[RFC PATCH 1/1] drm/xe: Add driver load error injection

[Jani Nikula]

On Sat, 10 Aug 2024, Matthew Brost <matthew.brost at intel.com> wrote:
> On Sat, Aug 10, 2024 at 12:16:32AM -0500, Lucas De Marchi wrote:
>> On Fri, Aug 09, 2024 at 03:44:24PM GMT, Matthew Brost wrote:
>> > Port over i915 driver load error injection.
>> > 
>> 
>> I don't like much the manual approach, but it's better to get the driver
>> not exploding. Then we can think of replacing this.  Some comments below
>
> Yep. I chatted with Rodrigo about this and we agreed their isn't a great
> way with the existing kernel error injection to easily get coverage like
> this plus a very simple test case [1]. Agree longterm we should not
> invent our own things and come up with either a kernel or drm level
> solution.
>
> In the short term, yes this better than our driver exploding. View this
> as a force probe blocker, so we need to get our driver fixed in a matter
> of weeks and this seems like the only viable path for now.
>
> [1] for i in {1..N}; do echo "Run $i"; modprobe xe inject_driver_load_error=$i; rmmod xe; done

*sad trombone*

It just pains me that we keep copy-pasting stuff from i915, especially
when it's the hacky less than stellar parts. Like this one.

Fixing this needs to go to some tracker somewhere, and get it assigned,
otherwise later means never.


BR,
Jani.


-- 
Jani Nikula, Intel