Re: [PATCH 2/2] edac: add support for Amazon's Annapurna Labs EDAC

[Benjamin Herrenschmidt]

On Wed, 2019-06-12 at 13:00 +0200, Borislav Petkov wrote:
> On Wed, Jun 12, 2019 at 07:42:42AM -0300, Mauro Carvalho Chehab wrote:
> > That's said, from the admin PoV, it makes sense to have a single
> > daemon that collect errors from all error sources and take the
> > needed actions.
> 
> Doing recovery actions in userspace is too flaky. Daemon can get killed
> at any point in time

So what ? If root kills your RAS daemon, then so be it. That has never
been a problem on POWER8/POWER9 server platforms and those have some of
the nastiest RAS in town.

You can kill PID 1 too you know ... 

>  and there are error types where you want to do recovery *before* you return to userspace.

Very few (precise examples please) and I yet have to see why those need
some kind of magic coordinator.

> Yes, we do have different error reporting facilities but I still think
> that concentrating all the error information needed in order to do
> proper recovery action is the better approach here. And make that part
> of the kernel so that it is robust. Userspace can still configure it and
> so on.

Ben.