Re: [PATCH] x86/traps: 'Fix' safety of read_registers() in #DF path

[Jan Beulich <jbeulich@xxxxxxxx>]

On 16.10.2020 14:07, Andrew Cooper wrote:
> On 16/10/2020 12:55, Jan Beulich wrote:
>> On 16.10.2020 13:24, Andrew Cooper wrote:
>>> On a tangent, what are your views WRT backport beyond 4.14?
>>>
>>> Back then, it was #DB which was adjacent to the guard frame (which was
>>> not present), but it doesn't use show_registers() by default, so I think
>>> the problem is mostly hidden.
>> I wasn't fully decided yet, but as long as it applies reasonably
>> cleanly I think I'm leaning towards also putting it on 4.13.
>> 4.12 closes anyway once 4.12.4 is out, and I don't think I want
>> to pick up not-really-urgent changes for putting there beyond
>> the few ones that I already have (and that I mean to put in
>> alongside the XSA fixes on Tuesday); I could be talked into it,
>> though.
> 
> The question I was asking was really "should I try and make an
> equivalent fix for 4.13 and older".

Oh, I see.

> While the base premise of the fix would be the same, the logic in
> load_system_tables() is different, and the commit message is completely
> wrong.
> 
> I only encountered this problem with added instrumentation in the #DB
> handler, which is why I'm questioning the utility of going to this effort.

Yeah, then probably not worth it.

Jan