drm/atomic-helper: Reject legacy flips on a disabled pipe

We want this for consistency with existing page_flip semantics.

Since this spurred quite a discussion on IRC also document why we
reject event generation when the pipe is off: It's not that it's hard
to implement, but userspace has a track recording which proves that it's
way too easy to accidentally abuse and cause havoc. We want to make
sure userspace doesn't get away with that.

v2: Somehow thought we do reject events already, but that code only
existed in my imagination ... Also suggestions from Thierry.

Cc: Daniel Stone <daniels@collabora.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Thierry Reding <thierry.reding@gmail.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1449564561-3896-4-git-send-email-daniel.vetter@ffwll.ch

3 files changed, 27 insertions, 1 deletions

diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
index 14b321580517..6050c80f1dbc 100644
--- a/drivers/gpu/drm/drm_atomic.c
+++ b/drivers/gpu/drm/drm_atomic.c
@@ -508,6 +508,22 @@ static int drm_atomic_crtc_check(struct drm_crtc *crtc,
                 return -EINVAL;
         }
 
+        /*
+         * Reject event generation for when a CRTC is off and stays off.
+         * It wouldn't be hard to implement this, but userspace has a track
+         * record of happily burning through 100% cpu (or worse, crash) when the
+         * display pipe is suspended. To avoid all that fun just reject updates
+         * that ask for events since likely that indicates a bug in the
+         * compositor's drawing loop. This is consistent with the vblank IOCTL
+         * and legacy page_flip IOCTL which also reject service on a disabled
+         * pipe.
+         */
+        if (state->event && !state->active && !crtc->state->active) {
+                DRM_DEBUG_ATOMIC("[CRTC:%d] requesting event but off\n",
+                                 crtc->base.id);
+                return -EINVAL;
+        }
+
         return 0;
 }
 
diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
index 26d258d0618b..738104b68d26 100644
--- a/drivers/gpu/drm/drm_atomic_helper.c
+++ b/drivers/gpu/drm/drm_atomic_helper.c
@@ -2284,6 +2284,15 @@ retry:
                 goto fail;
         drm_atomic_set_fb_for_plane(plane_state, fb);
 
+        /* Make sure we don't accidentally do a full modeset. */
+        state->allow_modeset = false;
+        if (!crtc_state->active) {
+                DRM_DEBUG_ATOMIC("[CRTC:%d] disabled, rejecting legacy flip\n",
+                                 crtc->base.id);
+                ret = -EINVAL;
+                goto fail;
+        }
+
         ret = drm_atomic_async_commit(state);
         if (ret != 0)
                 goto fail;
diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h
index e3c4d486e1af..c65a212db77e 100644
--- a/include/drm/drm_crtc.h
+++ b/include/drm/drm_crtc.h
@@ -551,7 +551,8 @@ struct drm_crtc_funcs {
          * ->page_flip() operation is already pending the callback should return
          * -EBUSY. Pageflips on a disabled CRTC (either by setting a NULL mode
          * or just runtime disabled through DPMS respectively the new atomic
-         * "ACTIVE" state) should result in an -EINVAL error code.
+         * "ACTIVE" state) should result in an -EINVAL error code. Note that
+         * drm_atomic_helper_page_flip() checks this already for atomic drivers.
          */
         int (*page_flip)(struct drm_crtc *crtc,
                          struct drm_framebuffer *fb,