On Mon, Aug 17, 2020 at 11:03:42AM +0200, Gerd Hoffmann wrote:

  Hi,

quoted

quoted

--- a/drivers/gpu/drm/virtio/virtgpu_display.c
+++ b/drivers/gpu/drm/virtio/virtgpu_display.c

@@ -100,6 +100,7 @@ static void virtio_gpu_crtc_atomic_enable(struct drm_crtc *crtc,
 	struct virtio_gpu_output *output = drm_crtc_to_virtio_gpu_output(crtc);
 
 	output->enabled = true;
+	output->need_update = true;

quoted

quoted

--- a/drivers/gpu/drm/virtio/virtgpu_plane.c
+++ b/drivers/gpu/drm/virtio/virtgpu_plane.c

@@ -163,7 +163,8 @@ static void virtio_gpu_primary_plane_update(struct drm_plane *plane,
 	    plane->state->src_w != old_state->src_w ||
 	    plane->state->src_h != old_state->src_h ||
 	    plane->state->src_x != old_state->src_x ||
-	    plane->state->src_y != old_state->src_y) {
+	    plane->state->src_y != old_state->src_y ||
+	    output->need_update) {

Uh instead of hand-rolling what's essentially a drm_crtc_needs_modeset
check, why not use that one? atomic helpers try to keep the usual suspects
for state transitions already handy, to avoid every driver rolling their
own. Or do I miss something here?

Well, the virtio-gpu virtual hardware can't do plane updates and crtc
updates independant from each other.  So the crtc callbacks handle
disable only (we don't need a fb for that) and leave the enable to the
plane update.

I suspect calling drm_atomic_crtc_needs_modeset() in plane update isn't
going to fly ...

Digged a bit more, seems crtc_state->*_changed is cleared after modeset
so the following plane update wouldn't see it.  Which I think means
there is no way around tracking that in need_update.

output->enabled is probably not needed though, seems I can replace that
by either output->crtc.state->enable or ->active.  Not fully sure which
one, probably ->active.

take care,
  Gerd

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization