From: Fred Kleinsorge [kleinsorge@star.zko.dec.com]
Sent: Monday, March 12, 2001 2:20 PM
To: Info-VAX@Mvb.Saic.Com
Subject: Re: VMS 7.2-1 Bugcheck In DECW$REINIT.EXE

The 4D10T (aka S3 Trio64 chip on a dimond video card) hasn't been touched in
a very long time.  It doesn't do DMA, and it doesn't use interrupts.  So it
should be relatively immune to the boot reset problems (the 500au doesn't
cleanly reset all PCI devices on a reboot anyway).

Do me a favor and try reducing WSMAX (sounds dumb I know).  Unfortunately I
will be going on vacation for about 10 days and won't be able to see the
results.  I'm assuming that this system has at least 64 MB of memory or that
it isn't running DECnet phase 5.

The one thing that can go wrong on systems - and I started to see this on
V7.2-1 - especially those with less than 64mb of memory (now the minimum
memory if you are running Motif and Dnet) - is that a IOC$MAP_IO call fails
when the driver try's to map the hardware.  The failure can be tracked back
to a failure to expand the system page tables, and that can be tracked to a
failed check for "fluid pages".  This check is trying to make sure that
there are enough physical pages available to allow inswap of the largest
possible process (or some such thing).  The failures here can be random -
but the graphics tend to initialize last - and so they end up being the
victim more often than not.

Also, check to see if INDRIVER/IK/IM are all loaded at the time of the
crash.  Try unplugging the mouse when you do the reboot.




Jerry Leslie wrote in message ...
>Fred Kleinsorge (kleinsorge@star.zko.dec.com) wrote:
>:
>: Jerry Leslie wrote in message <05Pn6.79$%L5.4561@insync>...
>:
>: >  2. Why does the system boot successfully from a power off
>: >     state, but crashes in an attempt to reboot
>: >
>:
>:
>: What type system and graphics is this again?
>
>The system is a Digital Personal Workstation 500au; the graphics card
>is a PowerStorm 4D10T.
>
>Is BOOT_RESET set to ON?
>
>It's OFF:
>
>  OLYWS3> write sys$output f$getenv("BOOT_RESET")
>  OFF
>
>:
>: A frequent problem with warm boots on some workstations is the failure to
>: re-initialize the PCI devices.  So a reboot can occur with a device still
>: active, and worse - sending interrupts.  What we have frequently seen is
the
>: device driver being loaded and getting interrupts before the data
structures
>: are all setup.  And poof.  A crash.
>:
>: This is/was a general problem with the ELSA graphics, and probably the
>: 3D30/4D20, on things like the DPW500 & 600.  There is new firmware that
>: explicitly resets the devices (because there isn't a general way to reset
>: the PCI on this system!) and at least on the ELSA a new driver (in the
most
>: recent patch kits) that does some belt-and-suspenders checks in the
>: interrupt path to make sure that initialization is complete.
>:
>:
>
>The following patches have been applied:
>
>  o vms721_update-v0100
>  o vms721_pcsi-v0100
>  o vms721_cpu2208-v0100
>  o vms721_graphics-v0400
>
>What's the version of the new firmware and where can it be obtained ?
>
>--Jerry Leslie   leslie@209-16-45-97.insync.net
>                 leslie@209-16-45-102.insync.net is invalid
>                 (my opinions are strictly my own)