Re: [PATCH] x86/vlapic: Don't reset APIC mode/ID when handling INIT signal

From: Juergen Gross <jgross@suse.com>
To: Boris Ostrovsky <boris.ostrovsky@oracle.com>,
	Jan Beulich <JBeulich@suse.com>, Chao Gao <chao.gao@intel.com>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>,
	Ankur Arora <ankur.a.arora@oracle.com>,
	xen-devel@lists.xen.org
Subject: Re: [PATCH] x86/vlapic: Don't reset APIC mode/ID when handling INIT signal
Date: Wed, 19 Apr 2017 16:13:07 +0200	[thread overview]
Message-ID: <4d5a9e10-f680-278a-af11-7aab5562c97c@suse.com> (raw)
In-Reply-To: <fb1476ed-33ec-2a07-bd53-022097ac1398@suse.com>

On 19/04/17 15:58, Juergen Gross wrote:
> On 19/04/17 15:43, Boris Ostrovsky wrote:
>>
>>
>> On 04/19/2017 09:21 AM, Jan Beulich wrote:
>>>>>> On 19.04.17 at 06:56, <chao.gao@intel.com> wrote:
>>>> On Wed, Apr 19, 2017 at 02:48:57AM -0600, Jan Beulich wrote:
>>>>>>>> On 18.04.17 at 23:51, <chao.gao@intel.com> wrote:
>>>>>> HVM guest can't enable x2apic as XENFEAT_hvm_pirqs is exposed to it.
>>>>>> Why we have this restriction? As a consequence, This patch isn't
>>>>>> tested in that case.
>>>>>
>>>>> HVM guests very well can use x2APIC, I'm seeing them use it all
>>>>> the time. Please be more specific with your question.
>>>>
>>>> OK. I failed to enable guest's x2apic, with the output: IRQ Remapping
>>>> doesn't
>>>> support X2APIC mode. Eventually, I found the xen_x2apic_para_available()
>>>> in linux kernel and it returns false when Xen exposes
>>>> XENFEAT_hvm_pirqs to
>>>> guest.
>>>
>>> Looks more like a Linux/pvops question, which I continue to not
>>> really be an expert on. Let's ask the maintainers (now Cc-ed)
>>> (but it may well be that this simply depends on whether you
>>> enable PVHVM mode)...
>>
>> Interestingly enough, I was looking at this last week and wondering why
>> it was there. Ankur (copied) found an issue related to x2apic he thought
>> was PVH-specific, but that was because HVM guests don't turn it on.
>>
>> Konrad couldn't remember why it was there neither.
> 
> I think this is the answer:
> 
> https://lists.xen.org/archives/html/xen-devel/2010-12/msg00373.html
> 
> Basically some very old assumption...
> 
> I think we can remove the offending test and do some more cleanup in
> this area (I think there are some #ifdef's and tests which can be
> removed, too).
> 
> I'll post some patches.

Thinking more about it: now is a bad time. This would make the planned
revert of 72a9b186292 and da72ff5bfcb0 even more difficult.

Juergen

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel