Mt. Jade AMI -> OpenBMC Issues

General Discussion
1

I have a Mt. Jade system of dubious origin (eBay; PVT system per the mobo silkscreen; single Q80-30) that I am trying to convert from AMI/MegaRAC to OpenBMC. Mainly, I want to bump the firmware versions on everything to try to get it off of (presumably) development firmware, and also possibly resolve some issues I am having with the graphics locking up. I built the OpenBMC Mt. Jade image and burned it onto the flash chip from the board, the BMC boots up fine but the host is having issues booting now. When I power the host on, the fans will stick at 100% and eventually the whole chassis will power back off after a minute or two. Checking logs, it seems that for some reason the SCP signature fail line is being asserted from the CPU/CPLD causing the BMC to try to switch to the backup EEPROM. It will continue doing this over and over until some watchdog runs out and the chassis powers itself back off. If I disable/mask the ampere_scp_failover service the system will boot fine, albeit the fans stay spun up longer than on MegaRAC. This did not happen on MegaRAC which would boot the system up basically immediately after I stuck in the power cord.

Logs from journalctl:

Jan 02 00:55:54 mtjade phosphor-multi-gpio-monitor[392]: GPIO Line s0-sys-auth-failure-n Deasserted
Jan 02 00:55:54 mtjade systemd[1]: Started Ampere Computing LLC Ampere failover event service.
-- Subject: A start job for unit ampere_scp_failover.service has finished successfully
-- Defined-By: systemd
-- Support: https://lists.freedesktop.org/mailman/listinfo/systemd-devel
-- 
-- A start job for unit ampere_scp_failover.service has finished successfully.
-- 
-- The job identifier is 9403.
Jan 02 00:55:54 mtjade ampere_scp_failover.sh[6229]: scp-failover: switch HOST to failover boot EEPROM
Jan 02 00:55:54 mtjade ampere_scp_failover.sh[6234]: Triggering sysreset pin
Jan 02 00:55:56 mtjade systemd[1]: ampere_scp_failover.service: Deactivated successfully.
-- Subject: Unit succeeded
-- Defined-By: systemd
-- Support: https://lists.freedesktop.org/mailman/listinfo/systemd-devel
-- 
-- The unit ampere_scp_failover.service has successfully entered the 'dead' state.
Jan 02 00:56:02 mtjade phosphor-multi-gpio-monitor[392]: GPIO Line s0-sys-auth-failure-n Deasserted
Jan 02 00:56:02 mtjade systemd[1]: Started Ampere Computing LLC Ampere failover event service.
-- Subject: A start job for unit ampere_scp_failover.service has finished successfully
-- Defined-By: systemd
-- Support: https://lists.freedesktop.org/mailman/listinfo/systemd-devel
-- 
-- A start job for unit ampere_scp_failover.service has finished successfully.
-- 
-- The job identifier is 9467.
Jan 02 00:56:02 mtjade ampere_scp_failover.sh[6246]: scp-failover: switch HOST to failover boot EEPROM
Jan 02 00:56:03 mtjade ampere_scp_failover.sh[6251]: Triggering sysreset pin
Jan 02 00:56:04 mtjade systemd[1]: ampere_scp_failover.service: Deactivated successfully.
-- Subject: Unit succeeded
-- Defined-By: systemd
-- Support: https://lists.freedesktop.org/mailman/listinfo/systemd-devel
-- 
-- The unit ampere_scp_failover.service has successfully entered the 'dead' state.
Jan 02 00:56:10 mtjade phosphor-multi-gpio-monitor[392]: GPIO Line s0-sys-auth-failure-n Deasserted
Jan 02 00:56:10 mtjade systemd[1]: Started Ampere Computing LLC Ampere failover event service.
-- Subject: A start job for unit ampere_scp_failover.service has finished successfully
-- Defined-By: systemd
-- Support: https://lists.freedesktop.org/mailman/listinfo/systemd-devel
-- 
-- A start job for unit ampere_scp_failover.service has finished successfully.
-- 
-- The job identifier is 9531.
Jan 02 00:56:11 mtjade ampere_scp_failover.sh[6267]: scp-failover: switch HOST to failover boot EEPROM
Jan 02 00:56:11 mtjade ampere_scp_failover.sh[6272]: Triggering sysreset pin
Jan 02 00:56:12 mtjade systemd-journald[128]: Forwarding to syslog missed 19 messages.
-- Subject: One or more messages could not be forwarded to syslog
-- Defined-By: systemd
-- Support: https://lists.freedesktop.org/mailman/listinfo/systemd-devel
-- 
-- One or more messages could not be forwarded to the syslog service
-- running side-by-side with journald. This usually indicates that the
-- syslog implementation has not been able to keep up with the speed of
-- messages queued.
Jan 02 00:56:12 mtjade systemd[1]: ampere_scp_failover.service: Deactivated successfully.

Any help is appreciated, and if I am doing something wrong or stepping on toes please tell me. Thanks!

1 Like
2

Where did you get the OpenBMC firmware image, and what version is it?

3

Where did you get the OpenBMC image

I tried both mainline openbmc and the Ampere repo which did the exact same thing.

Edit to specify: I pulled down the Git repo and built it locally.

what version is it?

Mainline is master as of Mon Dec 30 07:50:46 2024 -0600. The Ampere repo I tried a few months ago (I shelved this project for a while) so it would have been about August '24 vintage.

Cheers!

1 Like
4

Thanks, that’s what I would have suggested trying, too. Unfortunately I haven’t had much luck running OpenBMC on Mt Jade systems so I can’t help.

5

I have deduced this to be an issue with the SCP firmware on my unit that requires an update. Unfortunately since this is secondhand I don’t believe Ampere will let me join Customer Connect and sign the NDA, but I am willing to be proven wrong :slight_smile:

6

Unfortunately you’re correct about Customer Connect.
However, the SCP firmware is the same across all Altra-based systems I believe, so you may be able to find a copy elsewhere. For example Gigabyte, Supermicro, ADLINK, ASRock Rack etc. make Altra-based systems.