On 04/23/2018 10:30 AM, speck for Thomas Gleixner wrote:
> On Sun, 22 Apr 2018, speck for Borislav Petkov wrote:
> 
>> On Sun, Apr 22, 2018 at 05:53:50AM -0400, speck for Jon Masters wrote:
>>> I'd prefer that too. But we had this conversation with Thomas already
>>> and he wanted a "disable" option. It was fairly clearly spelled out that
>>> it was to be "mdd" or "ssbd" and therefore "spec_store_bypass_disable".
>>
>> No, he wants "on,off,auto" so that we're in-line with PTI's and
>> spectre_v2's options. So let's do:
>>
>> spec_store_bypass=[on|off|auto]
>>
>> and be done with the bikeshedding. The weather's too nice outside.
> 
> Nope. You're reverting the meaning again.
> 
> All mitigation options so far have on|off|auto where:
> 
>  - on:   enables the mitigation
>  - off:  disabled the mitigation
>  - auto: lets the kernel decided whether to enable/disable it
> 
> Now your name is: spec_store_bypass which is a performance feature and not
> the mitigation. So Jon is about right that spec_store_bypass_disable
> describes the mitigation and the on/off/auto have then consistent behaviour
> with kpti and spectre_v2, which admittedly could have named better.

Thanks.

So, let's all agree to keep spec_store_bypass_disable. It'll help to
keep the distro knobs properly consistent with upstream.

Jon.

-- 
Computer Architect | Sent from my Fedora powered laptop