sna/gen3+: Flush vertex buffer after computing resize

Upon aligning the buffer, we may enlarge the vbo to accomodate the
vertex alignment and push the current index past the end of the buffer.
Move the space check from before the alignment computation to
afterwards.

Reported-by: Jiri Slaby <jirislaby@gmail.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=47597
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

7 files changed, 124 insertions, 100 deletions

diff --git a/src/sna/gen3_render.c b/src/sna/gen3_render.c
index ebe108de..2445b91a 100644
--- a/src/sna/gen3_render.c
+++ b/src/sna/gen3_render.c
@@ -3123,20 +3123,29 @@ static void
 gen3_align_vertex(struct sna *sna,
 		  const struct sna_composite_op *op)
 {
-	if (op->floats_per_vertex != sna->render_state.gen3.last_floats_per_vertex) {
-		if (sna->render.vertex_size - sna->render.vertex_used < 2*op->floats_per_rect)
-			gen3_vertex_finish(sna);
-
-		DBG(("aligning vertex: was %d, now %d floats per vertex, %d->%d\n",
-		     sna->render_state.gen3.last_floats_per_vertex,
-		     op->floats_per_vertex,
-		     sna->render.vertex_index,
-		     (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex));
-		sna->render.vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex;
-		sna->render.vertex_used = sna->render.vertex_index * op->floats_per_vertex;
-		assert(sna->render.vertex_used < sna->render.vertex_size - op->floats_per_rect);
-		sna->render_state.gen3.last_floats_per_vertex = op->floats_per_vertex;
+	int vertex_index;
+
+	if (op->floats_per_vertex == sna->render_state.gen3.last_floats_per_vertex)
+		return;
+
+	DBG(("aligning vertex: was %d, now %d floats per vertex\n",
+	     sna->render_state.gen3.last_floats_per_vertex,
+	     op->floats_per_vertex));
+
+	assert(op->floats_per_rect == 3*op->floats_per_vertex);
+
+	vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex;
+	if ((int)sna->render.vertex_size - vertex_index * op->floats_per_vertex < 2*op->floats_per_rect) {
+		DBG(("%s: flushing vertex buffer: new index=%d, max=%d\n",
+		     __FUNCTION__, vertex_index, sna->render.vertex_size / op->floats_per_vertex));
+		if (gen3_vertex_finish(sna) < op->floats_per_vertex)
+			kgem_submit(&sna->kgem);
+
+		vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex;
 	}
+
+	sna->render.vertex_index = vertex_index;
+	sna->render.vertex_used = vertex_index * op->floats_per_vertex;
 }
 
 static bool
@@ -3806,8 +3815,8 @@ gen3_render_composite(struct sna *sna,
 			goto cleanup_mask;
 	}
 
-	gen3_emit_composite_state(sna, tmp);
 	gen3_align_vertex(sna, tmp);
+	gen3_emit_composite_state(sna, tmp);
 	return true;
 
 cleanup_mask:
@@ -5021,8 +5030,8 @@ gen3_render_composite_spans(struct sna *sna,
 			goto cleanup_src;
 	}
 
-	gen3_emit_composite_state(sna, &tmp->base);
 	gen3_align_vertex(sna, &tmp->base);
+	gen3_emit_composite_state(sna, &tmp->base);
 	return true;
 
 cleanup_src:
@@ -5665,8 +5674,8 @@ fallback_blt:
 	dst_dy += tmp.dst.y;
 	tmp.dst.x = tmp.dst.y = 0;
 
-	gen3_emit_composite_state(sna, &tmp);
 	gen3_align_vertex(sna, &tmp);
+	gen3_emit_composite_state(sna, &tmp);
 
 	do {
 		int n_this_time;
@@ -5803,8 +5812,8 @@ fallback:
 	tmp->blt  = gen3_render_copy_blt;
 	tmp->done = gen3_render_copy_done;
 
-	gen3_emit_composite_state(sna, &tmp->base);
 	gen3_align_vertex(sna, &tmp->base);
+	gen3_emit_composite_state(sna, &tmp->base);
 	return true;
 }
 
@@ -5943,8 +5952,8 @@ gen3_render_fill_boxes(struct sna *sna,
 		assert(kgem_check_bo(&sna->kgem, dst_bo, NULL));
 	}
 
-	gen3_emit_composite_state(sna, &tmp);
 	gen3_align_vertex(sna, &tmp);
+	gen3_emit_composite_state(sna, &tmp);
 
 	do {
 		int n_this_time;
@@ -6094,8 +6103,8 @@ gen3_render_fill(struct sna *sna, uint8_t alu,
 	tmp->boxes = gen3_render_fill_op_boxes;
 	tmp->done  = gen3_render_fill_op_done;
 
-	gen3_emit_composite_state(sna, &tmp->base);
 	gen3_align_vertex(sna, &tmp->base);
+	gen3_emit_composite_state(sna, &tmp->base);
 	return true;
 }
 
@@ -6172,8 +6181,8 @@ gen3_render_fill_one(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo,
 			return true;
 	}
 
-	gen3_emit_composite_state(sna, &tmp);
 	gen3_align_vertex(sna, &tmp);
+	gen3_emit_composite_state(sna, &tmp);
 	gen3_get_rectangles(sna, &tmp, 1);
 	DBG(("	(%d, %d), (%d, %d): %x\n", x1, y1, x2, y2, color));
 	OUT_VERTEX(x2);
diff --git a/src/sna/gen4_render.c b/src/sna/gen4_render.c
index 416452b4..48a8852d 100644
--- a/src/sna/gen4_render.c
+++ b/src/sna/gen4_render.c
@@ -827,16 +827,10 @@ gen4_align_vertex(struct sna *sna, const struct sna_composite_op *op)
 {
 	assert(op->floats_per_rect == 3*op->floats_per_vertex);
 	if (op->floats_per_vertex != sna->render_state.gen4.floats_per_vertex) {
-		if (sna->render.vertex_size - sna->render.vertex_used < 2*op->floats_per_rect)
-			gen4_vertex_finish(sna);
-
-		DBG(("aligning vertex: was %d, now %d floats per vertex, %d->%d\n",
+		DBG(("aligning vertex: was %d, now %d floats per vertex\n",
 		     sna->render_state.gen4.floats_per_vertex,
-		     op->floats_per_vertex,
-		     sna->render.vertex_index,
-		     (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex));
-		sna->render.vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex;
-		sna->render.vertex_used = sna->render.vertex_index * op->floats_per_vertex;
+		     op->floats_per_vertex));
+		gen4_vertex_align(sna, op);
 		sna->render_state.gen4.floats_per_vertex = op->floats_per_vertex;
 	}
 }
@@ -1396,8 +1390,8 @@ gen4_render_video(struct sna *sna,
 		assert(kgem_check_bo(&sna->kgem, tmp.dst.bo, frame->bo, NULL));
 	}
 
-	gen4_video_bind_surfaces(sna, &tmp);
 	gen4_align_vertex(sna, &tmp);
+	gen4_video_bind_surfaces(sna, &tmp);
 
 	/* Set up the offset for translating from the given region (in screen
 	 * coordinates) to the backing pixmap.
@@ -1987,8 +1981,8 @@ gen4_render_composite(struct sna *sna,
 			goto cleanup_mask;
 	}
 
-	gen4_bind_surfaces(sna, tmp);
 	gen4_align_vertex(sna, tmp);
+	gen4_bind_surfaces(sna, tmp);
 	return true;
 
 cleanup_mask:
@@ -2226,8 +2220,8 @@ gen4_render_composite_spans(struct sna *sna,
 			goto cleanup_src;
 	}
 
-	gen4_bind_surfaces(sna, &tmp->base);
 	gen4_align_vertex(sna, &tmp->base);
+	gen4_bind_surfaces(sna, &tmp->base);
 	return true;
 
 cleanup_src:
@@ -2428,8 +2422,8 @@ fallback_blt:
 	src_dx += tmp.src.offset[0];
 	src_dy += tmp.src.offset[1];
 
-	gen4_copy_bind_surfaces(sna, &tmp);
 	gen4_align_vertex(sna, &tmp);
+	gen4_copy_bind_surfaces(sna, &tmp);
 
 	do {
 		gen4_render_copy_one(sna, &tmp,
@@ -2558,8 +2552,8 @@ fallback:
 			return true;
 	}
 
-	gen4_copy_bind_surfaces(sna, &op->base);
 	gen4_align_vertex(sna, &op->base);
+	gen4_copy_bind_surfaces(sna, &op->base);
 
 	op->blt  = gen4_render_copy_blt;
 	op->done = gen4_render_copy_done;
@@ -2664,8 +2658,8 @@ gen4_render_fill_boxes(struct sna *sna,
 		assert(kgem_check_bo(&sna->kgem, dst_bo, NULL));
 	}
 
-	gen4_bind_surfaces(sna, &tmp);
 	gen4_align_vertex(sna, &tmp);
+	gen4_bind_surfaces(sna, &tmp);
 
 	do {
 		gen4_render_fill_rectangle(sna, &tmp,
@@ -2769,8 +2763,8 @@ gen4_render_fill(struct sna *sna, uint8_t alu,
 		assert(kgem_check_bo(&sna->kgem, dst_bo, NULL));
 	}
 
-	gen4_bind_surfaces(sna, &op->base);
 	gen4_align_vertex(sna, &op->base);
+	gen4_bind_surfaces(sna, &op->base);
 
 	op->blt   = gen4_render_fill_op_blt;
 	op->box   = gen4_render_fill_op_box;
@@ -2853,8 +2847,8 @@ gen4_render_fill_one(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo,
 		}
 	}
 
-	gen4_bind_surfaces(sna, &tmp);
 	gen4_align_vertex(sna, &tmp);
+	gen4_bind_surfaces(sna, &tmp);
 
 	gen4_render_fill_rectangle(sna, &tmp, x1, y1, x2 - x1, y2 - y1);
 
diff --git a/src/sna/gen4_vertex.c b/src/sna/gen4_vertex.c
index e2a48083..0585c4ce 100644
--- a/src/sna/gen4_vertex.c
+++ b/src/sna/gen4_vertex.c
@@ -38,6 +38,29 @@
 #define sse2
 #endif
 
+void gen4_vertex_align(struct sna *sna, const struct sna_composite_op *op)
+{
+	int vertex_index;
+
+	assert(op->floats_per_rect == 3*op->floats_per_vertex);
+
+	vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex;
+	if ((int)sna->render.vertex_size - vertex_index * op->floats_per_vertex < 2*op->floats_per_rect) {
+		DBG(("%s: flushing vertex buffer: new index=%d, max=%d\n",
+		     __FUNCTION__, vertex_index, sna->render.vertex_size / op->floats_per_vertex));
+		if (gen4_vertex_finish(sna) < op->floats_per_rect) {
+			kgem_submit(&sna->kgem);
+			_kgem_set_mode(&sna->kgem, KGEM_RENDER);
+		}
+
+		vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex;
+		assert(vertex_index * op->floats_per_vertex <= sna->render.vertex_size);
+	}
+
+	sna->render.vertex_index = vertex_index;
+	sna->render.vertex_used = vertex_index * op->floats_per_vertex;
+}
+
 void gen4_vertex_flush(struct sna *sna)
 {
 	DBG(("%s[%x] = %d\n", __FUNCTION__,
diff --git a/src/sna/gen4_vertex.h b/src/sna/gen4_vertex.h
index 1494ba14..03ba2f87 100644
--- a/src/sna/gen4_vertex.h
+++ b/src/sna/gen4_vertex.h
@@ -6,6 +6,7 @@
 #include "sna.h"
 #include "sna_render.h"
 
+void gen4_vertex_align(struct sna *sna, const struct sna_composite_op *op);
 void gen4_vertex_flush(struct sna *sna);
 int gen4_vertex_finish(struct sna *sna);
 void gen4_vertex_close(struct sna *sna);
diff --git a/src/sna/gen5_render.c b/src/sna/gen5_render.c
index 2bd818a4..ca41a9a8 100644
--- a/src/sna/gen5_render.c
+++ b/src/sna/gen5_render.c
@@ -796,16 +796,10 @@ gen5_align_vertex(struct sna *sna, const struct sna_composite_op *op)
 {
 	assert(op->floats_per_rect == 3*op->floats_per_vertex);
 	if (op->floats_per_vertex != sna->render_state.gen5.floats_per_vertex) {
-		if (sna->render.vertex_size - sna->render.vertex_used < 2*op->floats_per_rect)
-			gen4_vertex_finish(sna);
-
-		DBG(("aligning vertex: was %d, now %d floats per vertex, %d->%d\n",
+		DBG(("aligning vertex: was %d, now %d floats per vertex\n",
 		     sna->render_state.gen5.floats_per_vertex,
-		     op->floats_per_vertex,
-		     sna->render.vertex_index,
-		     (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex));
-		sna->render.vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex;
-		sna->render.vertex_used = sna->render.vertex_index * op->floats_per_vertex;
+		     op->floats_per_vertex));
+		gen4_vertex_align(sna, op);
 		sna->render_state.gen5.floats_per_vertex = op->floats_per_vertex;
 	}
 }
@@ -1394,8 +1388,8 @@ gen5_render_video(struct sna *sna,
 		assert(kgem_check_bo(&sna->kgem, tmp.dst.bo, frame->bo, NULL));
 	}
 
-	gen5_video_bind_surfaces(sna, &tmp);
 	gen5_align_vertex(sna, &tmp);
+	gen5_video_bind_surfaces(sna, &tmp);
 
 	/* Set up the offset for translating from the given region (in screen
 	 * coordinates) to the backing pixmap.
@@ -1969,8 +1963,8 @@ gen5_render_composite(struct sna *sna,
 			goto cleanup_mask;
 	}
 
-	gen5_bind_surfaces(sna, tmp);
 	gen5_align_vertex(sna, tmp);
+	gen5_bind_surfaces(sna, tmp);
 	return true;
 
 cleanup_mask:
@@ -2198,8 +2192,8 @@ gen5_render_composite_spans(struct sna *sna,
 			goto cleanup_src;
 	}
 
-	gen5_bind_surfaces(sna, &tmp->base);
 	gen5_align_vertex(sna, &tmp->base);
+	gen5_bind_surfaces(sna, &tmp->base);
 	return true;
 
 cleanup_src:
@@ -2388,8 +2382,8 @@ fallback_blt:
 	src_dx += tmp.src.offset[0];
 	src_dy += tmp.src.offset[1];
 
-	gen5_copy_bind_surfaces(sna, &tmp);
 	gen5_align_vertex(sna, &tmp);
+	gen5_copy_bind_surfaces(sna, &tmp);
 
 	do {
 		int n_this_time;
@@ -2553,8 +2547,8 @@ fallback:
 			return true;
 	}
 
-	gen5_copy_bind_surfaces(sna, &op->base);
 	gen5_align_vertex(sna, &op->base);
+	gen5_copy_bind_surfaces(sna, &op->base);
 
 	op->blt  = gen5_render_copy_blt;
 	op->done = gen5_render_copy_done;
@@ -2694,8 +2688,8 @@ gen5_render_fill_boxes(struct sna *sna,
 		assert(kgem_check_bo(&sna->kgem, dst_bo, NULL));
 	}
 
-	gen5_fill_bind_surfaces(sna, &tmp);
 	gen5_align_vertex(sna, &tmp);
+	gen5_fill_bind_surfaces(sna, &tmp);
 
 	do {
 		int n_this_time;
@@ -2864,8 +2858,8 @@ gen5_render_fill(struct sna *sna, uint8_t alu,
 		assert(kgem_check_bo(&sna->kgem, dst_bo, NULL));
 	}
 
-	gen5_fill_bind_surfaces(sna, &op->base);
 	gen5_align_vertex(sna, &op->base);
+	gen5_fill_bind_surfaces(sna, &op->base);
 
 	op->blt   = gen5_render_fill_op_blt;
 	op->box   = gen5_render_fill_op_box;
@@ -2959,8 +2953,8 @@ gen5_render_fill_one(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo,
 		assert(kgem_check_bo(&sna->kgem, bo, NULL));
 	}
 
-	gen5_fill_bind_surfaces(sna, &tmp);
 	gen5_align_vertex(sna, &tmp);
+	gen5_fill_bind_surfaces(sna, &tmp);
 
 	gen5_get_rectangles(sna, &tmp, 1, gen5_fill_bind_surfaces);
 
diff --git a/src/sna/gen6_render.c b/src/sna/gen6_render.c
index e5e07ce1..ee1686fc 100644
--- a/src/sna/gen6_render.c
+++ b/src/sna/gen6_render.c
@@ -1312,16 +1312,10 @@ gen6_align_vertex(struct sna *sna, const struct sna_composite_op *op)
 {
 	assert (sna->render.vertex_offset == 0);
 	if (op->floats_per_vertex != sna->render_state.gen6.floats_per_vertex) {
-		if (sna->render.vertex_size - sna->render.vertex_used < 2*op->floats_per_rect)
-			gen4_vertex_finish(sna);
-
-		DBG(("aligning vertex: was %d, now %d floats per vertex, %d->%d\n",
+		DBG(("aligning vertex: was %d, now %d floats per vertex\n",
 		     sna->render_state.gen6.floats_per_vertex,
-		     op->floats_per_vertex,
-		     sna->render.vertex_index,
-		     (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex));
-		sna->render.vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex;
-		sna->render.vertex_used = sna->render.vertex_index * op->floats_per_vertex;
+		     op->floats_per_vertex));
+		gen4_vertex_align(sna, op);
 		sna->render_state.gen6.floats_per_vertex = op->floats_per_vertex;
 	}
 	assert((sna->render.vertex_used % op->floats_per_vertex) == 0);
@@ -1649,8 +1643,8 @@ gen6_render_video(struct sna *sna,
 		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen6_emit_video_state(sna, &tmp);
 	gen6_align_vertex(sna, &tmp);
+	gen6_emit_video_state(sna, &tmp);
 
 	/* Set up the offset for translating from the given region (in screen
 	 * coordinates) to the backing pixmap.
@@ -2399,8 +2393,8 @@ gen6_render_composite(struct sna *sna,
 		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen6_emit_composite_state(sna, tmp);
 	gen6_align_vertex(sna, tmp);
+	gen6_emit_composite_state(sna, tmp);
 	return true;
 
 cleanup_mask:
@@ -2632,8 +2626,8 @@ gen6_render_composite_spans(struct sna *sna,
 		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen6_emit_composite_state(sna, &tmp->base);
 	gen6_align_vertex(sna, &tmp->base);
+	gen6_emit_composite_state(sna, &tmp->base);
 	return true;
 
 cleanup_src:
@@ -2916,8 +2910,8 @@ fallback_blt:
 		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen6_emit_copy_state(sna, &tmp);
 	gen6_align_vertex(sna, &tmp);
+	gen6_emit_copy_state(sna, &tmp);
 
 	do {
 		int16_t *v;
@@ -3075,8 +3069,8 @@ fallback:
 		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen6_emit_copy_state(sna, &op->base);
 	gen6_align_vertex(sna, &op->base);
+	gen6_emit_copy_state(sna, &op->base);
 
 	op->blt  = gen6_render_copy_blt;
 	op->done = gen6_render_copy_done;
@@ -3241,8 +3235,8 @@ gen6_render_fill_boxes(struct sna *sna,
 		assert(kgem_check_bo(&sna->kgem, dst_bo, NULL));
 	}
 
-	gen6_emit_fill_state(sna, &tmp);
 	gen6_align_vertex(sna, &tmp);
+	gen6_emit_fill_state(sna, &tmp);
 
 	do {
 		int n_this_time;
@@ -3421,8 +3415,8 @@ gen6_render_fill(struct sna *sna, uint8_t alu,
 		assert(kgem_check_bo(&sna->kgem, dst_bo, NULL));
 	}
 
-	gen6_emit_fill_state(sna, &op->base);
 	gen6_align_vertex(sna, &op->base);
+	gen6_emit_fill_state(sna, &op->base);
 
 	op->blt  = gen6_render_op_fill_blt;
 	op->box  = gen6_render_op_fill_box;
@@ -3505,8 +3499,8 @@ gen6_render_fill_one(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo,
 		}
 	}
 
-	gen6_emit_fill_state(sna, &tmp);
 	gen6_align_vertex(sna, &tmp);
+	gen6_emit_fill_state(sna, &tmp);
 
 	gen6_get_rectangles(sna, &tmp, 1, gen6_emit_fill_state);
 
@@ -3592,8 +3586,8 @@ gen6_render_clear(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo)
 		}
 	}
 
-	gen6_emit_fill_state(sna, &tmp);
 	gen6_align_vertex(sna, &tmp);
+	gen6_emit_fill_state(sna, &tmp);
 
 	gen6_get_rectangles(sna, &tmp, 1, gen6_emit_fill_state);
 
diff --git a/src/sna/gen7_render.c b/src/sna/gen7_render.c
index 4b60f531..30e1cc1b 100644
--- a/src/sna/gen7_render.c
+++ b/src/sna/gen7_render.c
@@ -1097,36 +1097,48 @@ gen7_emit_state(struct sna *sna,
 		const struct sna_composite_op *op,
 		uint16_t wm_binding_table)
 {
+	bool need_invalidate;
+	bool need_flush;
 	bool need_stall;
 
 	assert(op->dst.bo->exec);
 
-	gen7_emit_cc(sna, GEN7_BLEND(op->u.gen7.flags));
-	gen7_emit_sampler(sna, GEN7_SAMPLER(op->u.gen7.flags));
-	gen7_emit_sf(sna, GEN7_VERTEX(op->u.gen7.flags) >> 2);
-	gen7_emit_wm(sna, GEN7_KERNEL(op->u.gen7.flags));
-	gen7_emit_vertex_elements(sna, op);
+	need_invalidate = kgem_bo_is_dirty(op->src.bo) || kgem_bo_is_dirty(op->mask.bo);
+	if (ALWAYS_INVALIDATE)
+		need_invalidate = true;
 
-	need_stall = gen7_emit_binding_table(sna, wm_binding_table);
+	need_flush = sna->render_state.gen7.emit_flush;
+	if (ALWAYS_FLUSH)
+		need_flush = true;
+
+	need_stall = sna->render_state.gen7.surface_table != wm_binding_table;
 	need_stall &= gen7_emit_drawing_rectangle(sna, op);
 	if (ALWAYS_STALL)
 		need_stall = true;
 
-	if (ALWAYS_INVALIDATE || kgem_bo_is_dirty(op->src.bo) || kgem_bo_is_dirty(op->mask.bo)) {
+	if (need_invalidate) {
 		gen7_emit_pipe_invalidate(sna);
 		kgem_clear_dirty(&sna->kgem);
 		assert(op->dst.bo->exec);
 		kgem_bo_mark_dirty(op->dst.bo);
-		sna->render_state.gen7.emit_flush = false;
+
+		need_flush = false;
 		need_stall = false;
 	}
-	if (ALWAYS_FLUSH || (sna->render_state.gen7.emit_flush && GEN7_READS_DST(op->u.gen7.flags))) {
+	if (need_flush) {
 		gen7_emit_pipe_flush(sna, need_stall);
 		need_stall = false;
 	}
 	if (need_stall)
 		gen7_emit_pipe_stall(sna);
 
+	gen7_emit_cc(sna, GEN7_BLEND(op->u.gen7.flags));
+	gen7_emit_sampler(sna, GEN7_SAMPLER(op->u.gen7.flags));
+	gen7_emit_sf(sna, GEN7_VERTEX(op->u.gen7.flags) >> 2);
+	gen7_emit_wm(sna, GEN7_KERNEL(op->u.gen7.flags));
+	gen7_emit_vertex_elements(sna, op);
+	gen7_emit_binding_table(sna, wm_binding_table);
+
 	sna->render_state.gen7.emit_flush = GEN7_READS_DST(op->u.gen7.flags);
 }
 
@@ -1522,16 +1534,9 @@ static void
 gen7_align_vertex(struct sna *sna, const struct sna_composite_op *op)
 {
 	if (op->floats_per_vertex != sna->render_state.gen7.floats_per_vertex) {
-		if (sna->render.vertex_size - sna->render.vertex_used < 2*op->floats_per_rect)
-			gen4_vertex_finish(sna);
-
-		DBG(("aligning vertex: was %d, now %d floats per vertex, %d->%d\n",
-		     sna->render_state.gen7.floats_per_vertex,
-		     op->floats_per_vertex,
-		     sna->render.vertex_index,
-		     (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex));
-		sna->render.vertex_index = (sna->render.vertex_used + op->floats_per_vertex - 1) / op->floats_per_vertex;
-		sna->render.vertex_used = sna->render.vertex_index * op->floats_per_vertex;
+		DBG(("aligning vertex: was %d, now %d floats per vertex\n",
+		     sna->render_state.gen7.floats_per_vertex, op->floats_per_vertex));
+		gen4_vertex_align(sna, op);
 		sna->render_state.gen7.floats_per_vertex = op->floats_per_vertex;
 	}
 }
@@ -1865,8 +1870,8 @@ gen7_render_video(struct sna *sna,
 		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen7_emit_video_state(sna, &tmp);
 	gen7_align_vertex(sna, &tmp);
+	gen7_emit_video_state(sna, &tmp);
 
 	/* Set up the offset for translating from the given region (in screen
 	 * coordinates) to the backing pixmap.
@@ -2654,8 +2659,8 @@ gen7_render_composite(struct sna *sna,
 		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen7_emit_composite_state(sna, tmp);
 	gen7_align_vertex(sna, tmp);
+	gen7_emit_composite_state(sna, tmp);
 	return true;
 
 cleanup_mask:
@@ -2867,8 +2872,8 @@ gen7_render_composite_spans(struct sna *sna,
 		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen7_emit_composite_state(sna, &tmp->base);
 	gen7_align_vertex(sna, &tmp->base);
+	gen7_emit_composite_state(sna, &tmp->base);
 	return true;
 
 cleanup_src:
@@ -3141,8 +3146,8 @@ fallback_blt:
 		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen7_emit_copy_state(sna, &tmp);
 	gen7_align_vertex(sna, &tmp);
+	gen7_emit_copy_state(sna, &tmp);
 
 	do {
 		int16_t *v;
@@ -3295,8 +3300,8 @@ fallback:
 		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen7_emit_copy_state(sna, &op->base);
 	gen7_align_vertex(sna, &op->base);
+	gen7_emit_copy_state(sna, &op->base);
 
 	op->blt  = gen7_render_copy_blt;
 	op->done = gen7_render_copy_done;
@@ -3458,10 +3463,11 @@ gen7_render_fill_boxes(struct sna *sna,
 	if (!kgem_check_bo(&sna->kgem, dst_bo, NULL)) {
 		kgem_submit(&sna->kgem);
 		assert(kgem_check_bo(&sna->kgem, dst_bo, NULL));
+		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen7_emit_fill_state(sna, &tmp);
 	gen7_align_vertex(sna, &tmp);
+	gen7_emit_fill_state(sna, &tmp);
 
 	do {
 		int n_this_time;
@@ -3632,10 +3638,11 @@ gen7_render_fill(struct sna *sna, uint8_t alu,
 	if (!kgem_check_bo(&sna->kgem, dst_bo, NULL)) {
 		kgem_submit(&sna->kgem);
 		assert(kgem_check_bo(&sna->kgem, dst_bo, NULL));
+		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen7_emit_fill_state(sna, &op->base);
 	gen7_align_vertex(sna, &op->base);
+	gen7_emit_fill_state(sna, &op->base);
 
 	op->blt   = gen7_render_fill_op_blt;
 	op->box   = gen7_render_fill_op_box;
@@ -3713,10 +3720,11 @@ gen7_render_fill_one(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo,
 			kgem_bo_destroy(&sna->kgem, tmp.src.bo);
 			return false;
 		}
+		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen7_emit_fill_state(sna, &tmp);
 	gen7_align_vertex(sna, &tmp);
+	gen7_emit_fill_state(sna, &tmp);
 
 	gen7_get_rectangles(sna, &tmp, 1, gen7_emit_fill_state);
 
@@ -3797,10 +3805,11 @@ gen7_render_clear(struct sna *sna, PixmapPtr dst, struct kgem_bo *bo)
 			kgem_bo_destroy(&sna->kgem, tmp.src.bo);
 			return false;
 		}
+		_kgem_set_mode(&sna->kgem, KGEM_RENDER);
 	}
 
-	gen7_emit_fill_state(sna, &tmp);
 	gen7_align_vertex(sna, &tmp);
+	gen7_emit_fill_state(sna, &tmp);
 
 	gen7_get_rectangles(sna, &tmp, 1, gen7_emit_fill_state);