Re: [PATCH v3] KVM: x86/mmu: Warn on iTLB multi-hit for possible problems

[Sean Christopherson]

On Wed, Oct 20, 2021, Li Yu wrote:
> Warn for guest huge pages split if iTLB multi-hit bug is present
> and CPU mitigations is enabled.
> 
> Warn for possible CPU lockup if iTLB multi-hit bug is present but
> CPU mitigations is disabled.
> 
> Signed-off-by: Li Yu <liyu.yukiteru@xxxxxxxxxxxxx>
> ---

Gah, my last reply in the previous version was offlist.  I would like an answer
to the below question before we complicate KVM just to log a message.

On Wed, Oct 20, 2021 at 12:51 PM Sean Christopherson <seanjc@xxxxxxxxxx> wrote:
>
> On Wed, Oct 20, 2021, Li Yu wrote:
> > On 2021/10/20 00:23, Sean Christopherson wrote:
> > I think it is necessary to log something when `nx_huge_pages` is `auto` or
> > `off`, and the bug is present.
>
> Why is it necessary?  Specifically, what action can be taken based on KVM logging
> that can't reasonably be taken based on all the other information provided by the
> kernel?  We should teach userspace to go look at the vulnerabilites info, not to
> scrape the kernel log.