Messages in this thread | | | Date | Thu, 15 Aug 2013 18:11:53 +0200 | From | Borislav Petkov <> | Subject | Re: [PATCH 3/3] mce: acpi/apei: trace: Enable ghes memory error trace event |
| |
On Thu, Aug 15, 2013 at 11:14:07AM -0300, Mauro Carvalho Chehab wrote: > I don't see why should we have those two alternatives, as, at worse > case (e. g. if ghes_edac can't enrich the APEI data with labels), > they'll basically provide the very same data to userspace, and the > EDAC extra overhead is small, on its error report logic.
Well, a couple of reasons.
The first and foremost one is having another layer which needs registration, etc. because ghes_edac pulls the whole edac core stuff with it. The thinner we are, the less overhead we cause. And this is a must for RAS.
Actually, this is a must for all kernel code - the faster we can get done and the thinner we are, the better. We absolutely definitely don't want to have a useless layer in the error reporting path just because it is easier.
This short path will pay out later in error storms and other resources-constrained situations.
Furthermore, dealing with another edac driver is not trivial for distros, like going around and telling people that all of a sudden they need to enable ghes_edac. This is tangential to the issue which Naveen raised that on some machines, you want not only ghes_edac but the platform-specific one. Which doesn't work currently and we don't have a clear solution on how to get it working yet.
Finally, userspace doesn't care where it gets its TP data from as long as it is there.
> The risk of doing the very same thing on two different places is that > the logic to encapsulate APEI data into trace_mc_event() would be on > two separate places. It risks that someone would change one of the > drivers and forget to apply the very same change on the other, causing > parse errors on userspace, depending on the source.
We'll make sure that doesn't happen.
-- Regards/Gruss, Boris.
Sent from a fat crate under my desk. Formatting is fine. --
| |