Hi,

This is just an amusing quirk in installing OpenBSD on softraid on GPT:

The way to set up a softraid on the installer, is to go to the (S)hell boot 
option, create GPT partition tables yourself on the involved physical block 
devices e.g. by "fdisk -igy -b 960 sd0" to make sd0 a GPT disk that's actually 
EFI-bootable by also creating an EFI boot partition on it, then you set up your 
softraid using "bioctl" which will create out of your sd0 and possibly more 
devices, a new softraid sd device, let's call it sd1 .

You then proceed with the /install.sh script, tell the installer you want to 
install OpenBSD on sd1 , and you go on telling it that you want sd1 to have a 
GPT partition table.

The installer will pass on the "-b 960" arguments to its "fdisk -ig" on sd1, 
giving us a ~512KB EFI partition inside the softraid, which will never be 
needed.

You see this by doing "disklabel sd0 && disklabel sd1" in the installed system, 
which shows both the actually-used EFI boot partition on sd0, and the unused 
EFI boot partition inside the softraid.

This is superfluous, could be removed and save users of 512KB of disk space.

Tinker

Reply via email to