Re: xbd2 is not created during installation

[Sarton O'Brien <bsd-xen%roguewrt.org@localhost>]

On Thu, 8 May 2008 11:44:59 pm Michael Dexter wrote:
> I was experimenting with disk naming and multiple disk entries such as:
> 
> disk = 
[ 'file:/root/netbsd.img,hda,w' , 'file:/root/netbsd-swap.img,hdb,w' , 
'file:/root/i386cd-3.1.iso,cd0a,r' ]
> 
> disk = 
[ 'file:/root/netbsd.img,wd0d,w' , 'file:/root/netbsd-swap.img,wd1d,w' , 
'file:/root/i386cd-3.1.iso,cd0a,r' ]

The device entries generally don't apply to an nbsd domu. You could replace 
hda and wd0d with a '0' and all would still function. I simply use 0, 1 etc.

Linux is a different matter and the device names must align.

> These appear to work and the CD-ROM image, mapped to xbd2* is visible in 
dmesg:
> 
> xbd0: 1024 MB, 512 bytes/sect x 2097152 sectors
> xbd1: 97 MB, 512 bytes/sect x 200000 sectors
> xbd2: 207 MB, 512 bytes/sect x 425408 sectors
> 
> However, a device is not created by the installer for use with installation 
or on the system it is installing to. 'cd /dev && sh MAKEDEV xbd2' solves 
this problem but is this normal sysinst behavior?

This is probably more an oversight. I noticed you are mounting a separate swap 
image. Technically this isn't required. For a nbsd domu I'd disklabel the vnd 
to to include a swap area. For linux the same can be obtained via fdisk.

This limits your block devices to xbd0a, xbd0b and xbd1. I can confirm this 
scenario works just fine from sysinst.

You may want a separate swap image and this could be irrelevant. Including 
swap in the primary image keeps everything a little more contained though.

Sarton