My original question:
We are having trouble formatting a Seagate Wren VII 94601-1.2 drive.
The format.dat entry in SunOS 4.1 is wrong, it gives ncyl=1703 pcyl=1931.
The data sheet with the drive gives pcyl=1931 too. I have been collecting
format.dat entries from the net for some time but I don't have one for
this drive. Anyone out there like to send me one?
Very large and varied response. Most people said that the one they suggested
was actually in use at their site.
Abbreviated collection:
ncyl nhead nsect acyl pcyl total sects comments
? ? ? 2 1925 ???
1956 15 69 ? ? 2024460 from the nieusma collection
1925 15 70 2 1931 2021250 in use
1920 15 71 2 ? 2044800 preformatted by Artecon
1928 15 70 2 1930 2024400 in use
1925 15 70 2 ? 2021250 as the one 3 up
1068 31 61 1 1069 2019588 dagobah!mis see below
1703 15 80 2 1931 2043600 Sun 4.1 format.dat
1929 15 69 2 1931 1996515 in use
There was a rumour that the 80 blocks in the Sun 4.1 entry is too many for
the inner tracks hence the 1703 tracks used. The strangest one in the list
is the one from dagobah!mis (mis@seiden.com). I quizzed him on this and got the
following reply:
----- Begin Included Message -----
the theory is the number of sectors per track varies across the disk
in big wrens (constant within a zone).
so the idea of a cylinder group in the berkeley file system
is a joke anyway. there is no guarantee that with any assumption
about constant sized cylinders that cylinder groups are aligned on
a cylinder boundary.
scsi disks are addressed as a linear array of blocks. so the hack is
to *ignore* the number of cylinders in the manual, but take the
maximum logical block number and factor it into plausible factors
cyl*hd*sector = maximum lbn.
the embedded scsi controller has no idea of what a cylinder or a head
are anyway, it just knows how to address blocks.
"trust me" (my two favorite words in the english language...)
mis@seiden.com, berkeley.edu, usenix.org, seiden@eng.sun.com
----- End Included Message -----
This makes a lot of sense, and I'll give it a try.
The following comments were made by bit!markm@cse.ogi.edu :
>Also, once you format the drive, I reccommend that you perform a
>simple test or two using the 'c' partition. Make sure that you can
>fill the entire disk and also that you can fsck it. There is a
>limitation in SunOS when using scsi disks which limits the maximum
>disk blocks to 2097152 (or 1.0737GB). You may have to play some
>games with the cylinder count and sectors per track to make it work.
This does not effect the configurations above but is worth keeping in mind.
I suspect that the other factors in the format.dat such as bpt and rpm are
significant. Not everybody sent a full format.dat entry, but I did get these:
#:: Entry from Apunix: 2024400
disk_type = "Imprimis Wren VII 94601-12G" \
: ctlr = MD21 : fmt_time = 4 \
: cache = 0x11 : trks_zone = 1 : asect = 1 \
: ncyl = 1928 : acyl = 2 : pcyl = 1930 : nhead = 15 : nsect = 70 \
: rpm = 3600 : bpt = 33408
#:: dagobah 2021479
disk_type = "CDC Wren VII 94601-12Gphony" \
: ctlr = MD21 : fmt_time = 8 \
: cache = 0x11 : trks_zone = 9 : asect = 3 \
: ncyl = 1068 : acyl = 1 : pcyl = 1069 : nhead = 31 : nsect = 61 \
: rpm = 3600 : bpt = 20880
#:: sunos 4.1 2043600
disk_type = "CDC Wren VII 94601-12G" \
: ctlr = MD21 : fmt_time = 4 \
: cache = 0x11 : trks_zone = 15 : asect = 5 : atrks = 30 \
: ncyl = 1703 : acyl = 2 : pcyl = 1931 : nhead = 15 : nsect = 80 \
: rpm = 3597 : bpt = 41301
#:: 1996515
disk_type = "CDC Wren VII 94601-12G" \
: ctlr = MD21 : fmt_time = 4 \
: cache = 0x11 : trks_zone = 15 : asect = 3 \
: ncyl = 1929 : acyl = 2 : pcyl = 1931 : nhead = 15 : nsect = 69 \
: rpm = 3600 : bpt = 33408
>From all of this it is apparent that there is no "right" answer to all this.
Different sites do different things which work. So how come the Sun one
doesn't? This request was initiated by a friend at another site who is not
on the list, and I took it up as we are expecting delivery of a number of these
disks in the next few days. Our Wren V disks arrived preformatted, and I suspect
that the wren VIIs will too. However "format" will not be able to repair
bad blocks if the format.dat does not coincide with the format on the disk,
so reformatting, or putting in a format.dat which coincides with what is
on the disk will be necessary. I will take it as it comes.
Thanks all, and I hope you are all less confused about this than I am!
Alastair Young
Systems Supervisor
EuCAD
alastair@eucad.uucp
(alastair%eucad.uucp@ukc.ac.uk)
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:05:59 CDT