Any good alternative to Raspberry for Arm64?

[Mark Millard]

On 2021-Mar-28, at 17:36, Greg V <greg at unrelenting.technology> wrote:

> On March 28, 2021 11:44:25 PM UTC, Mark Millard via freebsd-arm <freebsd-arm at freebsd.org> wrote:
>> On 2021-Mar-28, at 15:52, Marcin Wojtas <mw at semihalf.com> wrote:
>> 
>>> H Mark,
>>> 
>>> niedz., 28 mar 2021 o 22:15 Mark Millard via freebsd-arm
>>> <freebsd-arm at freebsd.org> napisał(a):
>>>> 
>>>> On 2021-Mar-28, at 12:48, Klaus Küchemann via freebsd-arm <freebsd-arm at freebsd.org> wrote:
>>>>> 
>>>>>> . . .
>>>>> 
>>>>> Anyone successfully booting from nvme on any aarch64-board ??
>>>> 
>>>> As reported on the list earlier, I plugged a Optane 480 GiByte
>>>> into a MACCHIATObin Double Shot's PCIe in a system that's microsd
>>>> card had the EDK2-variant build from:
>>>> 
>>>> https://unrelentingtech.s3.dualstack.eu-west-1.amazonaws.com/flash-image-2020-07-01-mainline-tfa.bin
>>>> 
>>>> and the system just worked (/dev/nda0). I updated the FreeBSD in
>>>> this configuration and that booted too.
>>>> 
>>>> (The Optane was originally initialized while in a amd64 system
>>>> and so was boot-ready when first plugged into an aarch64.)
>>>> 
>>>> This is not the MACCHIATObin's normal configuration and things
>>>> were put back to normal after the experiment.
>>>> 
>>>> The MACCHIATObin Double Shot has 4 Cortex-A72 cores.
>>>> 
>>> 
>>> MacchiatoBin has fully opensource firmware, with the mainline
>>> edk2/TF-A one can boot FreeBSD with DT or ACPI. The referenced image
>>> from Greg should also be fine. The only significant interface missing
>>> is the built-in PP2 NIC.
> 
> Interestingly, OpenBSD already has a mvpp2 driver. I don't really have the motivation to port it right now, but hey, someone could…
> 
>> My memory was that there was some issue of "legacy"
>> PCIe devices showing up as duplicated and some change
>> was put in that avoided that --but prevented "modern"
>> PCIe devices from working. As I remember, disabling
>> the change was one reason Greg created his variant:
>> upstream choose which type of devices it wanted to
>> support and Greg went the other way. (No one has
>> support covering both contexts on the MACCHIATObin
>> Double Short?)
> 
> Last thing I remember, Marcin promised to make a setup menu toggle for the offset ;)

https://github.com/tianocore/edk2-platforms/blob/master/Silicon/Marvell/Armada7k8k/AcpiTables/Armada80x0McBin/Pcie.h

has not changed since 2019-Jul-23 . So if something was
done that covers this, it was done elsewhere without
changes here. (I've no general knowledge of the edk2
software structure to know where else to possibly look.)

>> removal of a headless flag in FADT 
> 
> I don't know if that even affects anything, I probably just copied the change from somewhere, but I like it out of a pure desire for correctness.
> 
>> There were also edk2 additions of 3 lines
>> involving something like:
>> 
>> MdeModulePkg/Bus/Pci/NvmExpressDxe/NvmExpressDxe.inf
> 
> This might've happened upstream too by now, but yes, this is necessary to boot from NVMe.

There are no longer separate *Ia32.dsc and *Ia32X64.dsc files
but UefiPayloadPkg.dsc does have such a line. The .fdf has
the related INF line as well.


===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)