git: 0b7472b3d8d2 - main - Mount the EFI system partition (ESP) on newly-installed systems.

[Kyle Evans]

On Mon, Mar 1, 2021 at 10:45 AM Nathan Whitehorn <nwhitehorn at freebsd.org> wrote:
>
>
>
> On 3/1/21 11:42 AM, Kyle Evans wrote:
> > On Mon, Mar 1, 2021 at 10:39 AM Nathan Whitehorn <nwhitehorn at freebsd.org> wrote:
> >>
> >>
> >> On 2/28/21 3:44 PM, Brandon Bergren wrote:
> >>> On Sun, Feb 28, 2021, at 2:25 PM, Warner Losh wrote:
> >>>> Yes. I agree as well. I was just hoping to say just that: EFI is barely
> >>>> theoretically possible, but in reality we'll likely never use it....
> >>>>
> >>>> The net effect is that we don't want to install efi on powerpc on freebsd.
> >>>>
> >>>> Warner
> >>>>
> >>> Yeah. The code before the change excluded mips and powerpc platforms, and it should continue to do so instead of using the existence of a /boot/efi directory as the only clue.
> >>>
> >>> Currently bsdinstall bails out and leaves powerpc* in a half-installed state because the die in the uname case propagates to the main script, so it never runs the bits after the bootconfig.
> >>>
> >> So that was a deliberate choice to keep the list of places that know
> >> about efi vs. non-EFI centralized. I'd prefer to just not make that
> >> directory on systems where it doesn't apply rather than messing with the
> >> installer. Do you know where it is being made?
> >> -Nathan
> > It's part of the hierarchy in ^/etc/mtree/BSD.root.dist
> >
>
> Is there a reason it needs to be? The installer bits all make it when
> needed already, so just removing it there seems like the simplest path.

I can't think of a reason, as long as both the release(7) scripts and
the installer create it as needed -- I note that vmimage.subr seems to
create it itself, but arm.subr seems to get it wrong atm. It'll need
to create /boot/efi for PART_SCHEME == GPT, and /boot/msdos should
probably be scoped down to PART_SCHEME == MBR where it's used at the
moment.