On 8/9/25 17:15, Peter Maydell wrote:
On Mon, 8 Sept 2025 at 16:09, Daniel P. Berrangé <[email protected]> wrote:
On Mon, Sep 08, 2025 at 01:50:57PM +0100, Peter Maydell wrote:
Currently our security policy defines a "virtualization use case"
where we consider bugs to be security issues, and a
"non-virtualization use case" where we do not make any security
guarantees and don't consider bugs to be security issues.
The rationale for this split is that much code in QEMU is older and
was not written with malicious guests in mind, and we don't have the
resources to audit, fix and defend it. So instead we inform users
about what the can in practice rely on as a security barrier, and
what they can't.
We don't currently restrict the "virtualization use case" to any
particular set of machine types. This means that we have effectively
barred ourselves from adding KVM support to any machine type that we
don't want to put into the "bugs are security issues" category, even
if it would be useful for users to be able to get better performance
with a trusted guest by enabling KVM. This seems an unnecessary
restriction, and in practice the set of machine types it makes
sense to use for untrusted-guest virtualization is quite small.
Specifically, we would like to be able to enable the use of
KVM with the imx8 development board machine types, but we don't
want to commit ourselves to having to support those SoC models
and device models as part of QEMU's security boundary:
https://lore.kernel.org/qemu-devel/[email protected]/
This patch updates the security policy to explicitly list the
machine types we consider to be useful for the "virtualization
use case".
This is an RFC partly to see if we have consensus that this change
makes sense, and partly because I was only able to identify the
machine types we want to cover for some of our target architectures.
If maintainers for the other architectures could clarify which
machine types work with KVM that would be helpful.
The split of "virtualization" vs "non-virtualization" use case
in the docs was always as rather a crude hack.
"Virtualization uses cases" was more or less a code phrase to
mean "the subset of QEMU that we traditionally shipped in RHEL"
as that is approximately what we have reasonable confidence
about.
Personally I wouldn't assign strict equivalence between "machine
can use KVM" and "virtualization use case".
I agree, but this is effectively what our docs are currently doing,
and what I'm trying to decouple with this patch...
Case in point - the "isapc" machine type can use KVM but I would
not consider that to be a virtualization use case, and would likely
reject a security report if it /only/ affected isapc, and not 'pc'
/ 'q35'.
We didn't want to undertake the work to annotate every QOM/QDev
impl in QEMU with info about whether we considered it in scope
for security fixes or not, which is what we really ought to do
at some point. The main challenge is someone having the time
to do the audit & annotation work.
diff --git a/docs/system/security.rst b/docs/system/security.rst
index f2092c8768b..395c2d7fb88 100644
--- a/docs/system/security.rst
+++ b/docs/system/security.rst
@@ -35,6 +35,34 @@ malicious:
Bugs affecting these entities are evaluated on whether they can cause damage
in
real-world use cases and treated as security bugs if this is the case.
+To be covered by this security support policy you must:
+
+- use a virtualization accelerator like KVM or HVF
If the "split-accel" (HW virtualization + SW emulation) effort lands,
we'll have a case of HVF + TCG. Do we want to consider it within our
security support policy? I doubt it (at least at the beginning). However
that would match "use a virtualization accelerator like KVM or HVF" (AND
"use one of the machine types listed below" with the 'virt' machine).
Should we reword? Or explicit "Any SW emulation use case is NOT
covered"?
+- use one of the machine types listed below
+
+It may be possible to use other machine types with a virtualization
+accelerator to provide improved performance with a trusted guest
+workload, but any machine type not listed here should not be
+considered to be providing guest isolation or security guarantees,
+and falls under the "non-virtualization use case".
+
+Supported machine types for the virtualization use case, by target
architecture:
+
+aarch64
+ ``virt``
+i386, x86_64
+ ``microvm``, ``xenfv``, ``xenpv``, ``xenpvh``, ``pc``, ``q35``, ``isapc``
+s390x
+ ``s390-ccw-virtio``
+loongarch64:
+ ``virt``
+ppc, ppc64:
+ TODO
+mips, mips64:
+ TODO
+riscv32, riscv64:
+ TODO
Currently 'virtualization use case' is reasonably vague such that we can
bend its scope as we desire, at the time it is questioned in a possible
security report.
Machine types are only one aspect of this. Devices are the other, and
the area where it gets significantly more fuzzy and difficult because
essentially any device can be used with KVM, and where we draw the
line is fairly arbitrary.
I think that being vague like this is a disservice to our users.
If I'm a user of QEMU, I'd like to know whether I'm inside the
line or outside of it before I put my config into production,
not later on when it turns out there was an exploitable bug
that wasn't classified as a security issue...
Most devices can't in fact be used with KVM, because they're
sysbus devices that aren't used in the machines that you can
use with KVM. Pluggable devices are rarer (and yes, under
our current policy random PCI devices are effectively
in-scope).
thanks
-- PMM
