Jason R Thorpe wrote:

>  > The following patch seems to fix this and now my 3000/300 can
>  > speak some Japanese :-) But I'm not sure how it works on pmax...
> 
> It looks like this patch will be okay on the DECstation, as well.  The
> new arch/pmax/tc/asc_ioasic.c clears the pointer reload interrupt bit
> for the SCSI as well...

I think this is a special case for doing polled I/O:

        /*
         * When doing polled I/O, the SCSI bits in the interrupt register won't
         * get cleared by the interrupt processing.  This will cause the DMA
         * address registers to not load on the next DMA transfer.
         * Check for these bits here, and clear them if needed.
         */

As the comment alludes, generally the correct place is to do it in the
machine-dependent 'interrupt processing'.

	-- Gregory McGarry <g.mcgarry@ieee.org>