diff mbox series

SUPPORT.md: add Dom0less as Supported

Message ID alpine.DEB.2.21.2107131734170.23286@sstabellini-ThinkPad-T480s (mailing list archive)
State Superseded
Headers show
Series SUPPORT.md: add Dom0less as Supported | expand

Commit Message

Stefano Stabellini July 14, 2021, 12:39 a.m. UTC
Add Dom0less to SUPPORT.md to clarify its support status. The feature is
mature enough and small enough to make it security supported.

Signed-off-by: Stefano Stabellini <stefano.stabellini@xilinx.com>

Comments

Julien Grall July 14, 2021, 6:01 p.m. UTC | #1
Hi Stefano,

On 14/07/2021 01:39, Stefano Stabellini wrote:
> Add Dom0less to SUPPORT.md to clarify its support status. The feature is
> mature enough and small enough to make it security supported.
> 
> Signed-off-by: Stefano Stabellini <stefano.stabellini@xilinx.com>
> 
> diff --git a/SUPPORT.md b/SUPPORT.md
> index 317392d8f3..c777f3da72 100644
> --- a/SUPPORT.md
> +++ b/SUPPORT.md
> @@ -832,6 +832,12 @@ OVMF firmware implements the UEFI boot protocol.
>   
>       Status, qemu-xen: Supported
>   
> +## Dom0less
> +
> +Guest creation from the hypervisor at boot without Dom0 intervention.
> +
> +    Status, ARM: Supported
> +

After XSA-372, we will not scrubbed memory assigned to dom0less DomU 
when bootscrub=on. Do we want to exclude this combination or mention 
that XSAs will not be issued if the domU can read secret from unscrubbed 
memory?

Cheers,
Stefano Stabellini July 14, 2021, 7:28 p.m. UTC | #2
On Wed, 14 Jul 2021, Julien Grall wrote:
> Hi Stefano,
> 
> On 14/07/2021 01:39, Stefano Stabellini wrote:
> > Add Dom0less to SUPPORT.md to clarify its support status. The feature is
> > mature enough and small enough to make it security supported.
> > 
> > Signed-off-by: Stefano Stabellini <stefano.stabellini@xilinx.com>
> > 
> > diff --git a/SUPPORT.md b/SUPPORT.md
> > index 317392d8f3..c777f3da72 100644
> > --- a/SUPPORT.md
> > +++ b/SUPPORT.md
> > @@ -832,6 +832,12 @@ OVMF firmware implements the UEFI boot protocol.
> >         Status, qemu-xen: Supported
> >   +## Dom0less
> > +
> > +Guest creation from the hypervisor at boot without Dom0 intervention.
> > +
> > +    Status, ARM: Supported
> > +
> 
> After XSA-372, we will not scrubbed memory assigned to dom0less DomU when
> bootscrub=on.

Do you mean *before* XSA-372, right? I thought the XSA-372 patches take
care of the problem?


> Do we want to exclude this combination or mention that XSAs will
> not be issued if the domU can read secret from unscrubbed memory?

I could say that if bootscrub=off then we won't issue XSAs for domUs reading
secrets from unscrubbed memory. But it is kind of obvious anyway? I am
happy to add it if you think it is good to clarify.
Julien Grall July 14, 2021, 7:43 p.m. UTC | #3
Hi Stefano,

On 14/07/2021 20:28, Stefano Stabellini wrote:
> On Wed, 14 Jul 2021, Julien Grall wrote:
>> Hi Stefano,
>>
>> On 14/07/2021 01:39, Stefano Stabellini wrote:
>>> Add Dom0less to SUPPORT.md to clarify its support status. The feature is
>>> mature enough and small enough to make it security supported.
>>>
>>> Signed-off-by: Stefano Stabellini <stefano.stabellini@xilinx.com>
>>>
>>> diff --git a/SUPPORT.md b/SUPPORT.md
>>> index 317392d8f3..c777f3da72 100644
>>> --- a/SUPPORT.md
>>> +++ b/SUPPORT.md
>>> @@ -832,6 +832,12 @@ OVMF firmware implements the UEFI boot protocol.
>>>          Status, qemu-xen: Supported
>>>    +## Dom0less
>>> +
>>> +Guest creation from the hypervisor at boot without Dom0 intervention.
>>> +
>>> +    Status, ARM: Supported
>>> +
>>
>> After XSA-372, we will not scrubbed memory assigned to dom0less DomU when
>> bootscrub=on.
> 
> Do you mean *before* XSA-372, right?

No, I really meant *after* XSA-372.

> I thought the XSA-372 patches take
> care of the problem?

It didn't. We have an open question for the bootscrub=on one. From the 
commit message of patch #1:

         2) The memory allocated for a domU will not be scrubbed anymore 
when an
         admin select bootscrub=on. This is not something we advertised, 
but if
         this is a concern we can introduce either force scrub for all 
domUs or
         a per-domain flag in the DT. The behavior for bootscrub=off and
         bootscrub=idle (default) has not changed.

>  >
>> Do we want to exclude this combination or mention that XSAs will
>> not be issued if the domU can read secret from unscrubbed memory?
> 
> I could say that if bootscrub=off then we won't issue XSAs for domUs reading
> secrets from unscrubbed memory. But it is kind of obvious anyway? I am
> happy to add it if you think it is good to clarify.

Right, it is pretty clear that bootscrub=off will not scrub the memory 
and the "problem" would not be specific to dom0less.

The one I asked to clarify is bootscrub=on because one may think the 
memory is scrubbed for dom0less domU for all the cases but bootscrub=off.

IIRC when we discussed about it on security@xen.org, we suggested that 
it would be fine to rely on the bootloader to scrub it. But I think this 
needs to be written down rather waiting for it to be re-discovered.

The other solution is to fix it properly.

Cheers,
Stefano Stabellini July 14, 2021, 11:46 p.m. UTC | #4
On Wed, 14 Jul 2021, Julien Grall wrote:
> Hi Stefano,
> 
> On 14/07/2021 20:28, Stefano Stabellini wrote:
> > On Wed, 14 Jul 2021, Julien Grall wrote:
> > > Hi Stefano,
> > > 
> > > On 14/07/2021 01:39, Stefano Stabellini wrote:
> > > > Add Dom0less to SUPPORT.md to clarify its support status. The feature is
> > > > mature enough and small enough to make it security supported.
> > > > 
> > > > Signed-off-by: Stefano Stabellini <stefano.stabellini@xilinx.com>
> > > > 
> > > > diff --git a/SUPPORT.md b/SUPPORT.md
> > > > index 317392d8f3..c777f3da72 100644
> > > > --- a/SUPPORT.md
> > > > +++ b/SUPPORT.md
> > > > @@ -832,6 +832,12 @@ OVMF firmware implements the UEFI boot protocol.
> > > >          Status, qemu-xen: Supported
> > > >    +## Dom0less
> > > > +
> > > > +Guest creation from the hypervisor at boot without Dom0 intervention.
> > > > +
> > > > +    Status, ARM: Supported
> > > > +
> > > 
> > > After XSA-372, we will not scrubbed memory assigned to dom0less DomU when
> > > bootscrub=on.
> > 
> > Do you mean *before* XSA-372, right?
> 
> No, I really meant *after* XSA-372.
> 
> > I thought the XSA-372 patches take
> > care of the problem?
> 
> It didn't. We have an open question for the bootscrub=on one. From the commit
> message of patch #1:
> 
>         2) The memory allocated for a domU will not be scrubbed anymore when
> an
>         admin select bootscrub=on. This is not something we advertised, but if
>         this is a concern we can introduce either force scrub for all domUs or
>         a per-domain flag in the DT. The behavior for bootscrub=off and
>         bootscrub=idle (default) has not changed.
>
> > > Do we want to exclude this combination or mention that XSAs will
> > > not be issued if the domU can read secret from unscrubbed memory?
> > 
> > I could say that if bootscrub=off then we won't issue XSAs for domUs reading
> > secrets from unscrubbed memory. But it is kind of obvious anyway? I am
> > happy to add it if you think it is good to clarify.
> 
> Right, it is pretty clear that bootscrub=off will not scrub the memory and the
> "problem" would not be specific to dom0less.
> 
> The one I asked to clarify is bootscrub=on because one may think the memory is
> scrubbed for dom0less domU for all the cases but bootscrub=off.
> 
> IIRC when we discussed about it on security@xen.org, we suggested that it
> would be fine to rely on the bootloader to scrub it. But I think this needs to
> be written down rather waiting for it to be re-discovered.

Ah right, I remember what you are talking about. I'll update the patch.
diff mbox series

Patch

diff --git a/SUPPORT.md b/SUPPORT.md
index 317392d8f3..c777f3da72 100644
--- a/SUPPORT.md
+++ b/SUPPORT.md
@@ -832,6 +832,12 @@  OVMF firmware implements the UEFI boot protocol.
 
     Status, qemu-xen: Supported
 
+## Dom0less
+
+Guest creation from the hypervisor at boot without Dom0 intervention.
+
+    Status, ARM: Supported
+
 # Format and definitions
 
 This file contains prose, and machine-readable fragments.