[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference marvin::sync_drivers

Title:VMS synchronous device drivers
Notice:Don't see UKCSSE::WANDD_UPDATE for anything
Moderator:MARVIN::CARLINI
Created:Wed Nov 04 1987
Last Modified:Tue Jun 03 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:656
Total number of notes:2530

655.0. "INVEXCEPTN @WANDD$CommonLdr+654E (VAX)" by 60549::SIMMONDS (loose canon) Fri May 09 1997 13:00

    | OpenVMS VAX V6.2
    | DECnet-Plus for OpenVMS VAX, V6.3 [no ECOs]
    | 
    
    Another crash in WANDD$COMMONLDR, associated with ZSDRIVER running
    DDCMP on DSW.. CANASTA failed to recognize it.
    
    Anybody know anything about this one?
    
    Thanks!
    John.
    
    CANASTA Summary:
    
    VMS VERSION        : 6.2
    CPU TYPE           : 3100
    BUGCHECK           : INVEXCEPTN
    PROCESS NAME       : GOPHER
    IMAGE NAME         : NONE
    SID REGISTER       : 0A000005
    ARGUMENT COUNT     : 00000005
    CONDITION CODE     : 0000000C
    REASON MASK        : 00000000
    VIRTUAL ADDRESS    : 00000018
    EXCEPTION PC       : 806C754E
    EXCEPTION PSL      : 00C80004
    MODULE NAME        : WANDD$COMMONLDR
    MODULE OFFSET      : 0000654E
    INSTRUCTION        : MOVZBL
    
    $ Anal/Image Sys$Loadable_Images:WANDD$CommonLdr
    [...]
    image name: "DNA PHASE V - WANDD COMMON LIBRARY"
    image file identification: "WANDDV V2.1-177"
    link date/time: 21-MAY-1994 05:12:30.70
    
T.RTitleUserPersonal
Name
DateLines
655.1Hmmm...MARVIN::CARLINIMon May 12 1997 14:2517
    ZSDRIVER running on a DSW would be a definite candidate for crashing -
    especially since it only knows about DST and DSH :-)
    
    Since the rest of your entry implies a CVAX-based CPU, I'll assume you
    actually have a DSH32 in a MicroVAX 3100 Model 10/20/10e/20e. In that
    case, there has been one bug fix applied to ZSDRIVER on 30th April
    1996. If your ZSDRIVER is older than that, you may wish to upgrade to
    V6.3-6; otherwise you will need to log an IPMT and provide a full dump.
    
    If V6.3-6 does fix it, please let CANASTA know that the problem is
    fixed and that the original IPMT case was CFS.27445.
    
    If by some chance you really do have a DSW, there have been no changes
    since 1993.
    
    Antonio
    
655.2now waiting for ECO06 to crash.. :)60549::SIMMONDSloose canonTue May 13 1997 06:0612
    Re: .1
    
    Antonio, you're right: not DSW but DSH32 in MicroVAX 3100 model 10..
    The Customer and I attempted to reproduce the crash before we installed
    ECO06 but of course we couldn't! :-}    Anyway, I've saved the dump and
    gone ahead and installed ECO06 for V6.3.. we'll try and stress it out..
    
    Would the failing instruction in WANDD$COMMONLDR correspond with the
    ZSDRIVER bug fix you mentioned?
    
    Thanks!
    John.
655.3MARVIN::CARLINITue May 13 1997 13:0515
I looked back at the original case and the effect seemed to be that we would
wind up in a driver common routine at IPL 21 instead of IPL 8. The usual effect
was to occasionally fail to allocate pool and therefore return SS$_BUFFEROVF
rather than crashing, but since we were left an interrupt routine at the wrong
IPL we could be messing up almost anything.

Looking at the code fix, there were two points at which IPL had to be restored
before returning. One was associated with BISYNC code (and does not apply in
your case) and the other was linked to a certain buffer not being queued up -
this might be what was happening in your case.

If V6.3-6 does not fix it, you'll need an IPMT and a valid crash dump (V7.1 has
no new fixes for DSH).

Antonio
655.460549::SIMMONDSloose canonFri May 16 1997 12:2612
    Re: .3
    
    Antonio, Thank you for your friendly informative reply!   We'll do as
    you instructed iff we crash with ECO06.. (I suspect we may never see
    it, though, as the Customer was getting the application to do very
    'abnormal' things at the time.. (of course,  it's indisputable that a
    device driver must be totally stable, _whatever_ the USER mode program
    attempts to make it do) ..and they can't remember what the magic
    sequence was which produced the crash.. :9)
    
    Thanks++
    John.
655.560549::SIMMONDSloose canonTue Jun 03 1997 06:584
    Customer can reproduce this even with ECO06.. IPMT will be launched,
    complete with small reproducer..
    
    John.