[Intel-gfx] [PATCH] drm/atomic-helper: Fix leak in disable_all

[Daniel Vetter]

On Mon, Jul 17, 2017 at 11:39:56AM +0200, Maarten Lankhorst wrote:
> Op 15-07-17 om 11:31 schreef Daniel Vetter:
> > The legacy plane->fb pointer is refcounted by calling
> > drm_atomic_clean_old_fb().
> >
> > In practice this isn't a real problem because:
> > - The caller in the i915 gpu reset code restores the original state
> >   again, which means the plane->fb pointer won't change, hence can't
> >   leak.
> > - Drivers using drm_atomic_helper_shutdown call the fbdev cleanup
> >   first, and that usually cleans up the fb through
> >   drm_remove_framebuffer, which does this correctly.
> > - Without fbdev the only framebuffers are from userspace, and those
> >   get cleaned up (again using drm_remove_framebuffer) befor the driver
> >   can even be unloaded.
> >
> > But in i915 I've switched the cleanup sequence around so that the
> > _shutdown() calls happens after the drm_remove_framebuffer(), which is
> > how I discovered this issue.
> >
> > v2: My analysis why the current code was ok for gpu reset and
> > suspend/resume was correct, but then I totally failed to realize that
> > we better keep this symmetric. Thanksfully CI noticed that for
> > balance, a refcounting bug must exist at 2 places if previously there
> > was no issue ...
> >
> > v3: Don't be lazy and compute the plane_mask in
> > commit_duplicated_state properly too, instead of just using ~0U.
> >
> > Cc: martin.peres at free.fr
> > Cc: chris at chris-wilson.co.uk
> > Acked-by: Dave Airlie <airlied at gmail.com> (v1)
> > Signed-off-by: Daniel Vetter <daniel.vetter at intel.com>
> > ---
> >  drivers/gpu/drm/drm_atomic_helper.c | 18 ++++++++++++++++--
> >  1 file changed, 16 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
> > index b07fc30372d3..545328a9a769 100644
> > --- a/drivers/gpu/drm/drm_atomic_helper.c
> > +++ b/drivers/gpu/drm/drm_atomic_helper.c
> > @@ -2726,6 +2726,7 @@ int drm_atomic_helper_disable_all(struct drm_device *dev,
> >  	struct drm_plane *plane;
> >  	struct drm_crtc_state *crtc_state;
> >  	struct drm_crtc *crtc;
> > +	unsigned plane_mask = 0;
> >  	int ret, i;
> >  
> >  	state = drm_atomic_state_alloc(dev);
> > @@ -2768,10 +2769,14 @@ int drm_atomic_helper_disable_all(struct drm_device *dev,
> >  			goto free;
> >  
> >  		drm_atomic_set_fb_for_plane(plane_state, NULL);
> > +		plane_mask |= BIT(drm_plane_index(plane));
> > +		plane->old_fb = plane->fb;
> >  	}
> >  
> >  	ret = drm_atomic_commit(state);
> >  free:
> > +	if (plane_mask)
> > +		drm_atomic_clean_old_fb(dev, plane_mask, ret);
> >  	drm_atomic_state_put(state);
> >  	return ret;
> >  }
> > @@ -2902,11 +2907,16 @@ int drm_atomic_helper_commit_duplicated_state(struct drm_atomic_state *state,
> >  	struct drm_connector_state *new_conn_state;
> >  	struct drm_crtc *crtc;
> >  	struct drm_crtc_state *new_crtc_state;
> > +	unsigned plane_mask = 0;
> > +	struct drm_device *dev = state->dev;
> > +	int ret;
> >  
> >  	state->acquire_ctx = ctx;
> >  
> > -	for_each_new_plane_in_state(state, plane, new_plane_state, i)
> > +	for_each_new_plane_in_state(state, plane, new_plane_state, i) {
> > +		plane_mask |= BIT(drm_plane_index(plane));
> We should really add a drm_plane_mask/drm_connector_mask at some point, to complement drm_crtc_mask.
> >  		state->planes[i].old_state = plane->state;
> > +	}
> >  
> >  	for_each_new_crtc_in_state(state, crtc, new_crtc_state, i)
> >  		state->crtcs[i].old_state = crtc->state;
> > @@ -2914,7 +2924,11 @@ int drm_atomic_helper_commit_duplicated_state(struct drm_atomic_state *state,
> >  	for_each_new_connector_in_state(state, connector, new_conn_state, i)
> >  		state->connectors[i].old_state = connector->state;
> >  
> > -	return drm_atomic_commit(state);
> > +	ret = drm_atomic_commit(state);
> > +	if (plane_mask)
> > +		drm_atomic_clean_old_fb(dev, plane_mask, ret);
> Kill the if () part, and make it unconditional? On second thought, I should have done the same in drm_framebuffer.c

I'd prefer to not bikeshed this stuff too much and just go with what we do
everywhere else (i.e. rmfb code and atomic commit) for consistency.
clean_old_fb is definitely a horrible function with misleading kerneldoc
on top, and I think the best way to clean that up would be to:

- Move the plane_mask computation in drm_atmic_commit and also put the
  clean_old_fb call in there. Maybe give it a more descriptive name like
  update_legacy_fb or whatever. Unexport them.

- This would break the compat helpers, where drm_atomic_commit must not
  update the legacy refcounts, because for set_config, page_flip and the
  plane hooks the core does that already. Create a
  drm_atomic_commit_legacy for these.

But since one of my patches caused an existing issue to pop up as a
regression, and this series fixes it, I'd like to first get the minimal
fix in through drm-intel. And then bikeshed it properly in drm-misc.

Typing the patches is also a bit annoying right now since I have a few
other atomic_commit patches in-flight ...

Thanks, Daniel

> > +	return ret;
> >  }
> >  EXPORT_SYMBOL(drm_atomic_helper_commit_duplicated_state);
> >  
> 
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch