Received-SPF: pass (google.com: domain of linux-kernel+bounces-84404-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
Date: Tue, 27 Feb 2024 18:11:59 -0800
In-Reply-To: <20240227232100.478238-1-pbonzini@redhat.com>
Precedence: bulk
Mime-Version: 1.0
References: <20240227232100.478238-1-pbonzini@redhat.com>
Message-ID: <Zd6Wb6eeuQ0XlSrb@google.com>
Subject: Re: [PATCH 00/21] TDX/SNP part 1 of n, for 6.9
From: Sean Christopherson <seanjc@google.com>
To: Paolo Bonzini <pbonzini@redhat.com>
Cc: linux-kernel@vger.kernel.org, kvm@vger.kernel.org, michael.roth@amd.com, 
	isaku.yamahata@intel.com, thomas.lendacky@amd.com
Content-Type: text/plain; charset="us-ascii"

I would strongly prefer we taret 6.10, not 6.9.  The TDX and SNP folks don't need
any of this code to be in Linus' tree, they just need it in _a_ KVM tree so that
they can develop on top.

And I will have limited availability the rest of this week (potentially very
limited), and I obviously have strong opinions about some of this code.  But even
if I had cycles to review this properly, I just don't see a reason to rush it in.

For the guest_memfd changes in particular, they're impossible to review in this
series.  Rather than prematurely shove them into mainline, we should create a
volatile topic branch and use that to enable TDX/SNP development.  That way we
can fixup patches if things need to change.

On Tue, Feb 27, 2024, Paolo Bonzini wrote:
> This is a first set of, hopefully non-controversial patches from the
> SNP and TDX series.  They cover mostly changes to generic code and new
> gmem APIs, and in general have already been reviewed when posted by
> Isaku and Michael.
> 
> One important change is that the gmem hook for initializing memory
> is designed to return -EEXIST if the page already exists in the
> guestmemfd filemap.  The idea is that the special case of
> KVM_SEV_SNP_LAUNCH_UPDATE, where __kvm_gmem_get_pfn() is used to
> return an uninitialized page and make it guest-owned, can be be done at
> most once per page unless the ioctl fails.
> 
> Of course these patches add a bunch of dead code.  This is intentional
> because it's the only way to trim the large TDX (and to some extent SNP)
> series to the point that it's possible to discuss them.  The next step is
> probably going to be the private<->shared page logic from the TDX series.