[igt-dev] [PATCH i-g-t] tests: Open DRM_MASTER for display operations

[Chris Wilson]

Quoting Petri Latvala (2019-12-18 11:49:40)
> On Wed, Dec 18, 2019 at 11:38:22AM +0000, Chris Wilson wrote:
> > Quoting Petri Latvala (2019-12-18 11:36:02)
> > > On Wed, Dec 18, 2019 at 11:20:12AM +0000, Chris Wilson wrote:
> > > > To configure the display, one is required to be the DRM_MASTER.
> > > > 
> > > > Closes: https://gitlab.freedesktop.org/drm/intel/issues/855
> > > > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > > > ---
> > > >  tests/kms_content_protection.c | 2 +-
> > > >  tests/kms_panel_fitting.c      | 2 +-
> > > >  tests/kms_plane_alpha_blend.c  | 2 +-
> > > >  3 files changed, 3 insertions(+), 3 deletions(-)
> > > > 
> > > > diff --git a/tests/kms_content_protection.c b/tests/kms_content_protection.c
> > > > index 8b361fdef..3b9cedcb4 100644
> > > > --- a/tests/kms_content_protection.c
> > > > +++ b/tests/kms_content_protection.c
> > > > @@ -654,7 +654,7 @@ static void test_content_protection_cleanup(void)
> > > >  igt_main
> > > >  {
> > > >       igt_fixture {
> > > > -             data.drm_fd = drm_open_driver(DRIVER_ANY);
> > > > +             data.drm_fd = drm_open_driver_master(DRIVER_ANY);
> > > >  
> > > >               igt_display_require(&data.display, data.drm_fd);
> > > >       }
> > > > diff --git a/tests/kms_panel_fitting.c b/tests/kms_panel_fitting.c
> > > > index ac24730f8..065fc2df8 100644
> > > > --- a/tests/kms_panel_fitting.c
> > > > +++ b/tests/kms_panel_fitting.c
> > > > @@ -252,7 +252,7 @@ igt_main
> > > >       data_t data = {};
> > > >  
> > > >       igt_fixture {
> > > > -             data.drm_fd = drm_open_driver(DRIVER_ANY);
> > > > +             data.drm_fd = drm_open_driver_master(DRIVER_ANY);
> > > >               igt_display_require(&data.display, data.drm_fd);
> > > >               igt_display_require_output(&data.display);
> > > >       }
> > > > diff --git a/tests/kms_plane_alpha_blend.c b/tests/kms_plane_alpha_blend.c
> > > > index d7dc44ab9..085099f81 100644
> > > > --- a/tests/kms_plane_alpha_blend.c
> > > > +++ b/tests/kms_plane_alpha_blend.c
> > > > @@ -565,7 +565,7 @@ igt_main
> > > >       enum pipe pipe;
> > > >  
> > > >       igt_fixture {
> > > > -             data.gfx_fd = drm_open_driver(DRIVER_ANY);
> > > > +             data.gfx_fd = drm_open_driver_master(DRIVER_ANY);
> > > >               igt_require_pipe_crc(data.gfx_fd);
> > > >               igt_display_require(&data.display, data.gfx_fd);
> > > >               igt_require(data.display.is_atomic);
> > > 
> > > 
> > > Don't they implicitly become masters if that's the first-and-only
> > > open() on the dev? Did that change?
> > 
> > That is the still the case. They aren't the only things running :)
> > 
> > By requesting master up front, we will get a complaint about what is
> > still running that prevents us acquiring master, rather than a belated
> > EACCES.
> 
> I see. Then claiming that this closes drm/intel#855 is a bit of a
> stretch.

It closes the EACESS. That stuff is still running is that not this
test's fault (not requiring the permission it needs is) and already
known, most recently #844.
 
> Getting the error happening earlier is good. However, surely these
> aren't the only tests that do master-requiring things (modeset? or any
> kind of commit?) on an fd opened with drm_open_driver()? Are there any
> tests that do igt_display_require and don't require master-y access?

I did a git grep and these 3 were the one's that I could identify that
did not request the privilege they required.
 
> Is checking for master in igt_display_require overkill?

Maybe, or maybe you want to allow readonly access and only require there
be display HW :)
-Chris