[CRIU] [PATCH 2/4] Make vDSO warning behave as such

[Christopher Covington]

Hi Cyrill,

On 07/11/2014 11:40 AM, Cyrill Gorcunov wrote:
> On Fri, Jul 11, 2014 at 11:19:52AM -0400, Christopher Covington wrote:
>> On 07/11/2014 04:03 AM, Cyrill Gorcunov wrote:
>>> On Thu, Jul 10, 2014 at 11:09:05AM -0400, Christopher Covington wrote:
>>>> This allows checkpoint dump on AArch64 to continue. Without this,
>>>> checkpoints on norandmaps systems that function fine cannot
>>>> proceed.
>>>>
>>>> Signed-off-by: Christopher Covington <cov at codeaurora.org>
>>>
>>> What about restore, does it pass with this patch applied?
>>
>> Workloads that don't use the VDSO work but workloads that do use the VDSO (or
>> at least the vdso00 test) segfault. This is a regression. vdso00 and vdso01
>> both pass with an earlier build of AArch64 CRIU.
> 
> So that this patch doesn't solve the main problem, right? With patch
> applied dump will pass but then srgfault on restore. That said I need
> some help in resolving issue on arm machine. Mind to take a look what
> exactly causes sigsegv once time permit?

One quick data point is that if I define VDSO for aarch64 and create symlinks
to the x86 vdso files from arch/aarch64, vdso00 and vdso01 pass on 3.15.0-rc7+.

Christopher

-- 
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by the Linux Foundation.