1 files changed, 88 insertions, 36 deletions

diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index d0ef50bf930a..fbeaec3994e7 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -69,19 +69,28 @@ unsigned int intel_ring_update_space(struct intel_ring *ring)
 static int
 gen2_render_ring_flush(struct i915_request *rq, u32 mode)
 {
+	unsigned int num_store_dw;
 	u32 cmd, *cs;
 
 	cmd = MI_FLUSH;
-
+	num_store_dw = 0;
 	if (mode & EMIT_INVALIDATE)
 		cmd |= MI_READ_FLUSH;
+	if (mode & EMIT_FLUSH)
+		num_store_dw = 4;
 
-	cs = intel_ring_begin(rq, 2);
+	cs = intel_ring_begin(rq, 2 + 3 * num_store_dw);
 	if (IS_ERR(cs))
 		return PTR_ERR(cs);
 
 	*cs++ = cmd;
-	*cs++ = MI_NOOP;
+	while (num_store_dw--) {
+		*cs++ = MI_STORE_DWORD_IMM | MI_MEM_VIRTUAL;
+		*cs++ = i915_scratch_offset(rq->i915);
+		*cs++ = 0;
+	}
+	*cs++ = MI_FLUSH | MI_NO_WRITE_FLUSH;
+
 	intel_ring_advance(rq, cs);
 
 	return 0;
@@ -91,6 +100,7 @@ static int
 gen4_render_ring_flush(struct i915_request *rq, u32 mode)
 {
 	u32 cmd, *cs;
+	int i;
 
 	/*
 	 * read/write caches:
@@ -127,12 +137,43 @@ gen4_render_ring_flush(struct i915_request *rq, u32 mode)
 			cmd |= MI_INVALIDATE_ISP;
 	}
 
-	cs = intel_ring_begin(rq, 2);
+	i = 2;
+	if (mode & EMIT_INVALIDATE)
+		i += 20;
+
+	cs = intel_ring_begin(rq, i);
 	if (IS_ERR(cs))
 		return PTR_ERR(cs);
 
 	*cs++ = cmd;
-	*cs++ = MI_NOOP;
+
+	/*
+	 * A random delay to let the CS invalidate take effect? Without this
+	 * delay, the GPU relocation path fails as the CS does not see
+	 * the updated contents. Just as important, if we apply the flushes
+	 * to the EMIT_FLUSH branch (i.e. immediately after the relocation
+	 * write and before the invalidate on the next batch), the relocations
+	 * still fail. This implies that is a delay following invalidation
+	 * that is required to reset the caches as opposed to a delay to
+	 * ensure the memory is written.
+	 */
+	if (mode & EMIT_INVALIDATE) {
+		*cs++ = GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE;
+		*cs++ = i915_scratch_offset(rq->i915) | PIPE_CONTROL_GLOBAL_GTT;
+		*cs++ = 0;
+		*cs++ = 0;
+
+		for (i = 0; i < 12; i++)
+			*cs++ = MI_FLUSH;
+
+		*cs++ = GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE;
+		*cs++ = i915_scratch_offset(rq->i915) | PIPE_CONTROL_GLOBAL_GTT;
+		*cs++ = 0;
+		*cs++ = 0;
+	}
+
+	*cs++ = cmd;
+
 	intel_ring_advance(rq, cs);
 
 	return 0;
@@ -178,8 +219,7 @@ gen4_render_ring_flush(struct i915_request *rq, u32 mode)
 static int
 intel_emit_post_sync_nonzero_flush(struct i915_request *rq)
 {
-	u32 scratch_addr =
-		i915_ggtt_offset(rq->engine->scratch) + 2 * CACHELINE_BYTES;
+	u32 scratch_addr = i915_scratch_offset(rq->i915) + 2 * CACHELINE_BYTES;
 	u32 *cs;
 
 	cs = intel_ring_begin(rq, 6);
@@ -212,8 +252,7 @@ intel_emit_post_sync_nonzero_flush(struct i915_request *rq)
 static int
 gen6_render_ring_flush(struct i915_request *rq, u32 mode)
 {
-	u32 scratch_addr =
-		i915_ggtt_offset(rq->engine->scratch) + 2 * CACHELINE_BYTES;
+	u32 scratch_addr = i915_scratch_offset(rq->i915) + 2 * CACHELINE_BYTES;
 	u32 *cs, flags = 0;
 	int ret;
 
@@ -282,8 +321,7 @@ gen7_render_ring_cs_stall_wa(struct i915_request *rq)
 static int
 gen7_render_ring_flush(struct i915_request *rq, u32 mode)
 {
-	u32 scratch_addr =
-		i915_ggtt_offset(rq->engine->scratch) + 2 * CACHELINE_BYTES;
+	u32 scratch_addr = i915_scratch_offset(rq->i915) + 2 * CACHELINE_BYTES;
 	u32 *cs, flags = 0;
 
 	/*
@@ -495,6 +533,13 @@ static int init_ring_common(struct intel_engine_cs *engine)
 
 	intel_engine_reset_breadcrumbs(engine);
 
+	if (HAS_LEGACY_SEMAPHORES(engine->i915)) {
+		I915_WRITE(RING_SYNC_0(engine->mmio_base), 0);
+		I915_WRITE(RING_SYNC_1(engine->mmio_base), 0);
+		if (HAS_VEBOX(dev_priv))
+			I915_WRITE(RING_SYNC_2(engine->mmio_base), 0);
+	}
+
 	/* Enforce ordering by reading HEAD register back */
 	I915_READ_HEAD(engine);
 
@@ -512,10 +557,11 @@ static int init_ring_common(struct intel_engine_cs *engine)
 	/* Check that the ring offsets point within the ring! */
 	GEM_BUG_ON(!intel_ring_offset_valid(ring, ring->head));
 	GEM_BUG_ON(!intel_ring_offset_valid(ring, ring->tail));
-
 	intel_ring_update_space(ring);
+
+	/* First wake the ring up to an empty/idle ring */
 	I915_WRITE_HEAD(engine, ring->head);
-	I915_WRITE_TAIL(engine, ring->tail);
+	I915_WRITE_TAIL(engine, ring->head);
 	(void)I915_READ_TAIL(engine);
 
 	I915_WRITE_CTL(engine, RING_CTL_SIZE(ring->size) | RING_VALID);
@@ -540,6 +586,12 @@ static int init_ring_common(struct intel_engine_cs *engine)
 	if (INTEL_GEN(dev_priv) > 2)
 		I915_WRITE_MODE(engine, _MASKED_BIT_DISABLE(STOP_RING));
 
+	/* Now awake, let it get started */
+	if (ring->tail != ring->head) {
+		I915_WRITE_TAIL(engine, ring->tail);
+		(void)I915_READ_TAIL(engine);
+	}
+
 	/* Papering over lost _interrupts_ immediately following the restart */
 	intel_engine_wakeup(engine);
 out:
@@ -574,7 +626,9 @@ static void skip_request(struct i915_request *rq)
 
 static void reset_ring(struct intel_engine_cs *engine, struct i915_request *rq)
 {
-	GEM_TRACE("%s seqno=%x\n", engine->name, rq ? rq->global_seqno : 0);
+	GEM_TRACE("%s request global=%d, current=%d\n",
+		  engine->name, rq ? rq->global_seqno : 0,
+		  intel_engine_get_seqno(engine));
 
 	/*
 	 * Try to restore the logical GPU state to match the continuation
@@ -606,7 +660,7 @@ static int intel_rcs_ctx_init(struct i915_request *rq)
 {
 	int ret;
 
-	ret = intel_ctx_workarounds_emit(rq);
+	ret = intel_engine_emit_ctx_wa(rq);
 	if (ret != 0)
 		return ret;
 
@@ -624,8 +678,6 @@ static int init_render_ring(struct intel_engine_cs *engine)
 	if (ret)
 		return ret;
 
-	intel_whitelist_workarounds_apply(engine);
-
 	/* WaTimedSingleVertexDispatch:cl,bw,ctg,elk,ilk,snb */
 	if (IS_GEN(dev_priv, 4, 6))
 		I915_WRITE(MI_MODE, _MASKED_BIT_ENABLE(VS_TIMER_DISPATCH));
@@ -707,9 +759,18 @@ static void cancel_requests(struct intel_engine_cs *engine)
 	/* Mark all submitted requests as skipped. */
 	list_for_each_entry(request, &engine->timeline.requests, link) {
 		GEM_BUG_ON(!request->global_seqno);
-		if (!i915_request_completed(request))
-			dma_fence_set_error(&request->fence, -EIO);
+
+		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT,
+			     &request->fence.flags))
+			continue;
+
+		dma_fence_set_error(&request->fence, -EIO);
 	}
+
+	intel_write_status_page(engine,
+				I915_GEM_HWS_INDEX,
+				intel_engine_last_submit(engine));
+
 	/* Remaining _unready_ requests will be nop'ed when submitted */
 
 	spin_unlock_irqrestore(&engine->timeline.lock, flags);
@@ -937,7 +998,7 @@ i965_emit_bb_start(struct i915_request *rq,
 }
 
 /* Just userspace ABI convention to limit the wa batch bo to a resonable size */
-#define I830_BATCH_LIMIT (256*1024)
+#define I830_BATCH_LIMIT SZ_256K
 #define I830_TLB_ENTRIES (2)
 #define I830_WA_SIZE max(I830_TLB_ENTRIES*4096, I830_BATCH_LIMIT)
 static int
@@ -945,7 +1006,9 @@ i830_emit_bb_start(struct i915_request *rq,
 		   u64 offset, u32 len,
 		   unsigned int dispatch_flags)
 {
-	u32 *cs, cs_offset = i915_ggtt_offset(rq->engine->scratch);
+	u32 *cs, cs_offset = i915_scratch_offset(rq->i915);
+
+	GEM_BUG_ON(rq->i915->gt.scratch->size < I830_WA_SIZE);
 
 	cs = intel_ring_begin(rq, 6);
 	if (IS_ERR(cs))
@@ -1021,8 +1084,7 @@ i915_emit_bb_start(struct i915_request *rq,
 int intel_ring_pin(struct intel_ring *ring)
 {
 	struct i915_vma *vma = ring->vma;
-	enum i915_map_type map =
-		HAS_LLC(vma->vm->i915) ? I915_MAP_WB : I915_MAP_WC;
+	enum i915_map_type map = i915_coherent_map_type(vma->vm->i915);
 	unsigned int flags;
 	void *addr;
 	int ret;
@@ -1403,7 +1465,6 @@ static int intel_init_ring_buffer(struct intel_engine_cs *engine)
 {
 	struct i915_timeline *timeline;
 	struct intel_ring *ring;
-	unsigned int size;
 	int err;
 
 	intel_engine_setup_common(engine);
@@ -1428,21 +1489,12 @@ static int intel_init_ring_buffer(struct intel_engine_cs *engine)
 	GEM_BUG_ON(engine->buffer);
 	engine->buffer = ring;
 
-	size = PAGE_SIZE;
-	if (HAS_BROKEN_CS_TLB(engine->i915))
-		size = I830_WA_SIZE;
-	err = intel_engine_create_scratch(engine, size);
-	if (err)
-		goto err_unpin;
-
 	err = intel_engine_init_common(engine);
 	if (err)
-		goto err_scratch;
+		goto err_unpin;
 
 	return 0;
 
-err_scratch:
-	intel_engine_cleanup_scratch(engine);
 err_unpin:
 	intel_ring_unpin(ring);
 err_ring:
@@ -1516,7 +1568,7 @@ static int flush_pd_dir(struct i915_request *rq)
 	/* Stall until the page table load is complete */
 	*cs++ = MI_STORE_REGISTER_MEM | MI_SRM_LRM_GLOBAL_GTT;
 	*cs++ = i915_mmio_reg_offset(RING_PP_DIR_BASE(engine));
-	*cs++ = i915_ggtt_offset(engine->scratch);
+	*cs++ = i915_scratch_offset(rq->i915);
 	*cs++ = MI_NOOP;
 
 	intel_ring_advance(rq, cs);
@@ -1625,7 +1677,7 @@ static inline int mi_set_context(struct i915_request *rq, u32 flags)
 			/* Insert a delay before the next switch! */
 			*cs++ = MI_STORE_REGISTER_MEM | MI_SRM_LRM_GLOBAL_GTT;
 			*cs++ = i915_mmio_reg_offset(last_reg);
-			*cs++ = i915_ggtt_offset(engine->scratch);
+			*cs++ = i915_scratch_offset(rq->i915);
 			*cs++ = MI_NOOP;
 		}
 		*cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;