kern/131037: Unable to create disklabel on .eli-Device

[Hannes Hauswedell]

>Number:         131037
>Category:       kern
>Synopsis:       Unable to create disklabel on .eli-Device
>Confidential:   no
>Severity:       non-critical
>Priority:       high
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Tue Jan 27 11:00:12 UTC 2009
>Closed-Date:
>Last-Modified:
>Originator:     Hannes Hauswedell
>Release:        7.0RC2
>Organization:
>Environment:
FreeBSD fbsdmain.lan.home 7.0-RC2 FreeBSD 7.0-RC2 #0: Fri Feb  8 00:09:57 UTC 2008     root at logan.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC  i386

>Description:
I am having serious problems creating a label on a geli'ed slice.

I have posted this on freebsd-geom, but havent gotten any repsonses so far. It is quite urgent for me to get this running though and it does look like a bug, so I posted here.

I can create the slice, create the geli container and write a standard label to it. As soon as I edit the label and save it (even if not changing anything from the default) I get:

WARNING: Expected rawoffset 487974, found 487974

(Or another number depending on where I created the slice...)


Which looks strange in itself - after all, what kind of messages does GEOM give if does NOT find the expected offset? ^^


I tried to ignore the warning and just started copying data on a 
partition inside the slice, but got a reboot with geli-write errors (=5)
in /var/log/messages.

I tried repartitioning and relabeling multiple times to no avail.

Thanks for any help!

Hannes


P.S: ar0 is a soft-raid of 1TB Samsung F1 EcoGreen and 1TB WD Cavier Green EADS. The errors also occur when working on each disk individually. 
>How-To-Repeat:
geli init -l 256 -K MYKEY -s 4096 /dev/ar0s2
geli attach -k MYKEY /dev/ar0s2
bsdlabel -w /dev/ar0s2.eli
bsdlabel -e /dev/ar0s2.eli
(change nothing and exit EDITOR)
>Fix:
I tried to read the part in sys/geom/geom_bsd.c that produces the warning, but don't understand it :(
Then again it is prefixed by:
/* Historical braindamage... */ 

So I don't know if anyone else does :/

>Release-Note:
>Audit-Trail:
>Unformatted: