From mboxrd@z Thu Jan 1 00:00:00 1970 From: Vitaly Kuznetsov Subject: Re: [PATCH] x86/kvm/vmx: don't read current->thread.{fs,gs}base of legacy tasks Date: Thu, 12 Jul 2018 13:23:09 +0200 Message-ID: <87h8l4y9ya.fsf@vitty.brq.redhat.com> References: <20180711173718.8850-1-vkuznets@redhat.com> <877em0ztoi.fsf@vitty.brq.redhat.com> Mime-Version: 1.0 Content-Type: text/plain Cc: kvm , Paolo Bonzini , Radim Krcmar , "the arch\/x86 maintainers" , Andy Lutomirski , ldv@altlinux.org, yamato@redhat.com, LKML To: Wanpeng Li Return-path: In-Reply-To: <877em0ztoi.fsf@vitty.brq.redhat.com> (Vitaly Kuznetsov's message of "Thu, 12 Jul 2018 11:31:41 +0200") Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org Vitaly Kuznetsov writes: > Wanpeng Li writes: > >> Test suite: MSR switch >> PASS: VM entry MSR load >> PASS: VM exit MSR store >> PASS: VM exit MSR load >> FAIL: VM entry MSR load: try to load FS_BASE >> SUMMARY: 4 tests, 1 unexpected failures >> >> kvm-unit-tests fails w/ and w/o the patch, maybe it is another issue, >> i didn't dig further, you can have a look if you are interested in. :) > > The patch only changes the behavior for legacy userspaces and I can > reproduce the failure on native x86_64, it is something different. I'm, > however, interested so stay tuned :-) Yes, the regression was introduced by commit e79f245ddec17bbd89d73cd0169dba4be46c9b55 Author: KarimAllah Ahmed Date: Sat Apr 14 05:10:52 2018 +0200 X86/KVM: Properly update 'tsc_offset' to represent the running guest basically, when nested_vmx_load_msr() fails we don't set exit_qualification accordingly. The fix is simple: @@ -11720,8 +11721,10 @@ static int enter_vmx_non_root_mode(struct kvm_vcpu *vcpu) msr_entry_idx = nested_vmx_load_msr(vcpu, vmcs12->vm_entry_msr_load_addr, vmcs12->vm_entry_msr_load_count); - if (msr_entry_idx) + if (msr_entry_idx) { + exit_qual = msr_entry_idx; goto fail; + } /* * Note no nested_vmx_succeed or nested_vmx_fail here. At this point I'll be sending a patch out shortly. But this is completely orthogonal to the 'legacy' issue ;-) -- Vitaly