LOADER_DISK_SUPPORT is also checked in stand/defs.mk, otherwise looks good to me.

Better wording

In D57568#1319455, @kib wrote:

In D57568#1319454, @vexeduxr wrote:

In D57568#1319451, @kib wrote:

why not change the word 'protected' to 'long'?

I'm not sure how the CPU mode is relevant to the rest of the text. Shouldn't the CPU be susceptible to faults regardless of the mode?

It is quite involved to make CPU fault in real mode. In VM86 mode it is somewhat easier but still not straight. The mention of the mode is to compare it with the old pre-EFI BIOS where loader practically cannot cause CPU faults.

Implement kib's suggestions.

Keep the sentence intact but change x86 protected mode to 64-bit long mode.

In D57568#1319451, @kib wrote:

why not change the word 'protected' to 'long'?

Can confirm that it boots fine with the things I mentioned below fixed.
Thanks!

Wouldn't the ia32 loader need this too?

I wanted to let this sit for a while since I didn't like that we had to build everything twice, but I can't seem to find a better solution for now.

Make the LIB*32 variable assignment more seamless.

Landed, but forgot the tag the review :/
Apparently I can't mark this as closed as it hasn't been accepted.

Small style tweak

Don't do anything until we know that all the functions we need are present.
Also invert the first if statment to lessen the indentation.

Bump.
I'd like to get this into stable/15 before releng/15.1 branches

Looks good. Sorry about the oversight!

In D55453#1269653, @brooks wrote:

@vexeduxr please commit with kib's suggestion

s/tmpdir/tmpmnt

In D55469#1269614, @jrtc27 wrote:

Can you please include an appropriate description of the effects in the commit message?

Add comments explaining the missing header guard and the use of SIZE_TYPE

In D55453#1269284, @brooks wrote:

I don't think the commit text is specific enough about the cause. You should point to a specific combination that fails, not claim its is was macOS only (I'm fairly sure I added this due to testing cross build on ubuntu).

Add comment noting that the ESP is mounted.
Rename ZFS mirror loop variable so mntpt can be reused.

Provide the full loader path as an argument, instead of just the file name.

In D55469#1269123, @jrtc27 wrote:

What is the effect of this bug?

It happens if anything other than __need_size_t is defined. Some headers include stddef.h but don't want the whole thing, so they define __need_*. The header can also be included multiple times with different __need_* macros defined.
This is all internal of course, simply including stddef.h with nothing defined gives you the whole thing.

Only pass -malign-double for the 32-bit build.
Suffix 32-bit EFI libraries with efi

Looks okay to me.

In D55140#1264432, @quentin.thebault_defenso.fr wrote:

In D55140#1264381, @vexeduxr wrote:

splash is set to /boot/images/freebsd-logo-rev.png by default. Should the same be done for the shutdown splash?

This is already the case since we call into the existing vtterm_splash code where this case is handled.

splash is set to /boot/images/freebsd-logo-rev.png by default. Should the same be done for the shutdown splash?