Google
 

Trailing-Edge - PDP-10 Archives - BB-H311B-RM - rsx20f-swskit/documentation/rsx20f-revision-history.mem
There are no other files named rsx20f-revision-history.mem in the archive.


  +---------------+
  | d i g i t a l |    I n t e r o f f i c e   M e m o r a n d u m
  +---------------+





  Subj: Current RSX-20F New Features


     Attached is a brief description of the new features and history  of
     RSX-20F  since  the  last  release.   We  would  like to attempt to
     release version 13-41 to both TOPS-10/7.00 and TOPS-20/4 if  it  is
     feasable.   Towards  this  end we will no longer be putting any new
     features  into  RSX-20F   for   these   releases   without   strong
     justification.


                      RSX-20F New System Features


     EXECUTIVE

       0. Features prior to V12-45
             - Security improvements to modem control
             - Terminal driver improvements
       1. Local processing of XON/XOFF on line enabled basis
       2. FE status block created
       3. Lines are shut down due to framing errors for 10 seconds
       4. Lines shut down are logged to SYSERR
       5. EBUS parity errors are retried and snapshot
       6. Power fail recovery improvements
       7. Autobaud to 1200 support
       8. I/O page saved in dump
       9. Bug fixes
             - interrupt enable lost on DM11-BB
             - KL10 memory written during protocol pause
             - boot task looping on protocol pause timeout
             - characters thrown away during protocol pause
             - sendall race condition

     KLINIT

       0. Features prior to V12-45
             - MOS memory support
             - Record if bootstrap loaded in configuration file
             - Selectively write configuration file
             - Execute KL address 71 on keep alive failure
       1. Tracking of hardware operations
       2. Reverse memory configuration
       3. CRAM/DRAM error reporting on parity errors
       4. Force memory reconfiguration for MOS
       5. Enviromental reporting or hardware
       6. Bug fixes
             - memory not configured correctly
             - configuration file "WRITTEN" not "ALTERED"
             - parity stops incorrectly initialized

     KLERR

       1. DTE-20 register print out
       2. FM registers sweep on FM parity errors
       3. Bug fixes
             - KL forced into halt loop corrupts snapshot

     PARSER

       0. Features prior to V12-45
             - Command to manipulate RETRY switch (XCT 71 flag)
       1. CLEAR DATE made a legal command


                          RSX-20F Edit History



     22-Jan-79

          V12-45    RSX20F    Executive
          V11-02    KLI       KL initialization program
          V2-02     KLE       KL error snapshot program
          V4-33     PAR       PARSER program

     This is the base level that was shipped with the first  field  test
     of TOPS-10/1090/1091 version 7.00 field test.


     16-Feb-79

          V12-46    RSX20F    Executive
          V2-03     KLE       KL error snapshot program

     The  exec  was  modified  to  do  local  processing   of   incoming
     XON/XOFF'S.   A  new  protocol  message  was  defined that the TOPS
     monitor  would  send  to  RSX20F  to  tell  it  to  start  or  stop
     recognizing  XON/XOFF'S  on a particular line, otherwise it acts as
     it always has (ie.  downward compatable with old  monitors).   Once
     local  recognition has been enabled for a line the old technique of
     sending the XON/XOFF through is  done  in  addition  to  immediatly
     processing the control character.  [TCO 4.2181]

     As part of the KL RAMP mini-project KLERR was modified to type  out
     the  DTE-20  register  contents when it is invoked and to just stop
     the KL clock rather than force it in  to  the  halt  loop  when  it
     starts up.  [TCO 4.2203]

     This is the initial TOPS-20 release 4 front end load test version.


     28-Feb-79

          V12-47    RSX20F    Executive

     Also as part of the KL RAMP mini-project the FE  status  block  was
     created.  The status block contains all of the data and status used
     by the executive while operating.  The  block  starts  at  a  fixed
     location  (1000)  and its first location contains its length.  [TCO
     4.2212]


          V12-50    RSX20F    Executive

     This release is a retroactive release.   At  the  time  of  release
     13-16  it  was necessary to take release 12-47 and put the patch in
     that fixed DH11'S.  This was necessary so we could  release  a  new
     front  end  with the second field test tape of TOPS-10 version 7.00
     that did not have all the untested KL mini  RAMP  project  changes.


     For more information on the patch for the DH11 see info on V13-16.

     This release will be shipped with the second field  test  tape  for
     TOPS-10 version 7.00.


          V13-01    RSX20F    Executive

     The terminal service handler was modified  to  check  the  DM-11/BB
     interrupt  enable  bit  during  time-out  servicing.   There  was a
     complaint that the interrupt enable could be lost  thus  preventing
     DM-11/BB datasets from being answered.  [TCO 4.2206]


          V13-02    RSX20F    Executive

     Several bugs dealing with the protocol pause state were fixed.  The
     boot  task  would loop on protocol pause timeout and the exec could
     write KL10 memory during protocol pause.  [TCO 4.2207 and 4.2208]


     13-Mar-79

          V13-03    RSX20F    Executive
          V11-03    KLI       KL initialization program

     Due to problems with open lines running wild on 2102  and  crashing
     the  front  end  a  change  was  made to the terminal service to no
     longer throw framing errors away but rather count them per line and
     shut  the  line  down  after  four  consecutive  framing errors are
     detected on any line till the next timeout servicing done  (max  10
     sec.).  [TCO 4.2205]

     As an aid to field service the internal debugging tracking aid  was
     permanently turned on in released versions of KLINIT.  [TCO 4.2210]


          V13-04    RSX20F    Executive

     There was a complaint that the front end  should  never  conciously
     throw away good data.  The particular case was that during protocol
     pause the front  end  through  away  characters  it  recieved  from
     terminals  with  out notifying the user.  The terminal servicer was
     changed to now queue the characters to the TOPS monitor  until  the
     free  pool  ran out in which case the front end would crash with an
     error code of B02.  [TCO 4.2211]


     20-Mar-79

          V13-05    RSX20F    Executive
          V11-04    KLI       KL initialization program
          V2-04     KLE       KL error snapshot program

     It was determined that as an aid to field service in  finding  open


     lines that when a line is shut down due to framing errors this fact
     should be logged and sent to SYSERR.  The error logging routine was
     modified  to  send the device address and the contents of the first
     two registers (control status register and next received  character
     register) of the device when a line is shut down.  [TCO 4.2245]

     As part of the KL RAMP mini-project the functionality of KLINIT was
     modified  to  allow an operator bringing up the KL to configure the
     memory controllers in reverse and then save this  configuration  in
     the  configuration  file.   Also  because  of  a  bug in the memory
     configurator compatability package the package has been updated  so
     systems  with  a  model A processor, external memory, and microcode
     212 or higher configure memory correctly.  [TCO 4.2204 and 4.2218]

     When  KLERR  is  run  it  dumps  the  DTE  registers  on  the  CTY.
     Unfortunately this is done after KLERR runs its initialization code
     which affects the contents  of  the  DTE  registers  and  therefore
     invalidates  the CTY dump.  KLE has been modified to not initialize
     itself before the registers are dumped.


          V13-06    RSX20F    Executive

     In some cases an open line  will  run  without  generating  framing
     errors.   Since this is undetectable as an open line RSX20F accepts
     the characters as legal data.  Since the data is coming  in  rather
     rapidly  the  free  pool  is quickly exhausted and the algorithm at
     this point is to slow the DTE down so as not to  recieve  any  more
     data from the 10.  This was a reasonable thing to do because it was
     usually the case that the free pool was used up by  the  TO11Q  and
     slowing  down  the  DTE  allowed this queue to run down.  With this
     open line problem if the DTE is slowed down few characters are sent
     to  the  10  and  the front end quickly crashes to due an exhausted
     free pool.  Now the DTE is never slowed down.  [TCO 4.2216]


          V13-07    RSX20F    Executive
          V11-05    KLI       KL initialization program

     As  part  of  the  KL  RAMP  mini-project  in  order  to  get  more
     information  on  EBUS parity errors a retry of the guilty operation
     is performed in an attempt to get the good data of  the  operation.
     The  DTE  registers  are saved in a buffer taken from the free pool
     before the retry for subsequent typeout  by  KLERR  and  after  the
     retry  the  contents  of the DTE registers are put into this buffer
     also.  The purpose for taking the buffer from the free pool  is  to
     make  it  easier  in the future to just log the error to SYSERR and
     just continue if the retry succeeds.   Currently  the  system  just
     reloads the KL.  [TCO 4.2217]

     As part of the KL RAMP mini-project KLI is  to  print  an  expanded
     CRAM/DRAM  error  report.  Now the microcode is verified and when a
     difference is detected the address is  typed,  the  good  value  is
     typed,  the  actual  contents are typed, and these values XOR'd are
     typed.  An attempt is then made to write the correct value  to  the


     CRAM/DRAM.  [TCO 4.2209]


          V13-10    RSX20F    Executive

     In order to make the front end more defensive two  new  stop  codes
     were  put in the exec.  The first is SAQ which occurs if a send all
     count is loaded as negative and the other is SAI  which  occurs  if
     the send all count goes negative at interrupt level.


     22-Mar-79

          V13-11    RSX20F    Executive

     It was discovered that  the  assembly  time  macros  for  the  exec
     generated  DH-11  vectors  for non-existent DH-11's.  These vectors
     were removed and replaced with vectors to the  non-existent  device
     interrupt routines.


          V13-12    RSX20F    Executive

     The terminal driver interrupt service was  modified  to  enter  the
     framing  error  log  information  into  a  queue  because the error
     logging routine could not be run at interrupt level 6.


          V13-13    RSX20F    Executive

     Inorder to reclaim some of the space that  all  the  new  code  has
     taken  up  the  code  used  to declare a significant event has been
     changed to call a subroutine rather than do it in line.


          V13-14    RSX20F    Executive

     The global status word .PRPSE (protocol pause state) was moved to a
     location in the FE status block in low core.


     26-Mar-79

          V13-15    RSX20F    Executive

     The framing error queue entries for logging are now  transfered  to
     the  10  during  the terminal time-out routine at task level.  This
     prevents deadlocking at interrupt level attempting to transfer  the
     data.


     27-Mar-79

          V13-16    RSX20F    Executive

     A bug in the DH-11 hardware was found that caused the front end  to
     be  reloaded  because  of keep-alive failure was programmed around.
     The problem was that when  a  line  was  shut  down  due  to  local
     processing of XON/XOFF's the silo of the DH-11 is not properly kept
     so the front end has trouble emptying it at interrupt  level.   The
     solution was to not dump the silo completely at interrupt level but
     just to remove one character, service it, return, then wait for the
     next  interrupt  to  get  the  next  character  in  the silo.  This
     solution adds some overhead in processing characters  but  it  gets
     around the DH-11 problem.  [TCO 4.2244]


          V13-17    RSX20F    Executive

     Several typographical errors were found in the  EBUS  parity  error
     retry code and fixed.


     28-Mar-79

          V13-20    RSX20F    Executive
          V2-05     KLE       KL error snapshot program

     A bug in the terminal framing error  logging  code  was  found  and
     fixed.

     As a result of the EBUS parity error retry the original contents of
     the  DTE  registers are lost.  Before the retry though the contents
     are saved in a buffer taken from the free pool.  The new KLERR  now
     runs  and  determines  where to get the values for the DTE type out
     when it is a EBUS parity error and whether the retry  succeeded  or
     not.


     3-Apr-79

          V13-21    RSX20F    Executive

     It was found that during error logging  it  was  not  necessary  to
     inhibit  interrupts  because  the  routine  was  always  called  at
     interrupt level with interrupts inhibitted.   Space  was  saved  by
     removing the inhibitting.


          V13-22    RSX20F    Executive

     Inorder to conserve space the 50 byte card  reader  overrun  buffer
     was changed to 2 bytes.


     4-Apr-79

          V13-23    RSX20F    Executive

     A bug was found in the DH11 framing error  logging  code.   When  a
     call is made to the routine to allocate a buffer for logging it was
     assumed that the routine stored the size of the returned buffer  in
     the  second  word  of the buffer.  As was the case the routine does
     not store the size so when the buffer is returned to the free space
     area  problems  occur.   The fix was to store the requested size in
     the second word of the buffer.


          V13-24    RSX20F    Executive

     As with the previous bug fix the EBUS parity error logging  routine
     had to be fixed.


     6-Apr-79

          V13-25    RSX20F    Executive

     When the front end reloads the KL a possible race conditions exists
     with  XOFF/XON  enabled  processing  where  the KL doesn't know the
     state of the line and assumes  it  is  disabled.   To  prevent  any
     misunderstanding  the enable bit is cleared by SETSPD when the line
     speeds are sent to the KL.


          V13-26    RSX20F    Executive
          V12-01    KLI       KL initialization program
          V2-06     KLE       KL error snapshot program

     Similar to the previous bug fix the enable bit is cleared during  a
     power up after a power fail.

     A new output format of the DTE  register  type  out  in  KLERR  was
     implemented.   This  output  is  a  more  condensed  version of the
     previous release.

     The final format of the CRAM/DRAM type out for the command FIX  and
     RAM  errors in KLINIT was implemented.  The type out corresponds to
     the type out done by KLDCP for the same functionality.  It was also
     discovered  that when a power fail occurs KLI attempts to "FIX" all
     the RAM locations.  This takes about 30 minutes  and  20  pages  of
     paper.  An edit was made so that on auto reload if the microcode is
     corrupted it is just reloaded if there was a power failure  instead
     of fixing all the RAM.

     This release will be sent with the first field test tape of TOPS-20
     release  4.   It  will also be shipped with the third field tape of
     TOPS-10 release 7.00.


          V13-27    RSX20F    Executive
          V5-00     PAR       PARSER program
          V2-00     RED       Device redirect program
          V2-00     REX       Device redirect program ignoring protocols

     While attempting to find a problem booting a system on 1042 it  was
     found  that  in the power fail recovery code when the eleven memory
     parity registers are initialized the priority could be lowered from
     7 to 0.  As a result of this discovery the power fail recovery code
     was  reorganized  to  not  1.)  save  the  registers,  restore  the
     registers,  and  then smash 2 registers, 2.) start the clock before
     initializing the memory parity registers, and  3.)  not  allow  the
     priority to be lowered during recovery.  [TCO 4.2249]

     The new version of the PARSER has no new functionality but has been
     deoverlayed so as to run faster from floppies and DECtapes.

     RED and REX were modified because it was found that redirecting SY:
     to  DT:   the  search  for F11ACP took to long so that the redirect
     timed-out and then failed.  There had been reports from  the  field
     of  this  problem  but  it was intermittent so it was assumed to be
     hardware problems but the case was that the time-out  loop  was  on
     the border line of success and failure.  [TCO 4.2250]


          V13-30    RSX20F    Executive

     Due to SAI stopcodes occuring on 2137 and the  hardware  apparently
     being  in  good condition the SENDALL facility was investigated for
     possible bugs.  A race condition was found where  if  sendalls  are
     supressed  while  a sendall is being queued up for a line and being
     sent to other lines the sendall count  can  be  decremented  twice,
     once during the queue up and once at interrupt level.  The code was
     changed so as to prevent this.  [TCO 4.2246]


     1-June-79

          V13-31    RSX20F    Executive
          V12-02    KLI       KL initialization program
          V5-01     PAR       PARSER program

     During a demostration of TOPS-10/7.00 SMP it  was  found  that  the
     KLINIK  line would not work in remote mode.  The problem turned out
     to be an editting error that occurred during the  terminal  handler
     reorganization.  The problem was patchable on V13-26.  The terminal
     input service was changed to dump the silo if the  input  character
     is not an xoff, this is a slight change from V13-16.  [TCO 4.2265]

     KLI was modified to type "CONFIGURATION FILE  WRITTEN"  instead  of
     "ALTERED"  because the file is always written, whether it exists or
     was used or not.  This was  requested  by  users  so  as  to  avoid
     confusion in semantics.  [TCO 4.2251]

     The CLEAR DATE command in the PARSER  was  modified  such  that  in


     maintenance  mode it would clear the date validity flag then prompt
     for a new date.

     This version will be shipped with the second  field  test  tape  of
     TOPS-20 release 4.  It will also go with the fourth field test tape
     of TOPS-10 version 7.00.


          V13-32    RSX20F    Executive

     It was found that it was possible that we  might  not  service  our
     keep-alive  fast  enough  because of I/O processing.  Therefore the
     keep-alive processing was moved  before  the  IODONE  servicing  to
     prevent keep-alive-cease.  [TCO 4.2247]


          V13-33    RSX20F    Executive

     Inorder to make the status block the same for all systems a pad  of
     10  words  was added to the DECtape driver to make it the same size
     as the floppy driver.  [TCO 4.2263]


     10-June-79

          V13-34    RSX20F    Executive

     As a result of a temporary carrier loss  (ie.   carrier  loss  then
     carrier  on due to noisy lines) the line is put into auto-baud wait
     when the carrier on is detected.  Unfortunately the line  speed  is
     not  reset  back to the base speed.  When a character comes at what
     ever speed the line was at it is compared to  the  auto-baud  table
     which  assumes  that the character came in at the base speed.  This
     causes a line to hang.  Now on  carrier  loss  the  line  speed  is
     resynced.  [TCO 4.2264]


          V13-35    RSX20F    Executive
          V12-03    KLI       KL initialization program
          V13-35    SAV       RSX-20F core image save program

     This version supercedes V13-34 in the way it handles carrier  loss.
     When  carrier  loss now occurs if carrier on is received before the
     time-out period then the line is not touched(ie.   the  user  never
     knows  that anything has happened).  This is done by adding another
     status bit in STSW1 that says the line is already connected so if a
     carrier on occurs don't auto-baud the line.  [TCO 4.2284]

     A QAR was received that stated that KLI didn't set the parity stops
     correctly  on  CPU1  of a 1099.  It was found that the parity stops
     are not set up if the KL bootstrap is not loaded, the parity  stops
     that are set up are due to a bug in the memory boot.  The fix is to
     do the final hardware  initialization  regardless  of  whether  the
     bootstrap  is loaded.  This QAR also brought to light the fact that
     on CPU1 after the load bootstrap  question  is  asked  some  people


     abort KLI by typing ^Z.  When this is done we can not guarentee the
     status of the KL  hardware.   The  KLI  dialogue  must  be  run  to
     completion  to  get  the KL into a known runnable state.  Also on a
     manual reload via the disk switch  the  default  operation  on  the
     microcode  is  to reload it, this a change from verifying it.  [TCO
     4.2275]

     A new bit in .KLIWD (KL.DEF) is set by SAV to specify that a reload
     occurred  from  operator  pushing  the  buttons.  This change is so
     microcode will be loaded not verified in this circumstance.


          V13-36    RSX20F    Executive

     A change was made to the modem servicer to allow a 22  second  base
     period  for  the  carrier detect timeout loop.  This is an increase
     from 10 seconds so as to allow BELL 212A modems  at  1200  baud  to
     cycle.  [TCO 4.2309]


     3-July-79

          V13-37    RSX20F    Executive
          V3-00     KLE       KL error snapshot program
          V12-04    KLI       KL initialization program

     To get more information from a front end crash about device  status
     the  I/O  page  is dumped by the crash routine into a 4K word block
     starting at the base of the GEN partition (100000).  If an  address
     in  the  I/O  page is non-existent then a 123456 value is placed in
     the corresponding position of the buffer.  [TCO 4.2314]

     The final part of the KL mini RAMP project is completed  with  this
     release  of  KLE.  The requirement was for the front end to do a FM
     sweep on the detection of a FM  parity  error.   This  sweep  would
     report  on  the CTY the block number and address of all AC's with a
     parity error.  [TCO 4.2308]

     A FORCE memory  configuration  command  was  added  to  the  memory
     dialogue of KLI.  This command will force the MOS memory code to do
     a double bit  error  scan.   Also  on  reloads  if  the  memory  is
     "inconsistent" then this code is run.  [TCO 4.2322]


     23-Jul-79

          V13-40    RSX20F    Executive
          V12-05    KLI       KL initialization program

     The KL serial number and hardware options information was added  to
     the  FE  status  block.   This  information  comes from KLI when it
     brings up the system.  Also added is that when a terminal  line  is
     in a carrier wait state incoming characters are rejected to prevent
     sending of invalid characters to KL.


     KLI was changed to do an enviromental report when it brings up  the
     system.   This  report  includes  the KL serial number, model type,
     line  frequency,  and   hardware   options   (cache,MOS,   extended
     addressing).  [TCO 4.2333]


          V13-41    RSX20F    Executive
          V12-07    KLI       KL initialization program
          V13-40    SAV       RSX-20F core image save program

     Because of timing problems on TOPS-10 the edit  for  version  13-32
     was  removed.   It turned out that when a ^\ was typed the terminal
     time-out routine where keep-alive is processed was activated enough
     times in less than the update period to cause a keep-alive-cease.

     The hardware enviromental reporting in KLI was moved so as that  it
     will  not  occur  if in responce to the "ENTER DIALOGUE" question a
     BOOT or EXIT is typed or on automatic reloads the reporting is  not
     done.   This  was  done  so  that  the  state of the machine is not
     changed unexpectedly and to save time during reloads.

     The data base for the hardware enviroment reporting is  cleared  by
     SAV  when  saving a new system.  This is so the saved front end can
     run on another system  without  mistaking  its  serial  number  and
     hardware options.