Re: [PATCH 3/3] EDAC: amd64_edac: decide if driver can load successfully early.

[Tejun Heo <tj@kernel.org>]

Hello, Borislav.

On Thu, Mar 19, 2015 at 10:40:54AM +0100, Borislav Petkov wrote:
> This is why I did the drv_instances hack - I don't want it to load if
> there's no ECC support on the system as it is confusing to people and
> tools.
> 
> It'd need to be able for the async probing to unload the driver if not a
> single node instance loads successfully.

I don't really think this is the type of things we want to be doing in
the specific drivers.  It makes the driver behave differnetly from
everything else.  If a feature like this is actually necessary, please
implement a proper abstraction at the driver layer (e.g. a flag to
indicate that if the initial probe fails, there's no point in keeping
the driver around) but frankly I don't think this matters enough to
warrant such extra complexities.

This is a gloss layering violation.  Please don't do things like this.
I'm all for ripping out the hack even w/o considering the async probe
issue.

Thanks.

-- 
tejun