Re: [PATCH v2] KVM: SEV: drop FOLL_WRITE for encrypted region registration

[Gupta, Pankaj]

> > > > Commit 7e066cb9b71a ("KVM: SEV: Use long-term pin when registering
> > > > encrypted memory regions") added FOLL_LONGTERM to
> > > > sev_mem_enc_register_region() so anonymous guest RAM is migrated out of
> > > > MIGRATE_CMA/ZONE_MOVABLE before a long term pin. It also kept
> > > > FOLL_WRITE on the pin.
> > > >
> > > > Combining FOLL_WRITE with FOLL_LONGTERM breaks registration of file-backed
> > > > guest memory, such as virtio-pmem host memory-backend-file mappings
> > > > (MAP_SHARED). GUP rejects long-term writable pins on dirty tracked file
> > > > mappings since:
> > > >
> > > > commit 8ac268436e6d ("mm/gup: disallow FOLL_LONGTERM GUP-nonfast writing to file-backed mappings")
> > > > commit a6e79df92e4a ("mm/gup: disallow FOLL_LONGTERM GUP-fast writing to file-backed mappings").
> > > >
> > > > Region registration only requires long-term pin to prevent page migration and
> > > > does not write through this GUP pin.
> > > >
> > > > Drop FOLL_WRITE and pin guest memory only with FOLL_LONGTERM.
> > > Worth mentioning here something like
> > >
> > > "In the past, FOLL_WRITE was required to trigger CoW unsharing, making sure that
> > > we don't end up replacing the page in the page tables during a later write fault
> > > after already having pinned a (shared) page in MAP_PRIVATE mappings.
> > > FOLL_LONGTERM does that nowadays (see gup_must_unshare()) even without FOLL_WRITE.
> Heh, this was going to be my exact question about why it was safe to drop FOLL_WRITE :-)
>
> > > Given that SEV only pins RAM for XYZ and doesn't actually write to the pinned
> > > pages, we can just drop the FOLL_WRITE"
> > >
> > > Fill out XYZ :)
> > Sure :)
> >
> > I will update the commit message in v3.
> No need for a v3, I'll add a blurb when applying.

Thanks, Sean!

Best regards,

Pankaj