[CRIU] [PATCH 00/11] tty: Add c/r for system console

[Cyrill Gorcunov]

On Tue, Sep 30, 2014 at 04:25:47PM -0500, Tycho Andersen wrote:
> > > 
> > > What I meant by the above is that this patchset actually fixes that
> > > error for me (somehow, not sure which patch does it :).
> > 
> > Actually this patchset only adds support of /dev/console, so in your
> > report we've had pty pair in first dump, then after restore (for some
> > unknown reason) instead of ptys we got /dev/console reference and that's
> > the question who is changed it, I suspect it might be systemd or something
> > like that.
> 
> Yes, I just saw it again today with this patchset applied, so it looks
> like I spoke too soon :(

I see, anyway, good to know!

> > 
> > Looks like this are external moutns which might require --ext-mount-map keys, how this mountinfo
> > represented on pid 1 of container (ie do "ps afx" on node and the choose pid for init of container,
> > then show /proc/$pid/mountinfo)
> 
> It looks like the same sort of confusing thing to me:

...
so they are bind mounts, that's fine

> > > > > I hit:
> > > > > 
> > > > > (00.247031) Error (tty.c:233): tty: Unexpected format on path /dev/lxc/tty4
> > > > > 
> > > > > which looks like it might be a bigger problem :)
> > 
> > Aha! We're expecting ptys to have a canonical forms, ie /dev/pts/N, where N is
> > pty number, need to thhink how to workaround it (you can fix up criu by now
> > changing parse_index() helper to parse not PTS_FMT but say "/dev/lxc/tty%d"
> 
> Yes, I was trying to think of a nice workaround too. I guess a hack
> would be to just parse an int at the end of the string, I'm not sure
> if there is a better way.

I'll think about this and ping you.

> > 
> > Thanks, this is indeed unix98 pty device, so need to fix this
> > problem somehow...
> 
> Ok, let me know if I can help.

Sure, thanks!