ParErrs-- frequency

Dave, 508-770-6877,-6721 17-Mar-1995 1016 cressman at elwood.enet.qntm.com
Fri Mar 17 07:56:14 PST 1995


From: David Cressman, Quantum Corp.
To:   SCSI folks

    Tak Asami wrote:

>Have anyone ever seen in this life an "occasional" occurance of SCSI parity
>error?  In my limited experience, it either never happens or happens ALL THE
>TIME.  In other words, all parity errors I've seen are hard errors.
>
>Where do parity errors come from?  Intermittent cable/connector connection,
>cold solder joints on your Heathkit SCSI drive, disintegrating PC board trace,
>a dead device on the bus (OJ did it!).

    Yes, I have seen intermittent/low frequency bus parity errors.  The
    worst ones were about 1/week and caused by improper system/cable
    grounding.  Another cause can be SCSI bus length being too long for
    the configuration/termination.  It can be within the spec limit,
    but signal levels don't have proper margin, and occasional failures
    can occur.  Noise can also cause intermittent failures, esp. with
    ribbon cable or improper grounding.  By definition, "occasional" PE's
    can be easy to miss or ignore as a "glitch" or "mystery", especially
    if the device doesn't handle/report it well, or the system doesn't
    handle/report/log it "properly."  I have been developing SCSI systems
    on and off for about 12 years, and have been in some high pressure
    situations caused by occasional parity errors, and the device not
    detecting and/or handling it properly, which made it *very* difficult
    to track down the problem(s).  (However, the main specific situation I
    am thinking of was in the mid-'80s, and I know SCSI interface chips 
    have come a long way since then.)

    Again I agree with the general sentiment of keeping PE handling code
    SIMPLE. I think it's a great idea to have a section in SIP that clearly
    lays out a minimum/recommended level of implementation for handling and
    *reporting* parity errors in all the different bus phases. (Sec 5.3, 5.4
    cover this area very briefly now.)  SCSI-2 does say a lot about this, 
    but 1) it's kind of scattered around, 2) sometimes it suggests multiple 
    ways, 3) it could probably be clearer.

    -Dave C.

    David C. Cressman
    Quantum Corp
    DLT (digital linear tape) development
    cressman at elwood.enet.qntm.com
-------------------------------------------------------------------




More information about the T10 mailing list