Received-SPF: pass (google.com: domain of linux-kernel+bounces-47658-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Precedence: bulk
MIME-Version: 1.0
References: <20240112055251.36101-1-vannapurve@google.com> <a06f9b81-38c1-4be8-9d07-6db31530741e@intel.com>
In-Reply-To: <a06f9b81-38c1-4be8-9d07-6db31530741e@intel.com>
From: Vishal Annapurve <vannapurve@google.com>
Date: Thu, 1 Feb 2024 11:14:28 +0530
Message-ID: <CAGtprH8HhCVvjARm3ZuHu04uxAENOLa--kHQhRtAfSz+4-23kg@mail.gmail.com>
Subject: Re: [RFC V1 0/5] x86: CVMs: Align memory conversions to 2M granularity
To: Dave Hansen <dave.hansen@intel.com>
Cc: x86@kernel.org, linux-kernel@vger.kernel.org, pbonzini@redhat.com, 
	rientjes@google.com, seanjc@google.com, erdemaktas@google.com, 
	ackerleytng@google.com, jxgao@google.com, sagis@google.com, oupton@google.com, 
	peterx@redhat.com, vkuznets@redhat.com, dmatlack@google.com, 
	pgonda@google.com, michael.roth@amd.com, kirill@shutemov.name, 
	thomas.lendacky@amd.com, dave.hansen@linux.intel.com, 
	linux-coco@lists.linux.dev, chao.p.peng@linux.intel.com, 
	isaku.yamahata@gmail.com, andrew.jones@linux.dev, corbet@lwn.net, hch@lst.de, 
	m.szyprowski@samsung.com, rostedt@goodmis.org, iommu@lists.linux.dev
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Wed, Jan 31, 2024 at 10:23=E2=80=AFPM Dave Hansen <dave.hansen@intel.com=
> wrote:
>
> There's a bunch of code in the kernel for TDX and SEV guests.  How much
> of it uses the "CVM" nomenclature?

Right, I see that "CoCo VMs" is a more accepted term in the kernel
codebase so far, will update the references in the next version.

>
> What do you do when you need to dynamically scale up the SWIOTLB size
> and can't allocate a 2M page?  I guess you're saying here that you'd
> rather run with a too-small 2M pool than a large-enough mixed 4k/2M pool.

I am not yet certain how to ensure 2M page is always available/made
available at runtime for CoCo VMs. Few options that I can think of:
1) Reserve additional memory for CMA allocations to satisfy runtime
requests of 2M allocations.
2) Pre-reserve SWIOTLB to a safe value just like it's done today and
not rely on dynamic scaling.

Any suggestions are welcome.

>
> I also had a really hard time parsing through the problem statement and
> solution here.  I'd really suggest cleaning up the problem statement and
> more clearly differentiating the host and guest sides in the description.

Thanks for taking a look at this series. I will reword the description
in the next version. The goal basically is to ensure private and
shared memory regions are always huge page aligned.