[CRIU] [PATCH] fsmptify: Filter out internal inotify bits when restoring marks

[Andrew Vagin]

On Tue, Sep 22, 2015 at 04:36:28PM +0300, Cyrill Gorcunov wrote:
> On Tue, Sep 22, 2015 at 04:15:43PM +0300, Andrew Vagin wrote:
> > > > 
> > > > I think we need only exclude FS_EVENT_ON_CHILD. Otherwise we will skip a
> > > > new flags, when they will add into the kernel.
> > > 
> > > New flags are gonna be defined in uapi/ as well so criu is depending on which
> > 
> > If you will exclude only FS_EVENT_ON_CHILD, the code will work for all
> > kernel. The current version depends on IN_ALL_EVENTS, which can be
> > changed in the next kernel.
> 
> You miss the point. It was a bug in the kernel that it has been exporting
> internal bit into userspace. The kernel has _all_ rights to
> 
> #define IN_SOME_NEW_BIT	 	0x08000000

Now I'm agree that my idea isn't good, but I don't like to use
IN_ALL_EVENTS too :-P
> 
> and export it into userspace as a valid one which would serve some
> new purpose and meaning. So we simply can't rip it off. That's
> why I propose to mask it with IN_ALL_EVENTS which is a part of
> userspace.
> 
> Another option -- don't modify criu at all but rip off the
> FS_EVENT_ON_CHILD in the test itself? As far as I understand
> the main problem is the tests are failing since the kernel
> masks the agrumens passing in with IN_ALL_EVENTS internally.

Tests will not fail, because they do dump and restore on the same
kernel.