[10/12] KVM: arm64: Save stage2 PTE dirty status if it is coverred

Commit Message

zhukeqian June 16, 2020, 9:35 a.m. UTC

There are two types of operations will change PTE and may cover
dirty status set by hardware.

1. Stage2 PTE unmapping: Page table merging (revert of huge page
table dissolving), kvm_unmap_hva_range() and so on.

2. Stage2 PTE changing: including user_mem_abort(), kvm_mmu_notifier
_change_pte() and so on.

All operations above will invoke kvm_set_pte() finally. We should
save the dirty status into memslot bitmap.

Question: Should we acquire kvm_slots_lock when invoke mark_page_dirty?
It seems that user_mem_abort does not acquire this lock when invoke it.

Signed-off-by: Keqian Zhu <[e-mail]>
---
 arch/arm64/kvm/mmu.c | 20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)

Patch

diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index 898e272a2c07..a230fbcf3889 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -294,15 +294,23 @@  static void unmap_stage2_ptes(struct kvm *kvm, pmd_t *pmd,
 {
 	phys_addr_t start_addr = addr;
 	pte_t *pte, *start_pte;
+	bool dirty_coverred;
+	int idx;
 
 	start_pte = pte = pte_offset_kernel(pmd, addr);
 	do {
 		if (!pte_none(*pte)) {
 			pte_t old_pte = *pte;
 
-			kvm_set_pte(pte, __pte(0));
+			dirty_coverred = kvm_set_pte(pte, __pte(0));
 			kvm_tlb_flush_vmid_ipa(kvm, addr);
 
+			if (dirty_coverred) {
+				idx = srcu_read_lock(&kvm->srcu);
+				mark_page_dirty(kvm, addr >> PAGE_SHIFT);
+				srcu_read_unlock(&kvm->srcu, idx);
+			}
+
 			/* No need to invalidate the cache for device mappings */
 			if (!kvm_is_device_pfn(pte_pfn(old_pte)))
 				kvm_flush_dcache_pte(old_pte);
@@ -1388,6 +1396,8 @@  static int stage2_set_pte(struct kvm *kvm, struct kvm_mmu_memory_cache *cache,
 	pte_t *pte, old_pte;
 	bool iomap = flags & KVM_S2PTE_FLAG_IS_IOMAP;
 	bool logging_active = flags & KVM_S2_FLAG_LOGGING_ACTIVE;
+	bool dirty_coverred;
+	int idx;
 
 	VM_BUG_ON(logging_active && !cache);
 
@@ -1453,8 +1463,14 @@  static int stage2_set_pte(struct kvm *kvm, struct kvm_mmu_memory_cache *cache,
 		if (pte_val(old_pte) == pte_val(*new_pte))
 			return 0;
 
-		kvm_set_pte(pte, __pte(0));
+		dirty_coverred = kvm_set_pte(pte, __pte(0));
 		kvm_tlb_flush_vmid_ipa(kvm, addr);
+
+		if (dirty_coverred) {
+			idx = srcu_read_lock(&kvm->srcu);
+			mark_page_dirty(kvm, addr >> PAGE_SHIFT);
+			srcu_read_unlock(&kvm->srcu, idx);
+		}
 	} else {
 		get_page(virt_to_page(pte));
 	}