util/u_trace/perfetto: add new env variable to enable perfetto

When using the Vulkan API, command buffers can be recorded way before
perfetto is enabled. This can be problematic if you want already
recorded command buffers to produce traces.

This new environment variable makes perfetto enabled internally so
that command buffers are recorded with timestamps, even though no
perfetto recording happens.

v2: rename to GPU_TRACE_INSTRUMENT (Rob)

v3: Move instrumentation check to generated headers (Danylo)
    Decouple instrumentation enabling from tracing (Danylo)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13911>

docs/perfetto.rst

@@ -104,6 +104,22 @@ the steps above :
 
    # Back in the perfetto tmux, press enter to start the capture
 
+Vulkan data sources
+~~~~~~~~~~~~~~~~~~~
+
+The Vulkan API gives the application control over recording of command
+buffers as well as when they are submitted to the hardware. As a
+consequence, we need to ensure command buffers are properly
+instrumented for the perfetto driver data sources prior to Perfetto
+actually collecting traces.
+
+This can be achieved by setting the ``GPU_TRACE_INSTRUMENT``
+environment variable before starting a Vulkan application :
+
+.. code-block:: console
+
+   GPU_TRACE_INSTRUMENT=1 ./build/my_vulkan_app
+
 Driver Specifics
 ~~~~~~~~~~~~~~~~
 

src/freedreno/vulkan/tu_drm.c

@@ -950,7 +950,7 @@ tu_queue_submit_create_locked(struct tu_queue *queue,
       }
    }
 
-   bool u_trace_enabled = u_trace_context_tracing(&queue->device->trace_context);
+   bool u_trace_enabled = u_trace_context_actively_tracing(&queue->device->trace_context);
    bool has_trace_points = false;
 
    uint32_t entry_count = 0;

src/util/perf/u_trace.c

@@ -38,6 +38,8 @@
 #define TIMESTAMP_BUF_SIZE 0x1000
 #define TRACES_PER_CHUNK   (TIMESTAMP_BUF_SIZE / sizeof(uint64_t))
 
+bool ut_trace_instrument;
+
 #ifdef HAVE_PERFETTO
 int ut_perfetto_enabled;
 
@@ -217,6 +219,7 @@ get_chunk(struct u_trace *ut, size_t payload_size)
    return chunk;
 }
 
+DEBUG_GET_ONCE_BOOL_OPTION(trace_instrument, "GPU_TRACE_INSTRUMENT", false)
 DEBUG_GET_ONCE_BOOL_OPTION(trace, "GPU_TRACE", false)
 DEBUG_GET_ONCE_FILE_OPTION(trace_file, "GPU_TRACEFILE", NULL, "w")
 
@@ -232,6 +235,8 @@ get_tracefile(void)
          tracefile = stdout;
       }
 
+      ut_trace_instrument = debug_get_option_trace_instrument();
+
       firsttime = false;
    }
 
@@ -281,7 +286,7 @@ u_trace_context_init(struct u_trace_context *utctx,
    list_add(&utctx->node, &ctx_list);
 #endif
 
-   if (!u_trace_context_tracing(utctx))
+   if (!u_trace_context_actively_tracing(utctx))
       return;
 
    queue_init(utctx);
@@ -425,7 +430,7 @@ u_trace_init(struct u_trace *ut, struct u_trace_context *utctx)
 {
    ut->utctx = utctx;
    list_inithead(&ut->trace_chunks);
-   ut->enabled = u_trace_context_tracing(utctx);
+   ut->enabled = u_trace_context_instrumenting(utctx);
 }
 
 void

src/util/perf/u_trace.h

@@ -267,6 +267,12 @@ void u_trace_disable_event_range(struct u_trace_iterator begin_it,
  */
 void u_trace_flush(struct u_trace *ut, void *flush_data, bool free_data);
 
+/**
+ * Whether command buffers should be instrumented even if not collecting
+ * traces.
+ */
+extern bool ut_trace_instrument;
+
 #ifdef HAVE_PERFETTO
 extern int ut_perfetto_enabled;
 
@@ -277,11 +283,17 @@ void u_trace_perfetto_stop(void);
 #endif
 
 static inline bool
-u_trace_context_tracing(struct u_trace_context *utctx)
+u_trace_context_actively_tracing(struct u_trace_context *utctx)
 {
    return !!utctx->out || (ut_perfetto_enabled > 0);
 }
 
+static inline bool
+u_trace_context_instrumenting(struct u_trace_context *utctx)
+{
+   return !!utctx->out || ut_trace_instrument || (ut_perfetto_enabled > 0);
+}
+
 #ifdef  __cplusplus
 }
 #endif

src/util/perf/u_trace.py

@@ -212,9 +212,9 @@ static inline void trace_${trace_name}(struct u_trace *ut, void *cs
 %    endfor
 ) {
 %    if trace.tp_perfetto is not None:
-   if (!unlikely(ut->enabled || ut_perfetto_enabled))
+   if (!unlikely(ut->enabled || ut_trace_instrument || ut_perfetto_enabled))
 %    else:
-   if (!unlikely(ut->enabled))
+   if (!unlikely(ut->enabled || ut_trace_instrument))
 %    endif
       return;
    __trace_${trace_name}(ut, cs