Re: [PATCH v7 00/15] x86/hvm: I/O emulation cleanup and fix | Full Backtrace of domU's X crash caused by SSE2 istruction in attachment

From: Fabio Fantoni <fabio.fantoni@m2r.biz>
To: Jan Beulich <JBeulich@suse.com>, Paul Durrant <Paul.Durrant@citrix.com>
Cc: Andrew Cooper <Andrew.Cooper3@citrix.com>,
	zhi.a.wang@intel.com,
	"xen-devel@lists.xen.org" <xen-devel@lists.xen.org>
Subject: Re: [PATCH v7 00/15] x86/hvm: I/O emulation cleanup and fix | Full Backtrace of domU's X crash caused by SSE2 istruction in attachment
Date: Fri, 10 Jul 2015 12:51:41 +0200	[thread overview]
Message-ID: <559FA3BD.70806@m2r.biz> (raw)
In-Reply-To: <559FB877020000780008F63C@mail.emea.novell.com>

Il 10/07/2015 12:20, Jan Beulich ha scritto:
>>>> On 10.07.15 at 12:09, <fabio.fantoni@m2r.biz> wrote:
>> Il 10/07/2015 11:54, Fabio Fantoni ha scritto:
>>> Il 10/07/2015 11:31, Paul Durrant ha scritto:
>>>>> -----Original Message-----
>>>>> From: Fabio Fantoni [mailto:fabio.fantoni@m2r.biz]
>>>>> Sent: 10 July 2015 10:28
>>>>> To: Paul Durrant; xen-devel@lists.xen.org
>>>>> Cc: Jan Beulich; Andrew Cooper
>>>>> Subject: Re: [Xen-devel] [PATCH v7 00/15] x86/hvm: I/O emulation
>>>>> cleanup
>>>>> and fix | Full Backtrace of domU's X crash caused by SSE2 istruction in
>>>>> attachment
>>>>>
>>>>> Il 09/07/2015 15:10, Paul Durrant ha scritto:
>>>>>> This patch series re-works much of the code involved in emulation
>>>>>> of port
>>>>>> and memory mapped I/O for HVM guests.
>>>>>>
>>>>>> The code has become very convoluted and, at least by inspection,
>>>>>> certain
>>>>>> emulations will apparently malfunction.
>>>>>>
>>>>>> The series is broken down into 15 patches (which are also available in
>>>>>> my xenbits repo:
>>>>> http://xenbits.xen.org/gitweb/?p=people/pauldu/xen.git
>>>>>> on the emulation34 branch).
>>>>> Yesterday I retried with this version and seems that you fixed
>>>>> something
>>>>> that make possible atleast debug in the domU.
>>>>>
>>>>> I taken gdb data of X crash inside Sid hvm domU:
>>>>> #0  sse2_blt (src_bits=<optimized out>, dst_bits=<optimized out>,
>>>>> src_stride=<optimized out>, dst_stride=<optimized out>,
>>>>> src_bpp=<optimized out>, src_x=<optimized out>, src_y=0, dest_x=0,
>>>>> dest_y=0, width=1024, height=<optimized out>, dst_bpp=<optimized out>,
>>>>> imp=<optimized out>) at ../../pixman/pixman-sse2.c:4773
>>>>>
>>>>> Latest istruction:
>>>>> => 0x7f70360ef8eb <sse2_blt+459>:    movaps %xmm0,-0x10(%rsi)
>>>>>
>>>>> Full log in attachment.
>>>>>
>>>>> If you need more informations/tests tell me and I'll post them.
>>>>>
>>>> I imagine you got a GP fault due to handle_mmio() returning
>>>> X86EMUL_UNHANDLEABLE, but that's only a guess.
>>>> I suggest you try to instrument Xen a little to find out why.
>>> Thanks for reply, sorry but I not understand exactly what I must do.
>>> Can you detail please?
>> I take a look in xen/arch/x86/x86_emulate/x86_emulate.c and the
>> istruction seems like this other found by xengt developers:
>> https://github.com/01org/XenGT-Preview-xen/commit/f2bad31f80f698a452c37cb39841
>> da8e4f69350f
> Other than MOVD, MOVAPS is already being supported by the
> insn emulator.
>
> Jan
>
Thanks for your reply.
Then why do you think MOVAPS fails in my test? If you need it, here is 
the full gdb output:
http://lists.xen.org/archives/html/xen-devel/2015-07/txtpTIgy5CpzU.txt
If you need more informations/tests tell me and I'll post them.
Sse2 was introduced in cpus 11 years ago, so I think it would be useful 
to add support for all missing instructions to avoid more cases like 
this (and it seems to improve performance too, comparing the same task 
executed with/without sse2).
I'm going to try updated xen staging with MOVD support patch added, any 
other useful information for other tests/debug is appreciated.

Thanks for any reply and sorry for my bad english.