Messages from MCA in the kernel log

[perryh at pluto.rain.com]

Frederic Perrin <fred at resel.fr> wrote:

> ... I don't know what MCA, DRD, SNOOP and {D,G}CACHE stand for...

MCA = Machine Check Architecture.
DRD here probably refers to a data read cycle.
SNOOP has to do with hardware-maintained cache coherency.
DCACHE = data cache.

Google and/or Wikipedia may help with the details.

> Is it a transient error?

I _think_ the COR refers to a corrected error, but that could be
either transient (a random bit-flip, possibly due to a cosmic
ray hit) or permanent (a bit in the cache has gone bad).  In the
latter case I'd expect ongoing error messages, rather than just
two isolated occurrences.

> Should I start screaming at my hosting provider so he changes my
> CPU?

I would say not, unless the errors become frequent or uncorrectable.

> +MCA: Bank 0, Status 0xcc00000120040189
> +MCA: Global Cap 0x0000000000180204, Status 0x0000000000000000
> +MCA: Vendor "GenuineIntel", ID 0xf49, APIC ID 0
> +MCA: CPU 0 COR OVER GCACHE L1 SNOOP error
> +MCA: Address 0x1015b00
> +MCA: Misc 0x140002d800aa0
> +MCA: Bank 1, Status 0x8000002000000135
> +MCA: Global Cap 0x0000000000180204, Status 0x0000000000000000
> +MCA: Vendor "GenuineIntel", ID 0xf49, APIC ID 0
> +MCA: CPU 0 COR DCACHE L1 DRD error