[RFC v2 PATCH 07/13] KVM: Handle page fault for fd based memslot

[Chao Peng <chao.p.peng@xxxxxxxxxxxxxxx>]

Current code assume the private memory is persistent and KVM can check
with backing store to see if private memory exists at the same address
by calling get_pfn(alloc=false).

Signed-off-by: Yu Zhang <yu.c.zhang@xxxxxxxxxxxxxxx>
Signed-off-by: Chao Peng <chao.p.peng@xxxxxxxxxxxxxxx>
---
 arch/x86/kvm/mmu/mmu.c | 75 ++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 73 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 40377901598b..cd5d1f923694 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -3277,6 +3277,9 @@ int kvm_mmu_max_mapping_level(struct kvm *kvm,
 	if (max_level == PG_LEVEL_4K)
 		return PG_LEVEL_4K;
 
+	if (memslot_is_memfd(slot))
+		return max_level;
+
 	host_level = host_pfn_mapping_level(kvm, gfn, pfn, slot);
 	return min(host_level, max_level);
 }
@@ -4555,6 +4558,65 @@ static bool kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
 				  kvm_vcpu_gfn_to_hva(vcpu, gfn), &arch);
 }
 
+static bool kvm_faultin_pfn_memfd(struct kvm_vcpu *vcpu,
+				  struct kvm_page_fault *fault, int *r)
+{	int order;
+	kvm_pfn_t pfn;
+	struct kvm_memory_slot *slot = fault->slot;
+	bool priv_gfn = kvm_vcpu_is_private_gfn(vcpu, fault->addr >> PAGE_SHIFT);
+	bool priv_slot_exists = memslot_has_private(slot);
+	bool priv_gfn_exists = false;
+	int mem_convert_type;
+
+	if (priv_gfn && !priv_slot_exists) {
+		*r = RET_PF_INVALID;
+		return true;
+	}
+
+	if (priv_slot_exists) {
+		pfn = slot->memfd_ops->get_pfn(slot, slot->priv_file,
+					       fault->gfn, false, &order);
+		if (pfn >= 0)
+			priv_gfn_exists = true;
+	}
+
+	if (priv_gfn && !priv_gfn_exists) {
+		mem_convert_type = KVM_EXIT_MEM_MAP_PRIVATE;
+		goto out_convert;
+	}
+
+	if (!priv_gfn && priv_gfn_exists) {
+		slot->memfd_ops->put_pfn(pfn);
+		mem_convert_type = KVM_EXIT_MEM_MAP_SHARED;
+		goto out_convert;
+	}
+
+	if (!priv_gfn) {
+		pfn = slot->memfd_ops->get_pfn(slot, slot->file,
+					       fault->gfn, true, &order);
+		if (fault->pfn < 0) {
+			*r = RET_PF_INVALID;
+			return true;
+		}
+	}
+
+	if (slot->flags & KVM_MEM_READONLY)
+		fault->map_writable = false;
+	if (order == 0)
+		fault->max_level = PG_LEVEL_4K;
+
+	return false;
+
+out_convert:
+	vcpu->run->exit_reason = KVM_EXIT_MEMORY_ERROR;
+	vcpu->run->mem.type = mem_convert_type;
+	vcpu->run->mem.u.map.gpa = fault->gfn << PAGE_SHIFT;
+	vcpu->run->mem.u.map.size = PAGE_SIZE;
+	fault->pfn = -1;
+	*r = -1;
+	return true;
+}
+
 static bool kvm_faultin_pfn(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault, int *r)
 {
 	struct kvm_memory_slot *slot = fault->slot;
@@ -4596,6 +4658,9 @@ static bool kvm_faultin_pfn(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault,
 		}
 	}
 
+	if (memslot_is_memfd(slot))
+		return kvm_faultin_pfn_memfd(vcpu, fault, r);
+
 	async = false;
 	fault->pfn = __gfn_to_pfn_memslot(slot, fault->gfn, false, &async,
 					  fault->write, &fault->map_writable,
@@ -4660,7 +4725,8 @@ static int direct_page_fault(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault
 	else
 		write_lock(&vcpu->kvm->mmu_lock);
 
-	if (fault->slot && mmu_notifier_retry_hva(vcpu->kvm, mmu_seq, fault->hva))
+	if (fault->slot && !memslot_is_memfd(fault->slot) &&
+			mmu_notifier_retry_hva(vcpu->kvm, mmu_seq, fault->hva))
 		goto out_unlock;
 	r = make_mmu_pages_available(vcpu);
 	if (r)
@@ -4676,7 +4742,12 @@ static int direct_page_fault(struct kvm_vcpu *vcpu, struct kvm_page_fault *fault
 		read_unlock(&vcpu->kvm->mmu_lock);
 	else
 		write_unlock(&vcpu->kvm->mmu_lock);
-	kvm_release_pfn_clean(fault->pfn);
+
+	if (memslot_is_memfd(fault->slot))
+		fault->slot->memfd_ops->put_pfn(fault->pfn);
+	else
+		kvm_release_pfn_clean(fault->pfn);
+
 	return r;
 }
 
-- 
2.17.1