linux-kernel - Re: [PATCH v4 3/3] KVM: nVMX: Emulate EPTP switching for the L1 hypervisor

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20170712134105.GG3326@potion>
Date:   Wed, 12 Jul 2017 15:41:05 +0200
From:   Radim Krčmář <rkrcmar@...hat.com>
To:     Bandan Das <bsd@...hat.com>
Cc:     Jim Mattson <jmattson@...gle.com>,
        David Hildenbrand <david@...hat.com>,
        kvm list <kvm@...r.kernel.org>,
        Paolo Bonzini <pbonzini@...hat.com>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v4 3/3] KVM: nVMX: Emulate EPTP switching for the L1
 hypervisor

2017-07-11 16:45-0400, Bandan Das:
> Radim Krčmář <rkrcmar@...hat.com> writes:
> 
> > 2017-07-11 15:38-0400, Bandan Das:
> >> Radim Krčmář <rkrcmar@...hat.com> writes:
> >> 
> >> > 2017-07-11 14:35-0400, Bandan Das:
> >> >> Jim Mattson <jmattson@...gle.com> writes:
> >> >> ...
> >> >> >>> I can find the definition for an vmexit in case of index >=
> >> >> >>> VMFUNC_EPTP_ENTRIES, but not for !vmcs12->eptp_list_address in the SDM.
> >> >> >>>
> >> >> >>> Can you give me a hint?
> >> >> >>
> >> >> >> I don't think there is. Since, we are basically emulating eptp switching
> >> >> >> for L2, this is a good check to have.
> >> >> >
> >> >> > There is nothing wrong with a hypervisor using physical page 0 for
> >> >> > whatever purpose it likes, including an EPTP list.
> >> >> 
> >> >> Right, but of all the things, a l1 hypervisor wanting page 0 for a eptp list
> >> >> address most likely means it forgot to initialize it. Whatever damage it does will
> >> >> still end up with vmfunc vmexit anyway.
> >> >
> >> > Most likely, but not certainly.  I also don't see a to diverge from the
> >> > spec here.
> >> 
> >> Actually, this is a specific case where I would like to diverge from the spec.
> >> But then again, it's L1 shooting itself in the foot and this would be a rarely
> >> used code path, so, I am fine removing it.
> >
> > Thanks, we're not here to judge the guest, but to provide a bare-metal
> > experience. :)
> 
> There are certain cases where do. For example, when L2 instruction emulation
> fails we decide to kill L2 instead of injecting the error to L1 and let it handle
> that. Anyway, that's a different topic, I was just trying to point out there
> are cases kvm does a somewhat policy decision...

Emulation failure is a KVM bug and we are too lazy to implement the
bare-metal behavior correctly, but avoiding the EPTP list bug is
actually easier than introducing it.  You can make KVM simpler and
improve bare-metal emulation at the same time.