On Wed, Jun 29 2022, Eric Auger <eauger@xxxxxxxxxx> wrote: > Hi Connie, > > On 6/14/22 10:40, Cornelia Huck wrote: >> On Fri, Jun 10 2022, Eric Auger <eauger@xxxxxxxxxx> wrote: >> >>> Hi Connie, >>> On 5/12/22 15:11, Cornelia Huck wrote: >>>> We need to disable migration, as we do not yet have a way to migrate >>>> the tags as well. >>> >>> This patch does much more than adding a migration blocker ;-) you may >>> describe the new cpu option and how it works. >> >> I admit this is a bit terse ;) The idea is to control mte at the cpu >> level directly (and not indirectly via tag memory at the machine >> level). I.e. the user gets whatever is available given the constraints >> (host support etc.) if they don't specify anything, and they can >> explicitly turn it off/on. > > Could the OnOffAuto property value be helpful? I completely forgot that this exists; I hacked up something (still untested), and it seems to be able to do what I want. I'll post it after I've verified that it actually works :) >> The big elefant in the room is how migration will end up >> working... after reading the disscussions in >> https://lore.kernel.org/all/CAJc+Z1FZxSYB_zJit4+0uTR-88VqQL+-01XNMSEfua-dXDy6Wg@xxxxxxxxxxxxxx/ >> I don't think it will be as "easy" as I thought, and we probably require >> some further fiddling on the kernel side. > Yes maybe the MTE migration process shall be documented and discussed > separately on the ML? Is Haibu Xu's address bouncing? Yes, that address is bouncing... I've piggybacked onto a recent kvm discussion in https://lore.kernel.org/all/875ykmcd8q.fsf@xxxxxxxxxx/ -- I guess there had not been any change for migration in the meantime, we need to find a way to tie page data + metadata together.
