Re: [PATCH] x86/hyper-v: use cheaper HVCALL_FLUSH_VIRTUAL_ADDRESS_{LIST,SPACE} hypercalls when possible

[Vitaly Kuznetsov <vkuznets@redhat.com>]

KY Srinivasan <kys@microsoft.com> writes:

>> -----Original Message-----
>> From: Michael Kelley (EOSG)
>> Sent: Tuesday, June 19, 2018 10:57 AM
>> To: Vitaly Kuznetsov <vkuznets@redhat.com>; x86@kernel.org
>> Cc: devel@linuxdriverproject.org; linux-kernel@vger.kernel.org; KY
>> Srinivasan <kys@microsoft.com>; Haiyang Zhang
>> <haiyangz@microsoft.com>; Stephen Hemminger
>> <sthemmin@microsoft.com>; Thomas Gleixner <tglx@linutronix.de>; Ingo
>> Molnar <mingo@redhat.com>; H. Peter Anvin <hpa@zytor.com>; Tianyu Lan
>> <Tianyu.Lan@microsoft.com>
>> Subject: RE: [PATCH] x86/hyper-v: use cheaper
>> HVCALL_FLUSH_VIRTUAL_ADDRESS_{LIST,SPACE} hypercalls when possible
>> 
>> ...
>>>
>> This is a good idea.  We should probably do the same with the hypercalls for
>> sending
>> IPIs -- try the simpler version first and move to the more complex _EX
>> version only
>> if necessary.
> I am not sure if this would work correctly. When I was developing the IPI enlightenment, 
> what I remember was that the guest is expected to use the API recommended by the Hypervisor.
>

I was under the same impression when I implemented PV TLB flush. Turns
out HV_X64_EX_PROCESSOR_MASKS_RECOMMENDED is a misnomer or at least
Windows treats it as HV_X64_EX_PROCESSOR_MASKS_AVAILABLE instead using
only when needed.

My guess would be that the situation with IPI is the same. In any case I
can try to implement Hyper-V style PV IPIs for Windows in KVM and we'll
see how Windows uses these hypercalls :-)

-- 
  Vitaly