Date: Mon, 24 Nov 1997 12:30:19 -0800 From: Greg Earle <earle@isolar.Tujunga.CA.US> To: bde@zeta.org.au Cc: freebsd-scsi@freebsd.org, gibbs@plutotech.com Subject: Re: FreeBSD 2.1-STABLE: Why can't I disklabel a SCSI disk? Message-ID: <9711242031.AA24312@isolar.Tujunga.CA.US> In-Reply-To: Your message of "Sun, 31 Aug 1997 11:08:48 %2B1000." <199708310108.LAA13107@godzilla.zeta.org.au>
next in thread | previous in thread | raw e-mail | index | archive | help
A few months ago Bruce Evans (and others) helped me with a disk
labelling problem I had trying to disk label a 4 Gb Seagate Barracuda to be
the same DOS-translated geometry labelled as my other existing disks.
Based on the other existing disks, I had come up with a prototype
/etc/disktab entry and had tried to use it, getting the following error:
partition c: partition extends past end of unit
Bruce Evans replied as follows:
>> xxxx1# tail -5 /etc/disktab
>>
>> seagate15150|Seagate ST15150N:\
>> :dt=SCSI:ty=winchester:ns#63:nt#255:nc#521:\
>> :pc#8385867:oc#0:
>>
>> (I changed the disktab entry to reflect the DOS remapped geometry)
>
> You didn't specify su#, so the disk size is ns*nt*nc = 63*255*521 = 8369865.
>
>> xxxx1# disklabel -w -r sd1 seagate15150
>> partition c: partition extends past end of unit
>
> oc+pc extends past end of unit.
I kind of dropped the ball on this (he said, sheepishly) as the going-bad
disk went into a lull where it wasn't having the same problems as before.
Well, sure enough, it went south over the weekend and now I have to deal
with it right away.
I tried adding Bruce's "su#" entry as follows:
nntp1# tail -4 /etc/disktab
seagate15150|Seagate ST15150N:\
:dt=SCSI:ty=winchester:ns#63:nt#255:nc#521:\
:su#8385867:pc#8385867:oc#0:
figuring that pc + oc == su.
Well, now I get a new error:
nntp1# disklabel -w -r sd1 seagate15150
disklabel: ioctl DIOCSDINFO: Open partition would move or shrink
I don't understand this at all. Which partition is "open" and why is it
considered "open"? Simply because the disk label has been read into memory?
This *seems* to be addressed by the "DIAGNOSTICS" section of disklabel(8):
DIAGNOSTICS
The kernel device drivers will not allow the size of a disk partition to
be decreased or the offset of a partition to be changed while it is open.
Some device drivers create a label containing only a single large parti-
tion if a disk is unlabeled; thus, the label must be written to the ``a''
partition of the disk while it is open. This sometimes requires the de-
sired label to be set in two steps, the first one creating at least one
other partition, and the second setting the label on the new partition
while shrinking the ``a'' partition.
but to be honest, I don't parse this paragraph at all. I tried creating an
"a" partition (same size as "c") which worked (disklabel -e) but I still get
the same thing when I try to use the disktab entry.
I tried to zero-out the present disklabel and this failed as well.
This worked back when the drive was new (i.e. had never been put into the
FreeBSD machine). Now it doesn't:
xxxx1# dd if=/dev/zero of=/dev/rsd1c bs=1 count=2
dd: /dev/rsd1c: Invalid argument
1+0 records in
0+0 records out
0 bytes transferred in 1 secs (0 bytes/sec)
xxxx1# dd if=/dev/zero of=/dev/rsd1 bs=1 count=2
dd: /dev/rsd1: Invalid argument
1+0 records in
0+0 records out
0 bytes transferred in 1 secs (0 bytes/sec)
Any enlightenment you can provide would be greatly appreciated.
(cc'd to Justin Gibbs and freebsd-scsi in case it's still too early in
Australia for Bruce to see this, also in hopes someone can shed some light
on the DIAGNOSTICS paragraph above vis a vis my DIOCSDINFO error. The box
in question is JPL's main Usenet newsfeed box, and I need to get this disk
disklabelled a.s.a.p. so I can try to back up the failed disk and get
the server back in service.)
Thanks in advance,
- Greg
P.S. This is on a Pentium 133 box running FreeBSD 2.1-STABLE.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9711242031.AA24312>