drm/i915: Record the position of the workarounds in the tail of the request

Rather than blindly assuming we need to advance the tail for
resubmitting the request via the ELSP, record the position.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-3-chris@chris-wilson.co.uk

2 files changed, 11 insertions, 8 deletions

diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index 91014de8bfbc..2faa3bb4c39b 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -88,20 +88,23 @@ struct drm_i915_gem_request {
          */
         u32 previous_seqno;
 
-        /** Position in the ringbuffer of the start of the request */
+        /** Position in the ring of the start of the request */
         u32 head;
 
         /**
-         * Position in the ringbuffer of the start of the postfix.
-         * This is required to calculate the maximum available ringbuffer
-         * space without overwriting the postfix.
+         * Position in the ring of the start of the postfix.
+         * This is required to calculate the maximum available ring space
+         * without overwriting the postfix.
          */
         u32 postfix;
 
-        /** Position in the ringbuffer of the end of the whole request */
+        /** Position in the ring of the end of the whole request */
         u32 tail;
 
-        /** Preallocate space in the ringbuffer for the emitting the request */
+        /** Position in the ring of the end of any workarounds after the tail */
+        u32 wa_tail;
+
+        /** Preallocate space in the ring for the emitting the request */
         u32 reserved_space;
 
         /**
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 9bfe304c5256..d7fa9b3a55c3 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -482,8 +482,7 @@ static void execlists_unqueue(struct intel_engine_cs *engine)
                  * resubmit the request. See gen8_emit_request() for where we
                  * prepare the padding after the end of the request.
                  */
-                req0->tail += 8;
-                req0->tail &= req0->ring->size - 1;
+                req0->tail = req0->wa_tail;
         }
 
         execlists_elsp_submit_contexts(req0, req1);
@@ -711,6 +710,7 @@ intel_logical_ring_advance(struct drm_i915_gem_request *request)
         intel_ring_emit(ring, MI_NOOP);
         intel_ring_emit(ring, MI_NOOP);
         intel_ring_advance(ring);
+        request->wa_tail = ring->tail;
 
         /* We keep the previous context alive until we retire the following
          * request. This ensures that any the context object is still pinned