Re: [PATCH v7 12/12] xen: clarify the security-support status of Kconfig options on ARM

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 25/07/2018 09:46, Julien Grall wrote:
>
>
> On 24/07/18 23:31, Stefano Stabellini wrote:
>> On Mon, 23 Jul 2018, Julien Grall wrote:
>>> Hi,
>>>
>>> On 07/07/18 00:14, Stefano Stabellini wrote:
>>>> Signed-off-by: Stefano Stabellini <sstabellini@kernel.org>
>>>> CC: George.Dunlap@eu.citrix.com
>>>> CC: Ian.Jackson@eu.citrix.com
>>>> CC: jbeulich@suse.com
>>>> CC: andrew.cooper3@citrix.com
>>>> ---
>>>>    SUPPORT.md | 10 ++++++++++
>>>>    1 file changed, 10 insertions(+)
>>>>
>>>> diff --git a/SUPPORT.md b/SUPPORT.md
>>>> index e3e49e2..151a63d 100644
>>>> --- a/SUPPORT.md
>>>> +++ b/SUPPORT.md
>>>> @@ -22,6 +22,16 @@ EXPERT and DEBUG Kconfig options are not security
>>>> supported. Other
>>>>    Kconfig options are supported, if the related features are
>>>> marked as
>>>>    supported in this document.
>>>>    +On ARM, a wider range of Kconfig configurations is available to
>>>> enable
>>>> +very small lines of code counts in the hypervisor. Not all possible
>>>> +combinations of kconfig options are security supported. Instead, a
>>>> few
>>>
>>> NIT: s/kconfig/Kconfig/
>>>
>>>> +pre-canned configurations have been added to xen/arch/arm/configs:
>>>> they
>>>> +are security suppored. Configurations derived from the pre-canned
>>>> files
>>>
>>> s/suppored/supported/
>>
>> I'll fix
>>
>>
>>>> +by adding non-listed options with their default values, or by
>>>> enabling
>>>> +any of the platform options under "Platform Support" (and their
>>>> +dependent options) are security supported, unless stated
>>>> +otherwise.
>>>
>>> I am not entirely sure to understand the implications the paragraph.
>>
>> It is meant to say:
>>
>> 1) xen/arch/arm/configs config files are security supported
>> 2) default values of any kconfig options are security supported
>> 3) if an option is marked as not security supported in SUPPORT.md, then
>>     it is not security supported, no matter the default value
>> 4) everything else is not security supported
>>   Should I try to clarify it? I guess I should make clear that a .config
>> with an unsupported option is unsupported as a whole. I can add:
>>
>>   "A configuration with one or more unsupported options, is not
>>   unsupported."
>>
>>
>>> For instance, if I choose arm64_defconfig, memaccess will be enabled by
>>> default but any use of it is not security supported. What will be
>>> the state of
>>> the security support for that .config?
>>
>> Yes, memaccess will default to enable. However, SUPPORT.md says it is
>> not security supported, hence, the result is that the .config is not
>> security supported, according to (3).
>
> We really don't want that. That arm64_defconfig is the default config
> for Xen. Anyone using it will not be security supported.
>
> Distros will likely use the default config as it enables everything.
> If I were a package maintainer, I would expect at minimum to security
> support the .config. This does not mean that using a specific feature
> will be supported.

Anything you can select in menuconfig without passing
XEN_CONFIG_EXPERT=y is security supported.  Anything hidden behind
XEN_CONFIG_EXPERT is security supported in its default configuration.

>
>>
>> There is a catch though. In the specific case of memaccess, SUPPORT.md
>> only states the following:
>>
>> ### Virtual Machine Introspection
>>
>>      Status, x86: Supported, not security supported
>>
>> Which doesn't say anything about ARM. It would be a good idea to do the
>> same that x86 is doing (Supported, not security supported)?
>
> Memaccess has never been considered to be supported on Arm, yet it is
> enabled by default as on x86. However, most of the code in that
> context is gated by "memaccess_enabled". So what is not security
> supported is the use of memaccess.

memaccess on x86 long predates any kind of Kconfig.

At some point, we're going to have to start saying that anything not
stated in Support.md is unsupported, but in practice, given the state of
the document, most things are actually "anything which isn't mentioned
needs a decision and a patch making to the file".

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel