Re: [PATCH 2/2] x86/speculation: Support "Enhanced IBRS" on future CPUs

From: Paolo Bonzini <pbonzini@redhat.com>
To: David Woodhouse <dwmw2@infradead.org>,
	tglx@linutronix.de, x86@kernel.org, kvm@vger.kernel.org,
	torvalds@linux-foundation.org, linux-kernel@vger.kernel.org,
	arjan.van.de.ven@intel.com, dave.hansen@intel.com
Subject: Re: [PATCH 2/2] x86/speculation: Support "Enhanced IBRS" on future CPUs
Date: Tue, 13 Feb 2018 10:58:00 +0100	[thread overview]
Message-ID: <27c85759-e662-d281-f8a0-0a80ca8ee18f@redhat.com> (raw)
In-Reply-To: <1518509708.12890.33.camel@infradead.org>

On 13/02/2018 09:15, David Woodhouse wrote:
>>>  
>>> -		if (!data)
>>> +		if (!data && !spectre_v2_ibrs_all())
>>>  			break;
>> This should check the value of IBRS_ALL in the VM, not in the host.
> No, it's host we want. If IBRS_ALL is set in the host, we set the
> actual hardware MSR once at boot time and never touch it again. The
> SPEC_CTRL MSR we expose to guests is purely a no-op fiction.
> 
> If spectre_v2_ibrs_all() is true then KVM should *never* actually pass
> through or touch the real MSR.

That would be nice but unfortunately it's not possible. :(

The VM might actually not have IBRS_ALL, as usual the reason is
migration compatibility.  In that case, that no-op fiction would be very
slow because the VM will actually do a lot of SPEC_CTRL writes.

So the right logic is:

- if the VM has IBRS_ALL, pass through the MSR when it is zero and
intercept writes when it is one (no writes should happen)

- if the VM doesn't have IBRS_ALL, do as we are doing now, independent
of what the host spectre_v2_ibrs_all() setting is.

Paolo