[systemd-devel] [PATCH V3] path-util.c: fix path_is_mount_point() for symlinks

[Lennart Poettering]

On Tue, 10.03.15 17:26, Lennart Poettering (lennart at poettering.net) wrote:

> On Fri, 20.02.15 13:25, harald at redhat.com (harald at redhat.com) wrote:
> 
> Sorry for the late review!
> 
> > --- a/src/shared/path-util.c
> > +++ b/src/shared/path-util.c
> > @@ -456,9 +456,9 @@ int path_is_mount_point(const char *t, bool allow_symlink) {
> >  
> >          union file_handle_union h = FILE_HANDLE_INIT;
> >          int mount_id = -1, mount_id_parent = -1;
> > -        _cleanup_free_ char *parent = NULL;
> >          struct stat a, b;
> >          int r;
> > +        _cleanup_close_ int fd = -1;
> >          bool nosupp = false;
> >  
> >          /* We are not actually interested in the file handles, but
> > @@ -468,7 +468,15 @@ int path_is_mount_point(const char *t, bool allow_symlink) {
> >          if (path_equal(t, "/"))
> >                  return 1;
> >  
> > -        r = name_to_handle_at(AT_FDCWD, t, &h.handle, &mount_id, allow_symlink ? AT_SYMLINK_FOLLOW : 0);
> > +        fd = openat(AT_FDCWD, t, O_RDONLY|O_NONBLOCK|O_DIRECTORY|O_CLOEXEC|(allow_symlink ? 0 : O_PATH));
> > +        if (fd < 0) {
> > +                if (errno == ENOENT)
> > +                        return 0;
> 
> We eat up this error? Why not pass the -ENOENT up?

Hmm, as I see know the existing code also eats up ENOENT... Not
entirely sure why though.

But patch looks good then, please commit!

Lennart

-- 
Lennart Poettering, Red Hat