Re: [PATCH v3] selftests/livepatch: introduce tests

[Miroslav Benes]

On Fri, 13 Apr 2018, Joe Lawrence wrote:

> On 04/13/2018 07:20 AM, Miroslav Benes wrote:
> > Hi,
> > 
> > On Thu, 12 Apr 2018, Joe Lawrence wrote:
> > 
> >> Add a few livepatch modules and simple target modules that the included
> >> regression suite can run tests against.
> > 
> > Could you include a brief description which features are tested?
> 
> I can add this to the commit msg:
> 
>       - basic livepatching (multiple patches, atomic replace)
>       - pre/post (un)patch callbacks
>       - shadow variable API
> 
> Or do you prefer a little more detail?

That would suffice. Thanks.
 
> >  
> >> Signed-off-by: Joe Lawrence <joe.lawrence@xxxxxxxxxx>
> >> ---
> > 
> >> diff --git a/lib/livepatch/test_klp_shadow_vars.c b/lib/livepatch/test_klp_shadow_vars.c
> >> new file mode 100644
> >> index 000000000000..18c75b21cb9e
> >> --- /dev/null
> >> +++ b/lib/livepatch/test_klp_shadow_vars.c
> >>
> >> +/*
> >> + * Shadow variable wrapper functions that echo the function and arguments
> >> + * to the kernel log for testing verification.  Don't display raw pointers,
> >> + * but use the ptr_id() value instead.
> >> + */
> >> +void *shadow_get(void *obj, unsigned long id)
> >> +{
> >> +	void *ret = klp_shadow_get(obj, id);
> >> +
> >> +	pr_info("klp_%s(obj=PTR%d, id=0x%lx) = PTR%d\n",
> >> +		__func__, ptr_id(obj), id, ptr_id(ret));
> >> +
> >> +	return ret;
> >> +}
> > 
> >> +void shadow_free(void *obj, unsigned long id, klp_shadow_dtor_t dtor)
> >> +{
> >> +	klp_shadow_free(obj, id, dtor);
> >> +	pr_info("klp_%s(obj=PTR%d, id=0x%lx, dtor=PTR%d)\n",
> >> +		__func__, ptr_id(obj), id, ptr_id(dtor));
> >> +}
> > 
> > Sparse (make C=1) would be happier with those two being static.
> 
> Ah right. I wonder why the kbuild test robot didn't complain about
> those, too.  Easy enough to fix up, thanks.
> 
> > Otherwise it works as expected. Good job!
> 
> Thanks for reviewing.  I'll hold off on posting v4 until Petr (and
> others) get a chance to comment.  Perhaps there are other tests that
> would be helpful?

I think it would be useful to have tests for a stack checking and a 
consistency. Nicolai has written some lately for our internal testing, but 
it would take some time to transform them appropriately, I think.

So I would start with this and we can add more later. I would also 
establish a (soft) rule to submit selftests with new features.

Miroslav