Re: [PATCH v8 4/4] kvm: vmx: virtualize split lock detection

From: Thomas Gleixner <tglx@linutronix.de>
To: Sean Christopherson <sean.j.christopherson@intel.com>
Cc: Xiaoyao Li <xiaoyao.li@intel.com>,
	Paolo Bonzini <pbonzini@redhat.com>,
	kvm@vger.kernel.org, linux-kernel@vger.kernel.org,
	x86@kernel.org, Ingo Molnar <mingo@redhat.com>,
	Borislav Petkov <bp@alien8.de>, Andy Lutomirski <luto@kernel.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Arvind Sankar <nivedita@alum.mit.edu>
Subject: Re: [PATCH v8 4/4] kvm: vmx: virtualize split lock detection
Date: Wed, 06 May 2020 14:12:48 +0200	[thread overview]
Message-ID: <87h7wtl0sf.fsf@nanos.tec.linutronix.de> (raw)
In-Reply-To: <20200505030736.GA20916@linux.intel.com>

Sean Christopherson <sean.j.christopherson@intel.com> writes:
> On Wed, Apr 15, 2020 at 02:43:18PM -0700, Sean Christopherson wrote:
>> On Wed, Apr 15, 2020 at 11:22:11PM +0200, Thomas Gleixner wrote:
>> > So we can go with the proposed mode of allowing the write but not
>> > propagating it. If the resulting split lock #AC originates from CPL != 3
>> > then the guest will be killed with SIGBUS. If it originates from CPL ==
>> > 3 and the guest has user #AC disabled then it will be killed as well.
>> 
>> An idea that's been floated around to avoid killing the guest on a CPL==3
>> split-lock #AC is to add a STICKY bit to MSR_TEST_CTRL that KVM can
>> virtualize to tell the guest that attempting to disable SLD is futile,
>> e.g. so that the guest can kill its misbehaving userspace apps instead of
>> trying to disable SLD and getting killed by the host.
>
> Circling back to this.  KVM needs access to sld_state in one form or another
> if we want to add a KVM hint when the host is in fatal mode.  Three options
> I've come up with:
>
>   1. Bite the bullet and export sld_state.  
>
>   2. Add an is_split_fatal_wrapper().  Ugly since it needs to be non-inline
>      to avoid triggering (1).
>
>   3. Add a synthetic feature flag, e.g. X86_FEATURE_SLD_FATAL, and drop
>      sld_state altogether.
>
> I like (3) because it requires the least amount of code when all is said
> and done, doesn't require more exports, and as a bonus it'd probably be nice
> for userspace to see sld_fatal in /proc/cpuinfo.

#3 makes sense and is elegant.

Thanks,

        tglx