On Sat, 16 Jan 2010 14:37 +0100, "Manuel Giraud" <manuel.giraud@univ-nantes.fr> wrote:
quoted text
> Otto Moerbeek <otto@drijf.net> writes:
> 
> >> Here's a probably stupid question: since the kernel can detect the "root
> >> on sd0a" why is there still a need for fstab entry for it?
> >
> > Because you might want to specify mount options, or an alternate root.
> 
> In fact, I was wondering because I have installed OpenBSD on an usb
> flash drive.
> 
> I use softraid and have a script to decrypt the RAID partition and setup
> a custom fstab with the correct 'sd?' for decrypted devices, it works
> alrigh. But if root is not sd0a, I have to 'boot bsd.rd' and 'ed
> /etc/fstab' before. Does anybody doing this and have a better solution?

Not sure I understand, but I have similar softraid crypto setups and there's no need to boot bsd.rd to edit /etc/fstab. When booting bsd or bsd.mp and you are dumped to sh to run bioctl, use ed to correct /etc/fstab there.

Also, in my experience, this is not an issue unless you are adding and removing sd devices. For example, the physical volume may be wd0 and the softraid volume may be sd0 at the moment, but when you insert a USB stick and reboot, then that USB stick will become sd0 and the softraid volume will become sd1. In that case /etc/fstab must be edited. I think Marco is working on a general fix for this.

Brad

quoted text
> 
> -- 
> Manuel Giraud