On Sun, Apr 20, 2003 at 10:13:48PM +0200, Manuel Bouyer wrote:
> Couldn't you do it at attach time ? Doing it a first access defeats the
> SCSI_DELAY delay before probe.
> Most others adapters do (or at last initiate) the reset at attach time.

What I would like to see, is that the scsipi layer does the reset.
I.e. that scsibus_config() uses a new ADAPTER_REQ_BUS_RESET request
to reset a specified channel, then waits SCSI_DELAY seconds, and
then starts probing.

Something like:

	* issue new req ADAPTER_REQ_GET_PARAMS to channel, which may
	  contain things like "should we reset this channel or not",
	  and perhaps other info

	* if reset is wanted, issue ADAPTER_REQ_BUS_RESET

	* wait SCSI_DELAY seconds.

	* start probing.

Having ADAPTER_REQ_BUS_RESET also means that you can move the bus reset
ioctl into the scsipi layer, and not have each driver implement it.

- Frank

-- 
Frank van der Linden                                            fvdl@netbsd.org
===============================================================================
NetBSD. Free, Unix-like OS. > 45 different platforms.    http://www.netbsd.org/