After system update, NMI fault caused by RAM parity error

tlecso

Hello!

I have a strange issue:

A freshly installed 2.6 from memstick. I want it to update to 22.01, but after installing the updates, the system hang with the following error message:

<2>NMI ISA b0, EISA ff
<2>RAM parity error, likely hardware failure.
panic: NMI indicates hardware failure
cpuid = 1
time = 1669208807
KDB: enter: panic

The system is an old Fujitsu Esprimo E9900 desktop computer, with Intel I3-540 CPU, 8GB RAM (4X2), 1x250GB Toshiba HDD, 2x4 port Broadcom NetXtreme BCM5719 and one integrated Intel 82578DM LAN.

Why i thought, something is strange? Because after the clearly indicated problem (RAM parity error), I've run Memtest for more than 16 hrs, and it is not found any problem in the memory. After the successfully tests (12 round passed), i was reinstalled the system. While i'm on 2.6, it was problem free. When I start the update procedure to 22.01 (first try was from WebConfig, the 2nd try from console), from this point the system is always crashed.

Do you have any idea about this problem?

Thank you!

Other logs:

https://pastebin.com/1dRJt6L6

https://pastebin.com/jyzMhXd6

Gertjan

@tlecso

Are you sure your device even has parity RAM ?
Parity RAM is an expensive option, available for server like devices, not something you would find in a 'off the shelf desktop PC'.

IMHO : FreeBSD sees a NMI comig in that isn't mapped to some functionality, so it throws out a default (?) "panic: NMI indicates hardware failure" message.

Your Memtest is probably very valid : your memory is ok.
You should consider your device for what it is : old .... maybe not 100 % FreeBSD compatible ?

Btw : get a copy of a native FreeBSD 12.3 install (www.freebsd.org) and install that. Chekc if you se the same error.

tlecso

@gertjan

Thank for the advice!

i'll check it now with the FreeBSD 12.3..

You're right, the system has built with non parity (ECC) RAM. It's a simple desktop RAM.

tlecso

@gertjan

I'm checked with BSD 12.3 memstick version. It is running flawless. I can login to use the system, etc. Networking is run too.

I don't have any clue what's wrong with the pFsense. Maybe somewhere the update process make a fault?

Thanks,
tlecso

Gertjan

@tlecso said in After system update, NMI fault caused by RAM parity error:

Maybe somewhere the update process make a fault?

When using FreeBSD 12.3 (not pfSense), do you see the same ACPI errors in the log :

Firmware Error (ACPI): Could not resolve symbol [\_SB.PCI0.EGBA], AE_NOT_FOUND (20200430/psargs-503)
ACPI Error: Aborting method \_SB.PCI0.LPC.MBRD._CRS due to previous error (AE_NOT_FOUND) (20200430/psparse-689)

Look here : ACPI errors all the time (Also a Fujitsu ).

tlecso

@gertjan

Yes, i have similar error messages, but i can't find the log file to insert here. So, I typing it, from a recorded video :)

Firmware Warning (ACPI): 32/64X lenght mismatch in FADT/Pm1aControlBlock: 16/32 (20200430/tbfadt-748)
Firmware Warning (ACPI): 32/64X lenght mismatch in FADT/Pm2ControlBlock: 8/32 (20200430/tbfadt-748)
Firmware Warning (ACPI): 32/64X lenght mismatch in FADT/PmTimerBlock: 32/24 (20200430/tbfadt-748)
Firmware Warning (ACPI): 32/64X lenght mismatch in FADT/Gpe0Block: 128/32 (20200430/tbfadt-748)

Firmware Warning (ACPI): Invalid lenght FADT/Pm1aControlBlock: 32, using default 16 (20200430/tbfadt-850)
Firmware Warning (ACPI): Invalid lenght FADT/Pm2ControlBlock: 32, using default 8 (20200430/tbfadt-850)
Firmware Warning (ACPI): Invalid lenght FADT/PmTimerBlock: 24, using default 32 (20200430/tbfadt-850)

[\_SB.PCI0.EGBA] AE_NOT_FOUND 
\_SB.PCI0.LPC.MBRD._CRS due to previous error
pcib0: could not evaluate _ADR - AE_NOT_FOUND

I hope, this messages are helpful. I found the first error message which is first in your post, with the exact same error, too.

Thank you,
tlecso

stephenw10

First thing I would do I would do here is disable any hardware you're not using in the BIOS.

Thiose ACPI errors are likely harmless and there's not much you can do about it anyway other than updating the BIOS.

2.6 and 22.01 are very similar. I wouldn't expect to find any additional hardware support between them. Especially on an older system like that.

Steve

tlecso

@stephenw10

All of not necessary hw is disabled (TPM, iATM, serial and parallel ports).
On the old machine after update I got some PHP and other package related errors: xxx.so: file not found.
For try, i install a win server os (2008 r2) onto the machine, it works without any problem. So, the disk is work normally. Tested with disk tool (HDSentinel).
I was tried to install and update in a virtual machine. It works. So i think, my problem is a little bit old machine :(
Maybe i can enable the iATM and then somehow can record the update process via RDP for analyzing the faults.

tlecso

stephenw10

Most BIOSes will pass a different set of ACPI values to Windows than any other OS. You can set FreeBSD to pass a different string so it pretends to be Windows, or some other OS, but I doubt that's the problem. Nothing there would have changed between 2.6 and 22.01.

Yes, some sort of log from the upgrade would help a lot here. A virtual com port would be ideal if it supports that.

Steve