"James C. McPherson" <[EMAIL PROTECTED]> wrote:

> A quick check of the copyright info on
>
> usr/closed/uts/common/io/scsi/targets/sgen.c
>
> reveals that another company developed the code. I don't know
> what the status of any negotiations are that may or may not
> have occurred about opening this file.
>
> Karyn - would you be able to shed some light on this from the
> legal point of view?

While I am reading this, it remimds me to the fact that in 8 weeks
we have the 20th anniversary of the first SCSI generic pass through
driver (scg). I did write it for SunOS-3.0 in the first week of August 
in 1986. Other drivers (like e.g. ASPI) did appear later (e.g. ASPI
appeared first in 1988).

While scg was written in a way that was optimized for low memory machines from 
1986 and thus only uses one instance per SCSI bus, it offers a better to use 
interface than sgen because it allows to access any SCSI drive in the system.

BTW: Sun did receive the scg source in December 1987 while I demonstrated my 
"sformat" program. At that time, Sun did not like disk formatting to be done
when the OS is running but short time later, Sun's "format" appeared ;-)

I am willing to make scg OpenSource.


The current problem with scg is that many recent Host adaptor drivers
do no longer follow the Sun SCSA whitepaper (Sun's SCSI driver architecture 
framework) and do not use the SCSI bus address cookie which results in the 
unability for scg to redirect SCSI commands to other targets than the one that
scg was "configured" for.

This could be fixed by either fixing all non-SCSA HBA drivers to again use the
official SCSI bus address cookies or by making scg create a new instance
for every possible target which would need some more kernel memory.

Another design problem of the whole SCSA framework is that you currently only
may write a generic driver for all "unowned" hardware. If you have a driver that
acts like scg, a possible specific driver will be confused by the fact that
another driver (scg) sends SCSI commands to the same target. What we really 
need 
is a generic driver (like scg) that is able to talk to _all_ targets and that 
is able to do this in a way that does not confuse the specific drivers.

I did already talk about this problem with FritS Vanderlinden and he seems to 
agree to my proposals.

Jörg

-- 
 EMail:[EMAIL PROTECTED] (home) Jörg Schilling D-13353 Berlin
       [EMAIL PROTECTED]                (uni)  
       [EMAIL PROTECTED]     (work) Blog: http://schily.blogspot.com/
 URL:  http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org

Reply via email to