Google
 

Trailing-Edge - PDP-10 Archives - bb-bt99g-bb - t1s702.d13
There are no other files named t1s702.d13 in the archive.
                 EDIT DESCRIPTIONS FOR TOPS-10-KS-V702                          
  
  
********************************************************************************
  
  
                             EDIT 11030  FOR 702
  
  
  
  
DMA NXMS, MEMORY OVERRUNS
  
[SYMPTOM]
  
     DMA NXMs and memory overruns occur  in  a  multi-CPU  environment
when using RP07s.
  
[DIAGNOSIS]
  
     Interlocks can tie up memory long enough doing  read-modify-write
instructions  to cause DMA NXMs when multiple CPUs and I/O devices are
contending for the memory.
  
[CURE]
  
     Give other CPUs and devices a chance to get at memory.  Don't try
to  get  an  interlock  unless  its potentially available.  Do this by
replacing the "stall" instruction (IMULI P,1) in the spin  locks  with
one  which causes an additional memory cycle (SKIPGE).  However, since
this instruction only causes a read and not a read-modify-write cycle,
other CPUs or devices will a stand better chance of accessing memory.
  
********************************************************************************
  
  
                             EDIT 12297  FOR 702
  
[SYMPTOM]
  
     IME from GETNMB in FILUUO.
  
[DIAGNOSIS]
  
     GETNMB has followed a zero NMB pointer in an ACC,  thus  trying  to
use  the left half of the stack pointer as an NMB/ACC pointer.  However,
the associated crash dump shows no such zero  pointer,  just  a  healthy
ACC/NMB  chain instead.  Another job apparently has delinked an ACC from
the chain while the current (crashing) job was searching it.
  
[CURE]
  
     Get the CB resource around the call to GETNMB.
********************************************************************************
  
  
                             EDIT 12346  FOR 702
  
[SYMPTOM]
If a NSP. "Read normal data" function is performed with a byte
pointer pointing to a write-protected page, an IME occurs.
  
[DIAGNOSIS]
The user's buffer is checked for readability, but not for writability.
When DECnet tries to copy data to the user's buffer, the system
dies with an IME.
  
[CURE]
While checking the buffer for readability, also check it for
writability.  If it is not writable, give the user an address
error.
********************************************************************************
  
  
                             EDIT 12348  FOR 702
  
  
  
  
[SYMPTOM]
It is possible for a TRMOP.  UUO to trash a few user memory
locations or use garbage data if the user has specified too
short an argument list.
  
[DIAGNOSIS]
The TRMOP.  action functions (those with a function code
which is less than 1000) do not check to see if the user has
supplied an argument list which is long enough to perform
the function.  These functions typically use the third value
from an argument list, even if the user only gave a 2 word
list.  If the function wants to write into the third word
(examples are .TOMGS, .TOSMS, .TOCLR), that word is written
without checking to see if it is part of the user's argument
list.
  
[CURE]
Add a field in table TOPTB0 in SCNSER.  This field contains
the minimum argument list length for each of the action
functions which requires more than 2 words.  If the user has
supplied too short a list, return error TOADB% ("Illegal
argument list address or length").
********************************************************************************
  
  
                             EDIT 12450  FOR 702
  
[SYMPTOM]
  
     MIC macros and batch jobs hang needlessly.
For example, the following sequence fails:
  
        .GET xxx
        .CSTART
        .ATTACH yyy
  
  
[DIAGNOSIS]
  
     Testing RUN bit even though we don't need to.
  
  
[CURE]
  
     Don't say user can't do input at monitor level just because  he's
running.  The line will be quite happy to take input.
  
  
********************************************************************************
  
  
                             EDIT 12622  FOR 702
  
[SYMPTOM]
  
	Can't send useful information to ACTDAE via QUEUE. UUO.  Have to
use regular IPCF (with all its asociated headaches) instead.
  
[DIAGNOSIS]
  
	Monitor was being to helpful in protecting QUASAR from users, no
matter how much it got in the way of other components, or how little
QUASAR needed the protectionism.
  
	Immediate mode arguments mistakenly don't allow more than one
word in the sub-block.
  
[CURE]
  
	Remove the restriction on the range of block codes from IPCSER.
  
********************************************************************************
  
  
                             EDIT 12735  FOR 702
  
[SYMPTOM]
  
	GALAXY version 5 LPTSPL doesn't run under a 7.02 monitor.  LPTSPL
claims printers do not exist.
  
[DIAGNOSIS]
  
	LPT device drivers do not set the printer class in the hardware
characteristics word of the DDB.
  
[CURE]
  
	Make sure the printer class is setup correctly.
  
********************************************************************************
  
  
                             EDIT 12807  FOR 702
  
  
  
  
[SYMPTOM]
  
     Stopcode PSF in IPCSER when attempting a page mode IPCF send to a
non-existant job.
  
[DIAGNOSIS]
  
     If the intended receiver of a page mode IPCF send  logs  off  the
system while the IPCFS.  UUO is being processed, the call to VALPID at
IPCS9A correctly fails.  At this point, J contains the job  number  of
the  job  which just logged off.  At IPCS11, we then try to insert the
page in the non-existant job's page map resulting in a PSF stopcode.
  
[CURE]
  
     Reset J to contain the sender's job number if the call to  VALPID
fails.
  
********************************************************************************
  
  
  
END OF  TOPS-10-KS-V702