[PATCH 41/45] KVM: x86/mmu: Obsolete all roots if memattr contains gPTEs

[Paolo Bonzini]

Note that the zapping of obsolete roots needs to happen after
mmu_lock is dropped, so introduce a separate kvm_arch callback.
The two map neatly to the "front" and "back" halves of
kvm_mmu_zap_all_fast().

Co-developed-by: Nicolas Saenz Julienne <nsaenz@xxxxxxxxxx>
Signed-off-by: Nicolas Saenz Julienne <nsaenz@xxxxxxxxxx>
Signed-off-by: Paolo Bonzini <pbonzini@xxxxxxxxxx>
Link: https://patch.msgid.link/20260918135030.171564-29-pbonzini@xxxxxxxxxx
Signed-off-by: Paolo Bonzini <pbonzini@xxxxxxxxxx>
---
 arch/x86/kvm/mmu/mmu.c   | 30 ++++++++++++++++++++++++++++++
 include/linux/kvm_host.h |  1 +
 virt/kvm/kvm_main.c      |  1 +
 3 files changed, 32 insertions(+)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index c08168595732..1cee8c3f07af 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -8359,6 +8359,7 @@ bool kvm_arch_post_set_vm_memory_attributes(struct kvm *kvm,
 {
 	unsigned long attrs = range->arg.attributes;
 	struct kvm_memory_slot *slot = range->slot;
+	struct kvm_mmu_page *sp;
 	int level;
 
 	lockdep_assert_held_write(&kvm->mmu_lock);
@@ -8414,9 +8415,38 @@ bool kvm_arch_post_set_vm_memory_attributes(struct kvm *kvm,
 				hugepage_set_mixed(slot, gfn, level);
 		}
 	}
+
+	/*
+	 * There are special considerations when applying an memory protection
+	 * attibute against a page containing guest PTEs, because virtual
+	 * addresses that requires traversing that page need to fault.
+	 *
+	 * When using shadow paging, the PTEs that the CPU actually walks
+	 * do not have any information about attributes (unlike TDP where
+	 * the attributes become part of the TDP permission bits).  Since
+	 * this should be a rare occasion, just invalidate all the mmu roots
+	 * if shadow paging is in use and attributes do cover a guest PTE.
+	 */
+	for (gfn_t gfn = range->start; gfn < range->end; gfn++) {
+		for_each_gfn_valid_sp_with_gptes(kvm, sp, gfn) {
+			__kvm_mmu_zap_all_fast_front_half(kvm);
+			return false;
+		}
+	}
 	return false;
 }
 
+void kvm_arch_post_set_vm_memory_attributes_unlocked(struct kvm *kvm)
+{
+	/*
+	 * This may also be called if no zap was done in
+	 * kvm_arch_post_set_memory_attributes().  It is not an issue because
+	 * __kvm_mmu_zap_all_fast_back_half() is cheap when it does not do
+	 * anything.
+	 */
+	 __kvm_mmu_zap_all_fast_back_half(kvm);
+}
+
 void kvm_mmu_init_memslot_memory_attributes(struct kvm *kvm,
 					    struct kvm_memory_slot *slot)
 {
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 4128cdb71869..c5defda4d9d4 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -2624,6 +2624,7 @@ bool kvm_arch_pre_set_vm_memory_attributes(struct kvm *kvm,
 					   struct kvm_gfn_range *range);
 bool kvm_arch_post_set_vm_memory_attributes(struct kvm *kvm,
 					    struct kvm_gfn_range *range);
+void kvm_arch_post_set_vm_memory_attributes_unlocked(struct kvm *kvm);
 bool kvm_range_has_any_vm_memory_attributes(struct kvm *kvm, gfn_t start, gfn_t end,
 					    unsigned long mask);
 u64 kvm_supported_vm_mem_attributes(struct kvm *kvm);
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 2d5d45e4ee71..061e952ea080 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -2773,6 +2773,7 @@ static int kvm_set_vm_mem_attributes(struct kvm *kvm, gfn_t start, gfn_t end,
 		KVM_MMU_LOCK(kvm);
 		kvm_mmu_invalidate_end(kvm);
 		KVM_MMU_UNLOCK(kvm);
+		kvm_arch_post_set_vm_memory_attributes_unlocked(kvm);
 	}
 
 out_unlock:
-- 
2.52.0