[lxc-users] CRIU with lxc.network.type empty fails
Tycho Andersen
tycho.andersen at canonical.com
Thu Aug 13 14:55:58 UTC 2015
Hi Dietmar,
On Thu, Aug 13, 2015 at 07:40:27AM +0200, Dietmar Maurer wrote:
> > > Just found this info: http://permalink.gmane.org/gmane.linux.kernel/1930352
> > >
> > > "Now when mounting /sys/kernel/debug, tracefs is automatically mounted
> > > in /sys/kernel/debug/tracing"
> > >
> > > Would that explain the behavior?
> >
> > Yep, I think that's it. Does the patch I sent work for you?
>
> Sorry for the delay, but it seems we are in different time zones.
No problem.
> I just tested your patch, and it seems to solve the tracefs
> problem - thanks.
>
> Now I no get another error:
>
> (00.005871) irmap: Refresh stat for /etc/group-
> (00.005873) irmap: Refresh stat for /etc/init
> (00.005875) irmap: Scanned /etc/init
> (00.005876) fsnotify: Dumping /etc/init as path for handle
> (00.005879) fsnotify: id 0x00000005 flags 0x00080800
> (00.005882) fdinfo: type: 0x 8 flags: 02004000/01 pos: 0x 0 fd: 6
> (00.005891) 13939 fdinfo 7: pos: 0x 0 flags: 2004002/0x1
> (00.005896) Searching for socket b17c7f (family 1)
> (00.005900) No filter for socket
> (00.005901) sk unix: Dumping unix socket at 7
> (00.005902) sk unix: Dumping: ino 0xb17c7f peer_ino 0 family 1 type 1
> state 10 name
> (00.005913) sk unix: Dumped: id 0x6 ino 0xb17c7f peer 0 type 1 state 10 name 20
> bytes
> (00.005916) fdinfo: type: 0x 5 flags: 02004002/01 pos: 0x 0 fd: 7
> (00.005924) 13939 fdinfo 10: pos: 0x 0 flags: 100002/0
> (00.005927) Error (files-ext.c:91): Can't dump file 10 of that type [60660]
> (unknown (null))
Looks like this is some sort of fd that has a block device open that
criu doesn't support. If you can figure out which block device it is
(or tell me how to reproduce), perhaps we can teach criu easily to
checkpoint it.
Also, thanks for your other mail for a reproducer. I don't have time
the next to days to look at it, but I'll play around with it next
week.
Tycho
> (00.005933) ----------------------------------------
>
More information about the lxc-users
mailing list