On Mon, 20 Sep 1999 09:50:06 -0400 
 Dan Winship <danw@MIT.EDU> wrote:

 > Assuming by "Adaptec 2930", he meant the 2930 under the name Adaptec
 > markets their Mac stuff under (which I suddenly can't remember...
 > something Domain... Future Domain maybe?), it works fine with the
 > existing Adaptec driver, but the macppc needs this patch to
 > sys/dev/ic/aic7xxx.c to make any adaptec card work:
 > 
 > *** aic7xxx.c.orig	Mon Sep 20 09:46:57 1999
 > --- aic7xxx.c	Mon Sep 20 09:47:53 1999
 > ***************
 > *** 522,527 ****
 > --- 522,528 ----
 >   
 >   	bus_space_write_1(st, sh, HCNTRL, CHIPRST | PAUSE);
 >   #endif
 > + 	DELAY(1000);
 >   	/*
 >   	 * Ensure that the reset has finished
 >   	 */
 > 
 > 
 > We haven't committed this because there's more consensus on the side
 > of "it's a bug somewhere else in the macppc bus code" than "it's an
 > bug in the Adaptec driver"

No, it's probably not a bug in the macppc bus code.

I think this change should be committed; on PCs and some Alpha chipsets,
it doesn't matter, but on other Alpha chipsets, and apparently on the
macppc core logic chipsets as well, you will get a machine check if the
PCI target doesn't respond.

I had to do something like this in the new Tulip driver, too; the Tulip
doesn't respond for 50 PCI cycles after a reset it initiated.

        -- Jason R. Thorpe <thorpej@nas.nasa.gov>