Subject: Re: New device buffer queue strategy
To: Jason R Thorpe <thorpej@wasabisystems.com>
From: Manuel Bouyer <bouyer@antioche.eu.org>
List: tech-perform
Date: 09/04/2002 22:07:32
On Wed, Sep 04, 2002 at 12:56:14PM -0700, Jason R Thorpe wrote:
> On Wed, Sep 04, 2002 at 09:40:42PM +0200, Manuel Bouyer wrote:
> 
>  > I think the right fix for this is to make MAXPHYS variable per driver.
>  > As I said, SCSI drivers could then use a larger MAXPHYS and this could
>  > help a lot with modern hardware.
> 
> In order to do this, you need a way to inherit from the parent bus.  I.e.
> there might be some bridge in the middle that clamps transfers at 32K, or
> something like that...

MAXPHYS has to stay probably, as a global variable. It should represent
the max transfert possible for the architecture (I suspect there are such
limits on some), not the lower of what all drivers can do as it is now.
Then drivers can clamp it down if needed.
Making this MAXPHYS a bus property would be a second step I guess, and
won't probably be done on all ports at once.

For example on i386 we could bump MAXPHYS to 256k if the IDE drivers could
tell that they don't want transfers larger than 64k. This would help for
modern SCSI adapters.

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
--