[CRIU] Migration ...

[Pavel Emelyanov]

On 01/27/2015 02:30 PM, Cyrill Gorcunov wrote:
> On Tue, Jan 27, 2015 at 02:03:17PM +0300, Cyrill Gorcunov wrote:
>>> [878588.364797] test2[28350]: segfault at 3431474570 ip
>>> 0000003431474570 sp 00007fffbf522cc8 error 14
>>>
>>> Further evidence of the segmentation violation suffered by the
>>> reinstated process. I have attached the checkpoint directory
>>> (compressed) ...
>>
>> Thanks. I'll ping you once I find any clue.
> 
> OK, from the images you've sent me
> 
> [878588.364797] test2[28350]: segfault at 3431474570 ip 0000003431474570 sp 00007fffbf522cc8 error 14
> 
> in mm image we've
> 
> {
>         start: 0x00003431400000
>         end: 0x0000343158a000
>         pgoff: 0000000000000000
>         shmid: 0x00000000000004
>         prot: 0x5
>         flags: 0x2
>         status: 0x41
>         fd: 0xffffffffffffffff
>         fdflags: 0
> }
> 
> 
> which in turn points to
> 
> id: 0x4 flags: 0 pos: 0000000000000000 fown: { uid: 0 euid: 0 signum: 0 pid_type: 0 pid: 0 }  name: "/lib64/libc-2.12.so" size: 0x000000001d6668 
> 
> so it looks like something is failing inside libc library upon restore, neeed to take
> a look into libc source code (if only I'm not missing something obvious).

I've seen such when the prelink was turned on on source and destination nodes.
Mark, can you compare (checksums or with cmp tool) the contents of the /lib64/libc-2.12.so
files? They should coincide on both nodes, otherwise the program may crash after restore
since it effectively gets its libc replaced on-the-fly :)

Thanks,
Pavel