Re: [PATCH][4.15] Performance regression due to XSA-336

[Ian Jackson <iwj@xxxxxxxxxxxxxx>]

Jan Beulich writes ("Re: [PATCH][4.15] Performance regression due to XSA-336"):
> On 26.03.2021 12:53, Ian Jackson wrote:
> > This amounts to a request to consider a release ack for "x86/vpt:
> > Replace per-guest pt_migrate lock with per pt lock".
> 
> Or, as already suggested by Roger, perhaps something simpler. First of
> all I wanted you to be aware.

Thank you, yes.  That is very helpful.

> > That patch fixes a regression due to XSA-336.  XSA-336 affected many
> > versions of Xen.  Therefore this is not a regression between 4.14 and
> > 4.15; rather it's a regresion from pre-XSA-336 to post-XSA-336.
> > 
> > I looked at the patch and I am really not comfortable with that kind
> > of change at this stage of the release.
> > 
> > So I'm afraid the answer is no, unless I have misunderstood something.
> 
> Understood.

I might consider "something simpler" but the bar is very high by now
and it seems unlikely to make the cut.

Thanks,
Ian.