. With mini-floppy drives the drive select line levels do not matter while the head is unloaded and so command &75 is also acceptable.
- ↑ Acorn Computers, source code to DNFS 3.00, file DOS03, l.349: EORIM &C0
- ↑ The 8271 does not know the purpose of the output and effectively assumes a single actuator for both sides of a double-sided drive. Bit 1 of the mode register applies when two drives share one actuator, such as the drives on a DEC RX50.
- ↑ http://mdfs.net/ROMs/Filing/Disk