Now that we have separate paths for the TDP MMU, it is trivial to only
grab rcu_read_lock() for the TDP MMU case. We do not need to grab it
for the shadow MMU, as pages are not RCU-freed in that case.

Signed-off-by: James Houghton <jthoughton@google.com>
---
 arch/x86/kvm/mmu/mmu.c | 33 ++++++++++++++++++---------------
 1 file changed, 18 insertions(+), 15 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 7df1b4ead705b..c8f7dd747d524 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -7577,17 +7577,18 @@ static void kvm_recover_nx_huge_pages(struct kvm *kvm,
 	nx_huge_pages = &kvm->arch.possible_nx_huge_pages[mmu_type].pages;
 
 	rcu_idx = srcu_read_lock(&kvm->srcu);
-	if (is_tdp_mmu)
+	if (is_tdp_mmu) {
 		read_lock(&kvm->mmu_lock);
-	else
+		/*
+		 * Zapping TDP MMU shadow pages, including the remote TLB flush,
+		 * must be done under RCU protection, because the pages are
+		 * freed via RCU callback.
+		 */
+		rcu_read_lock();
+	} else {
 		write_lock(&kvm->mmu_lock);
+	}
 
-	/*
-	 * Zapping TDP MMU shadow pages, including the remote TLB flush, must
-	 * be done under RCU protection, because the pages are freed via RCU
-	 * callback.
-	 */
-	rcu_read_lock();
 
 	for ( ; to_zap; --to_zap) {
 #ifdef CONFIG_X86_64
@@ -7641,25 +7642,27 @@ static void kvm_recover_nx_huge_pages(struct kvm *kvm,
 
 		if (need_resched() || rwlock_needbreak(&kvm->mmu_lock)) {
 			kvm_mmu_remote_flush_or_zap(kvm, &invalid_list, flush);
-			rcu_read_unlock();
 
-			if (is_tdp_mmu)
+			if (is_tdp_mmu) {
+				rcu_read_unlock();
 				cond_resched_rwlock_read(&kvm->mmu_lock);
-			else
+				rcu_read_lock();
+			} else {
 				cond_resched_rwlock_write(&kvm->mmu_lock);
+			}
 
 			flush = false;
-			rcu_read_lock();
 		}
 	}
 	kvm_mmu_remote_flush_or_zap(kvm, &invalid_list, flush);
 
-	rcu_read_unlock();
 
-	if (is_tdp_mmu)
+	if (is_tdp_mmu) {
+		rcu_read_unlock();
 		read_unlock(&kvm->mmu_lock);
-	else
+	} else {
 		write_unlock(&kvm->mmu_lock);
+	}
 	srcu_read_unlock(&kvm->srcu, rcu_idx);
 }
 
-- 
2.50.0.727.gbf7dc18ff4-goog

On Mon, Jul 07, 2025, James Houghton wrote:
> Now that we have separate paths for the TDP MMU, it is trivial to only
> grab rcu_read_lock() for the TDP MMU case.

Yeah, but it's also a largely pointless change.  For the overwhelming majority of
deployments, rcu_read_{un}lock() does literally nothing.  And when it does do
something, the cost is a single atomic.

I'm leaning quite strongly toward skipping this patch, as I find the code to be
much more readable if KVM grabs RCU unconditionally.

> We do not need to grab it for the shadow MMU, as pages are not RCU-freed in
> that case.

On Wed, Jul 23, 2025 at 1:38 PM Sean Christopherson <seanjc@google.com> wrote:
>
> On Mon, Jul 07, 2025, James Houghton wrote:
> > Now that we have separate paths for the TDP MMU, it is trivial to only
> > grab rcu_read_lock() for the TDP MMU case.
>
> Yeah, but it's also a largely pointless change.  For the overwhelming majority of
> deployments, rcu_read_{un}lock() does literally nothing.  And when it does do
> something, the cost is a single atomic.
>
> I'm leaning quite strongly toward skipping this patch, as I find the code to be
> much more readable if KVM grabs RCU unconditionally.

That's fine with me, thanks Sean.