Messages in this thread | | | Date | Wed, 6 Jun 2012 11:04:15 +0200 | From | Borislav Petkov <> | Subject | Re: [patch 2/2] x86: mce: Implement cmci poll mode for intel machines |
| |
On Wed, Jun 06, 2012 at 09:36:52AM +0800, Chen Gong wrote: > Yes, I am just afraid similar situation happening when injecting MCEs > or poll timer running too fast, which is like a pseduo CMCI storm. > Maybe similar logic behind CMCI_STORM can be used there.
Poll timer on !Intel fires once every 5 minutes and halves the interval when it detects a valid MCE.
I still don't think this is an issue if you poll only on single CPUs - I think you need the throttling because, as Tony explained earlier, the CMCI interrupt runs on _all_ CPUs and I can see how that can be hurting performance.
CMCI on AMD (or, rather, our version of it) runs only on the CPU which sees the error so I don't think we need throttling there, unless real life proves me otherwise.
So, the whole CMCI deal should be Intel-only (for now, at least) and it shouldn't be touching generic MCE code.
Thanks.
-- Regards/Gruss, Boris.
Advanced Micro Devices GmbH Einsteinring 24, 85609 Dornach GM: Alberto Bozzo Reg: Dornach, Landkreis Muenchen HRB Nr. 43632 WEEE Registernr: 129 19551
| |