On Mon, 7 Jul 1997 23:51:31 -0400 (EDT)
   "Matthew Schnierle" <pyld@grove.iup.edu> writes:

>On Mon, 7 Jul 1997, Jonathan Stone wrote:
>
>  > you should give up to use it with your unit.  Pull it off.  I think
>  > it's a safer way you go, I think.
> 
> Pulling the card is *required* on a 3KMIN (500/1xx), but should work
> on a 5000/25. *The interrupt logic on the two systems is different.)

>*ducks*  Are you sure?

``Nah, he's just making it up as he goes along!!'' :)

Yes, I'm sure. The device-drivers for a cfb (PMAG-B), sfb (PMAGB-BA)
and mfb (PMAG-AA) have interrupt handlers which get attached to the
vertical-blank interrupt which acknowledge that interrupt.

On all TC DECstations other than the 3MIN, the kernel could just leave
the interrupt unenabled. (*I* have never seen any video snow from
doing so.)

But on a 3MIN, TC slot interrupts are mapped direct onto CPU interrupt
lines. Given how SPLs work in 1.3, that means video cards in a 3MIN
*have* to have a driver, or the machine will hang.  (The board
requests a vertical-retrace interrupt, the CPU takes the interrupt, no
driver handles it, the CPU returns and the interrupt is still active,
so the CPU takes the interrupt again...)

There is a fix for this, it means redoing the spl machinery in the
kernel.  Jason Thorpe planned out most of the work and I'm still
waiting for him to finish it.  Meantime, if you have a 3MIN,
you *have* to pull unsupported video cards.

This too should go on the list Simon is apparently keeper of.


>My 5000/120 only required a setenv console s (albeit blindly from the
>standard keyboard) and a terminal jacked into the correct serial port.
>
>Having said that, I'm not completely convinced that the card in question
>(PMAG B) is not fried.

There's no reason to assume it's fried.  Do vmstat -i and see how many
interrupts are shown against the slot with the card. IF it's 60, the
card is probably fine.