Re: [PATCH v4 0/5] Add pidfd support to the fanotify API

[Matthew Bobrowski <repnop@google.com>]

Hey Jan,

On Wed, Aug 11, 2021 at 11:22:41AM +1000, Matthew Bobrowski wrote:
> On Tue, Aug 10, 2021 at 01:33:48PM +0200, Jan Kara wrote:
> > Hello Matthew!
> > 
> > On Sun 08-08-21 15:23:59, Matthew Bobrowski wrote:
> > > This is V5 of the FAN_REPORT_PIDFD patch series. It contains the minor
> > > comment/commit description fixes that were picked up by Amir in the
> > > last series review [0, 1].
> > > 
> > > LTP tests for this API change can be found here [2]. Man page updates
> > > for this change can be found here [3].
> > > 
> > > [0] https://lore.kernel.org/linux-fsdevel/CAOQ4uxhnCk+FXK_e_GA=jC_0HWO+3ZdwHSi=zCa2Kpb0NDxBSg@mail.gmail.com/
> > > [1] https://lore.kernel.org/linux-fsdevel/CAOQ4uxgO3oViTSFZ0zs6brrHrmw362r1C9SQ7g6=XgRwyrzMuw@mail.gmail.com/
> > > [2] https://github.com/matthewbobrowski/ltp/tree/fanotify_pidfd_v2
> > > [3] https://github.com/matthewbobrowski/man-pages/tree/fanotify_pidfd_v1
> > > 
> > > Matthew Bobrowski (5):
> > >   kernel/pid.c: remove static qualifier from pidfd_create()
> > >   kernel/pid.c: implement additional checks upon pidfd_create()
> > >     parameters
> > >   fanotify: minor cosmetic adjustments to fid labels
> > >   fanotify: introduce a generic info record copying helper
> > >   fanotify: add pidfd support to the fanotify API
> > 
> > Thanks! I've pulled the series into my tree. Note that your fanotify21 LTP
> > testcase is broken with the current kernel because 'ino' entry got added to
> > fdinfo. I think having to understand all possible keys that can occur in
> > fdinfo is too fragile. I understand why you want to do that but I guess the
> > test would be too faulty to be practical. So I'd just ignore unknown keys
> > in fdinfo for that test.
> 
> Excellent, for merging these changes!
> 
> In regards to the LTP test (fanotify21), at the time of writing I had also
> shared a similar thought in the sense that it was too fragile, but wrongly
> so I weighed up my decision based on the likelihood and frequency of fields
> being changed/added to fdinfo. I was very wrong...
> 
> Anyway, I will fix it so that any "unknown" fields are ignored.

FWIW, I've dropped that last else statement in the
parse_pidfd_fdinfo_line() helper in LTP fanotify21. An updated branch has
been pushed here [0].

[0] https://github.com/matthewbobrowski/ltp/commits/fanotify_pidfd_v3

/M