From:   Vitaly Kuznetsov <vkuznets@redhat.com>
To:     Maxim Levitsky <mlevitsk@redhat.com>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Igor Mammedov <imammedo@redhat.com>
Cc:     kvm@vger.kernel.org, Sean Christopherson <seanjc@google.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 2/2] KVM: x86: Forbid KVM_SET_CPUID{,2} after KVM_RUN
In-Reply-To: <7e7c7e22f8b1b1695d26d9e19a767b87c679df93.camel@redhat.com>
References: <20211122175818.608220-1-vkuznets@redhat.com>
 <20211122175818.608220-3-vkuznets@redhat.com>
 <16368a89-99ea-e52c-47b6-bd006933ec1f@redhat.com>
 <20211227183253.45a03ca2@redhat.com>
 <61325b2b-dc93-5db2-2d0a-dd0900d947f2@redhat.com>
 <87mtkdqm7m.fsf@redhat.com> <20220103104057.4dcf7948@redhat.com>
 <875yr1q8oa.fsf@redhat.com>
 <ceb63787-b057-13db-4624-b430c51625f1@redhat.com>
 <87o84qpk7d.fsf@redhat.com> <877dbbq5om.fsf@redhat.com>
 <5505d731-cf87-9662-33f3-08844d92877c@redhat.com>
 <20220111090022.1125ffb5@redhat.com> <87fsptnjic.fsf@redhat.com>
 <50136685-706e-fc6a-0a77-97e584e74f93@redhat.com>
 <87bl0gnfy5.fsf@redhat.com>
 <7e7c7e22f8b1b1695d26d9e19a767b87c679df93.camel@redhat.com>
Date:   Thu, 13 Jan 2022 15:36:08 +0100
Message-ID: <87zgnzn1nr.fsf@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

Maxim Levitsky <mlevitsk@redhat.com> writes:

> On Thu, 2022-01-13 at 10:27 +0100, Vitaly Kuznetsov wrote:
>> Paolo Bonzini <pbonzini@redhat.com> writes:
>> 
>> > On 1/12/22 14:58, Vitaly Kuznetsov wrote:
>> > > -	best = kvm_find_cpuid_entry(vcpu, 0xD, 1);
>> > > +	best = cpuid_entry2_find(entries, nent, 0xD, 1);
>> > >   	if (best && (cpuid_entry_has(best, X86_FEATURE_XSAVES) ||
>> > >   		     cpuid_entry_has(best, X86_FEATURE_XSAVEC)))
>> > >   		best->ebx = xstate_required_size(vcpu->arch.xcr0, true);
>> > >   
>> > > -	best = kvm_find_kvm_cpuid_features(vcpu);
>> > > +	best = __kvm_find_kvm_cpuid_features(vcpu, vcpu->arch.cpuid_entries,
>> > > +					     vcpu->arch.cpuid_nent);
>> > >   	if (kvm_hlt_in_guest(vcpu->kvm) && best &&
>> > 
>> > I think this should be __kvm_find_kvm_cpuid_features(vcpu, entries, nent).
>> > 
>> 
>> Of course.
>> 
>> > > +		case 0x1:
>> > > +			/* Only initial LAPIC id is allowed to change */
>> > > +			if (e->eax ^ best->eax || ((e->ebx ^ best->ebx) >> 24) ||
>> > > +			    e->ecx ^ best->ecx || e->edx ^ best->edx)
>> > > +				return -EINVAL;
>> > > +			break;
>> > 
>> > This XOR is a bit weird.  In addition the EBX test is checking the wrong 
>> > bits (it checks whether 31:24 change and ignores changes to 23:0).
>> 
>> Indeed, however, I've tested CPU hotplug with QEMU trying different
>> CPUs in random order and surprisingly othing blew up, feels like QEMU
>> was smart enough to re-use the right fd)
>> 
>> > You can write just "(e->ebx & ~0xff000000u) != (best->ebx ~0xff000000u)".
>> > 
>> > > +		default:
>> > > +			if (e->eax ^ best->eax || e->ebx ^ best->ebx ||
>> > > +			    e->ecx ^ best->ecx || e->edx ^ best->edx)
>> > > +				return -EINVAL;
>> > 
>> > This one even more so.
>> 
>> Thanks for the early review, I'm going to prepare a selftest and send
>> this out.
>> 
> I also looked at this recently (due to other reasons) and I found out that
> qemu picks a parked vcpu by its vcpu_id which is its initial apic id,
> thus apic id related features should not change.
>
> Take a look at 'kvm_get_vcpu' in qemu source.
> Maybe old qemu versions didn't do this?

I took Igor's word on this, I didn't check QEMU code :-)

In the v1 I've just sent [L,x2]APIC ids are allowed to change. This
shouldn't screw the MMU (which was the main motivation for forbidding
KVM_SET_CPUID{,2} after KVM_RUN in the first place) but maybe we don't
really need to be so permissive.

-- 
Vitaly