[CRIU] [PATCH] zdtm: unlinked socket with dentry alias

Pavel Emelyanov xemul at parallels.com
Tue Sep 29 03:16:24 PDT 2015


On 09/29/2015 11:47 AM, Stanislav Kinsburskiy wrote:
> 
> 
> 28.09.2015 20:46, Pavel Emelyanov пишет:
>> On 09/28/2015 05:38 PM, Stanislav Kinsburskiy wrote:
>>> There can be an unlinked UNIX socket, which dentry has a linked alias,
>>> pointing to the same inode.
>>> To check, that such socket is migrated successfully, test perform the
>>> following:
>>>
>>> Before suspend:
>>> 1) Creates UNIX socket and binds it.
>>> 2) Created hard link to the binded path.
>>> 3) Unlinks original UNIX bind path.
>>> 3) Forks a child, listen to unlinked socket and accepting connection.
>>>
>>> After resume:
>>> 1) Create UNIX socket and binds it to socket link.
>>> 2) Connects to socket.
>>> 3) Child have to accept the connection.
>>>
>>> Signed-off-by: Stanislav Kinsburskiy <skinsbursky at parallels.com>
>>> ---
>>>   test/zdtm/live/static/Makefile                   |    1
>>>   test/zdtm/live/static/deleted_unix_sock_linked.c |  196 ++++++++++++++++++++++
>> I don't see it in zdtm.sh list :) Does it fail? Why?
> 
> It fails for rhel6 kernel. So the test was created for it.
> But since zdtm in maintained in CRIU now, I sent it here.
> And, probably, it can be useful for CRIU as well, if you don't have such 
> a test case.

May I ask you to run this test with CRIU, what will it result in?



More information about the CRIU mailing list