KVM: MMU: simplify mmu_need_write_protect
authorXiao Guangrong <guangrong.xiao@linux.intel.com>
Wed, 24 Feb 2016 09:51:15 +0000 (17:51 +0800)
committerPaolo Bonzini <pbonzini@redhat.com>
Thu, 3 Mar 2016 13:36:23 +0000 (14:36 +0100)
Now, all non-leaf shadow page are page tracked, if gfn is not tracked
there is no non-leaf shadow page of gfn is existed, we can directly
make the shadow page of gfn to unsync

Signed-off-by: Xiao Guangrong <guangrong.xiao@linux.intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/mmu.c

index fe03d2a1d4d3370293b3292d02a13fbdea06fe08..d6b264b492c90ca7aa1a11142ef311eefcc798bf 100644 (file)
@@ -2444,7 +2444,7 @@ int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn)
 }
 EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page);
 
-static void __kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp)
+static void kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp)
 {
        trace_kvm_mmu_unsync_page(sp);
        ++vcpu->kvm->stat.mmu_unsync;
@@ -2453,39 +2453,24 @@ static void __kvm_unsync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp)
        kvm_mmu_mark_parents_unsync(sp);
 }
 
-static void kvm_unsync_pages(struct kvm_vcpu *vcpu,  gfn_t gfn)
-{
-       struct kvm_mmu_page *s;
-
-       for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
-               if (s->unsync)
-                       continue;
-               WARN_ON(s->role.level != PT_PAGE_TABLE_LEVEL);
-               __kvm_unsync_page(vcpu, s);
-       }
-}
-
 static bool mmu_need_write_protect(struct kvm_vcpu *vcpu, gfn_t gfn,
                                   bool can_unsync)
 {
-       struct kvm_mmu_page *s;
-       bool need_unsync = false;
+       struct kvm_mmu_page *sp;
 
        if (kvm_page_track_is_active(vcpu, gfn, KVM_PAGE_TRACK_WRITE))
                return true;
 
-       for_each_gfn_indirect_valid_sp(vcpu->kvm, s, gfn) {
+       for_each_gfn_indirect_valid_sp(vcpu->kvm, sp, gfn) {
                if (!can_unsync)
                        return true;
 
-               if (s->role.level != PT_PAGE_TABLE_LEVEL)
-                       return true;
+               if (sp->unsync)
+                       continue;
 
-               if (!s->unsync)
-                       need_unsync = true;
+               WARN_ON(sp->role.level != PT_PAGE_TABLE_LEVEL);
+               kvm_unsync_page(vcpu, sp);
        }
-       if (need_unsync)
-               kvm_unsync_pages(vcpu, gfn);
 
        return false;
 }