git: 2c26d77d989a - main - Remove /boot/efi from mtree, missed in 0b7472b3d8d2.

[Nathan Whitehorn]

On 3/2/21 1:24 PM, Rodney W. Grimes wrote:
>> The branch main has been updated by nwhitehorn:
>>
>> URL: https://cgit.FreeBSD.org/src/commit/?id=2c26d77d989abe48c662eeb6f52f7e4c9b81680c
>>
>> commit 2c26d77d989abe48c662eeb6f52f7e4c9b81680c
>> Author:     Nathan Whitehorn <nwhitehorn at FreeBSD.org>
>> AuthorDate: 2021-03-02 16:47:00 +0000
>> Commit:     Nathan Whitehorn <nwhitehorn at FreeBSD.org>
>> CommitDate: 2021-03-02 16:49:41 +0000
>>
>>      Remove /boot/efi from mtree, missed in 0b7472b3d8d2.
>>      
>>      This had prevented the bootconfig step from determining if an ESP exists,
>>      resulting in its unconditional setup. On BIOS-booted amd64, this wasn't
>>      harmful, just unnecessary, but it resulted in failed installations on
>>      non-EFI-supporting platforms like powerpc64.
>>      
>>      MFC after:      3 days
> We certainly must have a better way to detect EFI or not
> than the existance of /boot/efi????
>
> I would of thought that machdep.bootmethod == EFI was the
> defanitive way to detect this?
>

This is about the installer talking to itself internally. That sysctl 
doesn't exist on many platforms, and the decision of whether to set up 
an ESP is a policy decision only loosely related to whether the current 
system booted by EFI or not. As the commit message notes, this is to 
detect whether an EFI System Partition exists and is configured, which 
is *not* the same thing as whether the system booted by EFI. As such, 
this is a more direct check.
-Nathan