[963] in linux-scsi channel archive
Re: Adaptec -> NCR
daemon@ATHENA.MIT.EDU (Gerard Roudier)
Sun Nov 17 17:37:16 1996
Date: Mon, 18 Nov 1996 00:34:22 +0000 (GMT)
From: Gerard Roudier <groudier@club-internet.fr>
To: "Leonard N. Zubkoff" <lnz@dandelion.com>
cc: stephen@it.com.au, linux-scsi@vger.rutgers.edu
In-Reply-To: <199611172048.MAA01167@dandelion.com>
On Sun, 17 Nov 1996, Leonard N. Zubkoff wrote:
> Date: Sun, 17 Nov 1996 22:00:01 +0000 (GMT)
> From: Gerard Roudier <groudier@club-internet.fr>
>
> I am sure that the aha1542 driver of linux, at least until some 2.0.X
> revision does not probe correctly "extended BIOS translation" option for
> 1542CF controllers.
>
> How can the BT-948 guess?
> Does it assume that user has set "extentend translation" for disks
> capacities >1024x64x32 sectors?
>
> My opinion is the following:
>
> - linux ncr drivers use scsicam_bios_param() (drivers/scsi/scsicam.c).
> This function first tries to infer mapping from partition table. If that
> fails, it tries to apply SCSI2-CAM (X3T9 / 792D) etc... in order to
> assign a disk mapping in tracks, heads, sectors.
>
> - linux buslogic driver does some reverse engeniering of adaptec propriatary
> bioses to choose disk mapping.
>
> So, it is possible that users who used old adaptec controllers will
> be more fortunate with BusLogic driver.
>
> However, if drives have more than 1024x64x32 sectors they can have
> problems too.
>
> There's no reverse engineering involved. BusLogic's own BIOS uses three
> translations, 64/32, 128/32, and 255/63. If it finds one of those already
> present in the partition table, it will adopt and use the existing one, and
> ignore ths setting on the > 1GB disk translation switch in AutoSCSI. Since the
> Adaptec only uses 64/32 and 255/63 translations, which are a subset of the
> BusLogic ones, any disk formatted/fdisk'ed on an Adaptec should work, though it
> may lead to a warning from the driver that the partition table geometry does
> not match the setting of the > 1GB switch.
>
I am sorry Leonard. I was wrong about the reverse engineering from Adaptec.
I understand in your code that the driver reports only BusLogic translations:
64/32, 128/32 or 255/63, and obviously it is the right thing to do.
When existing partition table infers translation, Adaptec to
BusLogic compatibility can only be ok.
Otherwise, we only may have problems for capacities >= 2*1024*1024
and < 4*1024*1024 (1024 MB to 2048 MB).
Am I still wrong?
My Atlas 2.1GB is translated 70/62 1015 sectors. NCR BIOS seems to
be quite happy of that. Fortunately I will continue with NCR controller
family.
Just one question:
Did Adaptec copy BusLogic BIOS mappings, or is it the opposite?
Regards, Gerard.
