Re: Orange Pi 6 Plus works (15.1-RELEASE)

Michael Tuexen <[email protected]> Tue, 30 Jun 2026 20:53:49 +0200
Newsgroups gmane.os.freebsd.devel.arm
Message-ID <[email protected]>
> On 30. Jun 2026, at 20:13, Adrian Chadd <[email protected]> wrote:
>=20
> hi! do you have a recipe for building and flashing freebsd on the
> rockpro64 that I can try?
Doesn't dd'ing
=
https://download.freebsd.org/snapshots/ISO-IMAGES/16.0/FreeBSD-16.0-CURREN=
T-arm64-aarch64-ROCKPRO64-20260629-65251c43415a-287061.img.xz
on an SD card just work? I think this is what I did.

Best regards
Michael
>=20
> I have that hardware here (as well as bananapi and orangepi coming)
> and I'd really like to get freebsd-head builds running on these to
> help people who keep submitting us diffs. :-)
>=20
>=20
>=20
> a-
>=20
> On Tue, 30 Jun 2026 at 05:46, John F Carr <[email protected]> wrote:
>>=20
>>=20
>> On Jun 29, 2026, at 4:37=E2=80=AFPM, Sulev-Madis Silber =
<[email protected]> wrote:
>>=20
>>=20
>> i'm surprised it works at all
>>=20
>> iirc, no multiple speed cpu systems have support in fbsd so far, i =
haven't even heard it being planned, so booting current is not likely to =
help either
>>=20
>> the more hw experienced guys(/gals?) can add additional info here or =
correct me or add any roadmaps
>>=20
>>=20
>> The RockPro64 is the same sort of computer, with 4 A53 and 2 A72.  =
FreeBSD has been running on it for years.  Mostly it works without =
problems.  The issues are
>> 1. Scheduling does not know when to prefer slow and fast cores, as =
noted above.
>> 2. Performance counters don't work because the CPUs have different =
sets of counters.
>>  https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D284736
>> 3. Loading the vmm module sometimes hangs the system.  Nobody knows =
why.  Somebody speculated the difference in cores was responsible.
>>  https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D277559
>>=20
>> I remember BSD on one of the small ARM systems didn't support =
different clock speeds on different groups of cores.  You'd like to put =
the fast cores into a lower power mode until a process on a slow core =
keeps using up its time slice.  If the fast cores can't be throttled =
down the decision is harder.  There may still be a power consumption =
difference between maximum clock rate executing wait instructions and =
maximum clock rate doing real work.
>>=20
>> John Carr
>>=20
>>=20
>=20