4 files changed, 15 insertions, 9 deletions

diff --git a/src/gallium/drivers/radeon/r600_pipe_common.c b/src/gallium/drivers/radeon/r600_pipe_common.c
index 5b1ce04277b..95950c36804 100644
--- a/src/gallium/drivers/radeon/r600_pipe_common.c
+++ b/src/gallium/drivers/radeon/r600_pipe_common.c
@@ -80,7 +80,7 @@ void radeon_shader_binary_clean(struct radeon_shader_binary *b)
  * pipe_context
  */
 
-void r600_gfx_write_fence(struct r600_common_context *ctx,
+void r600_gfx_write_fence(struct r600_common_context *ctx, struct r600_resource *buf,
 			  uint64_t va, uint32_t old_value, uint32_t new_value)
 {
 	struct radeon_winsys_cs *cs = ctx->gfx.cs;
@@ -106,6 +106,8 @@ void r600_gfx_write_fence(struct r600_common_context *ctx,
 	radeon_emit(cs, (va >> 32) | EOP_DATA_SEL(1));
 	radeon_emit(cs, new_value); /* immediate data */
 	radeon_emit(cs, 0); /* unused */
+
+	r600_emit_reloc(ctx, &ctx->gfx, buf, RADEON_USAGE_WRITE, RADEON_PRIO_QUERY);
 }
 
 unsigned r600_gfx_write_fence_dwords(struct r600_common_screen *screen)
@@ -115,6 +117,9 @@ unsigned r600_gfx_write_fence_dwords(struct r600_common_screen *screen)
 	if (screen->chip_class == CIK)
 		dwords *= 2;
 
+	if (!screen->info.has_virtual_memory)
+		dwords += 2;
+
 	return dwords;
 }
 
diff --git a/src/gallium/drivers/radeon/r600_pipe_common.h b/src/gallium/drivers/radeon/r600_pipe_common.h
index f23f1c436a2..c6669ccfec2 100644
--- a/src/gallium/drivers/radeon/r600_pipe_common.h
+++ b/src/gallium/drivers/radeon/r600_pipe_common.h
@@ -700,7 +700,7 @@ r600_invalidate_resource(struct pipe_context *ctx,
 			 struct pipe_resource *resource);
 
 /* r600_common_pipe.c */
-void r600_gfx_write_fence(struct r600_common_context *ctx,
+void r600_gfx_write_fence(struct r600_common_context *ctx, struct r600_resource *buf,
 			  uint64_t va, uint32_t old_value, uint32_t new_value);
 unsigned r600_gfx_write_fence_dwords(struct r600_common_screen *screen);
 void r600_gfx_wait_fence(struct r600_common_context *ctx,
diff --git a/src/gallium/drivers/radeon/r600_query.c b/src/gallium/drivers/radeon/r600_query.c
index 429ea6a9565..4b6767dd38a 100644
--- a/src/gallium/drivers/radeon/r600_query.c
+++ b/src/gallium/drivers/radeon/r600_query.c
@@ -616,6 +616,7 @@ static void r600_query_hw_do_emit_stop(struct r600_common_context *ctx,
 				       uint64_t va)
 {
 	struct radeon_winsys_cs *cs = ctx->gfx.cs;
+	uint64_t fence_va = 0;
 
 	switch (query->b.type) {
 	case PIPE_QUERY_OCCLUSION_COUNTER:
@@ -626,8 +627,7 @@ static void r600_query_hw_do_emit_stop(struct r600_common_context *ctx,
 		radeon_emit(cs, va);
 		radeon_emit(cs, (va >> 32) & 0xFFFF);
 
-		va += ctx->max_db * 16 - 8;
-		r600_gfx_write_fence(ctx, va, 0, 0x80000000);
+		fence_va = va + ctx->max_db * 16 - 8;
 		break;
 	case PIPE_QUERY_PRIMITIVES_EMITTED:
 	case PIPE_QUERY_PRIMITIVES_GENERATED:
@@ -650,8 +650,7 @@ static void r600_query_hw_do_emit_stop(struct r600_common_context *ctx,
 		radeon_emit(cs, 0);
 		radeon_emit(cs, 0);
 
-		va += 8;
-		r600_gfx_write_fence(ctx, va, 0, 0x80000000);
+		fence_va = va + 8;
 		break;
 	case PIPE_QUERY_PIPELINE_STATISTICS: {
 		unsigned sample_size = (query->result_size - 8) / 2;
@@ -662,8 +661,7 @@ static void r600_query_hw_do_emit_stop(struct r600_common_context *ctx,
 		radeon_emit(cs, va);
 		radeon_emit(cs, (va >> 32) & 0xFFFF);
 
-		va += sample_size;
-		r600_gfx_write_fence(ctx, va, 0, 0x80000000);
+		fence_va = va + sample_size;
 		break;
 	}
 	default:
@@ -671,6 +669,9 @@ static void r600_query_hw_do_emit_stop(struct r600_common_context *ctx,
 	}
 	r600_emit_reloc(ctx, &ctx->gfx, query->buffer.buf, RADEON_USAGE_WRITE,
 			RADEON_PRIO_QUERY);
+
+	if (fence_va)
+		r600_gfx_write_fence(ctx, query->buffer.buf, fence_va, 0, 0x80000000);
 }
 
 static void r600_query_hw_emit_stop(struct r600_common_context *ctx,
diff --git a/src/gallium/drivers/radeonsi/si_perfcounter.c b/src/gallium/drivers/radeonsi/si_perfcounter.c
index d0c5392b441..0f5c28ce0f6 100644
--- a/src/gallium/drivers/radeonsi/si_perfcounter.c
+++ b/src/gallium/drivers/radeonsi/si_perfcounter.c
@@ -591,7 +591,7 @@ static void si_pc_emit_stop(struct r600_common_context *ctx,
 {
 	struct radeon_winsys_cs *cs = ctx->gfx.cs;
 
-	r600_gfx_write_fence(ctx, va, 1, 0);
+	r600_gfx_write_fence(ctx, buffer, va, 1, 0);
 	r600_gfx_wait_fence(ctx, va, 0, 0xffffffff);
 
 	radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));