Re: [PATCH for-4.21 1/2] x86/AMD: avoid REP MOVSB for Zen3/4

["Roger Pau Monné" <roger.pau@citrix.com>]

On Mon, Sep 29, 2025 at 07:35:53PM -0400, Jason Andryuk wrote:
> On 2025-09-25 08:18, Teddy Astie wrote:
> > Le 25/09/2025 à 12:48, Jan Beulich a écrit :
> > > Along with Zen2 (which doesn't expose ERMS), both families reportedly
> > > suffer from sub-optimal aliasing detection when deciding whether REP MOVSB
> > > can actually be carried out the accelerated way. Therefore we want to
> > > avoid its use in the common case (memset(), copy_page_hot()).
> > 
> > s/memset/memcpy (memset probably uses rep stosb which is not affected IIUC)
> > 
> > > 
> > > Reported-by: Andrew Cooper <andrew.cooper3@citrix.com>
> > > Signed-off-by: Jan Beulich <jbeulich@suse.com>
> 
> With Teddy's suggested change:
> 
> Reviewed-by: Jason Andryuk <jason.andryuk@amd.com>

Acked-by: Roger Pau Monné <roger.pau@citrix.com>

It would be nice to have some actual figures whether this makes any
difference though.  Teddy, I think Vates had been doing some testing
in this regard, do you think you could measure whether the patch makes
any noticeable difference in PV network traffic for example?  (as
that's a heavy user of grant copy).

Thanks, Roger.