Received-SPF: pass (google.com: domain of linux-kernel+bounces-182336-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Date: Fri, 17 May 2024 09:12:45 -0700
In-Reply-To: <ohvjbokpaxagc26kxmlrujab7cw3bekgi5ln7dt46cbsaxcqqh@crvqeohfazmf>
Precedence: bulk
Mime-Version: 1.0
References: <20240517121450.20420-1-jgross@suse.com> <ohvjbokpaxagc26kxmlrujab7cw3bekgi5ln7dt46cbsaxcqqh@crvqeohfazmf>
Message-ID: <Zkd28IN8KvyD4GIX@google.com>
Subject: Re: [PATCH] x86/kvm/tdx: Save %rbp in TDX_MODULE_CALL
From: Sean Christopherson <seanjc@google.com>
To: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Juergen Gross <jgross@suse.com>, linux-kernel@vger.kernel.org, x86@kernel.org, 
	linux-coco@lists.linux.dev, Dave Hansen <dave.hansen@linux.intel.com>, 
	Thomas Gleixner <tglx@linutronix.de>, Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>, 
	"H. Peter Anvin" <hpa@zytor.com>
Content-Type: text/plain; charset="us-ascii"

On Fri, May 17, 2024, Kirill A. Shutemov wrote:
> On Fri, May 17, 2024 at 02:14:50PM +0200, Juergen Gross wrote:
> > While testing TDX host support patches, a crash of the host has been
> > observed a few instructions after doing a seamcall. Reason was a
> > clobbered %rbp (set to 0), which occurred in spite of the TDX module
> > offering the feature NOT to modify %rbp across TDX module calls.
> > 
> > In order not having to build the host kernel with CONFIG_FRAME_POINTER,
> > save %rbp across a seamcall/tdcall.
> 
> There's a feature in TDX module 1.5 that prevents RBP modification across
> TDH.VP.ENTER SEAMCALL. See NO_RBP_MOD in TDX Module 1.5 ABI spec.

LOL, "feature".  How was clobbering RBP not treated as a bug?  I'm party joking,
but also quite serious.  Unless I'm missing something, the guest ABI changes
based on whether or not NO_RBP_MOD is enabled, as a TDVMCALL that was previously
valid would now fail if the guest attempts to expose RBP to the host.

The whole point of Intel defining a guest-host ABI is to allow interoperability
between hypervisors and guests.  Allowing the hypervisor to arbitrarily change the
ABI is asinine.

> I think it has to be enabled for all TDs and TDX modules that don't
> support it need to be rejected.

Yes, because as above, IIUC it's a breaking change for the guest ABI.