winsys/radeon: consolidate tracing into winsys v2

This move the tracing timeout and printing into winsys and add
an debug environement variable for it (R600_DEBUG=trace_cs).

Lot of file touched because of winsys API changes.

v2: Do not write lockup file if ib uniq id does not match last one

Signed-off-by: Jerome Glisse <[email protected]>
Reviewed-by: Marek Olšák <[email protected]>

5 files changed, 13 insertions, 57 deletions

diff --git a/src/gallium/drivers/r600/evergreen_compute.c b/src/gallium/drivers/r600/evergreen_compute.c
index 489629f4612..d6b0a47cf26 100644
--- a/src/gallium/drivers/r600/evergreen_compute.c
+++ b/src/gallium/drivers/r600/evergreen_compute.c
@@ -463,7 +463,7 @@ static void compute_emit_cs(struct r600_context *ctx, const uint *block_layout,
 		flush_flags |= RADEON_FLUSH_KEEP_TILING_FLAGS;
 	}
 
-	ctx->ws->cs_flush(ctx->rings.gfx.cs, flush_flags);
+	ctx->ws->cs_flush(ctx->rings.gfx.cs, flush_flags, ctx->screen->cs_count++);
 
 	ctx->flags = 0;
 
diff --git a/src/gallium/drivers/r600/r600_hw_context.c b/src/gallium/drivers/r600/r600_hw_context.c
index b4fb3bf7933..3abce1ebde2 100644
--- a/src/gallium/drivers/r600/r600_hw_context.c
+++ b/src/gallium/drivers/r600/r600_hw_context.c
@@ -137,21 +137,17 @@ void r600_need_cs_space(struct r600_context *ctx, unsigned num_dw,
 		for (i = 0; i < R600_NUM_ATOMS; i++) {
 			if (ctx->atoms[i] && ctx->atoms[i]->dirty) {
 				num_dw += ctx->atoms[i]->num_dw;
-#if R600_TRACE_CS
 				if (ctx->screen->trace_bo) {
 					num_dw += R600_TRACE_CS_DWORDS;
 				}
-#endif
 			}
 		}
 
 		/* The upper-bound of how much space a draw command would take. */
 		num_dw += R600_MAX_FLUSH_CS_DWORDS + R600_MAX_DRAW_CS_DWORDS;
-#if R600_TRACE_CS
 		if (ctx->screen->trace_bo) {
 			num_dw += R600_TRACE_CS_DWORDS;
 		}
-#endif
 	}
 
 	/* Count in queries_suspend. */
@@ -339,37 +335,7 @@ void r600_context_flush(struct r600_context *ctx, unsigned flags)
 	}
 
 	/* Flush the CS. */
-#if R600_TRACE_CS
-	if (ctx->screen->trace_bo) {
-		struct r600_screen *rscreen = ctx->screen;
-		unsigned i;
-
-		for (i = 0; i < cs->cdw; i++) {
-			fprintf(stderr, "[%4d] [%5d] 0x%08x\n", rscreen->cs_count, i, cs->buf[i]);
-		}
-		rscreen->cs_count++;
-	}
-#endif
-	ctx->ws->cs_flush(ctx->rings.gfx.cs, flags);
-#if R600_TRACE_CS
-	if (ctx->screen->trace_bo) {
-		struct r600_screen *rscreen = ctx->screen;
-		unsigned i;
-
-		for (i = 0; i < 10; i++) {
-			usleep(5);
-			if (!ctx->ws->buffer_is_busy(rscreen->trace_bo->buf, RADEON_USAGE_READWRITE)) {
-				break;
-			}
-		}
-		if (i == 10) {
-			fprintf(stderr, "timeout on cs lockup likely happen at cs %d dw %d\n",
-				rscreen->trace_ptr[1], rscreen->trace_ptr[0]);
-		} else {
-			fprintf(stderr, "cs %d executed in %dms\n", rscreen->trace_ptr[1], i * 5);
-		}
-	}
-#endif
+	ctx->ws->cs_flush(ctx->rings.gfx.cs, flags, ctx->screen->cs_count++);
 }
 
 void r600_begin_new_cs(struct r600_context *ctx)
diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c
index 008539d7c0c..5a0ea8407c5 100644
--- a/src/gallium/drivers/r600/r600_pipe.c
+++ b/src/gallium/drivers/r600/r600_pipe.c
@@ -45,6 +45,7 @@ static const struct debug_named_value debug_options[] = {
 	{ "texdepth", DBG_TEX_DEPTH, "Print texture depth info" },
 	{ "compute", DBG_COMPUTE, "Print compute info" },
 	{ "vm", DBG_VM, "Print virtual addresses when creating resources" },
+	{ "trace_cs", DBG_TRACE_CS, "Trace cs and write rlockup_<csid>.c file with faulty cs" },
 
 	/* shaders */
 	{ "fs", DBG_FS, "Print fetch shaders" },
@@ -212,7 +213,7 @@ static void r600_flush_dma_ring(void *ctx, unsigned flags)
 	}
 
 	rctx->rings.dma.flushing = true;
-	rctx->ws->cs_flush(cs, flags);
+	rctx->ws->cs_flush(cs, flags, 0);
 	rctx->rings.dma.flushing = false;
 }
 
@@ -430,14 +431,18 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void
 		goto fail;
 	}
 
-	rctx->rings.gfx.cs = rctx->ws->cs_create(rctx->ws, RING_GFX);
+	if (rscreen->trace_bo) {
+		rctx->rings.gfx.cs = rctx->ws->cs_create(rctx->ws, RING_GFX, rscreen->trace_bo->cs_buf);
+	} else {
+		rctx->rings.gfx.cs = rctx->ws->cs_create(rctx->ws, RING_GFX, NULL);
+	}
 	rctx->rings.gfx.flush = r600_flush_gfx_ring;
 	rctx->ws->cs_set_flush_callback(rctx->rings.gfx.cs, r600_flush_from_winsys, rctx);
 	rctx->rings.gfx.flushing = false;
 
 	rctx->rings.dma.cs = NULL;
 	if (rscreen->info.r600_has_dma && !(rscreen->debug_flags & DBG_NO_ASYNC_DMA)) {
-		rctx->rings.dma.cs = rctx->ws->cs_create(rctx->ws, RING_DMA);
+		rctx->rings.dma.cs = rctx->ws->cs_create(rctx->ws, RING_DMA, NULL);
 		rctx->rings.dma.flush = r600_flush_dma_ring;
 		rctx->ws->cs_set_flush_callback(rctx->rings.dma.cs, r600_flush_dma_from_winsys, rctx);
 		rctx->rings.dma.flushing = false;
@@ -958,12 +963,10 @@ static void r600_destroy_screen(struct pipe_screen* pscreen)
 		rscreen->ws->buffer_unmap(rscreen->fences.bo->cs_buf);
 		pipe_resource_reference((struct pipe_resource**)&rscreen->fences.bo, NULL);
 	}
-#if R600_TRACE_CS
 	if (rscreen->trace_bo) {
 		rscreen->ws->buffer_unmap(rscreen->trace_bo->cs_buf);
 		pipe_resource_reference((struct pipe_resource**)&rscreen->trace_bo, NULL);
 	}
-#endif
 	pipe_mutex_destroy(rscreen->fences.mutex);
 
 	rscreen->ws->destroy(rscreen->ws);
@@ -1308,9 +1311,8 @@ struct pipe_screen *r600_screen_create(struct radeon_winsys *ws)
 
 	rscreen->global_pool = compute_memory_pool_new(rscreen);
 
-#if R600_TRACE_CS
 	rscreen->cs_count = 0;
-	if (rscreen->info.drm_minor >= 28) {
+	if (rscreen->info.drm_minor >= 28 && (rscreen->debug_flags & DBG_TRACE_CS)) {
 		rscreen->trace_bo = (struct r600_resource*)pipe_buffer_create(&rscreen->screen,
 										PIPE_BIND_CUSTOM,
 										PIPE_USAGE_STAGING,
@@ -1320,7 +1322,6 @@ struct pipe_screen *r600_screen_create(struct radeon_winsys *ws)
 									PIPE_TRANSFER_UNSYNCHRONIZED);
 		}
 	}
-#endif
 
 	/* Create the auxiliary context. */
 	pipe_mutex_init(rscreen->aux_context_lock);
diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h
index 50861a12bce..39fd82be142 100644
--- a/src/gallium/drivers/r600/r600_pipe.h
+++ b/src/gallium/drivers/r600/r600_pipe.h
@@ -37,8 +37,6 @@
 
 #define R600_NUM_ATOMS 41
 
-#define R600_TRACE_CS 0
-
 /* the number of CS dwords for flushing and drawing */
 #define R600_MAX_FLUSH_CS_DWORDS	16
 #define R600_MAX_DRAW_CS_DWORDS		34
@@ -245,7 +243,8 @@ typedef boolean (*r600g_dma_blit_t)(struct pipe_context *ctx,
 /* logging */
 #define DBG_TEX_DEPTH		(1 << 0)
 #define DBG_COMPUTE		(1 << 1)
-#define DBG_VM                  (1 << 2)
+#define DBG_VM			(1 << 2)
+#define DBG_TRACE_CS		(1 << 3)
 /* shaders */
 #define DBG_FS			(1 << 8)
 #define DBG_VS			(1 << 9)
@@ -284,11 +283,9 @@ struct r600_screen {
 	 * XXX: Not sure if this is the best place for global_pool.  Also,
 	 * it's not thread safe, so it won't work with multiple contexts. */
 	struct compute_memory_pool *global_pool;
-#if R600_TRACE_CS
 	struct r600_resource		*trace_bo;
 	uint32_t			*trace_ptr;
 	unsigned			cs_count;
-#endif
 	r600g_dma_blit_t		dma_blit;
 
 	/* Auxiliary context. Mainly used to initialize resources.
@@ -654,19 +651,15 @@ static INLINE void r600_emit_command_buffer(struct radeon_winsys_cs *cs,
 	cs->cdw += cb->num_dw;
 }
 
-#if R600_TRACE_CS
 void r600_trace_emit(struct r600_context *rctx);
-#endif
 
 static INLINE void r600_emit_atom(struct r600_context *rctx, struct r600_atom *atom)
 {
 	atom->emit(rctx, atom);
 	atom->dirty = false;
-#if R600_TRACE_CS
 	if (rctx->screen->trace_bo) {
 		r600_trace_emit(rctx);
 	}
-#endif
 }
 
 static INLINE void r600_set_cso_state(struct r600_cso_state *state, void *cso)
diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index 8190873ce7b..87a2e2e55e7 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -1488,11 +1488,9 @@ static void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info
 					(info.count_from_stream_output ? S_0287F0_USE_OPAQUE(1) : 0);
 	}
 
-#if R600_TRACE_CS
 	if (rctx->screen->trace_bo) {
 		r600_trace_emit(rctx);
 	}
-#endif
 
 	/* Set the depth buffer as dirty. */
 	if (rctx->framebuffer.state.zsbuf) {
@@ -1764,7 +1762,6 @@ void r600_init_common_state_functions(struct r600_context *rctx)
 	rctx->context.draw_vbo = r600_draw_vbo;
 }
 
-#if R600_TRACE_CS
 void r600_trace_emit(struct r600_context *rctx)
 {
 	struct r600_screen *rscreen = rctx->screen;
@@ -1782,4 +1779,3 @@ void r600_trace_emit(struct r600_context *rctx)
 	r600_write_value(cs, PKT3(PKT3_NOP, 0, 0));
 	r600_write_value(cs, reloc);
 }
-#endif