1.2gig Fujitsu M2266SA SCSI info

[Eric A Pearce]

I would *welcome* any corrections you have.

Most systems available today use the SCSI "Group 0" command set.  It has
only 21 bits to store the absolute block number for a disk block.  This
means you have a maximum of 2097152 (2^21) for the number of blocks you
can address, which implies a maximum disk space of 1.073 gig (512k disk
block * 2097152) .   There is a newer SCSI command set (extended SCSI)
which allows 31 bits for the block number, but this doesn't seem to be
currently available from Sun, Dec or SGI.  Neither Sun nor SGI were able
to say anything definite about a new driver.  Dec claims they will have
extended SCSI in Ultrix 4.2, due out in May(?).  Sun said they only
support drives up to 669meg now, and will offer 1.2gig models later this
year, so it is not clear if they are in a hurry to fix the problem.

>From this information, it doesn't sound like you should buy SCSI drives
larger than 1.2gig until the workstation vendors catch up.

For the Fuji drive, under SunOS, this limitation wastes approx 5.4 meg of
the disk space, so this isn't really significant.    newfs will fail with
a write error if you try to use a number out of range.  Under Ultrix and
VMS, it will apparently let you format the drive, but when blocks over the
limit are accessed, it wraps around to 0 and overwrites the beginning of
the disk.   I haven't heard what happens on the SGI, but they do not
currently support anything bigger than 1.2 gig, and their format typically
eats 20% of the disk, so the problem isn't likely to occur with this
drive.  (I don't have hard experience with this)

I have several of these drives on a SparcStation II.  I formated them with
2093550 blocks, which in theory is 1.071 gig, but after newfs'ing the 'c'
partition is approx 983 meg.    This is still a good deal for the money
(under $3k)

Apparently there is firmware bug on the drive itself that prevents the
read-ahead cache from being used on the Sun (at least). The drive will
appear to format correctly, but will start spewing out read errors when
you try to use it.  You can make the drive work by disabling the cache -
remove jumper 5-6 on CN9.   Fujitsu says they will ship new proms to Fuji
distributors and you can contact them for the upgrade.  This is very new
information, Cranel (our distributor) had only heard of the problem
yesterday.

Fujitsu hinted that you could try to fool format with a magic entry that
would enable you to use more space, but I haven't been able to find out
what this is.

This is the Sun format.dat I'm using:

# offical Fujitsu blessed entry with some additions
disk_type = "CRANEL-M2266SA" \
	: ctlr = MD21 : fmt_time = 2 \
	: cache = 0x11 : trks_zone = 15 : asect = 3 \
	: ncyl = 1642 : acyl = 2 : pcyl = 1644 : nhead = 15 : nsect = 85 \
	: rpm = 3600 : bpt = 46635

I have CNH2 completely jumpered (SCSI-2 enabled) - this might not be
correct for the DecStation or SGI, but works on Sun Sparc platforms.

 Eric Pearce eap at bu-pub.bu.edu                             "Get Some!"
 Boston University Information Technology                  - Dispatches