R10-01 SCRs are:
21763.0 - Long BREAK timeout on block mode reads with CTD driver
21765.0 - Crash code 102 when issuing a write to a contig file on the CPU
21767.0 - No crash handling possible if ESCP board is sysgened but not present
21772.0 - Allocating and deleting large contig files does not always free all space used
21778.0 - CTM overwrite low memory at address x26
21782.0 - BACKUP stores its revision into the volume header on tape incorrectly
21791.0 - The D M/.SPL or D M/.MTM commands returns a PARM-ERR POS= error message
21793.0 - FASTBACK paused with an internal failure 6622 error code
21795.0 - FASTBACK terminated with a insufficient disk space message on a restore
21797.0 - Crash code 153 when marking on a disk and using SVC 7 xFF0D
21799.0 - Terminal hang problem with DASY driver using device code 156
21801.0 - FASTBACK paused with a FIXED POINT OVERFLOW ERROR
21803.0 - FASTBACK paused with a x8301 tape read error
21805.0 - DISPLAY command of COPY32 displays year as 1900
21808.0 - FASTBACK paused with a "NON-EXISTENT SEGMENT ERROR (PST) AT 818A10" error.
21810.0 - FASTBACK aborts with "HDR1 Label not found..." when backing up to a QIC 525 tape.
21811.0 - FASTCOPY displays incorrect data on non-verify of a record.
21814.0 - DUMPRINT shows a given IOB queued to more than one disk volume.
21815.0 - BACKUP may fail incorrectly if its confidence test detects a bad tape
21816.0 - CDS reports a "unsupported special memory operation" error on a 3200-850.
21818.0 - The CTD stream mode does not function if ones user application is
loaded into extended memory.
21820.0 - Attempting to use CTD stream mode with a device configured on a IOP will cause a 102 crash.
================================================================================
================================================================================
217630
Software Problem:
----------------
When hitting the BREAK key while a block mode read is outstanding to
the CTD terminal, it can take up to 35 seconds for the program to
receive the 0x8204 break status. The block mode read was issued with
a SVC 1 function code of 0x49 with extended options of 0x20000001
which is a read formatted with Request To Send (RQS). A sample
program has been provided which reproduces the problem.
Software Resolution:
-------------------
To handle problems with noisy dial-in lines when using Reliance, SCR
16899 was included in CTD on May of 1990 which became part of the OS/32
R09-01 release. When line noise caused a problem (normally parity or
framing errors), the old code terminated immediately. Reliance then
said "recoverable error - re-issue the read". This then got "satisfied"
by the data still coming in - HOWEVER - the driver (in response to its
Request To Send escape sequence) expected to get Esc R x (x any char).
Of course it did not - and the driver set 0x8426 status. Reliance now
said "unrecoverable error" and flagged the terminal as in error and
would wait for someone to use the ITC Controller to fix the problem.
The solution adopted by SCR 16899 was to have the driver wait until
after all the original data arrived, then terminate with a recoverable
error. This was handled by computing a time-out based upon the size
of the read buffer to allow enough time for all of the original data
to arrive. For a buffer size of 1000 bytes this resulted in a time
out value of 31 seconds. Hitting an actual BREAK key got wrapped up
in this error handling.
In our case the driver knows an actual BREAK key was hit because it
has seen the framing error status and a data character of 0x00 which
defines BREAK. Instead of setting a time-out to allow for additional
data, of which there is none, we should just return the 0x8204 break
status.
return to index
================================================================================
217650
Software Problem:
----------------
Crash code 102 when issuing a write to a contig file on the CPU.
The machine is a 3280MPS with one APU and one IOP.
Software Resolution:
-------------------
The user application task was linked with XSVC1. The SVC 1 function
code used was x25 which requests a write-random-XSVC1 with a extended
option word of x00708BE0. One of the bits set in this extended option
word requests a DMS checksum of the buffer. The user application did
not intend to request a DMS checksum, but did because of the garbage
in the extended options word.
When DMS issues an I/O to its data base file located on a disk under a
IOP it requests the IOP to perform the DMS checksum of the buffer. If
the data base file is located on a disk under the CPU, then the DMS
application performs the DMS checksum. The code within the contig file
manager (FMCO) relied upon this assumption. The functional variation of
FMCO for a IOP system contains the DMS checksum code, whereas a non-IOP
system does not contain the DMS checksum code. As a result FMCO in a
non-IOP system will ignore a garbage extended options word. If the user
application requests FMCO to use the provided extended options word,
then FMCO in a IOP configuration will try and use it.
Because the user application requested a IOP only function to a contig
file on the CPU, FMCO entered some IOP only code not supported by the
CPU, thus causing the 102 crash.
The user application can be corrected in three different ways.
1) Correct the extended options word in the SVC 1 parameter block.
2) Change the SVC 1 function code to not request XSVC1.
3) Re-link the task without XSVC1.
The following changes will protect the OS from a user application which
happens to request a DMS checksum on a contig file on the CPU if a IOP
is configured in the system.
return to index
================================================================================
217670
Software Problem:
----------------
If there is an ESCP board sysgenned but not present in the system,
in the "sysinit and power restore" routine, under the label ESCP.ERR,
the DCB fields DCB.WDN and DCB.RDN are used to index into the ISPTAB.
These two fields are not yet set at sysinit time, they will be set
later by the ITAM file manager when the device is assigned. As a
result, the first halfword of the ISPTAB (0xD0) is set to III, thus
no crash handling is possible.
Software Resolution:
-------------------
Instead of using the DCB fields DCB.WDN and DCB.RDN, one should use
the field DCB.DN which does contain the device address of the ESCP.
return to index
================================================================================
217720
Software Problem:
----------------
After allocating and deleting a large contig file, the space does not
always seem to be freed up. If one attempts to allocate the same
size file as was just deleted, a "ALLO-ERR TYPE=SIZE" error occurs.
Software Resolution:
-------------------
This problem in the Bit Map Manager (FMBM) will only occur under very
specific conditions. Using a disk configured as follows:
BARA:,1A9,89,SELCH=(1F6,FC,6),CONTR=1A8,DISK=(64,19,12057),VIRT=(0,862)
Initialized it with a directory for 1600 files and marked on with the
default bit map page size of 4 sectors. Finally allocate a contig
file of 900,000 sectors, delete this file and try to allocate it again.
A problem was found in the Bit Map Manager when it was performing
some of its final house keeping for a delete request. It had already
updated the bit map for the extent being released and was adding in
the free space immediately before this extent. If the extent that
was just released had started on some multiple of 256 sectors and
was near the beginning of the disk, the bit map manager could view
part of the extent just released a free space immediately before
this extent. This occurred because the bit map manager had used
the actual end address of the bit map, instead of computing a end
address within the bit map buffer. An additional side effect of
this error results in the resetting of a bit in memory between
128 kb to 256 kb above the end of the bit map buffer. The resetting
of this bit could cause a much more severe memory problem then the
allocate problem with the disk.
Because so many factors are involved, it is nearly impossible to
determine if this problem could occur for a given disk volume marked
on in a given fashion. It is recommended one install this patch if
one suspects a system space problem or has ever received a SIZE
error on a allocate of a large contig file they knew should have
succeeded.
return to index
================================================================================
217780
Software Problem:
----------------
CTM is writing to low memory address x26 in routines ZC.ICTM and
CTFA.ERR. In both cases they are trying to clear the status of the
current IOB in the DCB. The code used is the same.
STH R15,DCB.STAT(R1)
L R14,DCB.IOB(R1)
STH R15,IOB.STAT(R14)
The load of R14 followed by the store of R15 is totally unnecessary.
The code as written will try to fetch the IOB.NEXT pointer from the
DCB. This pointer is not valid and will contain zero. Executing the
next store instruction will cause a write to address x26. To get the
address of the IOB, it is necessary to do a LA R14,DCB.IOB(R1). Then
one can do a STH R15,IOB.STAT(R14). This is not needed because the
first store already wrote to this location.
Software Resolution:
-------------------
The statement of the problem is accurate. The Load and Store Halfword
instructions are unnecessary and do cause the overwriting of low memory.
As such, they will be deleted.
return to index
================================================================================
217820
Software Problem:
----------------
BACKUP R10-01 stores its revision into the volume header on tape
incorrectly. The revision being stored was R16-01 instead of R10-01.
BACKUP reports the revision of the BACKUP used to create the tape under
the Input: section when doing a restore or report only from tape.
Software Resolution:
-------------------
The macro in the BACKUPM.MLB macro library was modified to
create the revision string in hexadecimal instead of decimal. BACKUP
assumed the revision string was hexadecimal. This changes allows older
BACKUP tasks, such as the task included with the OS/32 R06-01 release,
to report the revision of BACKUP which created the tape correctly.
return to index
================================================================================
217910
Software Problem:
----------------
When performing a D M/.SPL or D M/.MTM the OS/32 command processor
returns a PARM-ERR POS= error message.
Software Resolution:
-------------------
When requesting a display map of a special system task, such as .SPL
or .MTM, the OS/32 command processor did not initialize register UF
properly before validating the task name. The display map command
handler uses a special entry point to the CHECKID routine that requies
register UF to be initialized to a value other than zero. Setting
register UF to one before validating the special task name resolves the
problem.
return to index
================================================================================
217930
Software Problem:
----------------
FASTBACK paused with an internal failure 6622 error code.
Software Resolution:
-------------------
FASTBACK will pause with an internal failure code of 6622 when it
detects it is about to queue or write a data block to tape that
contains illogical data links. If this block was actually written to
tape, one would likely fail on the verify or not be able to restore
the tapes back to disk. By having FASTBACK detect the problem before
it writes the data to tape prevents one from creating FASTBACK tapes
that one may not be able to restore.
All of the following conditions must be met before FASTBACK will create
a illogical data block:
o Doing a multi-reel backup from disk to tape without a verify
o FASTBACK was loaded with enough memory to allow it to have more
than two buffers
o The disk read process within FASTBACK was a number of buffers
ahead of the tape write process, ie. there were a number of write
requests queued to the tape drive
o Disk fragmentation existed on disk where the disk read process was
at the time the tape drive encountered end of tape
Other than the first condition, a multi-reel backup without verify, the
user has little or no control over the conditions required for FASTBACK
to create a illogical data block. All versions of FASTBACK from R08-03
thru R10-01 can create a illogical data block, only versions R09-02.1
thru R10-01 can detect a illogical data block was created and pause
with an internal failure. Therefore users of FASTBACK versions R08-03
thru R09-02 could create backup tapes that they may not be able to
restore.
The illogical data block was created by the disk read process because the
end of tape handling routine for the tape write process reset the current
disk LBA location used by the disk read process. The following change
removes the instruction in the tape write process which reset the current
disk location.
While regression testing the following change a problem was noticed with
the SEND STATUS command after using the SEND REWRITE command. The number
of megabytes verified displayed was incorrect. The change for this is
also included.
return to index
================================================================================
217950
Software Problem:
----------------
FASTBACK terminated with the following messages when attempting
to restore a 2.5 GB disk:
Number of Files Processed = 16928
Insufficient Disk Space for Restore:
Required Disk Space = 2318.11 MB
Available Disk Space = 2414.38 MB
Software Resolution:
-------------------
When the required disk space exceed 2048 MB the compare of required
disk space to available disk space will fail. The required disk
space is represented by two 31 bit integers, while the available
disk space is computed as two unsigned 32 bit integers. The problem
was resolved by converting the two unsigned 32 bit integers to two
31 bit integers before comparing the required disk space to the
available disk space.
return to index
================================================================================
217970
Software Problem:
----------------
Crash code 153 when marking on a disk.
Software Resolution:
-------------------
While the Command Processor was still in the process of marking on
the disk, another task issued a extended SVC 7 function to obtain
the amount of free space on the same disk. The extended SVC 7 xFF0D
function used a register save area in the ACB structure that was
being used by the mark on logic, thus causing the 153 crash in the
disk mark on logic.
The extended SVC 7 functions of xFF05, xFF06, xFF07, and xFF0D were
updated to include a call to the routine VOLAVAIL to check if the
disk is really marked on and available. If the disk is not available
each of these extended SVC 7 functions will now return an error status
of x03 which indicates the disk volume is not available. This will
prevent these extended SVC 7 functions from using the register save
area in the ACB which currently could be in use by the disk mark on
logic.
return to index
================================================================================
217990
Software Problem:
----------------
Terminal hang problem when using device code 156 or 157. At the
time of the terminal hang the Com Mux returned a status of x08
which means "busy".
Software Resolution:
-------------------
The ITAM async line driver (DASY) would return a status of x801C
which means "illogical status" when receiving a interrupt from a
Com Mux with the status x08 (busy) on a "read command". The
following change ignores a interrupt with a "busy" status and
waits for the next interrupt.
The DASY driver is used by ITAM device codes of 144, 145, 146, 147,
156, 157, and 158.
return to index
================================================================================
218010
Software Problem:
----------------
FASTBACK paused with a FIXED POINT - OVERFLOW ERROR
Software Resolution:
-------------------
The FIXED POINT OVERFLOW ERROR in FASTBACK was caused by the numerator
being to large for a DHR instruction. This will only occur during a
FASTBACK restore operation when the original disk which was backed up
contained a large number of very fragmented files. The following fix
changes the DHR instruction to a DR instruction.
NOTE: If the REV command shows 21801-SCR, then this SCR is already
applied to your task image and one must not apply this patch.
SCR's 21793 and 21795 must be applied before applying this
SCR.
return to index
================================================================================
218030
Software Problem:
----------------
When attempting to restore files from a FASTBACK tape created by
FASTBACK R08-03.2, on occasion FASTBACK pauses with a 8301 tape
read error.
Software Resolution:
-------------------
When support for the IBM 3480 cartridge tape was first introduced
into FASTBACK in R08-03.2, the tape buffer size used was computed
incorrectly. This resulted in FASTBACK writing larger blocks of
data to tape during a BACKUP operation than what was requested. As
a result tapes created by a R08-03.2 FASTBACK may not be able to be
read by another revision of FASTBACK. SCR 21502 was added to the
R09-02.1 version of FASTBACK to allow it to read tapes created by
a R08-03.2 version. Unfortunately the guess as to how much larger
the tape read buffer needed to be to read a R08-03.2 FASTBACK tape
was not quite large enough for all cases. The following change
increases the tape buffer size by an addition 8 bytes.
NOTE: If the REV command shows 21803-SCR, then this SCR is already
applied to you task image and one must not apply this patch.
NOTE: SCR's 21793, 21795, and 21801 must be applied before one
applies this image patch.
return to index
================================================================================
218050
Software Problem:
----------------
The DISPLAY command in COPY32 displays the year as 1900 instead
of 2000.
Software Resolution:
-------------------
The DISPLAY command in COPY32 displays an informational banner
composed of the file/device name, day of the week, month, day,
year, and time. In creating this informational banner COPY32
assumed the year to always be 19xx. With the following change
COPY32 will display the correct four digit year and day of the
week.
return to index
================================================================================
218080
Software Problem:
----------------
FASTBACK paused with a "NON-EXISTENT SEGMENT ERROR (PST) AT 818A10"
error.
Software Resolution:
-------------------
During a RESTORE operation FASTBACK creates a link table of LBA's
(logical block address or logical sector number) from the FAT (file
association table) on the beginning of the tape. This link table is
sorted into ascending order using the LBA location from the original
disk as the sort key. The size of this link table could be allocated
incorrectly by FASTBACK if it is run in a low memory condition, such
as being loaded without any memory segment size increment. The size
of the link table must be a multiple of 12 bytes. The following
change insures the link table size is a multiple of 12 bytes.
NOTE: If the REV command shows 21808-SCR, then this SCR is already
applied to your task image and one must not apply this patch.
SCR's 21793, 21795, 21801, and 21803 must be applied before
applying this SCR.
return to index
================================================================================
218100
Software Problem:
----------------
FASTBACK aborts with "HDR1 Label not found..." when backing up to a
QIC 525 tape drive.
Software Resolution:
-------------------
The QIC 525 SCSI tape drive does not handle tape positioning commands
such as back filemark and back record correctly all the time. As a
result FASTBACK may fail its verify of the tape labels it just wrote.
To verify the header labels FASTBACK issues one back filemark command
and five back record commands. Instead of backing over five header
labels, the QIC 525 drive backed over six header labels, placing the
tape at beginning of tape and the VOL1 label instead of the HDR1
label expected. The following change ignores the VOL1 label if it
was read first.
NOTE: If the REV command shows 21810-SCR, then this SCR is already
applied to your task image and one must not apply this patch.
SCR's 21793, 21795, 21801, 21803, and 21808 must be applied before
applying this SCR.
return to index
================================================================================
218110
Software Problem:
----------------
FASTCOPY may display incorrect contents of a record after the first
mismatch is displayed when performing a copy and verify or verify
only.
Software Resolution:
-------------------
The HEXASC assembly routine in FASTCOPY used a register that was
not saved on the stack. When FASTCOPY displays a record that did
not verify, it calls the routine HEXASC to display the record in
hexadecimal and ASCII. Because the routine HEXASC did not save
all the registers it used, it caused the corruption of the
register containing a buffer pointer which the verify process
used.
return to index
================================================================================
218140
Software Problem:
----------------
DUMPRINT shows a given IOB queued to more than one disk volume.
Software Resolution:
-------------------
IOB's are queued to the leaf for the physical disk drive. When two
or more virtual volumes exist on a physical disk drive, they all
point to the same leaf which is the leaf for the physical disk drive.
DUMPRINT must ensure the IOB being displayed is for the disk volume
it is currently working with.
return to index
================================================================================
218150
Software Problem:
----------------
If BACKUP detects a bad tape when performing its confidence test
at the beginning of a tape, it may not abort or pause as expected
based upon the setting of the ABORT option.
Software Resolution:
-------------------
BACKUP was testing the ABORT flag incorrectly. BACKUP must do
a compare immediate against the ABORT flag of MCT.IABT.
return to index
================================================================================
218160
Software Problem:
----------------
CDS reports a "unsupported special memory operation" error when one
boots a OS sysgened with MEMCHECK for a 3200-850.
Software Resolution:
-------------------
The XSTF instruction used within the memory test routine will causes
problems on a 3200-850 with local memory enabled. This instruction
was modified by the OS to a NOP instruction on S-Bus machines by the
OS from releases R08-02 thru R09-02. Therefore the three XSTF
instructions are being removed.
Source Patch:
------------
Module MCHK.MAC with update MCHKxxxx.UP
Using OSSELECT MCHK
DEL MCK00373
DEL MCK00389,MCK00390
DEL MCK00411,MCK00412
DEL MCK00437,MCK00438
ENDUP
return to index
================================================================================
218180
Software Problem:
----------------
The CTD stream mode does not function if ones user application is
loaded into extended memory. Also on occasion a 153 crash will
occur.
Software Resolution:
-------------------
The CTD driver must enter 32 bit mode prior to accessing the
circular lists for stream mode. The address of the write CCB
must be setup prior to calling the routine SETUPOUT.
return to index
================================================================================
218200
Software Problem:
----------------
Attempting to use CTD stream mode in ones user application with a
device configured on a IOP will cause a 102 crash.
Software Resolution:
-------------------
The CTD driver has code that functions as a preamble to the LFC
ISR (line frequency clock interrupt service routing) to monitor
the write queue within the user application. This ISR attempted
to schedule an ESR for each CTD device which data is present on
the user applications write queue. The ADDSQ macro was used to
do this, which works if the CTD device exists on the CPU, but
will cause a 102 crash if the CTD device is on a IOP.
To schedule the required ESR on the IOP, one must do the following:
1) First schedule a ESR on the CPU from the LFC ISR on the CPU.
2) From the ESR on the CPU schedule an ESR on each IOP in the system.
3) From the ESR on the IOP schedule the stream mode ESR for each CTD
device on the given IOP with data in the applications write queue.
When the CTD driver needs to add a trap to the users task for a
device on a IOP the driver can not use the macro ADDTTSKQ which
calls the routine TMATQ1N. This routine is nonreentrant (NS state)
and can only be executed by the CPU. Therefore the CTD driver
executing on a IOP must send a QCB to the CPU requesting the CPU
to add the trap to the users task queue.
As an additional correction all references to the SPT were change
from using the structure to extern's.
Along with this SCR 21820 one must also include SCR 21818.
return to index
================================================================================