Re: [PATCH 21/28] KVM: x86/mmu: Add TDP MMU helper to zap a root

[Ben Gardon]

On Mon, Nov 22, 2021 at 3:15 PM Sean Christopherson <seanjc@xxxxxxxxxx> wrote:
>
> On Mon, Nov 22, 2021, Ben Gardon wrote:
> > On Fri, Nov 19, 2021 at 8:51 PM Sean Christopherson <seanjc@xxxxxxxxxx> wrote:
> > >
> > > Add a small wrapper to handle zapping a specific root.  For now, it's
> > > little more than syntactic sugar, but in the future it will become a
> > > unique flow with rules specific to zapping an unreachable root.
> > >
> > > No functional change intended.
> > >
> > > Signed-off-by: Sean Christopherson <seanjc@xxxxxxxxxx>
> > > ---
> > >  arch/x86/kvm/mmu/tdp_mmu.c | 11 +++++++++--
> > >  1 file changed, 9 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> > > index 9449cb5baf0b..31fb622249e5 100644
> > > --- a/arch/x86/kvm/mmu/tdp_mmu.c
> > > +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> > > @@ -79,11 +79,18 @@ static void tdp_mmu_free_sp_rcu_callback(struct rcu_head *head)
> > >         tdp_mmu_free_sp(sp);
> > >  }
> > >
> > > +static bool tdp_mmu_zap_root(struct kvm *kvm, struct kvm_mmu_page *root,
> > > +                            bool shared)
> > > +{
> > > +       return zap_gfn_range(kvm, root, 0, -1ull, true, false, shared);
> >
> > Total aside:
> > Remembering the order of these three boolean parameters through all
> > these functions drives me nuts.
> > It'd be really nice to put them into a neat, reusable struct that tracks:
> > MMU lock mode (read / write / none)
> > If yielding is okay
> > If the TLBs are dirty and need to be flushed
> >
> > I don't know when I'll have time to do that refactor, but it would
> > make this code so much more sensible.
>
> Heh, I did exactly that, then threw away the code when I realized that I could
> break up zap_gfn_range() into three separate helpers and avoid control knob hell
> (spoiler alert for later patches in this series).
>
> There are still two booleans (to what ends up being tdp_mmu_zap_leafs()), but none
> none of the call sites pass true/false for _both_ params, so the call sites end up
> being quite readable.  At that point, using a struct ended up being a net negative,
> e.g. kvm_tdp_mmu_unmap_gfn_range() had to marshall from one struct to another.

Awesome! Disregard then! I'll review the remaining few tomorrow.