Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!samsung!uunet!mcsun!unido!uni-koeln!snert.ikp!se
From: se@snert.ikp (Stefan Esser)
Newsgroups: comp.unix.ultrix
Subject: Re: Help with FUJITSU SCSI disk drive
Message-ID: <1991May5.003915.115024@rrz.uni-koeln.de>
Date: 5 May 91 00:39:15 GMT
References: <1991May3.162032.5909@bnlux1.bnl.gov>
Sender: news@rrz.uni-koeln.de (Usenet News System)
Organization: Regional Computing Center, University of Cologne, F. R. Germany
Lines: 66

In article <1991May3.162032.5909@bnlux1.bnl.gov>, kneifel@bnlux1.bnl.gov (Charley Kneifel) writes:
|> Hello,
|> 
|> 	We just got a Fujitsu M2263 SCSI drive to hook up to a DS3100
|> running Ultrix (currently 3.1, going to 4.1 next week).  I am able to
|> format the disk using the rzdisk utility and have mounted it
|> successfully.  I made a guess at the disktab entry from the hardware
|> manual. 

|> Here is my disktab entry:
|> 
|> m2263|M2263|M2263S Fujitsu:\
|>         :ty=winchester:ns#54:nt#15:nc#1632:\
|>         pa#32768:ba#8192:fa#1024:\
|>         pb#131072:bb#8192:fb#1024:\
|>         pc#1312300:bc#8192:fc#1024:\
|>         pd#382820:bd#8192:fd#1024:\
|>         pe#382820:be#8192:fe#1024:\
|>         pf#382820:bf#8192:ff#1024:\
|>         pg#1148460:bg#8192:fg#1024:

Your hardware manual seems a bit different from mine, 
here are the values for my M2263 (partitions except 'c' deleted ...):

m2263|M2263|Fujitsu M2263 (660 MByte):\
        :ty=winchester:ns#53:nt#15:nc#1657:\
        :pc#1312344:bc#8192:fc#1024:

The manual says:
	  53 sectors per track
	  15 heads
	   3 reserved sectors per cylinder (for bad block replacement)
	1658 tracks
	   1 reserved cylinder (for bad block replacement)

==>	(53 * 15) - 3 = 792 sectors per cylinder
	(1658 - 1) * 792 = 1312344 sectors of 512 bytes = 656172 KByte

One problem with this is, that the Berkeley Fast File System expects 
the same number of sectors on each track of a cylinder. 
But SCSI drives usually have some number of alternate sectors on 
the last track of each cylinder. 
The M2263 typically uses sector 1 to 53 on heads 1 to 14 for data storage, 
but only sector 1 to 50 on head 15 !
That's the reason you can't simply multiply 'SecPerTrk * Heads * Cyl' when 
calculating the capacity of SCSI disks.
(And that's the reason, newfs will complain about unused blocks in the 
 last cylinder of the partitions. The partitions will become multiples of 
 795 sectors, while the disk size is a multiple of 792 sectors ...)

There is another problem with 'Heads * SecPerTrk != SecPerCyl', because 
the Berkeley FFS expects just this, when allocating blocks for new files.
That makes the file layout policies of Ultrix choose rotationally distant 
blocks, after the disk has been used for a while. This is also a result of 
other 'features' of SCSI (track skew, cylinder skew).
I've made a patch to the file system to avoid this problems, but I think 
its to much trouble for the average owner of a workstation.

I think that Sun has made the necessary changes to the file system 
to make good use of SCSI disks, but on Ultrix (even 4.1) the file system 
doesn't know of the special features of SCSI disks.

Stefan
-- 
 Stefan Esser, Institute of Nuclear Physics, University of Cologne, W. Germany
 se@ikp.Uni-Koeln.DE                                            [134.95.192.9]