Digital PDFs

XX-3288C-CB

February 1979

7 pages

Original

0.4MB

Document:	Known DEC X11 Bugs
Order Number:	XX-3288C-CB
Revision:	0
Pages:	7
Original Filename:	http://bitsavers.org/pdf/dec/pdp11/xxdp/Known_DEC_X11_Bugs_Feb79.pdf

OCR Text

+---------------+
I dig ita 1 I

I N T E R 0 F FIe E

to: distribution

suoject:

M E MaR AND U M

from:
dept:

22-feb-19
David R. Butenhof
Sys. Diag.
Engineering

ext:
loc/ms:

(22)3-2383

date:

ML21-4/E20

Known DEC/XIl bugs

These are all bugs, of which I am aware, which have been
found to date in the current (Jan 79) release of the new DEC/Xli
package. This status memo will go out semi-regularly to update
all interested parties on current DEC/XlI status from the point
of vie~ of the DEC/Xll Support Group. Feel free to update me on
problems (or solutions) or distribution.
Monitor (CXMONAO)
1)

Failure to turn
on
parity/ECC
CSR·s
due
to
two
auto-increment references to the "same" location. Fixed by
CXMONA3: patch relative locations in monitor module PONDF
(see load maP produced by enf/linker)
264
042731/ 000240
266
000002/ 000240
270
052731/ 012731
276
052731/ 012131
302
042731/ 000240
304
000001/ 000240

Fails to recover from powerfail after relocation, due to
attempted use of KT-mapped routine by the routine which sets
up KT. Fixed by CXMON!3:
patch relative locations in
monitor module KTSET [this patch must be inserted by the
DEC/Xll MOD command after RTE startup]
-

2
34

004767/ 000240
(global)/ 000240
004767/ 000240

(global)/ 000240

Fails to recover after ·C-powerfail-halt-powerup-restart
1000 sequence, due to restart routine assuming KT is set UP
properly. Fixed by CXMONA3: patch absolute RTE locations.
Do not patch if RTE is to run under APT or XXDP chain mode.

The KTSET patch must also be included.
2320
010546/ 000131
2322
012745/ 002712
2706
032767/ 000131
2710
. 0200001 002772

Page 2
2712

176102/ 010546

2714
2716
2720
2722
2724
2726

001031/ 012145
032767/ 001004
000010/ 004737
176074/ (KTSET entry point)
0014Q6/ 012605
010546/ 010546

2730

012745/ 012745
001004/ 001004

2732
2734

2736

004167/ 000131
(global)/ 002326

Modules of type IOMODR or IOMODP (e.g., modules GTAE and
DIAF), which can run only in certain address ranges, will be
dropped if they are unable to run for more than 15 minutes,
and declared "HUNGtI.
Fixed by CXMO'NA3: patcnre.latlve
locations in monitor module SYSCLK.
This will prevent
modules of the specified type from being declared HUNG, even
if they actually are.
Location 1602 protects NBKMOO's from being hung:
if
you have no NBKMOO's, then you can change this location.
Location 1610 protects SBKMOD's, so if you have none of
those, you may change this location.
Inserting a 102000 (into either of the locations) will
protect an IOMODP from being declared HUNG. Inserting a
112000 (into either of the locations) will protect an IOMODR
from being declared hung.
1602
0010001 (102000/112000)
1610
000000/ (102000/112000)

System error in soct.ing module list (SIZPL?
routine), due
to failure to turn off KT after system sizing. Fixed by
CIMONAi: patch relative locations in appropriate SIZPL?
routine as below; only if not under APT or XXDP chain mode.
This patch must be inserted before monitor start upl

S·IZPl.C, SIZP.LD: 1014

SIZPLE:

1016
1020
1114
1116

005737/ 00503'1
000042/ 171572
0014021 000402
005737/ 005031
000042/ 171512

1120

001402/ 000402

Inability to run without at least one BKMOD configured, due
to monitor ignoring possibility of null SKMOD pointer and
attempting to run the imaginary BKMOD at location O.
Fixed
by CXMONA2:
patch absolute RTE locations prior to monitor
RUN command. This patch will cause any configured BKMOD's
to be ig.nor'ed.
3146

4306

0167001 000416
032767/ 000443

Communications device modules (including nY-II,
DP-Il,
00-11, and DUP-l1) experience problems, apparently due to
fact that PIRQ monitor routine 1s long and runs at priority
1, locking out interrupts.
The old monitor works fine.
There is no certain fix at this t1me.

Page 3

In an 11/70, if the actual system size is smaller than the
value stored in the system size register, a reference to
non-existant memory above actual memory but below the size
register limit will trap through vector 114 rather than
vector 4. Since the monitor has not yet set UP the parity
vector when it attempts to size the system, it will halt if
the system size register exceeds actual memory size.
There
is no current fix for this problem.

Since parity trapping Is not enabled during system sizing
and setup of memory, some bad mefllorywords may not be found
until the RTE has relocated over them: which may result in
fatal and undiagnosed errors.

10)

~hen running on
Q-bus (LSI), some modules {particularly
DLAI)wlth t\1iO seperate interruPting units (e.g., receiver
and transmitter) may turn off one in the interrupt service
routine of the other. However, if an interrupt 1s pending
for the other, it has been latched already and will be
serviced.
This results 1n an unexpected trap through 0,
which can have dire consequences. The following patch will
correct the problem:

1050001 000002
0000001 000002

11) If a parity error is encountered, and the error is not in
the first parity/ECC CSR, the routine which searches for the
error will hang Inde.finateiy, as It never increments past
the first CSR.
This will be fixed in the next monitor
revision, but can not be patched. If the RTE hangs in a
relatively tight loop which the linker load map shows to be
in the routine PARERR, this Is your problem.
OPtion Modul~!s
1)

HMDB

A AC and restart may result in errors, due to
module'S failure to initialize SUM? counters to
Zero. Fixed by CXBMDCO.

CRAF

May fail after first pass, claiming incorrect
number of cards, dUe1Co mon.itor allowing modules
to continue iterations after end pass, until all
modules have finished a pass; this means some
cards are read, then RTE relocates and module is
rest~rted
with
deck
partially read.
Fix:
(temporary) do not re-Ioad cards until after
relocation.

DLAI

It seems that this

DNAG

Will loop if 16

module will sometimes fail
after a "'C and RUN. It is rarely necessary to do
this under the new mllni tor, and so the best
temporary avoidance i ) f the problem would be to
type JI"'e" only when te~3ting 1.5 complete.
dey ie !:!s

are

selected,

due

Page 4
using ASR instead of ROR, which results in device
bit map never being zero. Fix by CXDNIG1: patch
relative locations
252
370

0062001 006000
0062001 006000

DVAB

There is a problem with line selection in some
configurations, and monitor may retain control of
CPU long enough for DIll to complete message and
generate MARK signal, Ilhlch module will interpret
as error. It has been reported to hang in SCAN
loop (awaiting completion on all lines of all
DV11's). There are no fixes at present.

FPBC

In systems with IT and low OM! activity, FPBC
will collect errors at a slow rate after several
hours of running, and will eventually be dropped.
The
problem can be avoided by running the
exersizer locked at 20000 (i.e., RUNL command
from eMU) mode).

ICAC

Module does ENOl! call prior to testing, and
skips around ENOL! if current iteration count is
0; which it always will be since the call has
not been made. Fixed by CXICAC1: patch module
location
330
001402/ 000240

KMCB

Fails when read or write buffers overlap a 32k
boundary.
Uncertain whether Is problem with
hardware, KMeB module, or linker.
No present
fix.

LPHB

This module has several problems.

10) MNA-MNE

11)

f~CB

The DEC/Xll MINe modules were not released with
the January 79 release. Corrected by CXMNAAO CXMNEAO.
Module for the Net-It was not released
Jan 79 release. Corrected by CXNCBAO.

with

the

12) PL.1A

This module will not link due to discrepancy
between internal name and file name. Fixed by
CXPLABO.
PLAA will work, however, if
file
XPLAAO.OBJ
is [e-named to XPLABO.OBJ before
linking is attempted.

13) RKAG

Doesn#t error if drive goes off line.
no fix at present.

There

14) RKBD

Prints soft errors
CXRKBEO.

Fixed

15) RLAC

This module hangs if given reasonable amounts of
bus competition (com. modules such as DR-11,

hard

errors.

Page 5

DZ-ll and DUP-ll).
16) RMAC

Does not
CXRMADO.

drop

17) RMBC

Dual port modules communicate between port A and
port B modules via block 0 of the disk. One can
lock the other out from accessing the disk long
enough that it can not update in time; resulting
in one port dropping itself with a message that
the other Is not updating block O. There is no
fix.
This
problem
occurs
only
if
both
controllers are on the same system.

18) RPBH

Reports soft errors (transfer
errors. Fixed by CXRPBIO.

19) RPDa

Documentation error in instructions for running
single port (port A only).
It says to patch
location 2012 from 1740 to 240:
the location
should be 2222.
Also, for running multiple
drives not including drive 0, the following patch
should be included (patch CXRPOS1):
0036
0003001 000700
5010
0177001 000411

20) RXBB

This module has several problems.
It is not
maintained by the DEC/XlI Support Group, however:
problems should be directed to Tom Lawnsby or
Larry Prucha.

21) TMAI

Prints soft errors as hard errors, and will drop
itself if RL is load medium due to using hit test
instead of compare. Also, it will retry badspot
past retry limit. CXTMAIl and CXTMAI2 will fix
the load media problemD CXTMAJO ~ill fix all.

22) TMBI

Gets a single WRITE RETRY EXCEEDED error ~hen
writing from BOT on slave 0 if there are multiple
slaves on the TM03/TM02.
This is fixed by
CXTMBI1.
It also will fail if its read buffer
overlaps a 32k boundary. CXTMBKO will fix both
problems.
.

23) TRAC

Incorrect load media check, and an illegal global
reference to the monitor. It will also hang on a
media badspot.
Fixed
by
patches
CITRACl,
CXTRAC2, CXTRiC3:
0306
0308
0314
0674
0676
0700
0702
1022

load

medium

132731/
0400001
0014241
022771/
0020001
(global)/
001004/
000003/

drive.

122737
000100
001024

000240
000240
000240
000240
000002

errors)

Fixed

hard

Page 6

1056
2276

000003/ 000002
012777/ 000205

standard Linker (CZQOXHO)
Does not change module SR2, SR3, or SR4 at link
Patch linker location
22702 . to
001401

time.

Does not correctly save or print load map (except with.
LINK command switches IMP or IMLP). Patch linker
locations:
21260
0015601 016000
002454/ 016004
23142

NOTE
The DEC/XII linkers (CZQUX?O and
CZQUY?O) are maintained as XXDP
utilities, and not by the DEC/XlI
Support Group.

Page 7

Option modules listed in memo

.. • • • .. .. • • . • • • .. 3
CRAF • • • • .. .. • • • . • .. • .. 3
DLAI • • .. • • • • • • .. • • • .. 3
DNAG .. • • • • .. .. • • • • • • .. 3
DVAB • • • • • • • • • • • • • .. 4
FPBC .. • .. .. • • • • .. • . • .. • 4
ICAC • • .• • • • • • • • • • • . 4
KMCB • • .. • • • . • • • • • .. • 4
LPHB • • • • .. • • .. .. • • .. .. . 4
.. .. • • .. .. • • • .. • • 4
NCB • • • . • .. • .. .. • • .. .. • 4
PLAA . • • .. .. .. • • • • • .. • .. 4
i~.KAG • • • • .. • .. .. • .. • .. .. • 4
RKBD • • • • • • • • • • .. .. .. • 4
RLAC • .. • • • • .. .. • • • • • • 4
RMAC .. • .. • • .. • • • • .. .. . • 5
RMBC • .. • • • • • . • • .. .. • .. 5
ilPBH .. • • • .. • • • .. • • .. . .. 5
RPDB • • • . .. .. .. • • • • • • • 5
RXBB .. • • • .. .. .. • • • .. .. • • 5
TMAI • • • • • • . . • • .. . • • 5
TMBI • • .. • .. . .. .. .. .. .. .. • • 5
TRAC • • .. • • • • • • • .. .. • .. 5

Bt-1D 3 •

)