linux-kernel - Re: [RFC PATCH] KVM: TDX: Decouple TDX init mem region from kvm_gmem

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] [day] [month] [year] [list]

Message-ID: <CAGtprH9JifhhmTdseXLi9ax_imnY5b=K_+_bhkTXKSaW8VMFRQ@mail.gmail.com>
Date: Tue, 5 Aug 2025 17:09:39 -0700
From: Vishal Annapurve <vannapurve@...gle.com>
To: Sean Christopherson <seanjc@...gle.com>
Cc: Ira Weiny <ira.weiny@...el.com>, Yan Zhao <yan.y.zhao@...el.com>, 
	Michael Roth <michael.roth@....com>, pbonzini@...hat.com, kvm@...r.kernel.org, 
	linux-kernel@...r.kernel.org, rick.p.edgecombe@...el.com, kai.huang@...el.com, 
	adrian.hunter@...el.com, reinette.chatre@...el.com, xiaoyao.li@...el.com, 
	tony.lindgren@...el.com, binbin.wu@...ux.intel.com, dmatlack@...gle.com, 
	isaku.yamahata@...el.com, david@...hat.com, ackerleytng@...gle.com, 
	tabba@...gle.com, chao.p.peng@...el.com
Subject: Re: [RFC PATCH] KVM: TDX: Decouple TDX init mem region from kvm_gmem_populate()

On Tue, Aug 5, 2025 at 12:59 PM Sean Christopherson <seanjc@...gle.com> wrote:
>
> On Mon, Aug 04, 2025, Vishal Annapurve wrote:
> > On Mon, Aug 4, 2025 at 5:22 PM Sean Christopherson <seanjc@...gle.com> wrote:
> > > : 4) For SNP, if src != null, make the target pfn to be shared, copy
> > > : contents and then make the target pfn back to private.
> > >
> > > Copying from userspace under spinlock (rwlock) is illegal, as accessing userspace
> > > memory might_fault() and thus might_sleep().
> >
> > I would think that a combination of get_user_pages() and
> > kmap_local_pfn() will prevent this situation of might_fault().
>
> Yes, but if SNP is using get_user_pages(), then it looks an awful lot like the
> TDX flow, at which point isn't that an argument for keeping populate()?

Ack, I agree we can't ditch kvm_gmem_populate() for SNP VMs. I am ok
with using it for TDX/CCA VMs with the fixes discussed in this RFC.

>
> > Memory population in my opinion is best solved either by users asserting
> > ownership of the memory and writing to it directly or by using guest_memfd
> > (to be) exposed APIs to populate memory ranges given a source buffer. IMO
> > kvm_gmem_populate() is doing something different than both of these options.
>
> In a perfect world, yes, guest_memfd would provide a clean, well-defined API
> without needing a complicated dance between vendor code and guest_memfd.  But,
> sadly, the world of CoCo is anything but perfect.  It's not KVM's fault that
> every vendor came up with a different CoCo architecture.  I.e. we can't "fix"
> the underlying issue of SNP and TDX having significantly different ways for
> initializing private memory.
>
> What we can do is shift as much code to common KVM as possible, e.g. to minimize
> maintenance costs, reduce boilerplate and/or copy+paste code, provide a consistent
> ABI, etc.  Those things always need to be balanced against overall complexity, but
> IMO providing a vendor callback doesn't add anywhere near enough complexity to
> justify open coding the same concept in every vendor implementation.

Ack. My goal was to steer this implementation towards reusing existing
KVM synchronization to protect guest memory population within KVM
vendor logic rather than relying on guest_memfd filemap lock to
provide the needed protection here. That being said, I agree that we
can't solve this problem cleanly in a manner that works for all
architectures.