Re: [PATCH v4 linux-kselftest-test 5/6] kunit: allow kunit to be loaded as a module

[Alan Maguire]

On Tue, 19 Nov 2019, Stephen Boyd wrote:

> Quoting Alan Maguire (2019-11-15 02:16:11)
> > diff --git a/lib/kunit/test.c b/lib/kunit/test.c
> > index 87b5cf1..41ef71a 100644
> > --- a/lib/kunit/test.c
> > +++ b/lib/kunit/test.c
> > @@ -486,3 +486,16 @@ void kunit_cleanup(struct kunit *test)
> >         }
> >  }
> >  EXPORT_SYMBOL_GPL(kunit_cleanup);
> > +
> > +static int kunit_init(void)
> 
> Missing __init?
>

Oops, will fix.
 
> > +{
> > +       return 0;
> > +}
> > +late_initcall(kunit_init);
> 
> It looks pretty weird that this doesn't do anything in the module init
> or exit path. How does it work? And why does it need to be late init if
> nothing is called from here?
>

I used a late_initcall() here because if there is initialization code in 
the future (say setting up debugfs, sysctl tables etc), it will get run 
irrespective of whether kunit is built as a module or not.  For the module
case, we need init/exit to ensure the module can be unloaded 
successfully.

> > +
> > +static void __exit kunit_exit(void)
> > +{
> > +}
> > +module_exit(kunit_exit);
> > +
> > +MODULE_LICENSE("GPL");
> 
> I guess should be "GPL v2"?
> 

Good catch, will fix here and elsewhere, thanks!

BTW in replying to your review of patch 1 I made a mistake; we do in fact 
need to add the "#include <linux/kernel.h>" in assert.h as we are removing 
the "#include <kunit/string-stream.h>".  The latter indirectly includes 
linux/kernel.h which allows us to find definition for "struct va_format"
used in the struct kunit_assert definition. Apologies about this.

I'll have a go at generating v5 using -C as requested. Thanks again for 
the careful review!

Alan