[PATCH] Kernel Lockdown: Add an option to allow raw MSR access even, in confidentiality mode.

[Matt Parnell]

...possibly a bug on my older 4790k system, on my 8550u this doesn't
seem to happen. Strange.

Either way, I can stop bugging you if you'd like.

On 12/2/19 8:50 PM, Matt Parnell wrote:
> Correction: I'm out of caffeine, tired, and it has made me an idiot.
>
> That message triggers regardless, it seems. I apologize.
>
> On 12/2/19 8:24 PM, Matt Parnell wrote:
>> For what it is worth, this doesn't happen with lockdown disabled.
>>
>> That message and the code that checks for mitigations is in
>> arch/x86/kvm/vmx/vmx.c - for some reason locking down the MSRs is even
>> making the kernel think that the MSR for the mitigation isn't there,
>> meaning that it is also likely not mitigating the bug.
>>
>> On 12/2/19 8:16 PM, Matthew Garrett wrote:
>>> On Mon, Dec 2, 2019 at 6:01 PM Matt Parnell <mparnell at gmail.com> wrote:
>>>> I should also mention the kernel itself thinks it is vulnerable with the
>>>> MSRs locked down:
>>>>
>>>> [    7.367922] L1TF CPU bug present and SMT on, data leak possible. See
>>>> CVE-2018-3646 and
>>>> https://www.kernel.org/doc/html/latest/admin-guide/hw-vuln/l1tf.html for
>>>> details.
>>> The lockdown code doesn't touch any of the codepaths the kernel uses
>>> to access MSRs itself (a *lot* would break in that case), so if the
>>> kernel is asserting this inappropriately then that seems like a kernel
>>> bug.