[CRIU] [PATCH 1/4] [RFC] cr-dump: don't call parasite_fixup_vdso() on a non-x86 architecture

[Alexander Kartashov]

On 05/31/2013 09:55 AM, Cyrill Gorcunov wrote:
> On Fri, May 31, 2013 at 09:51:29AM +0400, Alexander Kartashov wrote:
>> On 05/31/2013 09:27 AM, Cyrill Gorcunov wrote:
>>> On Fri, May 31, 2013 at 09:20:45AM +0400, Alexander Kartashov wrote:
>>>> The call to parasite_fixup_vdso() fails on ARM that results
>>>> in a dump failure.
>>>>
>>>> Signed-off-by: Alexander Kartashov <alekskartashov at parallels.com>
>>>> Cc: Cyrill Gorcunov <gorcunov at openvz.org>
>>> I'm fine with the patch as it is, but could you please explain how
>>> it fails? Mind to provide more details?
>> Dumping fails with the following log:
>>
>> (00.055367) Error (parasite-syscall.c:740): Can't read pme for pid
>> 1775: Input/output error
>> (00.056384) Error (cr-dump.c:1423): Can't fixup vdso VMAs (pid: 1775)
>> (00.061399) Error (cr-dump.c:1718): Dumping FAILED.
>>
>> Offending VMA address is FFFF0000. It seems it's better to insert
>> the check
>>
>> if (vma->vma.start > TASK_SIZE)
>>
>> into the parasite_fixup_vdso() rather than using this macro code scissors.
> I see. Thanks. Lets go further. There is no vdso on arm, right? So I guess
> better to move this snippet to x86/ specifics thus arm won't have it
> at all. I'll try to cook a patch.
Yes, I think it would be even better.

-- 
Sincerely yours,
Alexander Kartashov

Intern
Core team

www.parallels.com

Skype: aleksandr.kartashov
Email: alekskartashov at parallels.com