I think we should add the attribute on the lowest-level code that
does the firmload operation.  Maybe one day we can write a validator?

Otherwise we'll manually through inspection catch the special cases
for kernels which miss it, which is rare enough...  probably such
effort is good enough since the target not really important -- it is
for shrinking kernel size / ability.

>On Tue, Nov 03, 2026 at 01:52:43PM -0000, Crystal Kolipe wrote:
>> The psp device relies on the loadfirmware() function, but doesn't include it,
>> so removing all other devices that do from the kernel config breaks the build.
>
>good point, thanks!  I wonder if we should handle this in sys/conf/files
>as the firmware loading is implemented in dev/ic/psp.c?  Like this:
>
>diff --git a/sys/conf/files b/sys/conf/files
>index 389e63cf1bc..f2a2cf8bee9 100644
>--- a/sys/conf/files
>+++ b/sys/conf/files
>@@ -480,7 +480,7 @@ device	ccp {}
> file	dev/ic/ccp.c			ccp
> 
> # AMD Platform Security Processor
>-device	psp: xcall
>+device	psp: xcall, firmload
> attach	psp at ccp
> file	dev/ic/psp.c			psp	needs-flag
> 
>
>