[Mesa-dev] [PATCH] i965: Use brw_bo_wait() for brw_bo_wait_rendering()

[Chris Wilson]

Currently, we use set_domain() to cause a stall on rendering. But the
set-domain ioctl has the side-effect of changing the kernel's cache
domain underneath the struct_mutex, which may perturb state if there was
no rendering to wait upon and in general is much heavier than the
lockless wait-ioctl. Historically libdrm used set-domain as we did not
have an explicit wait-ioctl (and the patches to teach it to use wait if
available were lost in the mists). Since mesa already depends upon a
kernel support the wait-ioctl, we do not need to supply a fallback.

Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
Cc: Kenneth Graunke <kenneth at whitecape.org>
---
 src/mesa/drivers/dri/i965/brw_bufmgr.c            | 8 +++++---
 src/mesa/drivers/dri/i965/brw_bufmgr.h            | 2 +-
 src/mesa/drivers/dri/i965/brw_context.c           | 2 +-
 src/mesa/drivers/dri/i965/brw_performance_query.c | 2 +-
 src/mesa/drivers/dri/i965/intel_batchbuffer.c     | 4 ++--
 5 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_bufmgr.c b/src/mesa/drivers/dri/i965/brw_bufmgr.c
index da12a13152..ee4a5cfa2c 100644
--- a/src/mesa/drivers/dri/i965/brw_bufmgr.c
+++ b/src/mesa/drivers/dri/i965/brw_bufmgr.c
@@ -831,10 +831,12 @@ brw_bo_get_subdata(struct brw_bo *bo, uint64_t offset,
 
 /** Waits for all GPU rendering with the object to have completed. */
 void
-brw_bo_wait_rendering(struct brw_context *brw, struct brw_bo *bo)
+brw_bo_wait_rendering(struct brw_bo *bo)
 {
-   set_domain(brw, "waiting for",
-              bo, I915_GEM_DOMAIN_GTT, I915_GEM_DOMAIN_GTT);
+   /* We require a kernel recent enough for WAIT_IOCTL support.
+    * See intel_init_bufmgr()
+    */
+   brw_bo_wait(bo, -1);
 }
 
 /**
diff --git a/src/mesa/drivers/dri/i965/brw_bufmgr.h b/src/mesa/drivers/dri/i965/brw_bufmgr.h
index 4d671b6aae..80c71825e8 100644
--- a/src/mesa/drivers/dri/i965/brw_bufmgr.h
+++ b/src/mesa/drivers/dri/i965/brw_bufmgr.h
@@ -227,7 +227,7 @@ int brw_bo_get_subdata(struct brw_bo *bo, uint64_t offset,
  * bo_subdata, etc.  It is merely a way for the driver to implement
  * glFinish.
  */
-void brw_bo_wait_rendering(struct brw_context *brw, struct brw_bo *bo);
+void brw_bo_wait_rendering(struct brw_bo *bo);
 
 /**
  * Tears down the buffer manager instance.
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index 0b3fdc6842..8a3ffab443 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -256,7 +256,7 @@ intel_finish(struct gl_context * ctx)
    intel_glFlush(ctx);
 
    if (brw->batch.last_bo)
-      brw_bo_wait_rendering(brw, brw->batch.last_bo);
+      brw_bo_wait_rendering(brw->batch.last_bo);
 }
 
 static void
diff --git a/src/mesa/drivers/dri/i965/brw_performance_query.c b/src/mesa/drivers/dri/i965/brw_performance_query.c
index 81389dbd3e..e4e1854bf2 100644
--- a/src/mesa/drivers/dri/i965/brw_performance_query.c
+++ b/src/mesa/drivers/dri/i965/brw_performance_query.c
@@ -1350,7 +1350,7 @@ brw_wait_perf_query(struct gl_context *ctx, struct gl_perf_query_object *o)
    if (brw_batch_references(&brw->batch, bo))
       intel_batchbuffer_flush(brw);
 
-   brw_bo_wait_rendering(brw, bo);
+   brw_bo_wait_rendering(bo);
 
    /* Due to a race condition between the OA unit signaling report
     * availability and the report actually being written into memory,
diff --git a/src/mesa/drivers/dri/i965/intel_batchbuffer.c b/src/mesa/drivers/dri/i965/intel_batchbuffer.c
index 62d2fe8ef3..28c2f474c0 100644
--- a/src/mesa/drivers/dri/i965/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/i965/intel_batchbuffer.c
@@ -497,7 +497,7 @@ throttle(struct brw_context *brw)
             /* Pass NULL rather than brw so we avoid perf_debug warnings;
              * stalling is common and expected here...
              */
-            brw_bo_wait_rendering(NULL, brw->throttle_batch[1]);
+            brw_bo_wait_rendering(brw->throttle_batch[1]);
          }
          brw_bo_unreference(brw->throttle_batch[1]);
       }
@@ -723,7 +723,7 @@ _intel_batchbuffer_flush_fence(struct brw_context *brw,
 
    if (unlikely(INTEL_DEBUG & DEBUG_SYNC)) {
       fprintf(stderr, "waiting for idle\n");
-      brw_bo_wait_rendering(brw, brw->batch.bo);
+      brw_bo_wait_rendering(brw->batch.bo);
    }
 
    /* Start a new batch buffer. */
-- 
2.13.2