Re: [PATCH 2/2 v2] drm/atomic: clear new_state pointers at hw_done

[Rob Clark]

On Tue, Nov 12, 2019 at 7:51 AM Enric Balletbo Serra
<eballetbo@xxxxxxxxx> wrote:
>
> Hi Rob,
>
> Missatge de Rob Clark <robdclark@xxxxxxxxx> del dia dl., 4 de nov.
> 2019 a les 18:42:
> >
> > From: Rob Clark <robdclark@xxxxxxxxxxxx>
> >
> > The new state should not be accessed after this point.  Clear the
> > pointers to make that explicit.
> >
> > Signed-off-by: Rob Clark <robdclark@xxxxxxxxxxxx>
>
> While looking to another issue I applied this patch on top of 5.4-rc7
> and my display stopped working. The system gets stuck with the
> messages below
>
> ...
> [   17.558689] rockchip-vop ff8f0000.vop: Adding to iommu group 1
> [   17.566014] rk3399-gru-sound sound: ASoC: failed to init link DP: -517
> [   17.567618] rockchip-vop ff900000.vop: Adding to iommu group 2
> [   17.580671] rk3399-gru-sound sound: ASoC: failed to init link DP: -517
> [   17.585996] rockchip-drm display-subsystem: bound ff8f0000.vop (ops
> vop_component_ops [rockchipdrm])
> [   17.589294] rk3399-gru-sound sound: ASoC: failed to init link DP: -517
> [   17.599899] rockchip-drm display-subsystem: bound ff900000.vop (ops
> vop_component_ops [rockchipdrm])
> [   17.615846] rockchip-dp ff970000.edp: no DP phy configured
> [   17.622495] rockchip-drm display-subsystem: bound ff970000.edp (ops
> rockchip_dp_component_ops [rockchipdrm])
> [   17.633688] rockchip-drm display-subsystem: bound fec00000.dp (ops
> cdn_dp_component_ops [rockchipdrm])
> [   17.644141] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
> [   17.651548] [drm] No driver support for vblank timestamp query.
>
> Not really useful information at this point, but I am wondering if
> could be that the rockchip driver is doing something wrong more than
> this patch is wrong?

I think we should drop this patch (but 1/2 is defn needed).. it turns
up some other problems in qemu, it seems.  So there is kind of a
bigger problem of things accessing new state after hw_done still.  But
I've not really had time to look into it.

BR,
-R

>
> Thanks,
>  Enric
>
> > ---
> >  drivers/gpu/drm/drm_atomic_helper.c | 30 +++++++++++++++++++++++++++++
> >  1 file changed, 30 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
> > index 648494c813e5..aec9759d9df2 100644
> > --- a/drivers/gpu/drm/drm_atomic_helper.c
> > +++ b/drivers/gpu/drm/drm_atomic_helper.c
> > @@ -2246,12 +2246,42 @@ EXPORT_SYMBOL(drm_atomic_helper_fake_vblank);
> >   */
> >  void drm_atomic_helper_commit_hw_done(struct drm_atomic_state *old_state)
> >  {
> > +       struct drm_connector *connector;
> > +       struct drm_connector_state *old_conn_state, *new_conn_state;
> >         struct drm_crtc *crtc;
> >         struct drm_crtc_state *old_crtc_state, *new_crtc_state;
> > +       struct drm_plane *plane;
> > +       struct drm_plane_state *old_plane_state, *new_plane_state;
> >         struct drm_crtc_commit *commit;
> > +       struct drm_private_obj *obj;
> > +       struct drm_private_state *old_obj_state, *new_obj_state;
> >         int i;
> >
> > +       /*
> > +        * After this point, drivers should not access the permanent modeset
> > +        * state, so we also clear the new_state pointers to make this
> > +        * restriction explicit.
> > +        *
> > +        * For the CRTC state, we do this in the same loop where we signal
> > +        * hw_done, since we still need to new_crtc_state to fish out the
> > +        * commit.
> > +        */
> > +
> > +       for_each_oldnew_connector_in_state(old_state, connector, old_conn_state, new_conn_state, i) {
> > +               old_state->connectors[i].new_state = NULL;
> > +       }
> > +
> > +       for_each_oldnew_plane_in_state(old_state, plane, old_plane_state, new_plane_state, i) {
> > +               old_state->planes[i].new_state = NULL;
> > +       }
> > +
> > +       for_each_oldnew_private_obj_in_state(old_state, obj, old_obj_state, new_obj_state, i) {
> > +               old_state->private_objs[i].new_state = NULL;
> > +       }
> > +
> >         for_each_oldnew_crtc_in_state(old_state, crtc, old_crtc_state, new_crtc_state, i) {
> > +               old_state->crtcs[i].new_state = NULL;
> > +
> >                 commit = new_crtc_state->commit;
> >                 if (!commit)
> >                         continue;
> > --
> > 2.23.0
> >
> > _______________________________________________
> > dri-devel mailing list
> > dri-devel@xxxxxxxxxxxxxxxxxxxxx
> > https://lists.freedesktop.org/mailman/listinfo/dri-devel