drm/i915: Use HWS for seqno tracking everywhere

By using the same address for storing the HWS on every platform, we can
remove the platform specific vfuncs and reduce the get-seqno routine to
a single read of a cached memory location.

v2: Fix semaphore_passed() to look at the signaling engine (not the
waiter's)

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-8-git-send-email-chris@chris-wilson.co.uk

9 files changed, 34 insertions, 86 deletions

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 309bb2f1a1e8..f5899b631c0e 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -662,7 +662,7 @@ static int i915_gem_pageflip_info(struct seq_file *m, void *data)
 					   engine->name,
 					   i915_gem_request_get_seqno(work->flip_queued_req),
 					   dev_priv->next_seqno,
-					   engine->get_seqno(engine),
+					   intel_engine_get_seqno(engine),
 					   i915_gem_request_completed(work->flip_queued_req));
 			} else
 				seq_printf(m, "Flip not associated with any ring\n");
@@ -792,7 +792,7 @@ static void i915_ring_seqno_info(struct seq_file *m,
 	struct rb_node *rb;
 
 	seq_printf(m, "Current sequence (%s): %x\n",
-		   engine->name, engine->get_seqno(engine));
+		   engine->name, intel_engine_get_seqno(engine));
 	seq_printf(m, "Current user interrupts (%s): %x\n",
 		   engine->name, READ_ONCE(engine->user_interrupts));
 
@@ -1420,7 +1420,7 @@ static int i915_hangcheck_info(struct seq_file *m, void *unused)
 
 	for_each_engine_id(engine, dev_priv, id) {
 		acthd[id] = intel_ring_get_active_head(engine);
-		seqno[id] = engine->get_seqno(engine);
+		seqno[id] = intel_engine_get_seqno(engine);
 	}
 
 	i915_get_extra_instdone(dev_priv, instdone);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 0ea69c5ecc8b..5a1e8e056ee5 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -3289,13 +3289,13 @@ i915_seqno_passed(uint32_t seq1, uint32_t seq2)
 
 static inline bool i915_gem_request_started(const struct drm_i915_gem_request *req)
 {
-	return i915_seqno_passed(req->engine->get_seqno(req->engine),
+	return i915_seqno_passed(intel_engine_get_seqno(req->engine),
 				 req->previous_seqno);
 }
 
 static inline bool i915_gem_request_completed(const struct drm_i915_gem_request *req)
 {
-	return i915_seqno_passed(req->engine->get_seqno(req->engine),
+	return i915_seqno_passed(intel_engine_get_seqno(req->engine),
 				 req->seqno);
 }
 
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 250f0b818099..1c15a74530cd 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -984,7 +984,7 @@ static void i915_record_ring_state(struct drm_i915_private *dev_priv,
 	ering->waiting = intel_engine_has_waiter(engine);
 	ering->instpm = I915_READ(RING_INSTPM(engine->mmio_base));
 	ering->acthd = intel_ring_get_active_head(engine);
-	ering->seqno = engine->get_seqno(engine);
+	ering->seqno = intel_engine_get_seqno(engine);
 	ering->last_seqno = engine->last_submitted_seqno;
 	ering->start = I915_READ_START(engine);
 	ering->head = I915_READ_HEAD(engine);
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index a11ab00cdee0..7c379afcff2f 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -2952,7 +2952,7 @@ static int semaphore_passed(struct intel_engine_cs *engine)
 	if (signaller->hangcheck.deadlock >= I915_NUM_ENGINES)
 		return -1;
 
-	if (i915_seqno_passed(signaller->get_seqno(signaller), seqno))
+	if (i915_seqno_passed(intel_engine_get_seqno(signaller), seqno))
 		return 1;
 
 	/* cursory check for an unkickable deadlock */
@@ -3140,7 +3140,7 @@ static void i915_hangcheck_elapsed(struct work_struct *work)
 			engine->irq_seqno_barrier(engine);
 
 		acthd = intel_ring_get_active_head(engine);
-		seqno = engine->get_seqno(engine);
+		seqno = intel_engine_get_seqno(engine);
 
 		/* Reset stuck interrupts between batch advances */
 		user_interrupts = 0;
diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
index 6768db032f84..3d13fde95fdf 100644
--- a/drivers/gpu/drm/i915/i915_trace.h
+++ b/drivers/gpu/drm/i915/i915_trace.h
@@ -558,7 +558,7 @@ TRACE_EVENT(i915_gem_request_notify,
 	    TP_fast_assign(
 			   __entry->dev = engine->i915->dev->primary->index;
 			   __entry->ring = engine->id;
-			   __entry->seqno = engine->get_seqno(engine);
+			   __entry->seqno = intel_engine_get_seqno(engine);
 			   ),
 
 	    TP_printk("dev=%u, ring=%u, seqno=%u",
diff --git a/drivers/gpu/drm/i915/intel_breadcrumbs.c b/drivers/gpu/drm/i915/intel_breadcrumbs.c
index a3bbf2d90dce..7cdb02d18c1f 100644
--- a/drivers/gpu/drm/i915/intel_breadcrumbs.c
+++ b/drivers/gpu/drm/i915/intel_breadcrumbs.c
@@ -146,7 +146,7 @@ static bool __intel_engine_add_wait(struct intel_engine_cs *engine,
 	first = true;
 	parent = NULL;
 	completed = NULL;
-	seqno = engine->get_seqno(engine);
+	seqno = intel_engine_get_seqno(engine);
 
 	 /* If the request completed before we managed to grab the spinlock,
 	  * return now before adding ourselves to the rbtree. We let the
@@ -296,7 +296,7 @@ void intel_engine_remove_wait(struct intel_engine_cs *engine,
 			 * the first_waiter. This is undesirable if that
 			 * waiter is a high priority task.
 			 */
-			u32 seqno = engine->get_seqno(engine);
+			u32 seqno = intel_engine_get_seqno(engine);
 
 			while (i915_seqno_passed(seqno, to_wait(next)->seqno)) {
 				struct rb_node *n = rb_next(next);
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 226bba22e4b4..c8411f8bb4ac 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -1783,16 +1783,6 @@ static int gen8_emit_flush_render(struct drm_i915_gem_request *request,
 	return 0;
 }
 
-static u32 gen8_get_seqno(struct intel_engine_cs *engine)
-{
-	return intel_read_status_page(engine, I915_GEM_HWS_INDEX);
-}
-
-static void gen8_set_seqno(struct intel_engine_cs *engine, u32 seqno)
-{
-	intel_write_status_page(engine, I915_GEM_HWS_INDEX, seqno);
-}
-
 static void bxt_a_seqno_barrier(struct intel_engine_cs *engine)
 {
 	/*
@@ -1808,14 +1798,6 @@ static void bxt_a_seqno_barrier(struct intel_engine_cs *engine)
 	intel_flush_status_page(engine, I915_GEM_HWS_INDEX);
 }
 
-static void bxt_a_set_seqno(struct intel_engine_cs *engine, u32 seqno)
-{
-	intel_write_status_page(engine, I915_GEM_HWS_INDEX, seqno);
-
-	/* See bxt_a_get_seqno() explaining the reason for the clflush. */
-	intel_flush_status_page(engine, I915_GEM_HWS_INDEX);
-}
-
 /*
  * Reserve space for 2 NOOPs at the end of each request to be
  * used as a workaround for not being allowed to do lite
@@ -1841,7 +1823,7 @@ static int gen8_emit_request(struct drm_i915_gem_request *request)
 				intel_hws_seqno_address(request->engine) |
 				MI_FLUSH_DW_USE_GTT);
 	intel_logical_ring_emit(ringbuf, 0);
-	intel_logical_ring_emit(ringbuf, i915_gem_request_get_seqno(request));
+	intel_logical_ring_emit(ringbuf, request->seqno);
 	intel_logical_ring_emit(ringbuf, MI_USER_INTERRUPT);
 	intel_logical_ring_emit(ringbuf, MI_NOOP);
 	return intel_logical_ring_advance_and_submit(request);
@@ -1987,12 +1969,8 @@ logical_ring_default_vfuncs(struct intel_engine_cs *engine)
 	engine->irq_get = gen8_logical_ring_get_irq;
 	engine->irq_put = gen8_logical_ring_put_irq;
 	engine->emit_bb_start = gen8_emit_bb_start;
-	engine->get_seqno = gen8_get_seqno;
-	engine->set_seqno = gen8_set_seqno;
-	if (IS_BXT_REVID(engine->i915, 0, BXT_REVID_A1)) {
+	if (IS_BXT_REVID(engine->i915, 0, BXT_REVID_A1))
 		engine->irq_seqno_barrier = bxt_a_seqno_barrier;
-		engine->set_seqno = bxt_a_set_seqno;
-	}
 }
 
 static inline void
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index af50aa01bcd9..02104fbf9045 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -1367,19 +1367,17 @@ static int gen8_rcs_signal(struct drm_i915_gem_request *signaller_req,
 		return ret;
 
 	for_each_engine_id(waiter, dev_priv, id) {
-		u32 seqno;
 		u64 gtt_offset = signaller->semaphore.signal_ggtt[id];
 		if (gtt_offset == MI_SEMAPHORE_SYNC_INVALID)
 			continue;
 
-		seqno = i915_gem_request_get_seqno(signaller_req);
 		intel_ring_emit(signaller, GFX_OP_PIPE_CONTROL(6));
 		intel_ring_emit(signaller, PIPE_CONTROL_GLOBAL_GTT_IVB |
 					   PIPE_CONTROL_QW_WRITE |
 					   PIPE_CONTROL_CS_STALL);
 		intel_ring_emit(signaller, lower_32_bits(gtt_offset));
 		intel_ring_emit(signaller, upper_32_bits(gtt_offset));
-		intel_ring_emit(signaller, seqno);
+		intel_ring_emit(signaller, signaller_req->seqno);
 		intel_ring_emit(signaller, 0);
 		intel_ring_emit(signaller, MI_SEMAPHORE_SIGNAL |
 					   MI_SEMAPHORE_TARGET(waiter->hw_id));
@@ -1408,18 +1406,16 @@ static int gen8_xcs_signal(struct drm_i915_gem_request *signaller_req,
 		return ret;
 
 	for_each_engine_id(waiter, dev_priv, id) {
-		u32 seqno;
 		u64 gtt_offset = signaller->semaphore.signal_ggtt[id];
 		if (gtt_offset == MI_SEMAPHORE_SYNC_INVALID)
 			continue;
 
-		seqno = i915_gem_request_get_seqno(signaller_req);
 		intel_ring_emit(signaller, (MI_FLUSH_DW + 1) |
 					   MI_FLUSH_DW_OP_STOREDW);
 		intel_ring_emit(signaller, lower_32_bits(gtt_offset) |
 					   MI_FLUSH_DW_USE_GTT);
 		intel_ring_emit(signaller, upper_32_bits(gtt_offset));
-		intel_ring_emit(signaller, seqno);
+		intel_ring_emit(signaller, signaller_req->seqno);
 		intel_ring_emit(signaller, MI_SEMAPHORE_SIGNAL |
 					   MI_SEMAPHORE_TARGET(waiter->hw_id));
 		intel_ring_emit(signaller, 0);
@@ -1450,11 +1446,9 @@ static int gen6_signal(struct drm_i915_gem_request *signaller_req,
 		i915_reg_t mbox_reg = signaller->semaphore.mbox.signal[id];
 
 		if (i915_mmio_reg_valid(mbox_reg)) {
-			u32 seqno = i915_gem_request_get_seqno(signaller_req);
-
 			intel_ring_emit(signaller, MI_LOAD_REGISTER_IMM(1));
 			intel_ring_emit_reg(signaller, mbox_reg);
-			intel_ring_emit(signaller, seqno);
+			intel_ring_emit(signaller, signaller_req->seqno);
 		}
 	}
 
@@ -1490,7 +1484,7 @@ gen6_add_request(struct drm_i915_gem_request *req)
 	intel_ring_emit(engine, MI_STORE_DWORD_INDEX);
 	intel_ring_emit(engine,
 			I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
-	intel_ring_emit(engine, i915_gem_request_get_seqno(req));
+	intel_ring_emit(engine, req->seqno);
 	intel_ring_emit(engine, MI_USER_INTERRUPT);
 	__intel_ring_advance(engine);
 
@@ -1628,7 +1622,9 @@ static int
 pc_render_add_request(struct drm_i915_gem_request *req)
 {
 	struct intel_engine_cs *engine = req->engine;
-	u32 scratch_addr = engine->scratch.gtt_offset + 2 * CACHELINE_BYTES;
+	u32 addr = engine->status_page.gfx_addr +
+		(I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
+	u32 scratch_addr = addr;
 	int ret;
 
 	/* For Ironlake, MI_USER_INTERRUPT was deprecated and apparently
@@ -1644,12 +1640,12 @@ pc_render_add_request(struct drm_i915_gem_request *req)
 		return ret;
 
 	intel_ring_emit(engine,
-			GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE |
+			GFX_OP_PIPE_CONTROL(4) |
+			PIPE_CONTROL_QW_WRITE |
 			PIPE_CONTROL_WRITE_FLUSH |
 			PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE);
-	intel_ring_emit(engine,
-			engine->scratch.gtt_offset | PIPE_CONTROL_GLOBAL_GTT);
-	intel_ring_emit(engine, i915_gem_request_get_seqno(req));
+	intel_ring_emit(engine, addr | PIPE_CONTROL_GLOBAL_GTT);
+	intel_ring_emit(engine, req->seqno);
 	intel_ring_emit(engine, 0);
 	PIPE_CONTROL_FLUSH(engine, scratch_addr);
 	scratch_addr += 2 * CACHELINE_BYTES; /* write to separate cachelines */
@@ -1668,9 +1664,8 @@ pc_render_add_request(struct drm_i915_gem_request *req)
 			PIPE_CONTROL_WRITE_FLUSH |
 			PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE |
 			PIPE_CONTROL_NOTIFY);
-	intel_ring_emit(engine,
-			engine->scratch.gtt_offset | PIPE_CONTROL_GLOBAL_GTT);
-	intel_ring_emit(engine, i915_gem_request_get_seqno(req));
+	intel_ring_emit(engine, addr | PIPE_CONTROL_GLOBAL_GTT);
+	intel_ring_emit(engine, req->seqno);
 	intel_ring_emit(engine, 0);
 	__intel_ring_advance(engine);
 
@@ -1702,30 +1697,6 @@ gen6_seqno_barrier(struct intel_engine_cs *engine)
 	spin_unlock_irq(&dev_priv->uncore.lock);
 }
 
-static u32
-ring_get_seqno(struct intel_engine_cs *engine)
-{
-	return intel_read_status_page(engine, I915_GEM_HWS_INDEX);
-}
-
-static void
-ring_set_seqno(struct intel_engine_cs *engine, u32 seqno)
-{
-	intel_write_status_page(engine, I915_GEM_HWS_INDEX, seqno);
-}
-
-static u32
-pc_render_get_seqno(struct intel_engine_cs *engine)
-{
-	return engine->scratch.cpu_page[0];
-}
-
-static void
-pc_render_set_seqno(struct intel_engine_cs *engine, u32 seqno)
-{
-	engine->scratch.cpu_page[0] = seqno;
-}
-
 static bool
 gen5_ring_get_irq(struct intel_engine_cs *engine)
 {
@@ -1856,7 +1827,7 @@ i9xx_add_request(struct drm_i915_gem_request *req)
 	intel_ring_emit(engine, MI_STORE_DWORD_INDEX);
 	intel_ring_emit(engine,
 			I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
-	intel_ring_emit(engine, i915_gem_request_get_seqno(req));
+	intel_ring_emit(engine, req->seqno);
 	intel_ring_emit(engine, MI_USER_INTERRUPT);
 	__intel_ring_advance(engine);
 
@@ -2675,7 +2646,9 @@ void intel_ring_init_seqno(struct intel_engine_cs *engine, u32 seqno)
 	memset(engine->semaphore.sync_seqno, 0,
 	       sizeof(engine->semaphore.sync_seqno));
 
-	engine->set_seqno(engine, seqno);
+	intel_write_status_page(engine, I915_GEM_HWS_INDEX, seqno);
+	if (engine->irq_seqno_barrier)
+		engine->irq_seqno_barrier(engine);
 	engine->last_submitted_seqno = seqno;
 
 	engine->hangcheck.seqno = seqno;
@@ -3021,8 +2994,6 @@ static void intel_ring_default_vfuncs(struct drm_i915_private *dev_priv,
 {
 	engine->init_hw = init_ring_common;
 	engine->write_tail = ring_write_tail;
-	engine->get_seqno = ring_get_seqno;
-	engine->set_seqno = ring_set_seqno;
 
 	engine->add_request = i9xx_add_request;
 	if (INTEL_GEN(dev_priv) >= 6)
@@ -3074,8 +3045,6 @@ int intel_init_render_ring_buffer(struct drm_device *dev)
 	} else if (IS_GEN5(dev_priv)) {
 		engine->add_request = pc_render_add_request;
 		engine->flush = gen4_render_ring_flush;
-		engine->get_seqno = pc_render_get_seqno;
-		engine->set_seqno = pc_render_set_seqno;
 		engine->irq_enable_mask = GT_RENDER_USER_INTERRUPT |
 					GT_RENDER_PIPECTL_NOTIFY_INTERRUPT;
 	} else {
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 6fd70a56e219..b03931f6dde5 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -207,9 +207,6 @@ struct intel_engine_cs {
 	 * monotonic, even if not coherent.
 	 */
 	void		(*irq_seqno_barrier)(struct intel_engine_cs *ring);
-	u32		(*get_seqno)(struct intel_engine_cs *ring);
-	void		(*set_seqno)(struct intel_engine_cs *ring,
-				     u32 seqno);
 	int		(*dispatch_execbuffer)(struct drm_i915_gem_request *req,
 					       u64 offset, u32 length,
 					       unsigned dispatch_flags);
@@ -485,6 +482,10 @@ int intel_init_blt_ring_buffer(struct drm_device *dev);
 int intel_init_vebox_ring_buffer(struct drm_device *dev);
 
 u64 intel_ring_get_active_head(struct intel_engine_cs *engine);
+static inline u32 intel_engine_get_seqno(struct intel_engine_cs *engine)
+{
+	return intel_read_status_page(engine, I915_GEM_HWS_INDEX);
+}
 
 int init_workarounds_ring(struct intel_engine_cs *engine);