[CRIU] Diagnosing runaway socket error

[Adrian Reber]

Sorry for the late reply. The mailing list is not the best place any
more to ask question. You have better chances of getting help if you
open a github issue.

In your case I would use 'lsof' to look which inode the socket has.

		Adrian

On Sun, Jun 12, 2022 at 12:54:56PM +0100, Dan Hatton wrote:
> 
> I have a particular process, embodying some numerical calculations in
> Octave, on which I'd like to perform a CRIU dump.  Having launched the
> process (as normal user my_user_name) with
> 
> (setsid nice -n +19 octave my_octave_script.m) >& log.log &
> 
> and found out its PID, PID_x, I run (as root)
> 
> criu dump -t PID_x -D ~my_user_name/a_dir/dump
> 
> where the directory ~my_user_name/a_dir/dump already exists and is
> owned by my_user_name.
> 
> The dump attempt consistently fails with an "External socket is used"
> error.
> 
> If I crank up the verbosity, what happens just before the failure is
> 
> (02.636547) unix: Dumping external sockets
> (02.636657) unix:       Dumping extern: ino 631338 peer_ino 632188 family
> 1 type    1 state  1 name (02.636787) unix:       Dumped extern: id 0x2a1
> ino 631338 peer 0 type 2 state 10 name 21 bytes
> (02.636873) unix:       Runaway socket: ino 631338 peer_ino 632188 family    1 type    1 state  1 name
> 
> If I then try (as root)
> 
> find / -inum 631338
> find / -inum 632188
> 
> Nothing is found.  I'm left with no idea what this socket is or
> whether it's of the type for which it's safe to use the --external
> option to criu.  What should be my next step, please?
> 
> --
> 
> Kind regards,
> 
> Daniel