[62] in linux-scsi channel archive
Re: While we're on the subject...
daemon@ATHENA.MIT.EDU (Drew Eckhardt)
Wed Feb 1 17:23:17 1995
To: Michael Adas <mja@telecom.wisc.edu>
Cc: linux-scsi@vger.rutgers.edu
In-reply-to: Your message of "Wed, 01 Feb 1995 12:44:29 CST."
<Pine.3.89.9502011229.A2855-0100000@venus.telecom.wisc.edu>
Date: Wed, 01 Feb 1995 14:02:12 MST
From: Drew Eckhardt <drew@boulder.openware.com>
--------
When can we expect to see support for disconnection in the NCR
driver?
Unknown. I haven't found much time to work on it (since about June),
and everyone who volunteers and gets a copy of my revised sources
seems to get scared off.
I'm not in a huge hurry <nudge, nudge>, but it seems
as though this would greatly improve performance with multiple
drives. If I had a clue, I'd implement it myself. Is this a
difficult thing to add for the NCR53c810,
Yes and no.
Yes : The Adaptec AIC-7{7,8}70 series and NCR53c{7,8}xx series
are intelligent, but require you to write what would have
been in firmware on a board like the 1540.
There are also some non-trivial interactions between the
host code and the code running on the SCSI chip when
error conditions are encoountered.
The driver is also not too small - about 4000 lines of
'C' and 1300 lines of NCR assembler.
No : I rewrote the disconnect/reconnect code and even debugged
it, and also rewrote the saved/active pointer code and handling
of phase changes when there's a residual left in the current
scatter/gather segment, so getting everything working is as simple as
debugging the new SCSI pointer code.
or just not a priority issue right now?
Yes. I don't have a tape drive which would anoy me enough to
fix it, and am currently more interested in playing with my antique
Toyota Landcruiser FJ40 (350 Chevy under the hood, 4.11 gears, Saginaw
power steering, and almost four wheel power disk brakes so the Woa matches
the Go).
One more thing.. I have an AHA1542C in addition to the on-board
NCR53c810 (486 Saturn II board), and I have had *no* luck in
getting linux to find both hosts. It sees the 1542, but not
the ncr.
I tried lilo command line options without success.
There was (is? I haven't looked at any kernel newer than about 1.1.40) an
alignment problem which would cause the NCR driver to fail and uninstall
itself (but you'd get messages to this effect). A patch was submitted
to Linus, but I don't know if it was ever included in the distribution kernel.
--------
