>>>>> "Chris" == Chris Torek <torek@BSDI.COM> writes:

>> It is really and truely an ie1.

Chris> Okay, because as I said, this problem could be expected on ie's
Chris> but not le's (abrown@eecs.harvard.edu mentioned le's).

>> If this is really a problem that should occasionally occur ...

Chris> The fact that it *is* occurring suggests that you may need an
Chris> splimp/splx around the code in ieintr, so that ie0 cannot
Chris> interrupt ie1 while it is in the middle of handling ie1's
Chris> interrupt.

	I don't know if this is the conflict, because I have found
that I can't use ie0 at all in my configuration (too many DMA
underruns), I have installed the ie1.  The ie0 port is not connected,
and is not ifconfig'd.

	I'm wondering if this is something that points to an overhaul
of most interupt driven devices in NetBSD/sparc and/or others.  This
was the worst problem I encountered with serial lockups awhile
ago... to hear it coming up again here is not pleasant.

Dave.

-- 
----------------------------------------------------------------------------
|David Gilbert, PCI, Richmond Hill, Ontario.  | Two things can only be     |
|Mail:      dgilbert@jaywon.pci.on.ca         |  equal if and only if they |
|http://www.pci.on.ca/~dgilbert               |   are precisely opposite.  |
---------------------------------------------------------GLO----------------