What about to use a flag for each scsi id ?
I think 16-255 flags for the scsi devices for a scsi controller
can be statically allocated at the SCSI controller's driver.
Then, the flags can be read and activated by scsipi layer.
I think these flags can be reconfigured by userconf (with some 
modification).

Sung-Won Chung

On Mon, 26 Aug 2002 at 19:54:37 +0200, Manuel Bouyer wrote:
>On Mon, Aug 26, 2002 at 10:30:02AM -0700, Andrey Petrov wrote:
> > On Mon, Aug 26, 2002 at 12:43:36PM +0200, Manuel Bouyer wrote:
> > > On Sun, Aug 25, 2002 at 11:40:58PM -0700, Andrey Petrov wrote:
> > > > >
> > > > > I really think these flags should be handled by the SCSI device 
>driver,
> > > > > not the SCSI HBA driver, much like is done for IDE.
> > > > >
> > > >
> > > > That might require 'wiring' everything (controllers and devices) in 
>config
> > > > file. Does it?
> > >
> > > Yes, it does. But this way you could say "use sync transfers for disks 
>but
> > > not CD", instead of assuming disks are id <= 3 and cds are id > 3
> > >
> > You can do it now because flags are 3 bits per target, and you can force
> > 'don't disconect', 'don't sync' and 'no tags', and because of that
> > there are no room for 8 more devices, so the question is where to find 
>that
> > place. Seems 'what Matt said'.
>
>What I'm saying is that instead of finding more bits here, we could move
>this to per-device flags. With scsipi, this would be trivial (exept the
>"don't disconnect", but we could add this too without much work).
>Doing it per-device would have some advantages (see what I said), the
>disavantage being that this would require hardwiring devices to
>scsibusses to controllers if we want to do it per-controllers.
>
>If we don't go that way, at last I think these flags should be moved to
>scsibus, so that we have an uniform way of controlling that.
>
>--
>Manuel Bouyer <bouyer@antioche.eu.org>
>--




_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com