28 Dec
2023
28 Dec
'23
10:39 p.m.
Ok, this is a long shot, but does anybody know the algorithm used by the
VMS (and presumably RSX) RX01/RX02 driver for interleaving sectors?
More specifically, logical blocks on VMS are always 512 bytes but RX01
sectors are 128 bytes and RX02 sectors are 256. So each logical block
consists of 4 RX01 sectors or 2 RX02 sectors. I'm presuming these sectors
are not contiguous, but rather the device driver uses some kind of
interleave to minimize the rotational delays. What exactly is that
algorithm?
Or even more detail, I have a bunch of RX01 and RX02 image files (as VMS
disk files) that were created using a program that dumps consecutive VMS
logical blocks (using SYS$QIOW(.IO$_READLBLK,.). This works great for
things where the logical to physical translation is hidden from the OS (like
all MSCP drives) for RX drives that's not the case. I need to figure out
how to unscramble the 512 byte blocks in my VMS file to re-create the
original sectors.
And no, I don't have access to the actual diskettes to dump them again.
Shouldn't matter - the data is in the file, I just have to figure out the
correct bit shuffling to get it back in the right order. I guess if I get
really desperate I'll try to find the source listings for the drivers, but
I'm hoping somebody just knows.
Bob
28 Dec
28 Dec
10:46 p.m.
Hi.
On 2023-12-28 22:39, Robert Armstrong wrote:
Ok, this is a long shot, but does anybody know the
algorithm used by
the VMS (and presumably RSX) RX01/RX02 driver for interleaving sectors?
Sure. RSX contains the sources for the device drivers, and this is in there.
Just check [11,10]DXDRV.MAC on any RSX kit. (I might have protected it
on MIM, but let me know if you want to get it from there.)
Anyway, the comment says:
; FUNCTIONS RECOGNIZED:
; IO.RPB - READ PHYSICAL BLOCK (MAY READ ANY SECTOR ON DISK)
; IO.WPB - WRITE PHYSICAL BLOCK (MAY WRITE ANY SECTOR ON DISK)
; IO.WDD - WRITE DELETED DATA - WRITE ANY SECTOR ON DISK
; SETTING THE DELETED DATA MARK IN SECTOR HEADER
; THE FOLLOWING TWO FUNCTION CODES OPERATE ON 256 WORD BLOCKS
; OF DATA (4 SECTORS) THE DRIVER AUTOMATICALLY INTERLEAVES
; AND SKEWS SECTORS TO OPTIMIZE THE TRANSFER OF DATA
; SO THAT A LONG READ/WRITE MAY BE DONE WITHOUT LOOSING
; A REVOLUTION OF THE DISK. THE INTERLEAVE FACTOR IS 2
; AND THE SKEW FACTOR IS 6.
; IO.RLB - READ LOGICAL BLOCKS OF 256. WORDS
; IO.WLB - WRITE LOGICAL BLOCKS OF 256. WORDS.
Also, the actual code is in function TRKSEC, starting at around line 500.
Johnny
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt(a)softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
29 Dec
29 Dec
1:07 a.m.
You might want to look at sim_disk.c's file system recognition logic. John Forecast
wrote the logic which walks through the data to determine the interleave and sector size
(as I recall).
From: Robert Armstrong <bob(a)jfcl.com>
Sent: Thursday, December 28, 2023 1:39 PM
To: 'The Hobbyist DECnet mailing list' <hecnet(a)lists.dfupdate.se>
Subject: [HECnet] Sector interleave for RX01/2
Ok, this is a long shot, but does anybody know the algorithm used by the VMS (and
presumably RSX) RX01/RX02 driver for interleaving sectors?
More specifically, logical blocks on VMS are always 512 bytes but RX01 sectors are 128
bytes and RX02 sectors are 256. So each logical block consists of 4 RX01 sectors or 2
RX02 sectors. I'm presuming these sectors are not contiguous, but rather the device
driver uses some kind of interleave to minimize the rotational delays. What exactly is
that algorithm?
Or even more detail, I have a bunch of RX01 and RX02 image files (as VMS disk files)
that were created using a program that dumps consecutive VMS logical blocks (using
SYS$QIOW(...IO$_READLBLK,...). This works great for things where the logical to physical
translation is hidden from the OS (like all MSCP drives) for RX drives that's not the
case. I need to figure out how to unscramble the 512 byte blocks in my VMS file to
re-create the original sectors.
And no, I don't have access to the actual diskettes to dump them again.
Shouldn't matter - the data is in the file, I just have to figure out the correct bit
shuffling to get it back in the right order. I guess if I get really desperate I'll
try to find the source listings for the drivers, but I'm hoping somebody just knows.
Bob
3:56 a.m.
Thanks for all the answers, and thanks to Johnny for providing a pointer
to the DX driver code. FWIW, the DX driver performs the logical to physical
translation using a "logical sector number", where each logical block
contains four consecutive logical sectors. The routine that converts an
LSN to a track and sector, and handles interleave and skew at the same time,
is below. I'm not even going to pretend I know how this works - it's a bit
obscure - but it's straight forward enough to translate it literally into C.
Bob
;+
; *** - TRKSEC - CONVERT LOGICAL OR PHYSICAL BLOCK NUMBER TO
; TRACK-SECTOR PAIR
; FROM ALGORITHM BY J GILBERT MODIFIED BY H. JACOBS
;
; INPUT:
; R3 - I/O PACKET ADDRESS
; I.PRM+10(R3) - LOGICAL OR PHYSICAL SECTOR
;
; OUTPUT:
; I.PRM+14(R3) - SECTOR (1-26.)
; I.PRM+15(R3) - TRACK (0-77.)
; R3 - UNCHANGED
; C CLEAR - VALID BLOCK
; C SET - BAD BLOCK NUMBER (PHYSICAL OR LOGICAL)
;
;-
TRKSEC: MOV I.PRM+10(R3),R1 ; GET LOGICAL OR
PHYSICAL BLOCK
MOV #8.,R0 ; SET LOOP COUNT
MOV #6400,R2 ; SET DIVISOR
1$: CMP R2,R1 ; DOES 26 GO INTO DIVIDEND?
BHI 2$ ; BRANCH IF NOT, C
CLEAR
SUB R2,R1 ; SUBTRACT 26 FROM
DIVIDEND
SEC ; SET CARRY
2$: ROL R1 ; SHIFT DIVIDEND AND
QUOTIENT
DEC R0 ; DONE?
BGT 1$ ; NO, LOOP
MOVB R1,R0 ; GET TRACK NUMBER
CLRB R1 ; CLEAR TRACK NUMBER
SWAB R1 ; SHIFT DONE SECTOR NUMBER
BITB #IO.RPB&377,I.FCN(R3) ; PHYSICAL BLOCK WANTED?
BNE 10$ ; YES
CMP #12.,R1 ; NO, C=1 IF 13<=R1<=25
ROL R1 ; DOUBLE FOR INTERLEAVE
FACTOR
ASL R0 ; ADD TRACK -TRACK
SKEW
ADD R0,R1 ; SKEW BY 2*TRACK
ADD R0,R1 ; SKEW BY 4*TRACK
ADD R0,R1 ; SKEW BY 6*TRACK
ASR R0 ; RESTORE TRACK NUMBER
MOV #26.,R2 ; SET MODULUS
5$: SUB R2,R1 ; MODULO SECTOR INTO RANGE
-26. TO -1.
BGE 5$ ; LOOP TILL REMAINDER
GOES NEG
ADD R2,R1 ; CONVERT TO RANGE 0-25.
INC R0 ; LBN0 STARTS ON TRACK
1
10$: INC R1 ; CONVERT TO RANGE 1-26.
MOV R1,I.PRM+14(R3) ; SAVE SECTOR NUMBER
MOVB R0,I.PRM+15(R3) ; SAVE TRACK NUMBER
CMP #77.*256.,I.PRM+14(R3) ; IS IT A VALID
TRACK/SECTOR?
RETURN
132
days inactive
133
days old
3 comments
3 participants
participants (3)
-
Johnny Billquist -
Mark Pizzolato - Info Comm -
Robert Armstrong