Mini-PC Morefine M9 Intel N100 on NetBSD

dpostolov

Hi! Sorry for my bad English...

I recently received from China a Morefine M9 Mini-PC with Intel SoC AlderLake-N N100. It works quite successfully under FreeBSD 13.2, but I also decided to test it under NetBSD 10.0_BETA.

NetBSD: https://bsd-hardware.info/?probe=d6f92a5ecc
FreeBSD: https://bsd-hardware.info/?probe=926ab149ae

The following problems were identified under NetBSD:

It is too early to talk about graphics support, support will only be in the ported video driver older than 5.18.
Network card Intel Ethernet Controller I225-V not supported yet.
And, most importantly, Kernel panic is observed because of one of the (internal) USB devices, most likely an audio output (internal) USB device.

Currently, the Morefine M9 has FreeBSD installed and works fine, I will do a more detailed investigation of the problems with NetBSD closer to version 10.0-RC.

dpostolov

Kernel panics images: https://disk.yandex.ru/d/z7Tm7UHww5YKOw

rvp

dpostolov This looks like PR #57255, and it may have been fixed in -HEAD recently.

dpostolov

RVP, this is Heisenbug https://en.wikipedia.org/wiki/Heisenbug

rvp

dpostolov RVP, this is Heisenbug

Try a latest -HEAD image. If there are no more panics after a few weeks of regular use, then post a positive reply to that PR.

dpostolov

All OK on -HEAD: fine9$ uptime 1:59AM up 10:17, 1 user, load averages: 0.73, 0.49, 0.27
after a few weeks I will be back! Thanks!

dpostolov

Sorry, I send wrong Info: Today was new Kernel Panic in Morefine M9 on -HEAD

fine9$ date
Wed Jun 28 00:24:30 +05 2023
fine9$ whoami
dmitrii
fine9$ uname -a
NetBSD fine9.localdomain 10.99.4 NetBSD 10.99.4 (GENERIC) #0: Thu Jun 22 14:04:54 +05 2023 root@fine9.localdomain:/root/sysbuild/amd64/obj/usr/src/sys/arch/amd64/compile/GENERIC amd64

https://disk.yandex.ru/d/EfmO03DvSc_e9g

rvp

dpostolov No idea about this. Possibly PR# 56654. Does the workaround mentioned there work?

dpostolov

rvp USB Legacy Support was already Enabled (workaround), now I manually set it value to Auto.

rvp

dpostolov The CSM or UEFI BIOS selection will be on a different page. Leave these USB settings at the default (F3).

dpostolov

rvp This is modern PC based on Intel AlderLake-N N100 January 2023 Model. No Legacy Bios setting, Uefi Bios by default and can't changed.

rvp

dpostolov No Legacy Bios setting, Uefi Bios by default and can't changed.

OK. If there's aren't any Legacy boot settings anywhere, then reply to that PR to let the kernel-folk know that this is still a problem.

dpostolov

fine$ uname -a
NetBSD fine.localnet 10.0_RC2 NetBSD 10.0_RC2 (GENERIC) #0: Mon Jan 1 14:04:52 UTC 2024 mkrepro@mkrepro.NetBSD.org:/usr/src/sys/arch/amd64/compile/GENERIC amd64

Network card Intel Ethernet Controller I225-V now is supported OK.
On NetBSD 10.0_RC2 [and Xfce] the computer still hangs, very often it happens when exiting from Firefox-ESR and Firefox (the probability of system halt on under FF 120.0.1 exiting is lower). Perhaps the next new versions of FF will fix the memory leak.

But what I don't understand is that under FreeBSD and licensed Windows 10/11 and FF the computer works fine! I think its hardware is fine.

I would really like to solve this halt problem under NetBSD 10.0_RC2 and future 10.0_RELEASE!

nettester

dpostolov the computer still hangs

That will be annoying to debug i fear. Wild suspicion: Maybe switching graphics modes screws up? What you could try is running a background loop that dumps dmesg into a file (something like C=0; while true; do dmesg | tail -n 100 > /root/dmesg.log.$C; sleep 10; C=$(( $C + 1 )); done) and see if anything comes up that might yield a clue. Sadly chances are that any possibly interesting entries will not make it to the log let alone the disk anymore when the PC freezes though. You could obviously lower the timeout but i guess it would just make the file spam worse and not change anything else.

rvp

dpostolov On NetBSD 10.0_RC2 [and Xfce] the computer still hangs, very often it happens when exiting from Firefox-ESR
[...]
under FreeBSD and licensed Windows 10/11 and FF the computer works fine! I think its hardware is fine.

Since you're running under wsfb I don't think it can be a graphics-driver issue. genfb+wsfb is the pretty much the safest Xorg display-driver combination possible.

My feeling is that this is a HW issue, but, as @nettester said it's gonna be very difficult to diagnose without any logs.

When the system hangs, can you ssh into your system? If yes, then you could run dmesg and see if there are any errors. If no, then you could try setting up a serial console and collect the logs that way.

I recommend running either memtest86 or memtest+ for a day or so to rule out RAM issues first.

dpostolov

BSD Hardware Info of Morefine M9 Mini-PC Intel AlderLake-N CPU

https://bsd-hardware.info/?probe=5deb235717

dpostolov

[Testing one idea, will take more time to test].

I think this problem is my fault. Right after installing NetBSD, I changed the volume level of the internal USB sound device to 170 to increase the volume on my headphones. If I'm not mistaken, the default setting was 149. The experiment showed that with the default volume and FF 121 the system does not hang when exiting FF. My guess is that the excessive volume setting of the internal USB audio device caused its electrical voltage to increase. Normal work of NetBSD has continued for several hours now, although I will need additional time to test.

Thanks to everyone who responded!

dpostolov

continuing testing...

rvp

dpostolov Make sure you've selected all the tests. Sometimes only the bit-fade test reveals RAM issues. And, that CPU temp. seems a bit high (for a PC running in Russia in winter).

dpostolov

Here is a summary of my observations:

The memory test showed no errors.
The system still stops sometimes even at lower volume levels.
The system stops only in graphics mode.
In physical console without logging into X and Xfce, or in CTRL-ALT-F1 mode, the system runs stably and can steadily build packages from source for 24 hours.
When the system hangs in graphical mode, M9 does not ping from the second computer ("host is down").
Hangs occur when exiting Firefox (including the latest version to date, 121), only when the sound is played when exiting the browser, for example in social networks.
That is, the problem is related to playing audio from an internal USB audio device and closing the application at that point.

fine$ dmesg | grep uaudio0
[ 3.184524] uaudio0 at uhub3 port 6 configuration 1 interface 0
[ 3.184524] uaudio0: CSCTEK (0x0573) USB Audio and HID (0x1573), rev 2.00/80.07, addr 3
[ 3.184524] uaudio0: audio rev 1.00
[ 3.184524] audio0 at uaudio0: playback, capture, full duplex, independent

If you do not playback sound in graphics mode, FF and the system work stably.
FreeBSD & Windows 10/11 have different drivers for this internal USB sound device... This explains why there is no system hang under them.
Workaround: do not use this M9 computer for multimedia under NetBSD, but use it as a build station in the console to build the system and packages ("sysbuild build" & future "pkg_comp -c default auto"), then it will work stable in the console.
For daily tasks, use my second Intel NUC7PJYH computer with a different audio chip on which NetBSD runs successfully in graphical mode and it can be used for multimedia purposes.

Thank you all!

rvp

dpostolov That is, the problem is related to playing audio from an internal USB audio device and closing the application at that point.

Hmm. This could be either:

a) the USB issue you reported before with this system, or
b) related to this hdaudio error:

hdaudio0 at pci0 dev 31 function 3: HD Audio Controller
hdaudio0: interrupting at msi5 vec 0
hdaudio0: HDA ver. 1.0, OSS 9, ISS 7, BSS 0, SDO 1, 64-bit
hdaudio0: autoconfiguration error: no codecs found
hdaudio0: autoconfiguration error: device driver failed to attach

Can you try the latest -HEAD for any USB issues?
For the hdaudio, can you just disable it?
In /boot.cfg add userconf=disable hdaudio* (make sure no hdaudio appears in the dmesg output afterwards.)

If the problem persists, you'll have to compile a custom kernel with:

options         HDAUDIO_DEBUG
options         HDAFG_DEBUG
options         USBVERBOSE
options         USB_DEBUG
options         UAUDIO_DEBUG
options         DEBUG
options         LOCKDEBUG

to get useful info for the kernel folks.

nettester

dpostolov Workaround: do not use this M9 computer for multimedia under NetBSD

Another workaround might be to use an external USB audio dongle. Not sure if it's worth it unless you already got something lying around though.