SMP and NMI errors (4.10)

[d***@gumbysoft.com]

I have a 4.10-p2 box with dual Xeon CPUs and Tyan S2727 motherboard.
Jul 11 18:28:41 server /kernel: NMI ISA 2c, EISA ff
Roughly about 10-40 times during the build.

NMIs sometimes get triggered for ECC corrections, which is why a memtester
wouldn't see it. I think there is a kernel option somewhere that hooks
NMI and attempts to get information from the platform as to what DIMM
triggered it. Otherwise you might check the Event Log in the BIOS for ECC
events.

If I boot a uniprocessor kernel this problem doesn't occur.

It might be temperature related then :)

[p***@xciv.org]

Actually, this sounds more like a weak/overloaded power supply problem.
Power supplies in many 1U servers are very inadequate for the job and often
have problems in dual processor configurations.

I know where you're coming from, but it's a proper dual xeon case, and
has a 400W EPS PSU.

I've now swapped the memory module out without success.

-Paul-