linux-kernel - Re: [PATCH] x86/CPU/AMD, mm: Extend with mem

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: <20171001171617.uzwfzps7sxowfram@pd.tnic>
Date:   Sun, 1 Oct 2017 19:16:17 +0200
From:   Borislav Petkov <bp@...e.de>
To:     Brijesh Singh <brijesh.singh@....com>
Cc:     Tom Lendacky <thomas.lendacky@....com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>,
        "H. Peter Anvin" <hpa@...or.com>,
        Paolo Bonzini <pbonzini@...hat.com>,
        Radim Krčmář <rkrcmar@...hat.com>,
        kvm@...r.kernel.org, x86@...nel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] x86/CPU/AMD, mm: Extend with mem_encrypt=sme option

On Sun, Oct 01, 2017 at 12:00:31PM -0500, Brijesh Singh wrote:
> When SEV feature is disabled, KVM will not be able to launch any SEV
> guests.  When SEV support is available, KVM can enable it in a specific
> VM by setting SEV bit before executing the VMRUN instruction.

So I want to be able to disable SEV and the whole code that comes with
it in the *host*.

> Guest OS:
> --------
> Checks the MSR_AMD64_SEV to determine if SEV feature is enabled. Please
> note that the MSR is a read-only. IOW, MSR is not intercepted by the
> hypervisor.
> 
> Currently, mem_encrypt=xxx and CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT
> is don't care. We can not depend on the command line because when SEV is
> enabled in a VM then instruction fetch will be decrypted by the
> hardware. If we want then we can perform the comparison between the SEV
> state obtained through MSR with user supplied command line and trigger
> BUG() if they don't match.

And when we have supplied mem_encrypt=sme to the *host* cmdline, it
should be impossible to start SEV guests. IOW, that feature mask test
should not happen and I should do instead:

        } else if (!strncmp(buffer, cmd_sme, sizeof(buffer))) {
                sme_only = true;
		sev_enabled = false;
        }

Or, respectively, not set it here as it is false already but set it at
the end of the function like this:

        if (sme_only)
                return;

	sev_enabled = true;
}

Hmmm?

-- 
Regards/Gruss,
    Boris.

SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
--