Hi Dan, The value of 'map_writable' shouldn't be used when the pfn is KVM_PFN_NOSLOT. That would be a bug. Can you show me the path on which it is used? On Thu, May 17, 2018 at 6:40 AM, Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote: > Hello Jim Mattson, > > The patch 3a2936dedd20: "kvm: mmu: Don't expose private memslots to > L2" from May 9, 2018, leads to the following static checker warning: > > arch/x86/kvm/mmu.c:3334 nonpaging_map() > error: uninitialized symbol 'map_writable'. > > arch/x86/kvm/mmu.c > 3315 if (fast_page_fault(vcpu, v, level, error_code)) > 3316 return RET_PF_RETRY; > 3317 > 3318 mmu_seq = vcpu->kvm->mmu_notifier_seq; > 3319 smp_rmb(); > 3320 > 3321 if (try_async_pf(vcpu, prefault, gfn, v, &pfn, write, &map_writable)) > ^^^^^^^^^^^^ > After that patch we return false but don't initialize *writable. > > 3322 return RET_PF_RETRY; > 3323 > 3324 if (handle_abnormal_pfn(vcpu, v, gfn, pfn, ACC_ALL, &r)) > 3325 return r; > 3326 > 3327 spin_lock(&vcpu->kvm->mmu_lock); > 3328 if (mmu_notifier_retry(vcpu->kvm, mmu_seq)) > 3329 goto out_unlock; > 3330 if (make_mmu_pages_available(vcpu) < 0) > 3331 goto out_unlock; > 3332 if (likely(!force_pt_level)) > 3333 transparent_hugepage_adjust(vcpu, &gfn, &pfn, &level); > 3334 r = __direct_map(vcpu, write, map_writable, level, gfn, pfn, prefault); > ^^^^^^^^^^^^ > > 3335 spin_unlock(&vcpu->kvm->mmu_lock); > > regards, > dan carpenter
