[CRIU] [PATCH 1/2] util: add a function to check output data in a file descriptor

[Cyrill Gorcunov]

On Tue, Sep 02, 2014 at 01:26:22AM +0400, Andrey Wagin wrote:
> 2014-09-02 1:21 GMT+04:00 Pavel Emelyanov <xemul at parallels.com>:
> > On 09/02/2014 01:15 AM, Andrew Vagin wrote:
> >> On Tue, Sep 02, 2014 at 12:40:39AM +0400, Cyrill Gorcunov wrote:
> >>> On Tue, Sep 02, 2014 at 12:25:13AM +0400, Andrey Vagin wrote:
> >>>> We can't dump netlink socket, inotify, fanotify, if they have queued
> >>>> data, so lets add a function to chech this.
> >>>>
> >>>> Signed-off-by: Andrey Vagin <avagin at openvz.org>
> >>>
> >>> We were dumping such fsnotify objects before and now we're refusing,
> >>> you propose to patch the kernel? I mean, how to deal with it?
> >>
> >> CRIU should not affect the application logic. If we want to ignore
> >> events, we need ask permission for this from a user.
> >
> > +1
> >
> > I'd then also check the restore -- if we restore an inotify
> > on deleted file with empty events queue, we should make sure
> > criu doesn't "generate" new ones when creating a file, putting
> > an inotify on it _and_ unlinking the file again :)
> 
> CRIU generates events for such files. Look at my patch for static/inotify00.

I'll take a look.