Re: HoneyComb first-boot notes

From: Mark Millard via freebsd-arm <freebsd-arm_at_freebsd.org>
Date: Thu, 24 Jun 2021 00:41:03 -0700
On 2021-Jun-22, at 02:15, Mark Millard <marklmi at yahoo.com> wrote:

> On 2021-Jun-22, at 01:40, Dan Kotowski <dan.kotowski at a9development.com> wrote:
>> 
>>>>>> I put an existing FreeBSD Optane into a just delivered
>>>>>> 
>>>>>> HoneyComb, put in the RAM and the matching UEFI/ACPI
>>>>>> 
>>>>>> image on a microsd card and put the card in the slot.
>>>>>> 
>>>>>> I plugged in a USB3 Ethernet dongle, like I use on some
>>>>>> 
>>>>>> other systems. Serial console via its USB port for such.
>>>>>> 
>>>>>> (Optane instead of video card.) The UEFI/ACPI image was
>>>>>> 
>>>>>> extracted from:
>>>>>> 
>>>>>> https://solid-run-images.sos-de-fra-1.exo.io/LX2k/lx2160a_uefi/lx2160acex7_2000_700_2900_8_5_2_sd_81b4bbe.img.xz
>>>>>> 
>>>>>> in order to match the RAM and being based on the
>>>>>> 
>>>>>> most current vintage at:
>>>>>> 
>>>>>> https://github.com/SolidRun/lx2160a_uefi/
>>>>>> 
>>>>>> It booted.
>>>>> 
>>>>> Could you share whole dmesg?
>>>> 
>>>> I've not done a "boot -v" . The below is actually from a
>>>> 
>>>> captured serial console output. It is lightly edited. I
>>>> 
>>>> have omitted nearly all loader output (and before).
>>> 
>>> Thanks for sharing. Have you seen somewhere in output from UEFI (before
>>> 
>>> the loader) that "DPAA2 Management Complex initialized"?
>>> 
>>> Regards,
>>> 
>>> Dmitry
>> 
>> stable/13 dmesg: https://gist.github.com/agrajag9/bf6325d4f3321d8d0ee566fee767e0d3
>> 
>> current dmesg: https://gist.github.com/6d07004fb3c6f0980d5e94627aaba060
>> 
>> _at_dmitry - just ignore all the DPAA2 stuff for now. We don't support it at all :'(
> 
> And UEFI/ACPI currently appears to not give anyone anything
> to support (yet).
> 
>> _at_mark - I was given a warning about i2c from SR's _at_jnettlet that it's possible to completely brick the system, although maybe they've solved that in the more recent SOC revisions?
> 
> FYI: Jon Nettleton recently reported:
> 
> QUOTE
> okay I have replicated the the i2c bus issues.  Now I have
> something I can fix
> END QUOTE
> 
>> The IORT warnings might be related to SMMU. I'll have to dig up my notes, I did a ton of work with Greg on this months ago and it seems that's been swapped out of brain memory onto cold storage somewhere. Possibly here? https://gist.github.com/agrajag9/ff3c52c11469dccb70e17a11cb9be2a9
> 
> FYI: Jon Nettleton reported:
> 
> QUOTE
> ACPI: IORT: Dropping unhandled type 6 This is ACPI RMR support
> which was just released in March 2020 and only recently updated
> as of 2 months ago.  I assume the other type 1 are the named
> components for the network packet processor.  Those will need
> to be fixed once driver support for the onboard networking is
> added
> END QUOTE
> 
> (Jon is not always clear about UEFI/ACPI vs. U-Boot vs.
> Linux kernel or whatever combination is involved.)
> 
>> Feel free to use/critique my KERNCONF: https://gist.github.com/agrajag9/e3991573a274e59f58a0b520d2dcd5d2
> 
> I've never been one to spend effort cutting things from
> the standard kernels. In a couple of cases, in dealing
> with old PowerMac issues, I've forced something to be
> in the kernel to avoid contexts where I could not load
> the same but it was always needed for what I was doing.
> 
>> Fans: I've had mixed success here. Lots of people went for a full watercooling loop with https://www.alphacool.com/shop/gpu-water-cooler/amd-gpu/11088/alphacool-hf-14-smart-motion-universal-copper-edition?c=20587
>> 
>> I'm just using https://www.radianheatsinks.com/product/inm37-5004-25pcu-2-6/ but others like http://www.enzotechnology.com/cnb_r1.htm
>> 
>> Either way, welcome to the party! I've been daily-driving mine as a desktop for a few months, although I still can't get drm-kmod's amdgpu to load properly - perhaps you can?
> 
> The PCIe slot has an Optane, not video. Very few of
> the machines I use have video engaged even rarely.
> One FreeBSD machine has a lower end (and low power)
> video card --but I do not use X11 or the like on it
> generally. Similarly for built-in HDMI (when it is
> used at all).
> 
> The same Optane can be used to boot the MACCHIATObin
> Double Shot, where it was previously in use. I do
> not have access to other aarch64 machines with a
> PCIe slot. These two machines only have one such
> slot.

Some more notes from Jon N. for the EDK2 UEFI/ACPI example
context:

QUOTE
Things got a bit more complicated, after I looked at things some more.
Basically BSD needs to grow I2C support for the LX2160a, which shouldn't
be a big deal since it is mostly compatible with the normal IMX driver.
END QUOTE

and:

QUOTE
However to support devices that don't support the I2C controller I am
writing a UEFI driver that can properly setup the PWM controller for
Auto mode.  This should allow the fan to ramp up and down without any
needed OS support at all.
END QUOTE

He wondered if I knew enough to "check on the i2c controller support for
imx in BSD, it could be a simple fix just by patching in ACPI support".
Unfortunately, that looks to involve way more than I'd be likely to figure
out any significant part of anytime soon.


===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)
Received on Thu Jun 24 2021 - 07:41:03 UTC

Original text of this message