Re: SCSI status

larz>>Note that the initial SCSI driver will have a number of limitations,
larz>>including:
larz>> 1) disk support only.
larz>>
larz>> 2) single server - we had talked in the past about having
larz>> seperate PDRV and XPT/SIM servers. For now, however,
larz>> everything will run in one process.

nick>I guess both of these could be changed easily later?

Yes, the CAM driver can be split into separate PDRV and XPT/SIM processes
later. I think this is one of the nice things about CAM - its division
of labor divided between 3 well defined layers. The middle layer, the
XPT, is responsible for delivering CCB's from the PDRV's to the SIM's.
In theory, the PDRV's and SIM's don't care how the CCB's are delivered.

In practice, the following issues will have to be addressed to implement
the split:

        1) how to get CCB's from the PDRV's to the SIM's (we touched
           upon this subject in our earlier discussions)

        2) asynchronous and I/O completion callbacks

        3) what XPT functionality resides in the PDRV's? At the very
           least, there needs to be function interfaces that reside
           in the PDRV's that send message requests from the PDRV to
           the XPT/SIM.

        4) data structures that may be shared among processes

        5) how to handle multiple XPT/SIM processes (for multiple
           types of host adaptor cards).

I don't think any of the issues are difficult ones, but they will take
time to implement. Right now, I'd rather spend my time working on a
solid, but somewhat less dynamically configurable, single server
solution and leave the multi server solution to another revision of the
driver.

                                        Mike Larson
Received on Wed Dec 1 08:53:46 1993