Re: [PATCH v3] drm/xe: Use fault injection infrastructure to find issues at probe time

[Rodrigo Vivi <rodrigo.vivi@intel.com>]

On Thu, Sep 26, 2024 at 04:32:08PM +0300, Jani Nikula wrote:
> On Thu, 26 Sep 2024, Lucas De Marchi <lucas.demarchi@intel.com> wrote:
> > On Thu, Sep 26, 2024 at 02:57:01PM GMT, Jani Nikula wrote:
> >>On Thu, 26 Sep 2024, Francois Dugast <francois.dugast@intel.com> wrote:
> >>> On Thu, Sep 26, 2024 at 12:43:59PM +0300, Jani Nikula wrote:
> >>>> On Wed, 25 Sep 2024, Francois Dugast <francois.dugast@intel.com> wrote:
> >>>> > +/*
> >>>> > + * The ALLOW_ERROR_INJECTION() macro is added to conditionally skip execution at
> >>>> > + * runtime and use a provided return value, in order to test errors paths in the
> >>>> > + * callers. The requirements for the error injectable functions are not strictly
> >>>> > + * fullfilled but this is acceptable because the caller only propagates the error
> >>>> > + * up the stack without cleanup of resources potentially allocated here.
> >>>> > + */
> >>>>
> >>>> I'm curious on the details of "The requirements for the error injectable
> >>>> functions are not strictly fullfilled". It's repeated many times, but
> >>>> not explained. Maybe I'd like the info spoon fed to me instead of having
> >>>> to figure it out for myself. ;)
> >>>
> >>> Understandable! I will make it more explicit in the next revision. Any
> >>> suggestion to avoid the duplication?
> >>
> >>All I can think of is adding a single, more thorough explanation comment
> >>about the approach to error injection somewhere suitable (*), and then
> >>have short comments referencing that.
> >>
> >>	/* See xxx for details on error injection. */
> >
> > https://docs.kernel.org/fault-injection/fault-injection.html#requirements-for-the-error-injectable-functions
> >
> > so... like this?
> >
> > 	/*
> > 	 * See "Requirements for the Error Injectable Functions" in
> > 	 * Documentation/fault-injection/fault-injection.rst
> > 	 */
> 
> All I wanted to know was what "not strictly fullfilled" means in "The
> requirements for the error injectable functions are not strictly
> fullfilled". What parts are we violating? What's the impact?

Yes, I believe that each of the new inject functions should have
individual comments explaining why exactly that case violates the
overal rule.

> 
> BR,
> Jani.
> 
> 
> -- 
> Jani Nikula, Intel