Gigabyte MJ11-EC1 EPYC 3151 Mystery

[RolloZ170]

Anyone know how to unbrick the BMC?

ch341a USB programmer. after flashing set BMC NIC to correct MAC (from local machine,ipmitool)

 

[PeterF]

Guess it doesn´t work. I can´t log on anymore. It´s stuck on login screen. Cheers.
Anyone know how to unbrick the BMC?View attachment 36069

Does the machine boot?
If you can get to the EFI command line you could try to flash it from there.

If not, connect a serial cable to the BMC JTAG pins and see if you can get an UBOOT prompt. there are flashing commands in uboot.
I have not used them so can give no further info on that.

The Firmware should have worked on the board. I downloaded and checked the documentation. It looks like Gigabyte uses the same BMC firmware for all their MBs with ast2500

BR
Peter

 

[Gizmo_Ger]

Does the machine boot?
If you can get to the EFI command line you could try to flash it from there.

If not, connect a serial cable to the BMC JTAG pins and see if you can get an UBOOT prompt. there are flashing commands in uboot.
I have not used them so can give no further info on that.

The Firmware should have worked on the board. I downloaded and checked the documentation. It looks like Gigabyte uses the same BMC firmware for all their MBs with ast2500

BR
Peter

Thanks @RolloZ170 and @PeterF.
Yes, I´ve checked the readme.txt and it stated that this board is supported by the firmware. Prior I updated a cpl of times the firmware. It always worked like a charm. Now I am stuck with the login page. I can type in username and password but it keeps the animation thingy "Processing" forever. The server itself works still like a charme. Yippie.

I just ordered the programmer on amazon and also ordered a new mainboard, just in case.
Cheap hardware is the silver lining in this case.
Does someone has a how-to for the programmer handy? Never done it before.

Could it be that the new authentication method is implemented? Not the standard admin/password but the serial number?

 

[RolloZ170]

I can type in username and password but it keeps the animation thingy "Processing" forever. The server itself works still like a charme. Yippie.

clear brower cache and allow browser all except unsafe stuff...

 

[Gizmo_Ger]

clear brower cache and allow browser all except unsafe stuff...

no luck. cleared cache, used different browser.

 

[PeterF]

no luck. cleared cache, used different browser.

Have you tried flashing from the UEFI prompt with the flashall option?
This would clear the stored settings

You could also try to set a known password from the host with ipmitool

Peter

 

[Gymnae]

Has anyone here experienced these motherboards not taking the manual memory clocksettings? I have two dimms per server that are on the qvl, that are 2666mhz but downclock on auto to 2400. If I follow the manual I read I have to run them at 1866, but when entering 933 the systems do a couple reboots and when I look in the bios they are back at 2400. All the while still chucking ecc errors

I currently have 4 dimms I can try over 2 boards and they both have the same issues. It does not matter much what sticks I put in what board. Have also memtested for a good 8 passes with no issues. What makes this a little difficult is that sometimes a system boots without throwing ecc errors and then after sitting a while, we are back at it.

HMA82GR7AFR8N-VKT3 are the modules im using and listed on the qvl so im not really sure what to do. Im still running the original bios.

Have you every figured this out? Stuck with the same issue and so far unwilling to buy four new RAM sticks...

 

[Grinchy]

Could there be a difference in 2Rx4 and 4Rx4? In the Compatibility Sheed are only 2Rx4. Someone got it working with 4Rx4?

Good you are on your way.

First a warning. This can potentially brick your system. I have had one side effect. The product name disappeared from the BIOS and instead there is a text "To be filled in by OEM" It does not seam to have any impact on running the system.

I did all my testing with the main CPU off, only the BMC powered. I suggest you do the same!

If you are logged in as sysadmin you should have root access (meaning you can do anything on the system)
you should have a normal Linux prompt.

type id to confirm that you are root

sysadmin [/conf/user_home/peter]# id
uid=0(sysadmin) gid=0(sysadmin) groups=0(sysadmin),109(kvm),501(ipmi),505(lanadmin),511(serialadmin),533(SmashCLP),539(vmedia)

now you need to change to the tmp directory

cd /tmp

It is the file SKU.xml in the tmp directory that we need to edit. First list its contents with

cat SKU.xml

In this there are 3 changes to make
G431-MM0-OT to MJ11-EC0-00 in 2 places
MJ11-EC1-OT to MJ11-EC0-00 in 2 places
G431_MM0 to MJ11

The xml file is just one long row so difficult to edit with vi etc. Best is to use sed.
I just tested these commands and they worked. Note the difference of Zero (0) and the letter O

sed -i s/G431-MM0-OT/MJ11-EC0-00/g SKU.xml
sed -i s/MJ11-EC1-OT/MJ11-EC0-00/g SKU.xml
sed -i s/G431_MM0/MJ11/g SKU.xml

now you can do

cat SKU.xml

to verify that the changes have been made
The next step is to run

bmcprog WS=FULL_AREA

this will create SKU.BIN
verify that it is created

cat SKU.BIN

should produce a lot of garbage (it is a compressed file)

Now this file should be flashed with the following command

skurw w flash SKU.BIN

to verify you should run

rm SKU.xml
skupioneer

This will produce some errors but give you a new SKU.xml

verify that this has the 5 changes by doing

cat SKU.xml

If this is OK that you should now reboot the BMC. But first turn on logging in Putty, it can be good to see what happens.
reboot by simply doing

reboot

During the reboot the BMC should generate the new FanPolicy and all 3 fans should be there!

If this is not working maybe the SKU.BIN file needs to be edited. It has an embedded zipped file mysku.xml.gz
that contains the SKU.xml file. If this is the case I will give instructions for that.

If everything works out start up the main CPU and watch the sensors page in the BMCGui

BR
Peter

Would this be possible using SSH into the IPMI? Or do you need this special adapter and plug it into the Board?

 

[RolloZ170]

Could there be a difference in 2Rx4 and 4Rx4? In the Compatibility Sheed are only 2Rx4. Someone got it working with 4Rx4?

4Rx4 should LRDIMM, not RDIMM.
LRDIMM has extra buffers to reduce load for the mem.controller but needs some extra timing/clock for the buffers.

 

[Grinchy]

4Rx4 should LRDIMM, not RDIMM.
LRDIMM has extra buffers to reduce load for the mem.controller but needs some extra timing/clock for the buffers.

So 4Rx4 wouldn't even work with this Board? Just ordered Samsung M386A4G40DM0 (seems to be 4Rx4) about 10 Minutes ago ...

 

[PeterF]

Would this be possible using SSH into the IPMI? Or do you need this special adapter and plug it into the Board?
[/QUOTE]

Only if you have an old version of the BMC firmware . I had that on my first board and could set a password for sysadmin under the maintenance tab in the BMC GUI. When I upgraded that functionality was gone and SSH access for sysadmin is not possible. Any other user is meaningless as you just get a prompt where you can do nothing!

It is not difficult to get the serial to JTAG working. When you are at that prompt you can log in with sysadmin. There you can create another root user and SSH into that

Peter

 

[RolloZ170]

So 4Rx4 wouldn't even work with this Board? Just ordered Samsung M386A4G40DM0 (seems to be 4Rx4) about 10 Minutes ago ...

sorry don't know. mobo BIOS and CPU has to support LRDIMM.
M386 is LRDIMM in any case.
if it was cheap it's mostly old worn out stuff.

 

[Grinchy]

sorry don't know. mobo BIOS and CPU has to support LRDIMM.
M386 is LRDIMM in any case.
if it was cheap it's mostly old worn out stuff.

Thank you for the info! Still new for me.

Someone wrote before that LRDIMM didn't worked for him.

So M393A4K40BB1 would be "normal" RDIMM? Maybe I can still change my order ...

 

[RolloZ170]

So M393A4K40BB1 would be "normal" RDIMM?

correct

 

[Grinchy]

Would this be possible using SSH into the IPMI? Or do you need this special adapter and plug it into the Board?

Only if you have an old version of the BMC firmware . I had that on my first board and could set a password for sysadmin under the maintenance tab in the BMC GUI. When I upgraded that functionality was gone and SSH access for sysadmin is not possible. Any other user is meaningless as you just get a prompt where you can do nothing!

It is not difficult to get the serial to JTAG working. When you are at that prompt you can log in with sysadmin. There you can create another root user and SSH into that

Peter
[/QUOTE]

I'm one the newest Version and was able to change it.




You mean this one, right?

 

[PeterF]

Only if you have an old version of the BMC firmware . I had that on my first board and could set a password for sysadmin under the maintenance tab in the BMC GUI. When I upgraded that functionality was gone and SSH access for sysadmin is not possible. Any other user is meaningless as you just get a prompt where you can do nothing!

It is not difficult to get the serial to JTAG working. When you are at that prompt you can log in with sysadmin. There you can create another root user and SSH into that

Peter

I'm one the newest Version and was able to change it.


View attachment 36094

You mean this one, right?
[/QUOTE]
Right just that one. But if you try to ssh into it with sysadm you will not get access, unfortunatly.
But the serial console will work with that user and password

Peter

 

[Niskon]

Thank you for the info! Still new for me.

Someone wrote before that LRDIMM didn't worked for him.

So M393A4K40BB1 would be "normal" RDIMM? Maybe I can still change my order ...

I got 4x 32GB 2Rx4 LRDIMMS -> M386A4K40BB0
No boot with either 1, 2 or 4 sticks. I think LRDIMMS aren't compatible.

 

[Gizmo_Ger]

Have you tried flashing from the UEFI prompt with the flashall option?
This would clear the stored settings

You could also try to set a known password from the host with ipmitool

Peter

root@pve:~# ipmitool
IANA PEN registry open failed: No such file or directory
Could not open device at /dev/ipmi0 or /dev/ipmi/0 or /dev/ipmidev/0: No such file or directory

It looks like the ipmi is f*%§ed. I´ll try the flashall on UEFI

 

[ilmari]

Not sure if this helps with the original problem, but that iana pen registry stuff is fixable.
[SOLVED] - IPMI Tool error after v8 upgrade
Scroll down the thread, the first fix is to install newer ipmitool, but it is kinda sketchy. The better fix, provided by LnxBil, is to install the missing file.

 

[Th0mas51]

This M.2 adapter incl. cable is working without any problems on the SFF8654 4i-Interface:

https://amzn.eu/d/4Rng46j

BIOS: F02

Good to know, thanks.

Do you know if it's possible to use the two different M.2 at the same time, or only one port is working ? (Likely due to the no PCIe bifurcation)