[igt-dev] [PATCH] tests: Add a new test for driver/device hot reload

Janusz Krzysztofik janusz.krzysztofik at linux.intel.com
Tue Apr 9 09:23:52 UTC 2019


On Tuesday, April 9, 2019 11:12:13 AM CEST Chris Wilson wrote:
> Quoting Janusz Krzysztofik (2019-04-09 09:54:16)
> 
> > Hi Chris,
> > 
> > On Friday, April 5, 2019 4:11:48 PM CEST Chris Wilson wrote:
> > > As you note, drm_open_driver() will autoreload the module. So will the
> > > next test.
> > 
> > Yes, but speaking of that, I'm still not sure if running a next test after
> > this one should be recommended.  I've been observing issues in other tests
> > which exhibited themselves after successful module reload.
> 
> In other words, this test is not complete yet! :) Realistically though,
> that is likely to be the memory corruption caused by the unclean module
> unload. You can say the same about any igt, they all have the danger
> that they may leave the system in an unstable state.

That's what I had on mind.  You don't have to use a test for module reloading, 
doing that manually was sufficient for me to observe issues under other tests.

Janusz

> -Chris






More information about the igt-dev mailing list