>It is possible that an optimization to the ahc driver at some point to use
>DMA more could have made deadlock much more likely.
>-ben

I'm well aware of the DEC 21050 bridge problem.  Using DMA actually reduces
the chance of a deadlock as it removes perhaps 95% of the host->adapter
accesses.  From my understanding of the problem, using I/O instead of mem
accesses ensures that the 21050 will perform correctly (no posted writes)
and that many (most?) motherboard BIOSes do not memory map the aic78X0
chips behind the bridge if it is a 21050.  You can, quite easily, force
the attach code in ahc_pci.c to use I/O accesses if you believe that this
may be your problem.

--
Justin T. Gibbs
===========================================
  FreeBSD: Turning PCs into workstations
===========================================