Subject: Re: ahc bug in current? (was: ccd changed in current?)
To: Jason Thorpe <>
From: Justin T. Gibbs <>
List: current-users
Date: 08/28/1997 20:59:28
>On Thu, 28 Aug 1997 18:29:59 -0700 
> "Michael L. VanLoon --" <> wrote:
> > The ccd driver seems to be exonerated for now... :-)
>*Whew*  :-)
> > Both running parallel fsck's, and parallel dd's locked up the system
> > tight.  I didn't even have ccd configured during the tests, so it's
> > not a factor.
>You are just the person I'm looking for :-)
>Here's the scoop: I have noticed that problem on an important production
>system, but can't debug it there since I can't have it crashing all the 
>time, so it's running an older kernel.
>I can't reproduce this problem on my i486 EISA system that has an ahc
>in it.
>I have two theories:
>	(a) The system gets stuck due to an unbalanced splbio() call.

You should be able to still break into DDB if you are stuck at splbio().

>	(b) The ahc driver sets up some sort of PCI DMA parameter that
>	    loses with 486 PCI chipsets.  (Note, I have not heard any
>	    reports of this problem on Pentiums or greater, and have
>	    not been able to reproduce the problem on an AlphaServer
>	    8200 with an ahc connected to several RAID boxes doing
>	    performance measurements).

PCI DMA paramters???  The only settings I know of that affect PCI transfers
are the latency timer and cache line size settings.  Neither of these 
settings are touched by the driver since "The BIOS Knows Best" and sets
these settings up just fine.

>Justin, were there any changes related to this?

The only change that comes anywhere near this type of problem was the 
general redesign of the driver to use DMA to pull in SCBs.  I wanted
to do this anyway for efficiency reasons, but it turned out that some
revs of the P6 had problems with the repeated accesses caused by 
"PIO'ing" down the SCBs.

Of course, there are lots of other bug fixes that are in the FreeBSD
version of the driver that NetBSD doesn't have yet, but nothing that
would cause these kinds of problems.

The other interesting thing about this problem is that the aic7xxx
driver in NetBSD hasn't changed much for a long time which makes
it even less likely that a recent change to the aic7xxx driver is
causing the problem.

>Jason R. Thorpe                             
>NASA Ames Research Center                            Home: +1 408 866 1912
>NAS: M/S 258-6                                       Work: +1 415 604 0935
>Moffett Field, CA 94035                             Pager: +1 415 428 6939

Justin T. Gibbs
  FreeBSD: Turning PCs into workstations