TDX MMU refactors

[PATCH 14/17] KVM: x86/mmu: Remove KVM_BUG_ON() that checks lock when removing PTs

Posted by Rick Edgecombe 5 days, 17 hours ago

As part of an ongoing effort to move TDX specific bits from the MMU into
the TDX code, drop the KVM_BUG_ON() that checks the MMU lock is held for
write while removing page tables.

Future changes forward PTE removal mirror EPT updates into the
set_private_spte() and let TDX code parse the PTE to decide what S-EPT
operations to take. This operations does not pass a shared bool for this
KVM_BUG_ON() to use in the logics future home.

But even today there are already MMU write lockdep asserts that mostly
cover the case. Since the KVM_BUG_ON() is already a bit redundant, just
remove it instead of trying to plumb the bool into TDX code.

Link: https://lore.kernel.org/kvm/aYUarHf3KEwHGuJe@google.com/
Suggested-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
---
 arch/x86/kvm/mmu/tdp_mmu.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index 991870789863..5dc9633c866e 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -473,10 +473,8 @@ static void handle_removed_pt(struct kvm *kvm, tdp_ptep_t pt, bool shared)
 		}
 		handle_changed_spte(kvm, sp, gfn, old_spte, FROZEN_SPTE, level, shared);
 
-		if (is_mirror_sp(sp)) {
-			KVM_BUG_ON(shared, kvm);
+		if (is_mirror_sp(sp))
 			remove_external_spte(kvm, gfn, old_spte, level);
-		}
 	}
 
 	if (is_mirror_sp(sp) &&
-- 
2.53.0

Re: [PATCH 14/17] KVM: x86/mmu: Remove KVM_BUG_ON() that checks lock when removing PTs

Posted by Yan Zhao 3 days, 7 hours ago

On Fri, Mar 27, 2026 at 01:14:18PM -0700, Rick Edgecombe wrote:
> As part of an ongoing effort to move TDX specific bits from the MMU into
> the TDX code, drop the KVM_BUG_ON() that checks the MMU lock is held for
> write while removing page tables.
> 
> Future changes forward PTE removal mirror EPT updates into the
> set_private_spte() and let TDX code parse the PTE to decide what S-EPT
> operations to take. This operations does not pass a shared bool for this
"operations" --> "operation" ?

> KVM_BUG_ON() to use in the logics future home.
By "in the logics future home", do you mean "when this logic is moved to its
future location"?
 
> But even today there are already MMU write lockdep asserts that mostly
> cover the case. Since the KVM_BUG_ON() is already a bit redundant, just
> remove it instead of trying to plumb the bool into TDX code.
> 
> Link: https://lore.kernel.org/kvm/aYUarHf3KEwHGuJe@google.com/
> Suggested-by: Sean Christopherson <seanjc@google.com>
> Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
> ---
>  arch/x86/kvm/mmu/tdp_mmu.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> index 991870789863..5dc9633c866e 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.c
> +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> @@ -473,10 +473,8 @@ static void handle_removed_pt(struct kvm *kvm, tdp_ptep_t pt, bool shared)
>  		}
>  		handle_changed_spte(kvm, sp, gfn, old_spte, FROZEN_SPTE, level, shared);
>  
> -		if (is_mirror_sp(sp)) {
> -			KVM_BUG_ON(shared, kvm);
> +		if (is_mirror_sp(sp))
>  			remove_external_spte(kvm, gfn, old_spte, level);
> -		}
>  	}
>  
>  	if (is_mirror_sp(sp) &&
> -- 
> 2.53.0
>

Re: [PATCH 14/17] KVM: x86/mmu: Remove KVM_BUG_ON() that checks lock when removing PTs

Posted by Huang, Kai 2 days, 3 hours ago

> 
> > KVM_BUG_ON() to use in the logics future home.
> By "in the logics future home", do you mean "when this logic is moved to its
> future location"?
> 

Maybe it just meant to be:

  the logic's future home

Re: [PATCH 14/17] KVM: x86/mmu: Remove KVM_BUG_ON() that checks lock when removing PTs

Posted by Edgecombe, Rick P 14 hours ago

On Tue, 2026-03-31 at 10:46 +0000, Huang, Kai wrote:
> > 
> > > KVM_BUG_ON() to use in the logics future home.
> > By "in the logics future home", do you mean "when this logic is moved to its
> > future location"?
> > 
> 
> Maybe it just meant to be:
> 
>   the logic's future home

Yea that is what I meant. I'll update it. Unless Yan do you find the "future
home" phrasing confusing?

Re: [PATCH 14/17] KVM: x86/mmu: Remove KVM_BUG_ON() that checks lock when removing PTs

Posted by Yan Zhao 12 hours ago

On Thu, Apr 02, 2026 at 08:08:54AM +0800, Edgecombe, Rick P wrote:
> On Tue, 2026-03-31 at 10:46 +0000, Huang, Kai wrote:
> > > 
> > > > KVM_BUG_ON() to use in the logics future home.
> > > By "in the logics future home", do you mean "when this logic is moved to its
> > > future location"?
> > > 
> > 
> > Maybe it just meant to be:
> > 
> >   the logic's future home
> 
> Yea that is what I meant. I'll update it. Unless Yan do you find the "future
> home" phrasing confusing?
Yeah, "future home" seems less clear to me. :)