gallium/radeon: remove old CS tracing

Cons:
- it was only integrated in r600g
- it doesn't work with GPUVM
- it records buffer contents at the end of IBs instead of at the beginning,
  so the replay isn't exact
- it lacks an IB parser and user-friendliness

A better solution is apitrace in combination with gallium/ddebug, which
has a complete IB parser and can pinpoint hanging CP packets.

Reviewed-by: Nicolai Hähnle <[email protected]>

13 files changed, 19 insertions, 84 deletions

diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 6fa892089ec..d100a9df55b 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -385,7 +385,7 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     if (!r300->ctx)
         goto fail;
 
-    r300->cs = rws->cs_create(r300->ctx, RING_GFX, r300_flush_callback, r300, NULL);
+    r300->cs = rws->cs_create(r300->ctx, RING_GFX, r300_flush_callback, r300);
     if (r300->cs == NULL)
         goto fail;
 
diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c
index 7a75b43a53e..63182cba2b2 100644
--- a/src/gallium/drivers/r300/r300_flush.c
+++ b/src/gallium/drivers/r300/r300_flush.c
@@ -53,7 +53,7 @@ static void r300_flush_and_cleanup(struct r300_context *r300, unsigned flags,
     }
 
     r300->flush_counter++;
-    r300->rws->cs_flush(r300->cs, flags, fence, 0);
+    r300->rws->cs_flush(r300->cs, flags, fence);
     r300->dirty_hw = 0;
 
     /* New kitchen sink, baby. */
@@ -88,11 +88,11 @@ void r300_flush(struct pipe_context *pipe,
              * and we cannot emit an empty CS. Let's write to some reg. */
             CS_LOCALS(r300);
             OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0);
-            r300->rws->cs_flush(r300->cs, flags, fence, 0);
+            r300->rws->cs_flush(r300->cs, flags, fence);
         } else {
             /* Even if hw is not dirty, we should at least reset the CS in case
              * the space checking failed for the first draw operation. */
-            r300->rws->cs_flush(r300->cs, flags, NULL, 0);
+            r300->rws->cs_flush(r300->cs, flags, NULL);
         }
     }
 
diff --git a/src/gallium/drivers/r600/r600_hw_context.c b/src/gallium/drivers/r600/r600_hw_context.c
index 4951297df42..7a6f957945b 100644
--- a/src/gallium/drivers/r600/r600_hw_context.c
+++ b/src/gallium/drivers/r600/r600_hw_context.c
@@ -57,18 +57,11 @@ void r600_need_cs_space(struct r600_context *ctx, unsigned num_dw,
 
 		/* The number of dwords all the dirty states would take. */
 		mask = ctx->dirty_atoms;
-		while (mask != 0) {
+		while (mask != 0)
 			num_dw += ctx->atoms[u_bit_scan64(&mask)]->num_dw;
-			if (ctx->screen->b.trace_bo) {
-				num_dw += R600_TRACE_CS_DWORDS;
-			}
-		}
 
 		/* 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 (ctx->screen->b.trace_bo) {
-			num_dw += R600_TRACE_CS_DWORDS;
-		}
 	}
 
 	/* Count in queries_suspend. */
@@ -273,7 +266,7 @@ void r600_context_gfx_flush(void *context, unsigned flags,
 	flags |= RADEON_FLUSH_KEEP_TILING_FLAGS;
 
 	/* Flush the CS. */
-	ctx->b.ws->cs_flush(cs, flags, fence, ctx->screen->b.cs_count++);
+	ctx->b.ws->cs_flush(cs, flags, fence);
 
 	r600_begin_new_cs(ctx);
 }
diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c
index 7018088d204..88c500a162a 100644
--- a/src/gallium/drivers/r600/r600_pipe.c
+++ b/src/gallium/drivers/r600/r600_pipe.c
@@ -187,9 +187,7 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen,
 	}
 
 	rctx->b.gfx.cs = ws->cs_create(rctx->b.ctx, RING_GFX,
-				       r600_context_gfx_flush, rctx,
-				       rscreen->b.trace_bo ?
-					       rscreen->b.trace_bo->buf : NULL);
+				       r600_context_gfx_flush, rctx);
 	rctx->b.gfx.flush = r600_context_gfx_flush;
 
 	rctx->allocator_fetch_shader = u_suballocator_create(&rctx->b.b, 64 * 1024, 256,
diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h
index f8a20398355..72aa64233a9 100644
--- a/src/gallium/drivers/r600/r600_pipe.h
+++ b/src/gallium/drivers/r600/r600_pipe.h
@@ -60,7 +60,6 @@
 /* the number of CS dwords for flushing and drawing */
 #define R600_MAX_FLUSH_CS_DWORDS	16
 #define R600_MAX_DRAW_CS_DWORDS		58
-#define R600_TRACE_CS_DWORDS		7
 
 #define R600_MAX_USER_CONST_BUFFERS 13
 #define R600_MAX_DRIVER_CONST_BUFFERS 3
@@ -571,15 +570,10 @@ static inline void r600_mark_atom_dirty(struct r600_context *rctx,
 	r600_set_atom_dirty(rctx, atom, true);
 }
 
-void r600_trace_emit(struct r600_context *rctx);
-
 static inline void r600_emit_atom(struct r600_context *rctx, struct r600_atom *atom)
 {
 	atom->emit(&rctx->b, atom);
 	r600_set_atom_dirty(rctx, atom, false);
-	if (rctx->screen->b.trace_bo) {
-		r600_trace_emit(rctx);
-	}
 }
 
 static inline void r600_set_cso_state(struct r600_context *rctx,
diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index 2211e07ceba..df41d3f028d 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -2029,10 +2029,6 @@ static void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info
 		cs->buf[cs->cdw++] = EVENT_TYPE(EVENT_TYPE_SQ_NON_EVENT);
 	}
 
-	if (rctx->screen->b.trace_bo) {
-		r600_trace_emit(rctx);
-	}
-
 	/* Set the depth buffer as dirty. */
 	if (rctx->framebuffer.state.zsbuf) {
 		struct pipe_surface *surf = rctx->framebuffer.state.zsbuf;
@@ -2927,22 +2923,3 @@ void r600_init_common_state_functions(struct r600_context *rctx)
 	rctx->b.set_occlusion_query_state = r600_set_occlusion_query_state;
 	rctx->b.need_gfx_cs_space = r600_need_gfx_cs_space;
 }
-
-void r600_trace_emit(struct r600_context *rctx)
-{
-	struct r600_screen *rscreen = rctx->screen;
-	struct radeon_winsys_cs *cs = rctx->b.gfx.cs;
-	uint64_t va;
-	uint32_t reloc;
-
-	va = rscreen->b.trace_bo->gpu_address;
-	reloc = radeon_add_to_buffer_list(&rctx->b, &rctx->b.gfx, rscreen->b.trace_bo,
-				      RADEON_USAGE_READWRITE, RADEON_PRIO_TRACE);
-	radeon_emit(cs, PKT3(PKT3_MEM_WRITE, 3, 0));
-	radeon_emit(cs, va & 0xFFFFFFFFUL);
-	radeon_emit(cs, (va >> 32UL) & 0xFFUL);
-	radeon_emit(cs, cs->cdw);
-	radeon_emit(cs, rscreen->b.cs_count);
-	radeon_emit(cs, PKT3(PKT3_NOP, 0, 0));
-	radeon_emit(cs, reloc);
-}
diff --git a/src/gallium/drivers/radeon/r600_pipe_common.c b/src/gallium/drivers/radeon/r600_pipe_common.c
index ea028272ccd..eed9d83ee49 100644
--- a/src/gallium/drivers/radeon/r600_pipe_common.c
+++ b/src/gallium/drivers/radeon/r600_pipe_common.c
@@ -229,7 +229,7 @@ static void r600_flush_dma_ring(void *ctx, unsigned flags,
 	struct radeon_winsys_cs *cs = rctx->dma.cs;
 
 	if (cs->cdw)
-		rctx->ws->cs_flush(cs, flags, &rctx->last_sdma_fence, 0);
+		rctx->ws->cs_flush(cs, flags, &rctx->last_sdma_fence);
 	if (fence)
 		rctx->ws->fence_reference(fence, rctx->last_sdma_fence);
 }
@@ -318,7 +318,7 @@ bool r600_common_context_init(struct r600_common_context *rctx,
 	if (rscreen->info.has_sdma && !(rscreen->debug_flags & DBG_NO_ASYNC_DMA)) {
 		rctx->dma.cs = rctx->ws->cs_create(rctx->ctx, RING_DMA,
 						   r600_flush_dma_ring,
-						   rctx, NULL);
+						   rctx);
 		rctx->dma.flush = r600_flush_dma_ring;
 	}
 
@@ -379,7 +379,6 @@ static const struct debug_named_value common_debug_options[] = {
 	{ "tex", DBG_TEX, "Print texture 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" },
 	{ "info", DBG_INFO, "Print driver information" },
 
 	/* shaders */
@@ -893,19 +892,6 @@ bool r600_common_screen_init(struct r600_common_screen *rscreen,
 	pipe_mutex_init(rscreen->aux_context_lock);
 	pipe_mutex_init(rscreen->gpu_load_mutex);
 
-	if (((rscreen->info.drm_major == 2 && rscreen->info.drm_minor >= 28) ||
-	     rscreen->info.drm_major == 3) &&
-	    (rscreen->debug_flags & DBG_TRACE_CS)) {
-		rscreen->trace_bo = (struct r600_resource*)pipe_buffer_create(&rscreen->b,
-										PIPE_BIND_CUSTOM,
-										PIPE_USAGE_STAGING,
-										4096);
-		if (rscreen->trace_bo) {
-			rscreen->trace_ptr = rscreen->ws->buffer_map(rscreen->trace_bo->buf, NULL,
-									PIPE_TRANSFER_UNSYNCHRONIZED);
-		}
-	}
-
 	if (rscreen->debug_flags & DBG_INFO) {
 		printf("pci_id = 0x%x\n", rscreen->info.pci_id);
 		printf("family = %i (%s)\n", rscreen->info.family,
@@ -951,9 +937,6 @@ void r600_destroy_common_screen(struct r600_common_screen *rscreen)
 	pipe_mutex_destroy(rscreen->aux_context_lock);
 	rscreen->aux_context->destroy(rscreen->aux_context);
 
-	if (rscreen->trace_bo)
-		pipe_resource_reference((struct pipe_resource**)&rscreen->trace_bo, NULL);
-
 	rscreen->ws->destroy(rscreen->ws);
 	FREE(rscreen);
 }
diff --git a/src/gallium/drivers/radeon/r600_pipe_common.h b/src/gallium/drivers/radeon/r600_pipe_common.h
index cf8dcf7ea88..a9de71a8734 100644
--- a/src/gallium/drivers/radeon/r600_pipe_common.h
+++ b/src/gallium/drivers/radeon/r600_pipe_common.h
@@ -61,7 +61,7 @@
 /* gap - reuse */
 #define DBG_COMPUTE		(1 << 2)
 #define DBG_VM			(1 << 3)
-#define DBG_TRACE_CS		(1 << 4)
+/* gap - reuse */
 /* shader logging */
 #define DBG_FS			(1 << 5)
 #define DBG_VS			(1 << 6)
@@ -303,10 +303,6 @@ struct r600_common_screen {
 	struct pipe_context		*aux_context;
 	pipe_mutex			aux_context_lock;
 
-	struct r600_resource		*trace_bo;
-	uint32_t			*trace_ptr;
-	unsigned			cs_count;
-
 	/* This must be in the screen, because UE4 uses one context for
 	 * compilation and another one for rendering.
 	 */
diff --git a/src/gallium/drivers/radeon/radeon_uvd.c b/src/gallium/drivers/radeon/radeon_uvd.c
index b8efc58eaab..233f46091a4 100644
--- a/src/gallium/drivers/radeon/radeon_uvd.c
+++ b/src/gallium/drivers/radeon/radeon_uvd.c
@@ -92,7 +92,7 @@ struct ruvd_decoder {
 /* flush IB to the hardware */
 static void flush(struct ruvd_decoder *dec)
 {
-	dec->ws->cs_flush(dec->cs, RADEON_FLUSH_ASYNC, NULL, 0);
+	dec->ws->cs_flush(dec->cs, RADEON_FLUSH_ASYNC, NULL);
 }
 
 /* add a new set register command to the IB */
@@ -1142,7 +1142,7 @@ struct pipe_video_codec *ruvd_create_decoder(struct pipe_context *context,
 	dec->stream_handle = rvid_alloc_stream_handle();
 	dec->screen = context->screen;
 	dec->ws = ws;
-	dec->cs = ws->cs_create(rctx->ctx, RING_UVD, NULL, NULL, NULL);
+	dec->cs = ws->cs_create(rctx->ctx, RING_UVD, NULL, NULL);
 	if (!dec->cs) {
 		RVID_ERR("Can't get command submission context.\n");
 		goto error;
diff --git a/src/gallium/drivers/radeon/radeon_vce.c b/src/gallium/drivers/radeon/radeon_vce.c
index 087d9422c04..2ab74e9eb6c 100644
--- a/src/gallium/drivers/radeon/radeon_vce.c
+++ b/src/gallium/drivers/radeon/radeon_vce.c
@@ -56,7 +56,7 @@
  */
 static void flush(struct rvce_encoder *enc)
 {
-	enc->ws->cs_flush(enc->cs, RADEON_FLUSH_ASYNC, NULL, 0);
+	enc->ws->cs_flush(enc->cs, RADEON_FLUSH_ASYNC, NULL);
 	enc->task_info_idx = 0;
 	enc->bs_idx = 0;
 }
@@ -429,7 +429,7 @@ struct pipe_video_codec *rvce_create_encoder(struct pipe_context *context,
 
 	enc->screen = context->screen;
 	enc->ws = ws;
-	enc->cs = ws->cs_create(rctx->ctx, RING_VCE, rvce_cs_flush, enc, NULL);
+	enc->cs = ws->cs_create(rctx->ctx, RING_VCE, rvce_cs_flush, enc);
 	if (!enc->cs) {
 		RVID_ERR("Can't get command submission context.\n");
 		goto error;
diff --git a/src/gallium/drivers/radeon/radeon_winsys.h b/src/gallium/drivers/radeon/radeon_winsys.h
index daa15db2812..d35e963133e 100644
--- a/src/gallium/drivers/radeon/radeon_winsys.h
+++ b/src/gallium/drivers/radeon/radeon_winsys.h
@@ -593,14 +593,12 @@ struct radeon_winsys {
      * \param ring_type The ring type (GFX, DMA, UVD)
      * \param flush     Flush callback function associated with the command stream.
      * \param user      User pointer that will be passed to the flush callback.
-     * \param trace_buf Trace buffer when tracing is enabled
      */
     struct radeon_winsys_cs *(*cs_create)(struct radeon_winsys_ctx *ctx,
                                           enum ring_type ring_type,
                                           void (*flush)(void *ctx, unsigned flags,
 							struct pipe_fence_handle **fence),
-                                          void *flush_ctx,
-                                          struct pb_buffer *trace_buf);
+                                          void *flush_ctx);
 
     /**
      * Destroy a command stream.
@@ -673,12 +671,10 @@ struct radeon_winsys {
      * \param flags,      RADEON_FLUSH_ASYNC or 0.
      * \param fence       Pointer to a fence. If non-NULL, a fence is inserted
      *                    after the CS and is returned through this parameter.
-     * \param cs_trace_id A unique identifier of the cs, used for tracing.
      */
     void (*cs_flush)(struct radeon_winsys_cs *cs,
                      unsigned flags,
-                     struct pipe_fence_handle **fence,
-                     uint32_t cs_trace_id);
+                     struct pipe_fence_handle **fence);
 
     /**
      * Return TRUE if a buffer is referenced by a command stream.
diff --git a/src/gallium/drivers/radeonsi/si_hw_context.c b/src/gallium/drivers/radeonsi/si_hw_context.c
index b5a4034cc12..8c900a4ecb6 100644
--- a/src/gallium/drivers/radeonsi/si_hw_context.c
+++ b/src/gallium/drivers/radeonsi/si_hw_context.c
@@ -118,8 +118,7 @@ void si_context_gfx_flush(void *context, unsigned flags,
 	}
 
 	/* Flush the CS. */
-	ws->cs_flush(cs, flags, &ctx->last_gfx_fence,
-		     ctx->screen->b.cs_count++);
+	ws->cs_flush(cs, flags, &ctx->last_gfx_fence);
 
 	if (fence)
 		ws->fence_reference(fence, ctx->last_gfx_fence);
diff --git a/src/gallium/drivers/radeonsi/si_pipe.c b/src/gallium/drivers/radeonsi/si_pipe.c
index 8b50a49cba0..042cfc764fd 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.c
+++ b/src/gallium/drivers/radeonsi/si_pipe.c
@@ -140,9 +140,8 @@ static struct pipe_context *si_create_context(struct pipe_screen *screen,
 		sctx->b.b.create_video_buffer = vl_video_buffer_create;
 	}
 
-	sctx->b.gfx.cs = ws->cs_create(sctx->b.ctx, RING_GFX, si_context_gfx_flush,
-				       sctx, sscreen->b.trace_bo ?
-					       sscreen->b.trace_bo->buf : NULL);
+	sctx->b.gfx.cs = ws->cs_create(sctx->b.ctx, RING_GFX,
+				       si_context_gfx_flush, sctx);
 	sctx->b.gfx.flush = si_context_gfx_flush;
 
 	/* Border colors. */