[4667] in linux-scsi channel archive
Re: New ncr53c8xx driver experimental version.
daemon@ATHENA.MIT.EDU (Richard Waltham)
Mon Sep 7 06:07:18 1998
From: Richard Waltham <dormouse@farsrobt.demon.co.uk>
To: groudier@club-internet.fr (Gerard Roudier)
Date: Mon, 7 Sep 1998 04:50:52 +0100 (BST)
Cc: linux-scsi@vger.rutgers.edu, ncr53c810@Colorado.EDU
In-Reply-To: <Pine.LNX.3.95.980905115541.1271A-100000@localhost> from "Gerard Roudier" at Sep 5, 98 11:59:11 am
Hi Gerard
Gerard Roudier wrote:
>
>
> Hello,
>
> I have made available a heavily changed ncr53c8xx driver at:
> ftp://ftp.tux.org/pub/roudier/896/
> (One tar file + 3 patches)
>
> This driver version is primarily intended for support of the new
> 64 bit PCI 53C896 Symbios controller.
> For now, I donnot know if this code actually works for the 896 since
> I donnot have any appropriate hardware for testing this.
>
> I donnot like to announce too incomplete code, but this one is very
> interesting in my opinion. The new driver uses LOAD/STORE SCRIPTS
> instructions. This makes it unable to work with normal 810, 815 and
> 825 controllers, but it supports all recent SYMBIOS chips, e.g:
>
> 810A (rev >= 16)
> 825A (rec >= 16)
> 860
> 875/876
> 895
> 896 once it will be tested.
>
> Generic NCR53C8XX controller SCRIPTS use MOVE MEMORY scripts instructions
> in order to support the whole 53C8XX family. This instruction is only able
> to use absolute addressing mode and is coded with 3 DWORDS.
> LOAD/STORE also support absolute mode, support DSA relative addressing
> mode and is coded using only 2 DWORDS.
> Most of things that were nasty to implement with MOVE MEMORY becomes easy
> and clean using LOAD/STORE. No need to implement self-modified SCRIPTS
> code for indirect addressing and coding SCRIPTS becomes a lot easier and
> clearer.
>
> Win against the stock ncr53c8xx driver is:
>
> - Simpler driver code.
> - Faster SCRIPTS code (12 micro-seconds / command actually measured).
> - No self-modifying SCRIPTS patterns.
>
> Obviously, this driver is not usable on systems that still have
> old-designed 810, 815 or 825 controllers.
>
> BTW, this driver works fine on my system:
>
> - 810A attached to 1 CD TOSH 3401B.
> - 875 1 Atlas I W + 2 ATlas II UW + 1 DDRS UW.
> - 895 1 Cheatah LVD + 1 Viking II LVD.
>
> It also seems not to have bad effects on Guinea pigs I use for
> experimenting new code. :-)
>
Works well here - with kernels 2.0.35, 2.0.36-pre6/7, 2.1.119/120. Don't
seem to be able to break it at all. I must try harder:)
I can force errors using my CD-Rom changer and it handles them very well. -
unexpected disconnects and timeouts on scsi bus hangs.
Would be nice to get the rest of Linux working as well! ;)
Bye for now
Richard
>
> Regards,
> Gerard.
>
> PS: Donnot ask me if the stock ncr53c8xx maintainance is dropped.
> I will drop support for old NCR53C8XX chips in the ncr53c8xx driver at
> the time Linux will drop support for 486 processors. :-)
>
--
Richard Waltham | Work: richard@digitalinterface.demon.co.uk
At home in | Home: dormouse@farsrobt.demon.co.uk
Southampton UK | 100421.1276@compuserve.com
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.rutgers.edu
