[RESEND PATCH v14 7/7] KVM: arm: page logging 2nd stage fault handling

[Mario Smarduch]

This patch adds support for handling 2nd stage page faults during migration,
it disables faulting in huge pages, and dissolves huge pages to page tables.
In case migration is canceled huge pages are used again.

Resending to addresse Marc's comments to simplify stage2_set_pte() handling 
of logging, and mapping device memory - flags.

Reviewed-by: Marc Zyngier <marc.zyngier at arm.com>
Reviewed-by: Christoffer Dall <christoffer.dall at linaro.org>
Signed-off-by: Mario Smarduch <m.smarduch at samsung.com>
---
 arch/arm/kvm/mmu.c |   61 +++++++++++++++++++++++++++++++++++++++++++++-------
 1 file changed, 53 insertions(+), 8 deletions(-)

diff --git a/arch/arm/kvm/mmu.c b/arch/arm/kvm/mmu.c
index 8137455..d29de77 100644
--- a/arch/arm/kvm/mmu.c
+++ b/arch/arm/kvm/mmu.c
@@ -47,6 +47,18 @@ static phys_addr_t hyp_idmap_vector;
 #define kvm_pmd_huge(_x)	(pmd_huge(_x) || pmd_trans_huge(_x))
 #define kvm_pud_huge(_x)	pud_huge(_x)
 
+#define KVM_S2PTE_FLAG_IS_IOMAP		(1UL << 0)
+#define KVM_S2PTE_FLAG_LOGGING_ACTIVE	(1UL << 1)
+
+static bool kvm_get_logging_state(struct kvm_memory_slot *memslot)
+{
+#ifdef CONFIG_ARM
+	return !!memslot->dirty_bitmap;
+#else
+	return false;
+#endif
+}
+
 static void kvm_tlb_flush_vmid_ipa(struct kvm *kvm, phys_addr_t ipa)
 {
 	/*
@@ -626,10 +638,13 @@ static int stage2_set_pmd_huge(struct kvm *kvm, struct kvm_mmu_memory_cache
 }
 
 static int stage2_set_pte(struct kvm *kvm, struct kvm_mmu_memory_cache *cache,
-			  phys_addr_t addr, const pte_t *new_pte, bool iomap)
+			  phys_addr_t addr, const pte_t *new_pte,
+			  unsigned long flags)
 {
 	pmd_t *pmd;
 	pte_t *pte, old_pte;
+	unsigned long iomap = flags & KVM_S2PTE_FLAG_IS_IOMAP;
+	unsigned long logging_active = flags & KVM_S2PTE_FLAG_LOGGING_ACTIVE;
 
 	/* Create stage-2 page table mapping - Level 1 */
 	pmd = stage2_get_pmd(kvm, cache, addr);
@@ -641,6 +656,18 @@ static int stage2_set_pte(struct kvm *kvm, struct kvm_mmu_memory_cache *cache,
 		return 0;
 	}
 
+	/*
+	 * While dirty memory logging, clear PMD entry for huge page and split
+	 * into smaller pages, to track dirty memory at page granularity.
+	 */
+	if (logging_active && kvm_pmd_huge(*pmd)) {
+		phys_addr_t ipa = pmd_pfn(*pmd) << PAGE_SHIFT;
+
+		pmd_clear(pmd);
+		kvm_tlb_flush_vmid_ipa(kvm, ipa);
+		put_page(virt_to_page(pmd));
+	}
+
 	/* Create stage-2 page mappings - Level 2 */
 	if (pmd_none(*pmd)) {
 		if (!cache)
@@ -693,7 +720,8 @@ int kvm_phys_addr_ioremap(struct kvm *kvm, phys_addr_t guest_ipa,
 		if (ret)
 			goto out;
 		spin_lock(&kvm->mmu_lock);
-		ret = stage2_set_pte(kvm, &cache, addr, &pte, true);
+		ret = stage2_set_pte(kvm, &cache, addr, &pte,
+						KVM_S2PTE_FLAG_IS_IOMAP);
 		spin_unlock(&kvm->mmu_lock);
 		if (ret)
 			goto out;
@@ -908,6 +936,10 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
 	struct vm_area_struct *vma;
 	pfn_t pfn;
 	pgprot_t mem_type = PAGE_S2;
+	unsigned long logging_active = 0;
+
+	if (kvm_get_logging_state(memslot))
+		logging_active = KVM_S2PTE_FLAG_LOGGING_ACTIVE;
 
 	write_fault = kvm_is_write_fault(kvm_vcpu_get_hsr(vcpu));
 	if (fault_status == FSC_PERM && !write_fault) {
@@ -918,7 +950,7 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
 	/* Let's check if we will get back a huge page backed by hugetlbfs */
 	down_read(&current->mm->mmap_sem);
 	vma = find_vma_intersection(current->mm, hva, hva + 1);
-	if (is_vm_hugetlb_page(vma)) {
+	if (is_vm_hugetlb_page(vma) && !logging_active) {
 		hugetlb = true;
 		gfn = (fault_ipa & PMD_MASK) >> PAGE_SHIFT;
 	} else {
@@ -964,7 +996,7 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
 	spin_lock(&kvm->mmu_lock);
 	if (mmu_notifier_retry(kvm, mmu_seq))
 		goto out_unlock;
-	if (!hugetlb && !force_pte)
+	if (!hugetlb && !force_pte && !logging_active)
 		hugetlb = transparent_hugepage_adjust(&pfn, &fault_ipa);
 
 	if (hugetlb) {
@@ -978,16 +1010,21 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
 		ret = stage2_set_pmd_huge(kvm, memcache, fault_ipa, &new_pmd);
 	} else {
 		pte_t new_pte = pfn_pte(pfn, mem_type);
+		unsigned long flags = logging_active;
+
+		if (mem_type == PAGE_S2_DEVICE)
+			flags |= KVM_S2PTE_FLAG_IS_IOMAP;
+
 		if (writable) {
 			kvm_set_s2pte_writable(&new_pte);
 			kvm_set_pfn_dirty(pfn);
 		}
 		coherent_cache_guest_page(vcpu, hva, PAGE_SIZE);
-		ret = stage2_set_pte(kvm, memcache, fault_ipa, &new_pte,
-				     mem_type == PAGE_S2_DEVICE);
+		ret = stage2_set_pte(kvm, memcache, fault_ipa, &new_pte, flags);
 	}
 
-
+	if (write_fault)
+		mark_page_dirty(kvm, gfn);
 out_unlock:
 	spin_unlock(&kvm->mmu_lock);
 	kvm_release_pfn_clean(pfn);
@@ -1137,7 +1174,15 @@ static void kvm_set_spte_handler(struct kvm *kvm, gpa_t gpa, void *data)
 {
 	pte_t *pte = (pte_t *)data;
 
-	stage2_set_pte(kvm, NULL, gpa, pte, false);
+	/*
+	 * We can always call stage2_set_pte with logging_active == false,
+	 * because MMU notifiers will have unmapped a huge PMD before calling
+	 * ->change_pte() (which in turn calls kvm_set_spte_hva()) and therefore
+	 * stage2_set_pte() never needs to clear out a huge PMD through this
+	 * calling path.
+	 */
+
+	stage2_set_pte(kvm, NULL, gpa, pte, 0);
 }
 
 
-- 
1.7.9.5