Google
 

Trailing-Edge - PDP-10 Archives - bb-jr93e-bb - 7,6/ap018/mon703.d18
There is 1 other file named mon703.d18 in the archive. Click here to see a list.
MCO: 13584		Name: RCB		Date:  3-Sep-87:16:16:10


[Symptom]
STOPCD ANFSBA (secondary buffer allocated) at unpredictable times.

[Diagnosis]
When calling GETWDS to allocate a PCB, we call CLNPCB, which doesn't clear
enough locations in the PCB.  Since GETWDS doesn't zero core, this has been a
time-bomb waiting to explode for quite some time.

[Cure]
Clear yet more locations in CLNPCB.

[Comments]
Sigh.

[Keywords]
ANFSBA
Dirty core

[Related MCOs]
None

[Related SPRs]
36045

[MCO status]
Checked

[MCO attributes]
PCO required

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
704	310	NETSER	CLNPCB
703A	


[End of MCO 13584]

MCO: 13588		Name: DPM/LWS		Date: 10-Sep-87:09:43:58


[Symptom]
If a user has a tape mounted and then logs out, PULSAR
complains that the drive is still assigned to the job logging out
when QUASAR tells him to unload the drive.

[Diagnosis]
In the above situation, QUASAR gets 2 ICPF msgs from the
monitor - a LOGOUT message and then a DISMOUNT message. During
the processing of the LOGOUT message, QUASAR tells PULSAR to unload
the tape drive, but PULSAR says "no, the drive's still assigned".
The problem is that the LOGOUT message is sent before the drive
is deassigned and QUASAR and PULSAR both execute before DEASTY is
called in JOBKL. The subsequent DISMOUNT message is ignored by
QUASAR since its work was done processing the LOGOUT message.

[Cure]
Move call to IPCLGO after call to DEASTY.

[Comments]

[Keywords]
DISMOUNT
LOGOUT

[Related MCOs]
None

[Related SPRs]
35760

[MCO status]
None

[MCO attributes]
PCO required

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
704	311	COMCON	JOBKL1

703A	


[End of MCO 13588]

MCO: 13590		Name: KDO		Date: 11-Sep-87:00:45:28


[Symptom]
LATNSC's, IME's, UIL's

[Diagnosis]
Accumulator CB is not initialized if MAXIMUM ACTIVE CIRCUITS
is exceeded.

[Cure]
Include TCO 7234 in 7.03 as per SPR 20-21077 of 27-Jan-1987.

[Comments]
Paul Mead says that this has happened at R and R.
The code is already in 7.04 (common sources).

Yes, Paul the TOPS-10 and TOPS-20 groups <<DO>> talk to each other.
But one group speaks only in arcane Murphy macros, so we don't always
understand each other.

[Keywords]

[Related MCOs]
None

[Related SPRs]
None

[MCO status]
None

[MCO attributes]
None

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
703A		LATSER	LARSCH,HMSTRT


[End of MCO 13590]

MCO: 13591		Name: KDO		Date: 11-Sep-87:02:18:40


[Symptom]
The official object type (.OBDQS=^D66) is not defined in UUOSYM.

[Diagnosis]
YES

[Cure]
PLEASE

[Comments]

[Keywords]

[Related MCOs]
None

[Related SPRs]
None

[MCO status]
None

[MCO attributes]
UUOSYM change

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
704	311	UUOSYM

703A		UUOSYM


[End of MCO 13591]

MCO: 13610		Name: TL		Date: 24-Sep-87:21:13:45


[Symptom]
NML does not behave according to the network management spec.

[Diagnosis]
Tarl.

The spec says that if a user directs a command to a node using a numeric
specifier (it's address, as in 123, or 7.123), the connection is to be 
made to the address.  

Because DECnet-36 does not allow connections to a numeric address, NML
translates the address to a node name, then connects.  This fails if
the address has no name defined (by SET NODE n NAME foo).  For example,
this prevents a system manager from saying TELL 7.123 SHOW EXECUTOR SUMMARY,
which would tell him (her) the proper name to SET.  This is unreasonable.
Further, if a system wants to obtain a network database from a name
server, that name must be defined before it can connect.  But, you need
the name (from the server) in order to connect.  Catch-36.

The original rationale for not allowing numeric connects was that a user
should not be able to see or hard-code a node number, as this is contrary
to the Session Control layer specification.  In order to preserve upward
compatibility of DNA, it is extremely desirable that the Session Control
specification's intent be upheld.  The format of a node number has changed
before, and will again.  User applications really should never know.

However, our NML implementation uses Session Control to establish its
connections.  That's all there is.  Catch-1296. (Barb, that's 36^2.)

[Cure]
Tim.

Add an undocumented, hard-to-use escape clause to Session Control.
It works as follows:

[text deleted]

This interface is NOT TO BE DOCUMENTED OR SUPPORTED FOR CUSTOMERS.  To do
so would subvert the DECnet architecture.

[Comments]
I didn't mean for it to be quite so arcane - it just turned out
to be convenient.  We've discussed this for quite some time, but no one else
had the nerve do the work.

If I ever find non-network management applications using this interface,
the culprit will be reported to the Better Networks Bureau, who has the
authority to chop fingers.  Or, I'll have to make it privileged.

But, if it's used in the spirit in which it was implemented, it's a
very cheap way to allow NML to be fixed.

The DOC ATTENTION is to make sure that this isn't documented by mistake.

SCCd

[Keywords]
NML
NODE ADDRESSING
WATCH YOUR FINGERS

[Related MCOs]
None

[Related SPRs]
None

[MCO status]
Restricted distribution

[MCO attributes]
New development MCO
Documentation change

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
703A		SCMUUO	SCCCBL,SCCCB1
704	


[End of MCO 13610]

MCO: 13614		Name: RCB		Date: 29-Sep-87:04:41:55


[Symptom]
Yet another problem with entry vectors:
	.R DDT/U:1
	?No start address
	.

[Diagnosis]
Not doing the right thing with a JOBDAT-in-section-N sort of entry vector.

[Cure]
Handle it correctly.

[Comments]
So what's a millstone, anyway?

[Keywords]
ENTVC.
RUN/USE

[Related MCOs]
13478, 13269

[Related SPRs]
None

[MCO status]
Checked

[MCO attributes]
None

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
704	313	COMCON	REENTE,SNOAR1,GSADDR,URUN3
703A	


[End of MCO 13614]

MCO: 13616		Name: RCB		Date: 29-Sep-87:06:11:18


[Symptom]
When typing the first character of a switch sequence to CTHNRT, doubling it
is supposed to get one instance through SCNSER.  However, it takes another
(different) character after that before it works.

[Diagnosis]
Not enough special-casing, leading to considering the second SW1 character to
be just as interesting and special as the first was.

[Cure]
Bypass a bit more of the normal RECINT logic in DCSSWI.

[Comments]
I finally got bugged enough....

[Keywords]
CTERM
 .TOSWI

[Related MCOs]
None

[Related SPRs]
None

[MCO status]
Checked

[MCO attributes]
None

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
704	313	SCNSER	DCSSWI
703A	


[End of MCO 13616]

MCO: 13646		Name: JMF		Date: 13-Oct-87:07:05:42


[Symptom]

RUN BIG[31,6061,FOR]
%mumble mumble mumble ...

EXIT
 .DDT
STOPCD KAF

[Diagnosis]
PAGLOW takes to long. Also, PAGOUT tries to page out ABZ pages.

[Cure]
PUSHJ P,SCDCHK, have PAGOUT also check PA.OUT when trying to decide
if a page can be paged out.

[Comments]
Frankly, I think its a bug that page accessibility doesn't turn on
PA.CPO for an ABZ page. Afterall, it can't be.

[Keywords]
KAF

[Related MCOs]
None

[Related SPRs]
None

[MCO status]
Restricted distribution

[MCO attributes]
New development MCO
PCO required

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
704	315	COMCON	PAGOU0,PGLOW4
703A	


[End of MCO 13646]

MCO: 13675		Name: DPM		Date:  2-Nov-87:03:40:03


[Symptom]
AP 17 doesn't.

[Diagnosis]
Patch to prevent IMEs during reel switches for 704 didn't make
it into 703.

[Cure]
 ...

[Comments]

[Keywords]
REEL SWITCH

[Related MCOs]
None

[Related SPRs]
None

[MCO status]
Restricted distribution

[MCO attributes]
New development MCO

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
703A		TAPUUO	TPMLSU


[End of MCO 13675]

MCO: 13701		Name: JAD		Date: 12-Nov-87:13:11:32


[Symptom]
Possible IME from FILOP. if user specifies an illegal "returned path"
block or "file specfication" block.

[Diagnosis]
Call to check LOOKUP/ENTER/RENAME block (FCLERB) steps on T4;
FOPEN0 expects T4 to still contain the length of the FILOP. UUO
argument block.

[Cure]
HRRZ T4,FOPFLG at FOPEN0

[Comments]
Hey, who'd DELIBERATELY try to crash the system, huh?

[Keywords]
FCLERB
IME

[Related MCOs]
None

[Related SPRs]
36066

[MCO status]
None

[MCO attributes]
New development MCO
PCO required

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
704	322	UUOCON	FOPEN0
703A	


[End of MCO 13701]

MCO: 13702		Name: JAD		Date: 12-Nov-87:14:14:46


[Symptom]
None observed here, but possible death doing a 96. page transfer
on an RH10/DF10C disk.

[Diagnosis]
The monitor checks for a GOTO CCW by testing whether the word
count field in the LH of the CCW is non-zero.  The mask used is
the tests is 577777.  Bit 1 (200000) is excluded since that is
set in an RH20 JUMP CCW.  Prior to MCO 12454, MAPIO went out of
its way to avoid generating an IOWD whose LH containing 200000.
12454 changed the test to only apply to internal channels.  This
can allow MAPIO to generate an IOWD which looks like a GOTO CCW.
SETL15 in FILIO can wander off into the weeds chasing user buffers
out of cache when it stumbles across one of these CCWs.

[Cure]
Delete the test on internal channels, and never generate a
CCW of the form under discussion.  Comment this restriction in
MAPIO (original SPR complained of reduced performance).

[Comments]
Gee, if we had the RH10/DF10C connected I could actually TEST
my hypothesis, but since we don't, trust my instincts and Jeff's
inquisitive nature.

[Keywords]
MAPIO

[Related MCOs]
12454

[Related SPRs]
35336

[MCO status]
None

[MCO attributes]
New development MCO
KL10 only

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
704	322	APRSER	MAPIO3

703A		KLSER


[End of MCO 13702]

MCO: 13720		Name: RJF		Date: 30-Nov-87:10:31:35


[Symptom]
DIEBAD STOPCDs.

[Diagnosis]
When a non-logged-in command gets an error, the stack is
not always restored correctly on return from the command.

[Cure]
Retrofit 704 COMCON code into 703 to restore the stack from .CPISF
on return from a command error.

[Comments]

[Keywords]
DIEBAD

[Related MCOs]
None

[Related SPRs]
None

[MCO status]
None

[MCO attributes]
None

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
703A		COMCON


[End of MCO 13720]

MCO: 13735		Name: JAD/RJF		Date:  4-Dec-87:10:13:09


[Symptom]
DAU stopcode kinda maybe after a TMDELI stopcode kinda maybe after
another CPU gets a KAF if a cosmic ray comes in the right direction.

[Diagnosis]
For INFORM stopcodes, DIE types out using the "normal" SCNSER
TTY output routines.  These just happen to do "SCNOFF"s to keep
the TTY chunks intact.  If someone already owns the SCNSER lock,
DIE will hang on the interlock, and maybe get a KAF, and maybe
not, but it will wind up eventually croaking with a DAU stop or
some reasonable facsimile.

[Cure]
Make the SCNSER interlock multi-level, but to keep Spider
happy, only allow recursion when DIEFLG is non-zero, which
indicates (probably) this CPU is in the middle of a stopcode.

[Comments]
Unprintable, mostly; what a hassle to prevent the DTEs from
typing "%DECsystem-10 not running".  Wonder why we're getting
those TMDELIs in the first place?  Guess we'll never know . . .

Ya know, maybe we could use protocol pause in this case, but
since your average stopcode takes a few minutes to type out,
it might take too long for the FE.  Comments, DPM?

[Keywords]
DAU
SCNSER LOCK

[Related MCOs]
None

[Related SPRs]
None

[MCO status]
None

[MCO attributes]
Multi CPU only

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
704	325	COMMON	LOKSCI,ULKSCI


[End of MCO 13735]

MCO: 13765		Name: KDO		Date: 23-Dec-87:14:30:14


[Symptom]
DECnet End Nodes don't work.

[Diagnosis]
We are dropping adjacency to the designated router.

[Cure]
Fix RHMASE, which parses Hello messages for end nodes, so that it takes the
Hello Timer.  There is code to do this in RHMASR, which does the parsing for
routing nodes.  But it is missing for RHMASE.

Also, unwanted multicast messages are using up the DECnet receive buffers.  The
result is lots of User Buffer Unavailable errors.  Fix ETHSER so that it returns
error code UNIMA% (Invalid Mulicast).  Then fix DNADLL to post a new receive
buffer when one of these events occurs.

[Comments]
Many thanks to Brian Lilja at Colorado for this timely QAR.

[Keywords]

[Related MCOs]
None

[Related SPRs]
None

[MCO status]
None

[MCO attributes]
None

[Validity]

Monitor	 Load	Module	 Tags
-------	------	------	------
703A		ROUTER	RHMASE
		DNADLL	NIIRCE
		ETHSER	ETCRDG

704	330	ROUTER	RHMASE
		DNADLL	NIIRCE
		ETHSER	ETCRDG


[End of MCO 13765]