From: Brijesh Singh Subject: Re: [Part2 PATCH v9 00/38] x86: Secure Encrypted Virtualization (AMD) Date: Thu, 21 Dec 2017 10:09:44 -0600 Message-ID: References: <20171205010438.5773-1-brijesh.singh@amd.com> <51fe0851-14d7-cb37-613c-0eaf8a7d2a65@amd.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit Cc: brijesh.singh@amd.com, bp@alien8.de, Thomas Gleixner , Ingo Molnar , "H. Peter Anvin" , =?UTF-8?B?UmFkaW0gS3LDhMKNbcODwqHDhQ==?= =?UTF-8?B?4oSi?= , Joerg Roedel , Borislav Petkov , Tom Lendacky , Herbert Xu , "David S . Miller" , Gary Hook , linux-crypto@vger.kernel.org To: Paolo Bonzini , kvm@vger.kernel.org, linux-kernel@vger.kernel.org, x86@kernel.org Return-path: In-Reply-To: <51fe0851-14d7-cb37-613c-0eaf8a7d2a65@amd.com> Content-Language: en-US Sender: linux-crypto-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On 12/21/17 9:51 AM, Brijesh Singh wrote: > > On 12/21/17 7:06 AM, Paolo Bonzini wrote: > .... > > Hi Paolo, >   > >> Hi Brijesh, >> >> I have a couple comments: >> >> 1) how is MSR_AMD64_SEV's value passed to the guest, and where is it in >> the manual? > It is a non interceptable read-only MSR set by the HW when SEV feature > is enabled in VMRUN instruction. I just checked both PPR and Family 17 manual and it seems both are still missing this MSR definition. I will ping doc team to get it updated. thanks >> 2) ECX should be 0 in the guest's 0x8000_001f leaf, because we don't >> support nested SEV guests. Likewise, EAX bit 2 should be 0 since you >> don't emulate the page flush MSR. > IIRC, I do clear both EAX Page_Flush and nested virtualization case from > Qemu SEV feature is enabled. >> Both can be fixed on top (and I can do the second myself of course), so >> there should be no need for a v10. > Thanks >