Re: loader warnings with LOADER_FDT_SUPPORT

On Sun, Jan 4, 2026 at 6:53 PM Warner Losh <[email protected]> wrote:

>
>
> On Sun, Jan 4, 2026 at 6:31 PM Colin Percival <[email protected]>
> wrote:
>
>> On 1/4/26 17:25, Warner Losh wrote:
>> > On Sun, Jan 4, 2026 at 6:16 PM Colin Percival <[email protected]
>> > <mailto:[email protected]>> wrote:
>> >      > No valid device tree blob found!
>> >      > WARNING! Trying to fire up the kernel, but no device tree blob
>> found!
>> >
>> >     Since many systems legitimately don't have DTBs, can we remove this
>> warning
>> >     and/or make it conditional on some check as to whether we think a
>> DTB might
>> >     be needed?
>> >
>> > diff --git a/stand/efi/loader/bootinfo.c b/stand/efi/loader/bootinfo.c
>> > index 67b36313c26f..e43eebb4bd5c 100644
>> > --- a/stand/efi/loader/bootinfo.c
>> > +++ b/stand/efi/loader/bootinfo.c
>> > @@ -432,9 +432,12 @@ bi_load(char *args, vm_offset_t *modulep,
>> vm_offset_t
>> > *kernendp, bool exit_bs)
>> >   #if defined(LOADER_FDT_SUPPORT)
>> >          if (dtb_size)
>> >                  file_addmetadata(kfp, MODINFOMD_DTBP, sizeof(dtbp),
>> &dtbp);
>> > -       else
>> > -               printf("WARNING! Trying to fire up the kernel, but no "
>> > -                   "device tree blob found!\n");
>> > +       else {
>> > +               if (getenv("acpi.revision") == NULL) {
>> > +                       printf("WARNING! Trying to fire up the kernel,
>> but no "
>> > +                           "device tree blob found!\n");
>> > +               }
>> > +       }
>> >   #endif
>> >          file_addmetadata(kfp, MODINFOMD_KERNEND, sizeof(kernend),
>> &kernend);
>> >   #ifdef MODINFOMD_MODULEP
>> >
>> > ???
>>
>> Yeah, I was wondering if it was as simple as "if we have ACPI we don't
>> need FDT".  (There's two printfs to change though.  If people agree with
>> the concept I'll put together a patch, test, and post it for review.)
>>
>
> It's complicated by the fact that sometimes you need both (the LinuxBoot
> interface requires both). So we have to be careful. We want to warn when we
> have NEITHER and I think the above does that. But BOTH is OK, even though
> the kernel side of both is still buggy.
>

Also https://reviews.freebsd.org/D54510 so you don't have to do it.

Warner