Re: [PATCH RFC] x86/mce: Move MCE sysfs attributes out of the per-cpulocation

[Naveen N. Rao]

On 08/29/2012 04:10 PM, Borislav Petkov wrote:
> On Wed, Aug 29, 2012 at 03:56:04PM +0530, Naveen N. Rao wrote:
> > Hmmm.. Can't we just deprecate these? ;) Perhaps we can consider
> > adding newer tunables in the right place.
>
> In case you haven't noticed yet: I'm all on your side.

Yup, I know :)

I had my doubts when I sent this patch (hence the RFC tag) and I was 
only wondering above if it'll be a good idea to limit such tunables 
going forward. We could force all _new_ MCE tunables to be global, 
except where they actually apply on a per-processor basis.

> But let me ask you this: these attributes grow to a large number with
> a large number of cores but why is this a problem? We have a bunch of
> redundant attributes in sysfs, so what?
>
> See what I mean?

Well, it's ugly and does not make much sense, as I'm sure you noticed. 
On a 10-core, 8-socket machine with HT, we'll end up with nearly a 
thousand such entries!

I don't know how much resource this takes up (if any) and like you said, 
this may just be a "so what?", but I wanted to bring this up and see if 
we could/want to do anything about this. I'm certainly fine if we want 
to ignore this.


Thanks,
Naveen

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/