[igt-dev] [PATCH i-g-t] tests/device_reset: remove artificial timeout

[Zbigniew Kempczyński]

On Mon, Nov 08, 2021 at 05:27:46PM +0100, Janusz Krzysztofik wrote:
> On Monday, 8 November 2021 14:47:46 CET Marcin Bernatowicz wrote:
> > Do not use artificial timeout. Let the test fail or finish.
> > 
> > Signed-off-by: Marcin Bernatowicz <marcin.bernatowicz at linux.intel.com>
> > ---
> >  tests/device_reset.c | 3 ---
> >  1 file changed, 3 deletions(-)
> > 
> > diff --git a/tests/device_reset.c b/tests/device_reset.c
> > index 982ba5ef2a..e60d4c7fde 100644
> > --- a/tests/device_reset.c
> > +++ b/tests/device_reset.c
> > @@ -301,8 +301,6 @@ igt_main
> >  		set_device_filter(dev_path);
> >  
> >  		igt_skip_on(!is_sysfs_reset_supported(dev.fds.dev));
> > -
> > -		igt_set_timeout(60, "device reset tests timed out after 
> 60s");
> >  	}
> >  
> >  	igt_describe("Unbinds driver from device, initiates reset"
> > @@ -319,7 +317,6 @@ igt_main
> >  	}
> >  
> >  	igt_fixture {
> > -		igt_reset_timeout();
> >  		cleanup_device_fds(&dev);
> >  	}
> >  }
> > 
> 
> I think it could still make sense from the CI point of view to time out if 
> some uncommon steps preceding the scope of the exercise, e.g., driver unbind 
> before the reset operation itself, would take too much time, but anyway, 
> interrupting the exercised reset operation in the middle and not collecting as 
> much of driver output as possible if the driver response is slow doesn't sound 
> like a good idea, then
> 
> Acked-by: Janusz Krzysztofik <janusz.krzysztofik at linux.intel.com>
> 

What is my concern is that test is not only for i915, see:

dev->fds.dev = __drm_open_driver(DRIVER_ANY);

Are we sure we won't break anything?

--
Zbigniew

> 
>