Re: port-amd64/58982: NetBSD 10.1 install image (amd64) failed to boot on J3455-ITX board (ASRock)

[wgstuken%grimme-reitenspezial.de@localhost]

Hi,

just as remark at the beginning : today it was nessesary to reset the 
CMOS after trying to boot of 10.1 in SMP mode (by accident) ....
VGA monitor output was dark - even for bios setup menus.
There must have something thrown some bombs around ...
This has not happend yesterday - do not ask me why ...


I had a try with 10.1 and 'boot -12'
This bring up the kernel but failed to initialize USB - keyboard lost, 
no input possible.

next try 'boot -2'
This bring up the kernel, but failed to setup sd0 (the USB-stick), 
keyboard not working anymore -> nothing to boot from ...

next try 'boot -1'
This bring up the kernel and allows to boot from USB-stick. (two times 
due to USB initialisation problem when USB-stick attached to same 
hardware as keyboard also connected.)
It enters system setup. I dropped to shell from there.

the boot -1 and boot -2 messages are inside attached file - system not 
powered down inbetween so all messages are there ...

The 'boot -2 try failed to setup 'all' interrupt handler - no devices 
active ...

The 'boot -1' looks better ....

The next boot sequence found in dmesg output after system comes up looks 
much better:
ACPI setup looks OK to me
the CPU 1,2,3 are not started - multiprocessor boot disabled
GP00, 01, 02, 03 and PEPD at acpi0 are not configured - OK, may be
acpifan0 failed to configure - there is no fan connected - I assume this 
is the cause here.

ichsmb0 at cpi0 dev 31 function 1: Intel Apollo Lake SMBUS (rev 0x0c)
pci_int_map: no mapping for pin A (line=ff)
..
ichsmb0ß: polling

Hmm - there seems to be somethin very special here ...

uhidev0 at uhub1 port 5 configuration 1 interface 0
...
ukdb0 at uhidev0
wskbd0 at ukbd0: console keyboard

But then in first try with USB-stick placed at second rear USB2 port:
uhub1: port 6, set config at addr 2 failed
uhub1: autoconfiguration error: device prolem, disabling port 6

Here nothing to boot mount was there ....
I choose 'reboot' and had next try with 'boot -1' again with USB-stick 
placed at other USB port (USB3-port this time)
The kernel synced the disks and detache all devices again
and the rebooted

Same messages as before

the usb stick was found at
umass0 at uhub1 port 3 configuration 0 interface 0

"cpuctl -a cpu0" runs into usage message
"cat /proc/cpuinfo' - output attached too.

I run 'cpuctl identify 0' - this looks like what you like to get with 
above cpuctl call.
See output attached.

So it is definitly something related to the SMP part for this CPU ...


While waiting for the 8.0 download and writing USB-stick I have another 
reboot retry with 'boot -1'.
This one now failes every second reboot - USB-stick in same slot and 
initializations is failing every second time ...
It looks like something is not reset completely again.
I switched back to USB2 port that had failed in first boot try with -1.
Same is there - every second reboot the stick is not initialized 
correctly again.
There seems to be another bug in USB-sybsystem related to at least this 
board/bios that no full/stable inititalization is done.

Last thing I've done today: try with NetBSD-8.0-amd64-uefi-install.img 
...

Boot failed too with 'failed to start CPU1' ....
'boot -1' comes up again.

So looks like this is not covered by whatever workaround in 8.0.

Regards,

W. Stukenbrock


Am 11.01.2025 23:20, schrieb Andrius V:
> The following reply was made to PR port-amd64/58982; it has been noted 
> by GNATS.
>
> From: Andrius V <vezhlys%gmail.com@localhost>
> To: gnats-bugs%netbsd.org@localhost
> Cc:
> Subject: Re: port-amd64/58982: NetBSD 10.1 install image (amd64) failed 
> to
>  boot on J3455-ITX board (ASRock)
> Date: Sun, 12 Jan 2025 00:15:22 +0200
>
>  On Sat, Jan 11, 2025 at 3:29=E2=80=AFPM 
> <wgstuken%grimme-reitenspezial.de@localhost> =
>  wrote:
>  >
>  > >Number:         58982
>  > >Category:       port-amd64
>  > >Synopsis:       NetBSD 10.1 install image (amd64) failed to boot on 
> J345=
>  5-ITX board (ASRock)
>  > >Confidential:   no
>  > >Severity:       serious
>  > >Priority:       medium
>  > >Responsible:    port-amd64-maintainer
>  > >State:          open
>  > >Class:          sw-bug
>  > >Submitter-Id:   net
>  > >Arrival-Date:   Sat Jan 11 13:00:01 +0000 2025
>  > >Originator:     Wolfgang Stukenbrock
>  > >Release:        10.1
>  > >Organization:
>  > >Environment:
>  > sorry, not possible - crash/hangup while booting kernel
>  > >Description:
>  > I tried to install 10.1 from installation image
>  > NetBSD-10.1-amd64-install.img
>  > from USB stick.
>  > The stick is found and system start booting.
>  > While kernel is initialing, it reports that it fails to start CPU1.
>  >
>
>  Apollo Lake CPUs has an errata with the workaround implemented  (see
>  PR port-amd64/53420) before NetBSD 8.0 release, which seemingly had
>  same symptoms. Out of curiosity you can try to boot 8.0 release and
>  see if it fails in the same manner (thus to confirm that workaround
>  was or wasn't working at that time). I believe your model may have a
>  different stepping though, still affected by errata, as a result
>  workaround is not applied by netbsd. cat /proc/cpuinfo or "cpuctl -a
>  cpu0" information would be useful here I believe (you can try "boot
>  -12" to disable acpi and smp) or just get this info using linux or
>  windows (cpu-z for example). Thank you.
>
>  Regards,
>  Andrius V