
 Runstream (R) IBM MCA SCSI SPOCK-206 Windows NT Miniport Driver
 Copyright (C) Unal Z, 2006 for Runstream & PS/2 MCABASE
 ----------------------------------------------------------------
 Beta Test SPOCK-206                            unalz@mail333.com

          http://www.members.aon.at/mcabase/spock206.htm

  This document complements the SPOCK206 installation instructions
  and presents authentic test reports.

  The driver installation procedure is described in INSTALL.TXT.
  

    *****  Read INSTALL.TXT before you install the driver *****


  CONTENTS

  1. Before You Start
        Driver ZIP contents
        LED Display PS/2 Server 95  	
        LED Panel SCSI-II Operation Codes
  2. Driver Update and Performance Test
  3. History of the Driver     
  4. Test Reports


 1. BEFORE YOU START

   You need a MicroChannel machine with an IBM SCSI host adapter and the
   Windows NT operating system. 
   
   SPOCK-206 supports the following IBM MCA SCSI Host Adapters in high
   performance busmastering mode with 32-bit addressing:
  
   ID     Adapter                                      Code
   -------------------------------------------------------------
   8EF8   IBM Expansion Unit SCSI Controller           --  
   8EFC   IBM Fast/Wide SCSI-2 Adapter/A               Corvette
   8EFD   IBM 7568 Industrial Computer SCSI Adapter    -- 
   8EFE   IBM PS/2 SCSI Adapter                        Tribble
   8EFF   IBM PS/2 SCSI Adapter w/Cache                Spock
   FFFF   Planar SCSI IBM PS/2 Models 9556/9557
  
   The eventually available onboard cache does not affect the performance
   of the adapter when used with SPOCK206. Commands, natively supported by
   the adapter are bypassed and SCSI-II commands are issued directly to the
   addressed device. Thus, all of the above adapters are SCSI-II compliant
   when operated under SPOCK206. SCSI messaging is handled by the adapter.

   Planar SCSI subsystems are detected and supported by SPOCK26, even when
   the subsystem identifies itself with an FFFF ID.
   
   NOTE: At each boot, Windows NT might log a SPOCK-206 Controller Error,
   Event-ID 11, in the System Event Log. This has error has been traced
   to the command SCSIOP_REQUEST_BLOCK_ADDR which either could not be
   completed or rejected by the host adapter. This command is issued after
   finishing the driver init phase and its failure does not impair the
   correct operation of the SCSI subsystem resp. SPOCK206.
   
  
   DRIVER ZIP CONTENTS

   Unpack the contents of SPOCK2NT.ZIP in a directory or on a diskette.
   You should have the following files:

   SPOCK2NT   <DIR>  -- Windows NT SPOCK-206 drivers
      SPOCK2   SYS   -- Binary, IBM MCA SCSI SPOCK-206
      SPOCK2B  SYS   -- Binary, IBM MCA SCSI SPOCK-206 DISK LIGHT
      SPOCK2C  SYS   -- Binary, IBM MCA SCSI SPOCK-206 LED DISPLAY
      SPOCK2NT       -- Driver setup tag file
      TXTSETUP OEM   -- Driver setup file
      SPOCK206 INF   -- Windows NT driver installation INF file
      README   TXT   -- Driver information and test reports
      INSTALL  TXT   -- Installation instructions

   To differentiate between the Windows NT SPOCK driver and the SPOCK-206
   drivers, the driver descriptions have been defined as follows:

   IBM MCA SCSI SPOCK-206             = IBM MCA SCSI Host Adapter on SPOCK-206
   IBM MCA SCSI SPOCK-206 DISK LIGHT  = Same as above, with disk light enabled
   IBM MCA SCSI SPOCK-206 LED DISPLAY = Same as above, with LED Display Panel
   IBM MCA SCSI Host Adapter          = This is the Windows NT driver.
   
   NOTE: SPOCK-206 can manage the disk light on your machine to indicate
   communications with the SCSI host adapter. The disk activity light in
   does not necessarily imply your fixed disks are being accessed, it rather
   indicates an activity on the SCSI bus. 
      
   Select "IBM MCA SCSI SPOCK-206 DISK LIGHT" at installation time if you wish
   the driver to manage the disk activity light.
   
   Select "IBM MCA SCSI SPOCK-206" at installation time if you wish to disable
   the disk activity light. This version will NOT manage the disk light on your
   machine, and it is the fastest driver version.
   
   Select "IBM MCA SCSI SPOCK-206 LED DISPLAY" at installation time if you
   wish to monitor the communication between the host adapter and SPOCK-206
   on the SCSI bus on the LED Display Panel of the IBM PS/2 Server 95. This
   version of the driver will indicate READ, WRITE and other SCSI commands
   along with their command codes as defined in the ANSI SCSI-II Standard.
    
    
   SPOCK-206 LED DISPLAY on PS/2 SERVER 95  	
   
   Windows NT boot times will be slightly delayed when using the LED Panel
   option. A time delay is inserted between the driver initialization messages
   for the better readability. Without delays, messages are simply unreadable.

   No delays are inserted between normal operation messages.
      
   You may not be able to see and detect all messages in the normal course of
   operation. Most of the time you will see READ or WRT messages along with 
   an occasionally blinking asterisk '*' indicating interrupt handling.
   
   Following are the driver messages for the LED Display Panel. For the
   xx codes, see the list further below.
   
   Message      Description
   ---------------------------------------------------------------------------
   *          = Interrupt handler entered
   READ xx    = Data in from device to system, xx is the SCSI-II code of the
                command
                
   WRT  xx    = Data out from system to device, xx is the SCSI-II code of the
                command
                
   SEND xx    = No data transfer involved, xx is the SCSI-II code of the
                command
                
   FAIL xx    = Command failed, xx is the SCSI-II code of the command.

   INIT xxxx  = Status code xxxx of SCSI PortInitialize on leaving DriverEntry
                0000 is evidently the success code (Windows 95)
                0001 might indicate initialization failure (Windows 98SE)
   INIT OK    = Init success

   PORT IO:X  = Probing I/O ports for IBM MCA SCSI adapters, X is the port index
   CORVETTE   = Corvette detected
   ID FAIL    = No IBM MCA SCSI adapters found on the known I/O ports
   BAD FW:F   = Bad firmware, adapter revision level 0xF.

   RESETBUS   = Reset Bus Command
   IMMD       = Issuing Immediate Command to adapter
   TSB        = Checking Termination Status Block

   EOI TOUT   = Wait on busy to send EOI timed out
   SCB TOUT   = Wait on busy to send SCB timed out
   IMC TOUT   = Wait on busy to send immediate command timed out

   GPOS FAIL  = Get POS SCB command to adapter failed
   TSB ERR    = TSB Error, SCB command completed with error

   INIT E:FC  = Feature Control immediate command to adapter failed
   UNCX E:00  = System call failed, UncachedExtension cannot be allocated
   PHYS E:00  = System call failed, cannot obtain PhysicalAddress.
   ---------------------------------------------------------------------------


   LED PANEL SCSI-II OPERATION CODES
 
   Following is the of the SCSI command codes displayed on the LED Panel
   when using the SPOCK-206 LED Panel version. This table is taken from
   SCSI.H of the Windows NT 4.0 DDK:
   
   Command         LED xx
   ------------------------
   TEST_UNIT_READY     00
   REZERO_UNIT         01
   REWIND              01
   REQUEST_BLOCK_ADDR  02
   REQUEST_SENSE       03
   FORMAT_UNIT         04
   READ_BLOCK_LIMITS   05
   REASSIGN_BLOCKS     07
   READ6               08
   RECEIVE             08
   WRITE6              0A
   PRINT               0A
   SEND                0A
   SEEK6               0B
   TRACK_SELECT        0B
   SLEW_PRINT          0B
   SEEK_BLOCK          0C
   PARTITION           0D
   READ_REVERSE        0F
   WRITE_FILEMARKS     10
   FLUSH_BUFFER        10
   SPACE               11
   INQUIRY             12
   VERIFY6             13
   RECOVER_BUF_DATA    14
   MODE_SELECT         15
   RESERVE_UNIT        16
   RELEASE_UNIT        17
   COPY                18
   ERASE               19
   MODE_SENSE          1A
   START_STOP_UNIT     1B
   STOP_PRINT          1B
   LOAD_UNLOAD         1B
   RECEIVE_DIAGNOSTIC  1C
   SEND_DIAGNOSTIC     1D
   MEDIUM_REMOVAL      1E
   READ_CAPACITY       25
   READ                28
   WRITE               2A
   SEEK                2B
   LOCATE              2B
   WRITE_VERIFY        2E
   VERIFY              2F
   SEARCH_DATA_HIGH    30
   SEARCH_DATA_EQUAL   31
   SEARCH_DATA_LOW     32
   SET_LIMITS          33
   READ_POSITION       34
   COMPARE             39
   COPY_COMPARE        3A
   WRITE_DATA_BUFF     3B
   READ_DATA_BUFF      3C
   READ_LONG           3E
   CHANGE_DEFINITION   40
   READ_SUB_CHANNEL    42
   READ_TOC            43
   READ_HEADER         44
   PLAY_AUDIO_10       45
   PLAY_AUDIO_MSF      47
   PLAY_TRACK_INDEX    48
   PLAY_TRACK_RELATIVE 49
   PAUSE_RESUME        4B
   LOG_SELECT          4C
   LOG_SENSE           4D
   PLAY_AUDIO          A5
   READ_CDDA           D8
  
   
 2. DRIVER UPDATE AND PERFORMANCE TEST
   
   Driver updates are announced in comp.sys.ibm.ps2.hardware. Normally, 
   the latest update of the driver is at the following link:
   
      http://www.members.aon.at/mcabase/pub/files/spock2nt.zip

   The driver homepage located at the following link:

      http://www.members.aon.at/mcabase/spock206.htm

   You can use any SCSI disk performance test program to benchmark your
   with SPOCK-206 updated Windows NT system. HDTach delivers nice results,
   you can obtain it from here:
   
       http://www.simplisoftware.com/Public/index.php?request=HdTach2.7
       
   More tools are listed here, see also the page "Hard Disk Drive 2"
   at the top menu:

       http://www.benchmarkhq.ru/english.html?/be_hdd.html


 3. HISTORY OF THE DRIVER
 
   SPOCK-206 is an acronym for SPOCK Miniport Driver 2006. The driver is
   based on code samples published by Microsoft Corp. in the Windows NT DDK.
   You can find the code samples in the MSDN Subscription Collection 1995-96.
   
   Driver testing has been done by the members of comp.sys.ibm.ps2.hardware.
   I'd like to express my sincere thanks to all testers for their invaluable
   work and reported problems. Test reports are included in the next section
   of this document.
     
   The excerpts below from the discussion in W9xNT MCA Adapters - SPOCK
   (this the thread from comp.sys.ibm.ps2.hardware) highlight some of the
   interesting moments in the development phase. 
   
   First driver builds managed to increase the disk transfer rate by more
   than 60% over the Windows NT SPOCK driver.


   CHARLES LASITTER (July 4, 2006)
 
   UZnal's free programming labor for the SPOCK206 driver update has
   again netted Doctors Without Borders a contribution in his honor.

   Thanks UZnal!


 4. TEST REPORTS

   Following are excerpts from authentic SPOCK206 test reports, lightly edited
   for brevity. Reports are presented in chronological order.


   UNAL Z (July 20, 2006)
   
   SPOCK206 on NT at the speed of light, Buzz Lightyear hardcore fans only:
   
   M$ = 6100 KB/sec
   UZ = 9800 KB/sec

   WARNING: You must remove the previous Spock driver after you have installed
   Spock206. That was the only problem with my first tests, I did not remove
   it and deadlocked the system with two competing drivers loaded. Remember to
   hit NO when asked to restart the computer. Remove the previous Spock and
   only then restart:

   - DO NOT DARE SAY YES to "Restart Computer" prompt of NT
   - Hit ***NO***
   - Stay in the SCSI driver window
   - Click on "IBM MCA SCSI Host Adapter"
   - Hit REMOVE button
   - You must have only SPOCK206 left in the SCSI Adapters of SPOCK.
   - Reboot
   
   Remove always the previous Spock, no matter which one it is !!!
   
   
   UNAL Z (July 21, 2006)
      
   Spock206 is very fast on Windows NT:
   
   Test setup: NT4/SP3, Mod. 95 on P60, Wide IBM DDRS-34560W on Corvette,
   Plextor PX-12CS and Yamaha CRW-4416S on Spock with cache. Timings with
   SCSIBench (16-bit) of EZ SCSI 4.0
   
   Here my results with a 64K transfer size for the wide IBM DDRS (SEQ =
   Sequential Access, RND = Random Access and SEC = Same Sector Access),
   average values in KB/sec:

   IBM DDRS-34560W
   Read    M$ SPOCK    SPOCK-206   DISKLIGHT   LED PANEL
   ----------------------------------------------------------                        
   SEQ     6100        9800          9600        9300  KB/sec
   RND     2400        3100          3100        3000  KB/sec
   SEC     2600        3900          3900        3800  KB/sec
   
   At times it hit the 11000 KB/sec marks !!!
           
   Plextor 12xPlex caddy CD-ROM Drive:
       1800 KB/Sec SEQ
   
   Yamaha 4416S CD-ROM Drive/CD Writer:
       1600 - 1700 KB/sec SEQ
   
   The 16x Yamaha is slower than the 12x Plextor ....!
   
 
   CASOLAI (July 23, 2006)
   
   Here are the benchmarks from SiSoft Sandra 2001te Professional for
   SPOCK-206 Disk Light:
   
   Buffered Read        12 MB/s
   Sequential Read      12 MB/s
   Random Read           5 MB/s
   Buffered Write        8 MB/s
   Sequential Write     12 MB/s
   Random Write          7 MB/s
   Average Access Time   7 ms (estimated)
   
   These are a considerable improvement over the NT default driver, it
   never got more than 9 MB/s before with these hard disks.
  
   Here are the same benchmarks using the M$ scsi driver:
   
   Buffered Read         8 MB/s
   Sequential Read       8 MB/s
   Random Read           4 MB/s
   Buffered Write        4 MB/s
   Sequential Write      4 MB/s
   Random Write          4 MB/s
   Average Access Time   7 ms (estimated)
   
   Pretty impressive boost ! 200% to sequential write speed !
   
   Damn fine effort UZ !!

   Test setup: 9595-1NT: 256 MB ram, P90, 3x 9.1 GB DPSS Ultra160
               2x IBM 8x CDROM, Corvette, Reply SB-16, Yamaha DB50XG
               PCMCIA, Dual Etherstreamer, Cheetah with SideCard, GUP
               Digi MC16i, NT4 Server
  
   The benchmarks I listed were run with a 68-pin cable from the Corvette
   to the hard disks, and a 50-pin cable also on the same Corvette to the
   two cdrom's. I reran the benchmark with the 50-pin cable removed, but
   the results were exactly the same.
   
   
   LOUIS OHLAND (July 23, 2006)

   Empirical evidence will be presented!

   9595A, Turbo-T BIOS 10 at 180MHz, 128MB ECC, two Corvette, one Patriot.
   IBM DDRS-34560W, rev S97B [wide disk].

   M$ Spock
      Sequent  :  6,400
      Random   :  2,500
      Same Sect:  3,800

   UZ Spock
      Sequent  : 11,000
      Random   :  3,100
      Same Sect:  7,600

   For those that use Adaptec's SCSIBENCH, you know it swings back and 
   forth a bit. I guesstimated the happy medium and these are it.
   
   UZ > I've got the same disk, the results are similar, though yours are
       better. The sequential access rate is amazing, the read cache of the
       DDRS is well exploited.

   UZ > The picture stabilizes - you get a big improvement with modern, wide
       and fast disks having a reasonably sized cache. I don't suppose older
       disks will show the same preference for the higher transfer rates.


   UNAL Z (July 24, 2006)
   
   HDTach 2.70 results:
  
   Easy and simple, but tests only sequential read. Below the results for the
   wide DDRS on Corvette, disk access time reported for all about 13 ms. Note
   the much lower CPU utilization by Spock206, a factor of about 4. The max
   improvement over M$ is 60%, confirming SCSIBench, the average figure is 40%.
   
   IBM DDRS-34560W
   M$ Spock:              Spock206            Spock206 LED Panel
   ------------------------------------------------------------------
   Read speed:                                    
      max 7.0 MB/sec        max 11.1 MB/sec     max 10.8 MB/sec
      min 4.5 MB/sec        min  6.4 MB/sec     min  6.8 MB/sec
      avg 6.9 MB/sec        avg  9.9 MB/sec     avg  9.7 MB/sec
   Read Burst:                                    
      7.0 MB/sec            11.2 MB/sec         11.1 MB/sec
   CPU Utilization:                                
      46%                   11.6%               12.4%
   
   Repeat the test a few times, the figures can be slightly different.
   
   
   JIM SHORNEY (July 24, 2006)
   
   At last, something that works. My M$ Spock numbers are from SCSIBench32
   4.01, I didn't feel like going back to M$ Spock and doing it all over
   after banging my head all afternoon. I know, I know, don't even say it....
   
   IBM 0661-467
   M$ Spock (SCSIBench32)     Spock206 Disk Light (HD Tach)
   -----------------------    -----------------------------
   Read speed:                 Read speed:
       1.21 MB/sec                  max 1.4 MB/sec
                                    min 1.3 MB/sec
                                    avg 1.0 MB/sec
                               Read Burst:
                                    3.3 MB/sec
                               CPU Utilization:
                                    1.0%

   IBM 0661-371
   M$ Spock (SCSIBench32)    Spock206 Disk Light (HD Tach)
   ----------------------    -----------------------
   Read speed:                 Read speed:
       1.24 MB/sec                  max 1.7 MB/sec
                                    min 1.6 MB/sec
                                    avg 2.0 MB/sec (huh?)
                             Read Burst:
                                    2.1 MB/sec
                             CPU Utilization:
                                    1.1%

  UZ > Still about a 40% increase. However, the older disks perform better
       with smaller block transfers, like 8 KB, possibly due to the smaller
       cache. HDTach does not allow to play with the block size, but SCSIBench
       did it.

   IBM DPES-31080
   M$ Spock (SCSIBench32)    Spock206 Disk Light (HD Tach)       
   ----------------------    ----------------------
   Read speed:                 Read speed:
       1.72 MB/sec                  max 4.4 MB/sec
                                    min 2.7 MB/sec
                                    avg 3.8 MB/sec
                               Read Burst:
                                    6.8 MB/sec
                               CPU Utilization:
                                    2.4%
   
   UZ > This looks much better, it hits the 250% mark...?
   
   UZ > I for myself look at the change rate and see a commonly confirming
        average pattern of 40% to 60% improvement. This will be of course less
        felt with slower disks, but is dramatic with fast/wide disks.
      

   For comparison, here are the numbers on the drives residing on the
   BT646S before and after.
   
   Imprimis 94601-15
   Living with M$ Spock (SCSIBench32)    Spock206 Disk Light (HD Tach)
   ---------------------------------     ----------------------
   Read speed:                            Read speed:
       1.72 MB/sec                             max 2.2 MB/sec
                                               min 1.6 MB/sec
                                               avg 2.0 MB/sec
                                          Read Burst:
                                                2.9 MB/sec
                                          CPU Utilization:
                                                8.0%
   HP D2645
   Living with M$ Spock (SCSIBench32)  Living with Spock206 Disk Light (HD Tach)
   --------------------------------    ----------------------
   Read speed:                           Read speed:
       1.63 MB/sec                            max 2.3 MB/sec
                                              min 2.2 MB/sec
                                              avg 2.0 MB/sec
                                          Read Burst:
                                              4.9 MB/sec
                                          CPU Utilization:
                                              9.5%
   
   So getting bad M$ drivers out of the way can improve the performance of
   other devices as well.
   
   System: 9595-0PT Turbo-Y complex @200 MHz, 96MB RAM, 
           NTws 4.0 1381 SP6a, ASPI Layer 4.71.2 (was 4.6 (1021) )
           IBM SCSI-2 F/W (Corvette), wait states disabled, 100ns
           streaming enabled, combined bus mode.
           BusLogic BT646S with streaming and sync negotiation enabled
           Cornerstone ImageAccel, SoundPiper 16, USR Courier 2400,
           512K SVGA/A, HP Scanjet (parallel) adapter, IBM Etherstreamer MC32
           
   Hard disks as listed above, all SCSI-2 narrow except the Imprimis SCSI-1
   narrow, according to Adaptec SCSI Explorer.
   
   Man, this could be a real screamer if I stuck some Fast/Wide drives in it.
   
      
   WILLIAM WALSH (July 24, 2006)
 
   I finally settled on dskspeed.exe from
   http://www.benchmarkhq.ru/english.html?/be_hdd.html
  
   Benchmarking results:
   
   9585-0XF/ST12550N
   
   M$ Spock                  Spock-206
   --------------------------------------
   Linear Read:                 
        1.5355 MB/sec         1.4960 MB/sec
   Random Read:                 
        0.5293 MB/sec         0.5199 MB/sec
   Access Time:                 
       20.98ms               20.37ms
   
   
   The computer feels faster by an ever so slight amount, but data transfers
   with this benchmark are awfully slow. I don't know why.
   
   9595-0PT "Q-Q-Clock"
   This machine is actually running a Type 4 P90 complex with a Pentium 200MMX
   underclocked to 180MHz. It has a Corvette SCSI adapter with four wide disks
   and a narrow (with wide>narrow adapter) CD-ROM II attached to it on the wide
   cable. I would tend to think that this represents nearly the fastest
   possible PS/2 setup, save for those few who have dared clock their machines
   up to a full 200 or 233MHz.
   
   The hard disk types are, in order as best I can tell (I did not set up the
   disk subsystem on this machine, nor have I attempted to optimize it or get
   some more reliable drives...):
   
   C = ID6: 2255 MB DFHS    D = ID 4: 4512 MB DCHS
   E = ID2: 1052 MB DFHS    F = ID 9: 4512 MB DFHS
   
   The disks scored as follows:
   
   M$                                    Spock-206/UZ with LED
   -----------------------------------------------------------------
   C Linear Read   3.6031 MB/sec         C Linear Read   4.654  MB/sec  
   C Random Read   1.228  MB/sec         C Random Read   1.1674 MB/sec
   C Access Time  13.92 ms               C Access Time  15.17 ms
                                           
   D Linear Read  4.2044 MB/sec          D Linear Read   4.1861 MB/sec
   D Random Read  1.1895 MB/sec          D Random Read   1.2547 MB/sec
   D Access Time 14.01 ms                D Access Time 14.36 ms
                                           
   E Linear Read  5.0989 MB/sec          E Linear Read   6.3397 MB/sec
   E Random Read  0.9991 MB/sec          E Random Read   0.9989 MB/sec
   E Access Time 12.92 ms                E Access Time 13.41 ms
                                           
   F Linear Read  6.6181 MB/sec          F Linear Read   7.0931 MB/sec (!!)
   F Random Read  0.9588 MB/sec          F Random Read   0.9696 MB/sec
   F Access Time 14.36 ms                F Access Time 14.62 ms
   										
   
   Finally, for further comparison, a 9595-1NT with P90 complex, Windows NT 4.0
   SP6a, and two hard disks--one IBM 0663 at ID6/C (newer quarter height
   version) and a customized for Compaq DPES-31080 at ID5/D. Both drives and a
   Teac 6X CD-ROM are being driven by a late, three oscillator cached SCSI/A.
   
   M$                                    Spock-206/UZ
   -------------------------------------------------------------------
   C Linear Read  1.5525 MB/sec          C Linear Read  1.9030 MB/sec
   C Random Read  0.5393 MB/sec          C Random Read  0.5573 MB/sec
   C Access Time 22.26 ms                C Access Time 23.39 ms
   
   D Linear Read  1.8553 MB/sec          D Linear Read  1.8759 MB/sec
   D Random Read  0.5553 MB/sec          D Random Read  0.5598 MB/sec
   D Access Time 20.00 ms                D Access Time 19.87 ms
   
   
   What do I see here? Well, perhaps the fact that the DskSpeed.exe benchmark
   gives somewhat more conservative readings than do the others mentioned in
   this thread. I also see that I have slow hard disks in all machines, with
   the possible exception of the 9585-0XF and its ST12550N Barracuda. Why the
   9585 runs so slowly compared to Spock-206 on Win95 and CORETEST, I am not
   sure. However, I have two variants of the otherwise identical ST12550N
   drive. The one in this machine is one of the few I have that is made in the
   USA. It has an 80C96 microcontroller onboard that is rated for and clocked
   at 16MHz. The Singapore made variant has the same 80C96, but it is clocked
   at (and rated for) 20MHz.
   
   
   WILLIAM WALSH (July 28, 2006)
   
   HDTach reports the following from the Server 95 with the P90...the one that
   was mentioned earlier with the 0663 and DPES-31080 drives being driven by a
   triple-oscillator SCSI/A with cache. Now the SCSI adapter is the short
   uncached SCSI/A.
   
   Read Speed: 
        Maximum 2.3 MB/sec
        Minimum 2.1 MB/sec, 
        Average 2.0 MB/sec
   CPU Utilization 
        4.4%
  
  
   LOUIS OHLAND (July 28, 2006)

   "The driver detected a controller error on \Device\ScsiPort2."
   This happens once during boot. There is no specific error reported.

   UZ > The error is due to a SCSI command error. 

        2 = SP_INTERNAL_ADAPTER_ERROR
                SCB_STATUS_COMMAND_ERROR
   


   UNAL Z (July 28, 2006)
  
   Windows NT Error Log format:
  
   http://support.microsoft.com/default.aspx?scid=kb;EN-US;Q182335
  
   Let us look at the error log entry format, it might be useful to you. Click
   on an entry in the event log and view hex data as WORDS:
  
   0000: 0010000f 00680001 00000000 c004000b
   0010: 0001000e 00000000 00000000 00000000
   0020: 00000000 00000000 00000000 00000007
   0030: 00000000 00000006
  
   At offset 0000 is the internal logger data, c004000b is the error type
   (controller error here).
  
   At offset 0010 we have above 0001000e. Spock miniport error codes are in the
   upper two bytes, this is 0001 in this case (forget 000e, it is something
   else).
  
   At the end of the offset 0020 line we see 00000007, this is the SCSI ID
   which is the adapter.
  
   At the end of offset 0030 line is the miniport logged error type 00000006,
   which is the code for the internal controller error.
  
   Since error types are limited in NT, "internal controller error" can mean
   many things, it is used where no other type is available. 
   
   
   UNAL Z (July 30, 2006)
   
   Cache or not cache? Easy to answer - no cache. When you look at the table of
   directly supported SCSI commands by Spock, posted recently, you will notice
   that support is basicly built in around read and write commands. These are
   the operations that affect the cache, all other commands can be directly
   passed to the device with the adapter still managing DMA, SCSI messages and
   status codes, but leaving out the cache for this device.
   
   The READ command of Spock is a basic read command intended for rather dumb
   devices. It cannot meet all needs at all times, for instance, you might want
   to tell that disk device something special before read starts. If you ask
   the adapter to take care of the READ command (to manage the cache) but still
   want to directly talk to the device before, things are getting too
   complicated. As always, a simple unified common strategy works best.
   
   
  

END.