[LTP] [PATCH 4/5] syscalls/fanotify03: add test for FAN_MARK_FILESYSTEM permission events

[Amir Goldstein]

On Wed, Nov 21, 2018 at 6:36 PM Amir Goldstein <amir73il@gmail.com> wrote:
>
> On Wed, Nov 21, 2018 at 3:16 PM Cyril Hrubis <chrubis@suse.cz> wrote:
> >
> > Hi!
> > > -static void setup_mark(unsigned int n)
> > > +static int setup_mark(unsigned int n)
> > >  {
> > >       struct tcase *tc = &tcases[n];
> > >       struct fanotify_mark_type *mark = &tc->mark;
> > > @@ -144,7 +149,12 @@ static void setup_mark(unsigned int n)
> > >       if (fanotify_mark(fd_notify, FAN_MARK_ADD | mark->flag,
> > >                         FAN_ACCESS_PERM | FAN_OPEN_PERM,
> > >                         AT_FDCWD, fname) < 0) {
> > > -             if (errno == EINVAL) {
> > > +             if (errno == EINVAL && support_perm_events &&
> > > +                 mark->flag == FAN_MARK_FILESYSTEM) {
> > > +                     tst_res(TCONF,
> > > +                             "FAN_MARK_FILESYSTEM not supported in kernel?");
> > > +                     return -1;
> > > +             } else if (errno == EINVAL) {
> > >                       tst_brk(TCONF | TERRNO,
> > >                               "CONFIG_FANOTIFY_ACCESS_PERMISSIONS not "
> > >                               "configured in kernel?");
> > > @@ -155,9 +165,16 @@ static void setup_mark(unsigned int n)
> > >                               "AT_FDCWD, %s) failed.",
> > >                               fd_notify, mark->name, fname);
> > >               }
> > > +     } else {
> > > +             /*
> > > +              * To distigouish between perm event not supported and
> > > +              * filesystem mark not supported.
> > > +              */
> > > +             support_perm_events = 1;
> >
> > I'm a bit puzzled here, so we attempted to cache if perm_events are
> > supported here?
> >
>
> Yes.
>
> > I guess that we depend on the order of the tcases[] array here, which is
> > not very nice.
> >
>
> Indeed. It is a simplification. All test cases are permission events,
> so the order of failure types is not likely to change, but I could add a
> comment about the order near test cases definition.
> Would you rather that I did an explicit test for permission support at setup()
> time?
>

Nevermind. this discussion is moot because no permission events support
breaks out of the test. I'll drop this logic altogether.

It will have to make a comeback though with Mark's patches to add tests for
FAN_OPEN_EXEC_PERM. (need to distinguish between to support for
FAN_MARK_FILESYSTEM and no support for FAN_OPEN_EXEC_PERM)
so please state your preference for the best way to handle this issue.

Thanks,
Amir.