Re: linux-next: Tree for October 29 (x86 posttest)

[Masami Hiramatsu <mhiramat@redhat.com>]

Stephen Rothwell wrote:
> Hi all,
> 
> On Mon, 09 Nov 2009 19:40:11 -0500 Masami Hiramatsu<mhiramat@redhat.com>  wrote:
>>
>> Randy Dunlap wrote:
>>> On Thu, 29 Oct 2009 12:48:12 -0400 Masami Hiramatsu wrote:
>>>
>>>> Randy Dunlap wrote:
>>>>>
>>>>> I'm getting this on x86_64:
>>>>>
>>>>>      TEST    posttest
>>>>> Error: ffffffff810299e2:	f2 41 0f 28 52 b0    	repnz movaps -0x50(%r10),%xmm2
>>>>> Error: objdump says 6 bytes, but insn_get_length() says 4 (attr:0)
>>>>> make[2]: *** [posttest] Error 2
>>>>
>>>> Ah, that's should be fixed by AVX support patch which I've posted.
>>>> http://git.kernel.org/tip/e0e492e99b372c6990a5daca9e4683c341f1330e
>>>
>>> Has that been merged into linux-next?
>>> I'm still seeing this failure in linux-next-20091106.
>>
>> I checked it has been merged.
>> So, it should disappear from linux-next-20091109. :-)
> 
> That commit (e0e492e99b372c6990a5daca9e4683c341f1330e "x86: AVX
> instruction set decoder support") has been in linux-next since
> next-20091104, so if Randy's problem still existed in next-20091106, then
> it may be a different problem.

Hmm, this seems a bad sequence, movaps can't have repnz prefix,
since movaps is not string operation... It might happen since
AVX opcodes supporting processors are not released yet.

Anyway, I think we'd better change the posttest to report it
as warning instead of errors.

Thank you,

-- 
Masami Hiramatsu

Software Engineer
Hitachi Computer Products (America), Inc.
Software Solutions Division

e-mail: mhiramat@redhat.com