nv50: fix drawing from tfb, direct-to-pushbuf submits

The stride was being set to 0, which is illegal (and also non-sensical).
Also we must wait for the buffer to become available for reading as
otherwise a wrong value may be prefetched. Since we must wait for the
buffer anyways, and it's mapped and in GART, we may as well avoid the
annoyance of the indirect pushbuf submit.

Signed-off-by: Ilia Mirkin <[email protected]>
Cc: [email protected]

4 files changed, 15 insertions, 14 deletions

diff --git a/src/gallium/drivers/nouveau/nv50/nv50_context.h b/src/gallium/drivers/nouveau/nv50/nv50_context.h
index e7adf472ed0..69c121274a9 100644
--- a/src/gallium/drivers/nouveau/nv50/nv50_context.h
+++ b/src/gallium/drivers/nouveau/nv50/nv50_context.h
@@ -197,7 +197,7 @@ extern struct draw_stage *nv50_draw_render_stage(struct nv50_context *);
 
 /* nv50_query.c */
 void nv50_init_query_functions(struct nv50_context *);
-void nv50_query_pushbuf_submit(struct nouveau_pushbuf *,
+void nv50_query_pushbuf_submit(struct nouveau_pushbuf *, uint16_t method,
                                struct pipe_query *, unsigned result_offset);
 void nv84_query_fifo_wait(struct nouveau_pushbuf *, struct pipe_query *);
 void nva0_so_target_save_offset(struct pipe_context *,
diff --git a/src/gallium/drivers/nouveau/nv50/nv50_query.c b/src/gallium/drivers/nouveau/nv50/nv50_query.c
index f4adbf8c653..5368ee73750 100644
--- a/src/gallium/drivers/nouveau/nv50/nv50_query.c
+++ b/src/gallium/drivers/nouveau/nv50/nv50_query.c
@@ -266,6 +266,7 @@ nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
       nv50_query_get(push, q, 0, 0x1000f010);
       break;
    case NVA0_QUERY_STREAM_OUTPUT_BUFFER_OFFSET:
+      q->sequence++;
       nv50_query_get(push, q, 0, 0x0d005002 | (q->index << 5));
       break;
    case PIPE_QUERY_TIMESTAMP_DISJOINT:
@@ -451,18 +452,18 @@ nv50_render_condition(struct pipe_context *pipe,
 }
 
 void
-nv50_query_pushbuf_submit(struct nouveau_pushbuf *push,
+nv50_query_pushbuf_submit(struct nouveau_pushbuf *push, uint16_t method,
                           struct pipe_query *pq, unsigned result_offset)
 {
    struct nv50_query *q = nv50_query(pq);
 
-   /* XXX: does this exist ? */
-#define NV50_IB_ENTRY_1_NO_PREFETCH (0 << (31 - 8))
+   nv50_query_update(q);
+   if (q->state != NV50_QUERY_STATE_READY)
+      nouveau_bo_wait(q->bo, NOUVEAU_BO_RD, push->client);
+   q->state = NV50_QUERY_STATE_READY;
 
-   PUSH_REFN(push, q->bo, NOUVEAU_BO_RD | NOUVEAU_BO_GART);
-   nouveau_pushbuf_space(push, 0, 0, 1);
-   nouveau_pushbuf_data(push, q->bo, q->offset + result_offset, 4 |
-                        NV50_IB_ENTRY_1_NO_PREFETCH);
+   BEGIN_NV04(push, SUBC_3D(method), 1);
+   PUSH_DATA (push, q->data[result_offset / 4]);
 }
 
 void
diff --git a/src/gallium/drivers/nouveau/nv50/nv50_shader_state.c b/src/gallium/drivers/nouveau/nv50/nv50_shader_state.c
index b033ce5c6dc..fdde11f4cd5 100644
--- a/src/gallium/drivers/nouveau/nv50/nv50_shader_state.c
+++ b/src/gallium/drivers/nouveau/nv50/nv50_shader_state.c
@@ -641,12 +641,12 @@ nv50_stream_output_validate(struct nv50_context *nv50)
       PUSH_DATA (push, so->num_attribs[i]);
       if (n == 4) {
          PUSH_DATA(push, targ->pipe.buffer_size);
-
-         BEGIN_NV04(push, NVA0_3D(STRMOUT_OFFSET(i)), 1);
          if (!targ->clean) {
             assert(targ->pq);
-            nv50_query_pushbuf_submit(push, targ->pq, 0x4);
+            nv50_query_pushbuf_submit(push, NVA0_3D_STRMOUT_OFFSET(i),
+                                      targ->pq, 0x4);
          } else {
+            BEGIN_NV04(push, NVA0_3D(STRMOUT_OFFSET(i)), 1);
             PUSH_DATA(push, 0);
             targ->clean = false;
          }
@@ -655,6 +655,7 @@ nv50_stream_output_validate(struct nv50_context *nv50)
             (so->stride[i] * nv50->state.prim_size);
          prims = MIN2(prims, limit);
       }
+      targ->stride = so->stride[i];
       BCTX_REFN(nv50->bufctx_3d, SO, buf, WR);
    }
    if (prims != ~0) {
diff --git a/src/gallium/drivers/nouveau/nv50/nv50_vbo.c b/src/gallium/drivers/nouveau/nv50/nv50_vbo.c
index 6324726acec..ca51ea1dc7e 100644
--- a/src/gallium/drivers/nouveau/nv50/nv50_vbo.c
+++ b/src/gallium/drivers/nouveau/nv50/nv50_vbo.c
@@ -736,9 +736,8 @@ nva0_draw_stream_output(struct nv50_context *nv50,
       BEGIN_NV04(push, NVA0_3D(DRAW_TFB_BASE), 1);
       PUSH_DATA (push, 0);
       BEGIN_NV04(push, NVA0_3D(DRAW_TFB_STRIDE), 1);
-      PUSH_DATA (push, 0);
-      BEGIN_NV04(push, NVA0_3D(DRAW_TFB_BYTES), 1);
-      nv50_query_pushbuf_submit(push, so->pq, 0x4);
+      PUSH_DATA (push, so->stride);
+      nv50_query_pushbuf_submit(push, NVA0_3D_DRAW_TFB_BYTES, so->pq, 0x4);
       BEGIN_NV04(push, NV50_3D(VERTEX_END_GL), 1);
       PUSH_DATA (push, 0);