Merge branch 'drm-radeon-next' of ../drm-radeon-next into drm-core-next

* 'drm-radeon-next' of ../drm-radeon-next:
  drm/radeon: introduce a sub allocator and convert ib pool to it v4
  drm/radeon/kms: add support for per-ring fence interrupts
  drm/radeon/kms: add cayman specific fence_ring_emit
  drm/radeon/kms: add some new ring params to better handle other ring types
  drm/radeon: improve radeon_test_syncing function
  drm/radeon: precompute fence cpu/gpu addr once v3
  drm/radeon: move ring debugfs into radeon_ring.c
  drm/radeon: rename struct radeon_cp to radeon_ring
  drm/radeon: disable compute rings on cayman for now
  drm/radeon: add radeon_fence_count_emited function
  drm/radeon: make some asic pointers per ring
  drm/radeon: Add radeon_test_syncing function v2
  drm/radeon: make cp variable an array
  drm/radeon: make ring rptr and wptr register offsets variable
  drm/radeon: make all functions work with multiple rings.
  drm/radeon/kms: add support for semaphores v3
  drm/radeon/kms: add support for multiple fence queues v2
  drm/radeon: fix a spelling mistake
  drm/radeon: no need to check all relocs for duplicates
  drm/radeon: fix debugfs handling v3

drivers/gpu/drm/radeon/Makefile

@@ -70,7 +70,8 @@ radeon-y += radeon_device.o radeon_asic.o radeon_kms.o \
 	r200.o radeon_legacy_tv.o r600_cs.o r600_blit.o r600_blit_shaders.o \
 	r600_blit_kms.o radeon_pm.o atombios_dp.o r600_audio.o r600_hdmi.o \
 	evergreen.o evergreen_cs.o evergreen_blit_shaders.o evergreen_blit_kms.o \
-	radeon_trace_points.o ni.o cayman_blit_shaders.o atombios_encoders.o
+	radeon_trace_points.o ni.o cayman_blit_shaders.o atombios_encoders.o \
+	radeon_semaphore.o radeon_sa.o
 
 radeon-$(CONFIG_COMPAT) += radeon_ioc32.o
 radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o
@@ -78,4 +79,4 @@ radeon-$(CONFIG_ACPI) += radeon_acpi.o
 
 obj-$(CONFIG_DRM_RADEON)+= radeon.o
 
-CFLAGS_radeon_trace_points.o := -I$(src)
\ No newline at end of file
+CFLAGS_radeon_trace_points.o := -I$(src)

drivers/gpu/drm/radeon/nid.h

@@ -42,6 +42,9 @@
 #define CAYMAN_MAX_TCC_MASK          0xFF
 
 #define DMIF_ADDR_CONFIG  				0xBD4
+#define	SRBM_GFX_CNTL				        0x0E44
+#define		RINGID(x)					(((x) & 0x3) << 0)
+#define		VMID(x)						(((x) & 0x7) << 0)
 #define	SRBM_STATUS				        0x0E50
 
 #define VM_CONTEXT0_REQUEST_RESPONSE			0x1470
@@ -394,6 +397,12 @@
 #define	CP_RB0_RPTR_ADDR				0xC10C
 #define	CP_RB0_RPTR_ADDR_HI				0xC110
 #define	CP_RB0_WPTR					0xC114
+
+#define CP_INT_CNTL                                     0xC124
+#       define CNTX_BUSY_INT_ENABLE                     (1 << 19)
+#       define CNTX_EMPTY_INT_ENABLE                    (1 << 20)
+#       define TIME_STAMP_INT_ENABLE                    (1 << 26)
+
 #define	CP_RB1_BASE					0xC180
 #define	CP_RB1_CNTL					0xC184
 #define	CP_RB1_RPTR_ADDR				0xC188
@@ -411,6 +420,10 @@
 #define	CP_ME_RAM_DATA					0xC160
 #define	CP_DEBUG					0xC1FC
 
+#define VGT_EVENT_INITIATOR                             0x28a90
+#       define CACHE_FLUSH_AND_INV_EVENT_TS                     (0x14 << 0)
+#       define CACHE_FLUSH_AND_INV_EVENT                        (0x16 << 0)
+
 /*
  * PM4
  */
@@ -494,7 +507,27 @@
 #define		PACKET3_ME_INITIALIZE_DEVICE_ID(x) ((x) << 16)
 #define	PACKET3_COND_WRITE				0x45
 #define	PACKET3_EVENT_WRITE				0x46
+#define		EVENT_TYPE(x)                           ((x) << 0)
+#define		EVENT_INDEX(x)                          ((x) << 8)
+                /* 0 - any non-TS event
+		 * 1 - ZPASS_DONE
+		 * 2 - SAMPLE_PIPELINESTAT
+		 * 3 - SAMPLE_STREAMOUTSTAT*
+		 * 4 - *S_PARTIAL_FLUSH
+		 * 5 - TS events
+		 */
 #define	PACKET3_EVENT_WRITE_EOP				0x47
+#define		DATA_SEL(x)                             ((x) << 29)
+                /* 0 - discard
+		 * 1 - send low 32bit data
+		 * 2 - send 64bit data
+		 * 3 - send 64bit counter value
+		 */
+#define		INT_SEL(x)                              ((x) << 24)
+                /* 0 - none
+		 * 1 - interrupt only (DATA_SEL = 0)
+		 * 2 - interrupt when data write is confirmed
+		 */
 #define	PACKET3_EVENT_WRITE_EOS				0x48
 #define	PACKET3_PREAMBLE_CNTL				0x4A
 #              define PACKET3_PREAMBLE_BEGIN_CLEAR_STATE     (2 << 28)

drivers/gpu/drm/radeon/r200.c

@@ -87,6 +87,7 @@ int r200_copy_dma(struct radeon_device *rdev,
 		  unsigned num_gpu_pages,
 		  struct radeon_fence *fence)
 {
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
 	uint32_t size;
 	uint32_t cur_size;
 	int i, num_loops;
@@ -95,33 +96,33 @@ int r200_copy_dma(struct radeon_device *rdev,
 	/* radeon pitch is /64 */
 	size = num_gpu_pages << RADEON_GPU_PAGE_SHIFT;
 	num_loops = DIV_ROUND_UP(size, 0x1FFFFF);
-	r = radeon_ring_lock(rdev, num_loops * 4 + 64);
+	r = radeon_ring_lock(rdev, ring, num_loops * 4 + 64);
 	if (r) {
 		DRM_ERROR("radeon: moving bo (%d).\n", r);
 		return r;
 	}
 	/* Must wait for 2D idle & clean before DMA or hangs might happen */
-	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
-	radeon_ring_write(rdev, (1 << 16));
+	radeon_ring_write(ring, PACKET0(RADEON_WAIT_UNTIL, 0));
+	radeon_ring_write(ring, (1 << 16));
 	for (i = 0; i < num_loops; i++) {
 		cur_size = size;
 		if (cur_size > 0x1FFFFF) {
 			cur_size = 0x1FFFFF;
 		}
 		size -= cur_size;
-		radeon_ring_write(rdev, PACKET0(0x720, 2));
-		radeon_ring_write(rdev, src_offset);
-		radeon_ring_write(rdev, dst_offset);
-		radeon_ring_write(rdev, cur_size | (1 << 31) | (1 << 30));
+		radeon_ring_write(ring, PACKET0(0x720, 2));
+		radeon_ring_write(ring, src_offset);
+		radeon_ring_write(ring, dst_offset);
+		radeon_ring_write(ring, cur_size | (1 << 31) | (1 << 30));
 		src_offset += cur_size;
 		dst_offset += cur_size;
 	}
-	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
-	radeon_ring_write(rdev, RADEON_WAIT_DMA_GUI_IDLE);
+	radeon_ring_write(ring, PACKET0(RADEON_WAIT_UNTIL, 0));
+	radeon_ring_write(ring, RADEON_WAIT_DMA_GUI_IDLE);
 	if (fence) {
 		r = radeon_fence_emit(rdev, fence);
 	}
-	radeon_ring_unlock_commit(rdev);
+	radeon_ring_unlock_commit(rdev, ring);
 	return r;
 }
 

drivers/gpu/drm/radeon/r300.c

@@ -175,37 +175,40 @@ void rv370_pcie_gart_fini(struct radeon_device *rdev)
 void r300_fence_ring_emit(struct radeon_device *rdev,
 			  struct radeon_fence *fence)
 {
+	struct radeon_ring *ring = &rdev->ring[fence->ring];
+
 	/* Who ever call radeon_fence_emit should call ring_lock and ask
 	 * for enough space (today caller are ib schedule and buffer move) */
 	/* Write SC register so SC & US assert idle */
-	radeon_ring_write(rdev, PACKET0(R300_RE_SCISSORS_TL, 0));
-	radeon_ring_write(rdev, 0);
-	radeon_ring_write(rdev, PACKET0(R300_RE_SCISSORS_BR, 0));
-	radeon_ring_write(rdev, 0);
+	radeon_ring_write(ring, PACKET0(R300_RE_SCISSORS_TL, 0));
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, PACKET0(R300_RE_SCISSORS_BR, 0));
+	radeon_ring_write(ring, 0);
 	/* Flush 3D cache */
-	radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
-	radeon_ring_write(rdev, R300_RB3D_DC_FLUSH);
-	radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
-	radeon_ring_write(rdev, R300_ZC_FLUSH);
+	radeon_ring_write(ring, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
+	radeon_ring_write(ring, R300_RB3D_DC_FLUSH);
+	radeon_ring_write(ring, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
+	radeon_ring_write(ring, R300_ZC_FLUSH);
 	/* Wait until IDLE & CLEAN */
-	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
-	radeon_ring_write(rdev, (RADEON_WAIT_3D_IDLECLEAN |
+	radeon_ring_write(ring, PACKET0(RADEON_WAIT_UNTIL, 0));
+	radeon_ring_write(ring, (RADEON_WAIT_3D_IDLECLEAN |
 				 RADEON_WAIT_2D_IDLECLEAN |
 				 RADEON_WAIT_DMA_GUI_IDLE));
-	radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
-	radeon_ring_write(rdev, rdev->config.r300.hdp_cntl |
+	radeon_ring_write(ring, PACKET0(RADEON_HOST_PATH_CNTL, 0));
+	radeon_ring_write(ring, rdev->config.r300.hdp_cntl |
 				RADEON_HDP_READ_BUFFER_INVALIDATE);
-	radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0));
-	radeon_ring_write(rdev, rdev->config.r300.hdp_cntl);
+	radeon_ring_write(ring, PACKET0(RADEON_HOST_PATH_CNTL, 0));
+	radeon_ring_write(ring, rdev->config.r300.hdp_cntl);
 	/* Emit fence sequence & fire IRQ */
-	radeon_ring_write(rdev, PACKET0(rdev->fence_drv.scratch_reg, 0));
-	radeon_ring_write(rdev, fence->seq);
-	radeon_ring_write(rdev, PACKET0(RADEON_GEN_INT_STATUS, 0));
-	radeon_ring_write(rdev, RADEON_SW_INT_FIRE);
+	radeon_ring_write(ring, PACKET0(rdev->fence_drv[fence->ring].scratch_reg, 0));
+	radeon_ring_write(ring, fence->seq);
+	radeon_ring_write(ring, PACKET0(RADEON_GEN_INT_STATUS, 0));
+	radeon_ring_write(ring, RADEON_SW_INT_FIRE);
 }
 
 void r300_ring_start(struct radeon_device *rdev)
 {
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
 	unsigned gb_tile_config;
 	int r;
 
@@ -227,44 +230,44 @@ void r300_ring_start(struct radeon_device *rdev)
 		break;
 	}
 
-	r = radeon_ring_lock(rdev, 64);
+	r = radeon_ring_lock(rdev, ring, 64);
 	if (r) {
 		return;
 	}
-	radeon_ring_write(rdev, PACKET0(RADEON_ISYNC_CNTL, 0));
-	radeon_ring_write(rdev,
+	radeon_ring_write(ring, PACKET0(RADEON_ISYNC_CNTL, 0));
+	radeon_ring_write(ring,
 			  RADEON_ISYNC_ANY2D_IDLE3D |
 			  RADEON_ISYNC_ANY3D_IDLE2D |
 			  RADEON_ISYNC_WAIT_IDLEGUI |
 			  RADEON_ISYNC_CPSCRATCH_IDLEGUI);
-	radeon_ring_write(rdev, PACKET0(R300_GB_TILE_CONFIG, 0));
-	radeon_ring_write(rdev, gb_tile_config);
-	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
-	radeon_ring_write(rdev,
+	radeon_ring_write(ring, PACKET0(R300_GB_TILE_CONFIG, 0));
+	radeon_ring_write(ring, gb_tile_config);
+	radeon_ring_write(ring, PACKET0(RADEON_WAIT_UNTIL, 0));
+	radeon_ring_write(ring,
 			  RADEON_WAIT_2D_IDLECLEAN |
 			  RADEON_WAIT_3D_IDLECLEAN);
-	radeon_ring_write(rdev, PACKET0(R300_DST_PIPE_CONFIG, 0));
-	radeon_ring_write(rdev, R300_PIPE_AUTO_CONFIG);
-	radeon_ring_write(rdev, PACKET0(R300_GB_SELECT, 0));
-	radeon_ring_write(rdev, 0);
-	radeon_ring_write(rdev, PACKET0(R300_GB_ENABLE, 0));
-	radeon_ring_write(rdev, 0);
-	radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
-	radeon_ring_write(rdev, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
-	radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
-	radeon_ring_write(rdev, R300_ZC_FLUSH | R300_ZC_FREE);
-	radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
-	radeon_ring_write(rdev,
+	radeon_ring_write(ring, PACKET0(R300_DST_PIPE_CONFIG, 0));
+	radeon_ring_write(ring, R300_PIPE_AUTO_CONFIG);
+	radeon_ring_write(ring, PACKET0(R300_GB_SELECT, 0));
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, PACKET0(R300_GB_ENABLE, 0));
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
+	radeon_ring_write(ring, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
+	radeon_ring_write(ring, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
+	radeon_ring_write(ring, R300_ZC_FLUSH | R300_ZC_FREE);
+	radeon_ring_write(ring, PACKET0(RADEON_WAIT_UNTIL, 0));
+	radeon_ring_write(ring,
 			  RADEON_WAIT_2D_IDLECLEAN |
 			  RADEON_WAIT_3D_IDLECLEAN);
-	radeon_ring_write(rdev, PACKET0(R300_GB_AA_CONFIG, 0));
-	radeon_ring_write(rdev, 0);
-	radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
-	radeon_ring_write(rdev, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
-	radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
-	radeon_ring_write(rdev, R300_ZC_FLUSH | R300_ZC_FREE);
-	radeon_ring_write(rdev, PACKET0(R300_GB_MSPOS0, 0));
-	radeon_ring_write(rdev,
+	radeon_ring_write(ring, PACKET0(R300_GB_AA_CONFIG, 0));
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
+	radeon_ring_write(ring, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
+	radeon_ring_write(ring, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
+	radeon_ring_write(ring, R300_ZC_FLUSH | R300_ZC_FREE);
+	radeon_ring_write(ring, PACKET0(R300_GB_MSPOS0, 0));
+	radeon_ring_write(ring,
 			  ((6 << R300_MS_X0_SHIFT) |
 			   (6 << R300_MS_Y0_SHIFT) |
 			   (6 << R300_MS_X1_SHIFT) |
@@ -273,8 +276,8 @@ void r300_ring_start(struct radeon_device *rdev)
 			   (6 << R300_MS_Y2_SHIFT) |
 			   (6 << R300_MSBD0_Y_SHIFT) |
 			   (6 << R300_MSBD0_X_SHIFT)));
-	radeon_ring_write(rdev, PACKET0(R300_GB_MSPOS1, 0));
-	radeon_ring_write(rdev,
+	radeon_ring_write(ring, PACKET0(R300_GB_MSPOS1, 0));
+	radeon_ring_write(ring,
 			  ((6 << R300_MS_X3_SHIFT) |
 			   (6 << R300_MS_Y3_SHIFT) |
 			   (6 << R300_MS_X4_SHIFT) |
@@ -282,16 +285,16 @@ void r300_ring_start(struct radeon_device *rdev)
 			   (6 << R300_MS_X5_SHIFT) |
 			   (6 << R300_MS_Y5_SHIFT) |
 			   (6 << R300_MSBD1_SHIFT)));
-	radeon_ring_write(rdev, PACKET0(R300_GA_ENHANCE, 0));
-	radeon_ring_write(rdev, R300_GA_DEADLOCK_CNTL | R300_GA_FASTSYNC_CNTL);
-	radeon_ring_write(rdev, PACKET0(R300_GA_POLY_MODE, 0));
-	radeon_ring_write(rdev,
+	radeon_ring_write(ring, PACKET0(R300_GA_ENHANCE, 0));
+	radeon_ring_write(ring, R300_GA_DEADLOCK_CNTL | R300_GA_FASTSYNC_CNTL);
+	radeon_ring_write(ring, PACKET0(R300_GA_POLY_MODE, 0));
+	radeon_ring_write(ring,
 			  R300_FRONT_PTYPE_TRIANGE | R300_BACK_PTYPE_TRIANGE);
-	radeon_ring_write(rdev, PACKET0(R300_GA_ROUND_MODE, 0));
-	radeon_ring_write(rdev,
+	radeon_ring_write(ring, PACKET0(R300_GA_ROUND_MODE, 0));
+	radeon_ring_write(ring,
 			  R300_GEOMETRY_ROUND_NEAREST |
 			  R300_COLOR_ROUND_NEAREST);
-	radeon_ring_unlock_commit(rdev);
+	radeon_ring_unlock_commit(rdev, ring);
 }
 
 void r300_errata(struct radeon_device *rdev)
@@ -375,26 +378,26 @@ void r300_gpu_init(struct radeon_device *rdev)
 		 rdev->num_gb_pipes, rdev->num_z_pipes);
 }
 
-bool r300_gpu_is_lockup(struct radeon_device *rdev)
+bool r300_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring)
 {
 	u32 rbbm_status;
 	int r;
 
 	rbbm_status = RREG32(R_000E40_RBBM_STATUS);
 	if (!G_000E40_GUI_ACTIVE(rbbm_status)) {
-		r100_gpu_lockup_update(&rdev->config.r300.lockup, &rdev->cp);
+		r100_gpu_lockup_update(&rdev->config.r300.lockup, ring);
 		return false;
 	}
 	/* force CP activities */
-	r = radeon_ring_lock(rdev, 2);
+	r = radeon_ring_lock(rdev, ring, 2);
 	if (!r) {
 		/* PACKET2 NOP */
-		radeon_ring_write(rdev, 0x80000000);
-		radeon_ring_write(rdev, 0x80000000);
-		radeon_ring_unlock_commit(rdev);
+		radeon_ring_write(ring, 0x80000000);
+		radeon_ring_write(ring, 0x80000000);
+		radeon_ring_unlock_commit(rdev, ring);
 	}
-	rdev->cp.rptr = RREG32(RADEON_CP_RB_RPTR);
-	return r100_gpu_cp_is_lockup(rdev, &rdev->config.r300.lockup, &rdev->cp);
+	ring->rptr = RREG32(RADEON_CP_RB_RPTR);
+	return r100_gpu_cp_is_lockup(rdev, &rdev->config.r300.lockup, ring);
 }
 
 int r300_asic_reset(struct radeon_device *rdev)
@@ -1396,6 +1399,12 @@ static int r300_startup(struct radeon_device *rdev)
 	if (r)
 		return r;
 
+	r = radeon_fence_driver_start_ring(rdev, RADEON_RING_TYPE_GFX_INDEX);
+	if (r) {
+		dev_err(rdev->dev, "failed initializing CP fences (%d).\n", r);
+		return r;
+	}
+
 	/* Enable IRQ */
 	r100_irq_set(rdev);
 	rdev->config.r300.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
@@ -1405,11 +1414,18 @@ static int r300_startup(struct radeon_device *rdev)
 		dev_err(rdev->dev, "failed initializing CP (%d).\n", r);
 		return r;
 	}
-	r = r100_ib_init(rdev);
+
+	r = radeon_ib_pool_start(rdev);
+	if (r)
+		return r;
+
+	r = r100_ib_test(rdev);
 	if (r) {
-		dev_err(rdev->dev, "failed initializing IB (%d).\n", r);
+		dev_err(rdev->dev, "failed testing IB (%d).\n", r);
+		rdev->accel_working = false;
 		return r;
 	}
+
 	return 0;
 }
 
@@ -1434,11 +1450,14 @@ int r300_resume(struct radeon_device *rdev)
 	r300_clock_startup(rdev);
 	/* Initialize surface registers */
 	radeon_surface_init(rdev);
+
+	rdev->accel_working = true;
 	return r300_startup(rdev);
 }
 
 int r300_suspend(struct radeon_device *rdev)
 {
+	radeon_ib_pool_suspend(rdev);
 	r100_cp_disable(rdev);
 	radeon_wb_disable(rdev);
 	r100_irq_disable(rdev);
@@ -1539,7 +1558,14 @@ int r300_init(struct radeon_device *rdev)
 			return r;
 	}
 	r300_set_reg_safe(rdev);
+
+	r = radeon_ib_pool_init(rdev);
 	rdev->accel_working = true;
+	if (r) {
+		dev_err(rdev->dev, "IB initialization failed (%d).\n", r);
+		rdev->accel_working = false;
+	}
+
 	r = r300_startup(rdev);
 	if (r) {
 		/* Somethings want wront with the accel init stop accel */

drivers/gpu/drm/radeon/r420.c

@@ -199,6 +199,8 @@ static void r420_clock_resume(struct radeon_device *rdev)
 
 static void r420_cp_errata_init(struct radeon_device *rdev)
 {
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+
 	/* RV410 and R420 can lock up if CP DMA to host memory happens
 	 * while the 2D engine is busy.
 	 *
@@ -206,22 +208,24 @@ static void r420_cp_errata_init(struct radeon_device *rdev)
 	 * of the CP init, apparently.
 	 */
 	radeon_scratch_get(rdev, &rdev->config.r300.resync_scratch);
-	radeon_ring_lock(rdev, 8);
-	radeon_ring_write(rdev, PACKET0(R300_CP_RESYNC_ADDR, 1));
-	radeon_ring_write(rdev, rdev->config.r300.resync_scratch);
-	radeon_ring_write(rdev, 0xDEADBEEF);
-	radeon_ring_unlock_commit(rdev);
+	radeon_ring_lock(rdev, ring, 8);
+	radeon_ring_write(ring, PACKET0(R300_CP_RESYNC_ADDR, 1));
+	radeon_ring_write(ring, rdev->config.r300.resync_scratch);
+	radeon_ring_write(ring, 0xDEADBEEF);
+	radeon_ring_unlock_commit(rdev, ring);
 }
 
 static void r420_cp_errata_fini(struct radeon_device *rdev)
 {
+	struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+
 	/* Catch the RESYNC we dispatched all the way back,
 	 * at the very beginning of the CP init.
 	 */
-	radeon_ring_lock(rdev, 8);
-	radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
-	radeon_ring_write(rdev, R300_RB3D_DC_FINISH);
-	radeon_ring_unlock_commit(rdev);
+	radeon_ring_lock(rdev, ring, 8);
+	radeon_ring_write(ring, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
+	radeon_ring_write(ring, R300_RB3D_DC_FINISH);
+	radeon_ring_unlock_commit(rdev, ring);
 	radeon_scratch_free(rdev, rdev->config.r300.resync_scratch);
 }
 
@@ -254,6 +258,12 @@ static int r420_startup(struct radeon_device *rdev)
 	if (r)
 		return r;
 
+	r = radeon_fence_driver_start_ring(rdev, RADEON_RING_TYPE_GFX_INDEX);
+	if (r) {
+		dev_err(rdev->dev, "failed initializing CP fences (%d).\n", r);
+		return r;
+	}
+
 	/* Enable IRQ */
 	r100_irq_set(rdev);
 	rdev->config.r300.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
@@ -264,11 +274,18 @@ static int r420_startup(struct radeon_device *rdev)
 		return r;
 	}
 	r420_cp_errata_init(rdev);
-	r = r100_ib_init(rdev);
+
+	r = radeon_ib_pool_start(rdev);
+	if (r)
+		return r;
+
+	r = r100_ib_test(rdev);
 	if (r) {
-		dev_err(rdev->dev, "failed initializing IB (%d).\n", r);
+		dev_err(rdev->dev, "failed testing IB (%d).\n", r);
+		rdev->accel_working = false;
 		return r;
 	}
+
 	return 0;
 }
 
@@ -297,11 +314,14 @@ int r420_resume(struct radeon_device *rdev)
 	r420_clock_resume(rdev);
 	/* Initialize surface registers */
 	radeon_surface_init(rdev);
+
+	rdev->accel_working = true;
 	return r420_startup(rdev);
 }
 
 int r420_suspend(struct radeon_device *rdev)
 {
+	radeon_ib_pool_suspend(rdev);
 	r420_cp_errata_fini(rdev);
 	r100_cp_disable(rdev);
 	radeon_wb_disable(rdev);
@@ -414,7 +434,14 @@ int r420_init(struct radeon_device *rdev)
 			return r;
 	}
 	r420_set_reg_safe(rdev);
+
+	r = radeon_ib_pool_init(rdev);
 	rdev->accel_working = true;
+	if (r) {
+		dev_err(rdev->dev, "IB initialization failed (%d).\n", r);
+		rdev->accel_working = false;
+	}
+
 	r = r420_startup(rdev);
 	if (r) {
 		/* Somethings want wront with the accel init stop accel */

drivers/gpu/drm/radeon/r520.c

@@ -187,6 +187,12 @@ static int r520_startup(struct radeon_device *rdev)
 	if (r)
 		return r;
 
+	r = radeon_fence_driver_start_ring(rdev, RADEON_RING_TYPE_GFX_INDEX);
+	if (r) {
+		dev_err(rdev->dev, "failed initializing CP fences (%d).\n", r);
+		return r;
+	}
+
 	/* Enable IRQ */
 	rs600_irq_set(rdev);
 	rdev->config.r300.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL);
@@ -196,9 +202,15 @@ static int r520_startup(struct radeon_device *rdev)
 		dev_err(rdev->dev, "failed initializing CP (%d).\n", r);
 		return r;
 	}
-	r = r100_ib_init(rdev);
+
+	r = radeon_ib_pool_start(rdev);
+	if (r)
+		return r;
+
+	r = r100_ib_test(rdev);
 	if (r) {
-		dev_err(rdev->dev, "failed initializing IB (%d).\n", r);
+		dev_err(rdev->dev, "failed testing IB (%d).\n", r);
+		rdev->accel_working = false;
 		return r;
 	}
 	return 0;
@@ -223,6 +235,8 @@ int r520_resume(struct radeon_device *rdev)
 	rv515_clock_startup(rdev);
 	/* Initialize surface registers */
 	radeon_surface_init(rdev);
+
+	rdev->accel_working = true;
 	return r520_startup(rdev);
 }
 
@@ -292,7 +306,14 @@ int r520_init(struct radeon_device *rdev)
 	if (r)
 		return r;
 	rv515_set_safe_registers(rdev);
+
+	r = radeon_ib_pool_init(rdev);
 	rdev->accel_working = true;
+	if (r) {
+		dev_err(rdev->dev, "IB initialization failed (%d).\n", r);
+		rdev->accel_working = false;
+	}
+
 	r = r520_startup(rdev);
 	if (r) {
 		/* Somethings want wront with the accel init stop accel */

drivers/gpu/drm/radeon/r600_cp.c

@@ -1815,7 +1815,7 @@ static void r600_cp_init_ring_buffer(struct drm_device *dev,
 		     dev_priv->ring.size_l2qw);
 #endif
 
-	RADEON_WRITE(R600_CP_SEM_WAIT_TIMER, 0x4);
+	RADEON_WRITE(R600_CP_SEM_WAIT_TIMER, 0x0);
 
 	/* Set the write pointer delay */
 	RADEON_WRITE(R600_CP_RB_WPTR_DELAY, 0);

drivers/gpu/drm/radeon/r600d.h

@@ -831,6 +831,8 @@
 #define	PACKET3_STRMOUT_BUFFER_UPDATE			0x34
 #define	PACKET3_INDIRECT_BUFFER_MP			0x38
 #define	PACKET3_MEM_SEMAPHORE				0x39
+#              define PACKET3_SEM_SEL_SIGNAL	    (0x6 << 29)
+#              define PACKET3_SEM_SEL_WAIT	    (0x7 << 29)
 #define	PACKET3_MPEG_INDEX				0x3A
 #define	PACKET3_WAIT_REG_MEM				0x3C
 #define	PACKET3_MEM_WRITE				0x3D

drivers/gpu/drm/radeon/radeon_asic.h

@@ -58,17 +58,20 @@ void r100_fini(struct radeon_device *rdev);
 int r100_suspend(struct radeon_device *rdev);
 int r100_resume(struct radeon_device *rdev);
 void r100_vga_set_state(struct radeon_device *rdev, bool state);
-bool r100_gpu_is_lockup(struct radeon_device *rdev);
+bool r100_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *cp);
 int r100_asic_reset(struct radeon_device *rdev);
 u32 r100_get_vblank_counter(struct radeon_device *rdev, int crtc);
 void r100_pci_gart_tlb_flush(struct radeon_device *rdev);
 int r100_pci_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr);
-void r100_cp_commit(struct radeon_device *rdev);
 void r100_ring_start(struct radeon_device *rdev);
 int r100_irq_set(struct radeon_device *rdev);
 int r100_irq_process(struct radeon_device *rdev);
 void r100_fence_ring_emit(struct radeon_device *rdev,
 			  struct radeon_fence *fence);
+void r100_semaphore_ring_emit(struct radeon_device *rdev,
+			      struct radeon_ring *cp,
+			      struct radeon_semaphore *semaphore,
+			      bool emit_wait);
 int r100_cs_parse(struct radeon_cs_parser *p);
 void r100_pll_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
 uint32_t r100_pll_rreg(struct radeon_device *rdev, uint32_t reg);
@@ -83,7 +86,7 @@ int r100_set_surface_reg(struct radeon_device *rdev, int reg,
 void r100_clear_surface_reg(struct radeon_device *rdev, int reg);
 void r100_bandwidth_update(struct radeon_device *rdev);
 void r100_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib);
-int r100_ring_test(struct radeon_device *rdev);
+int r100_ring_test(struct radeon_device *rdev, struct radeon_ring *cp);
 void r100_hpd_init(struct radeon_device *rdev);
 void r100_hpd_fini(struct radeon_device *rdev);
 bool r100_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd);
@@ -101,12 +104,12 @@ void r100_pci_gart_disable(struct radeon_device *rdev);
 int r100_debugfs_mc_info_init(struct radeon_device *rdev);
 int r100_gui_wait_for_idle(struct radeon_device *rdev);
 void r100_gpu_lockup_update(struct r100_gpu_lockup *lockup,
-			    struct radeon_cp *cp);
+			    struct radeon_ring *cp);
 bool r100_gpu_cp_is_lockup(struct radeon_device *rdev,
 			   struct r100_gpu_lockup *lockup,
-			   struct radeon_cp *cp);
+			   struct radeon_ring *cp);
 void r100_ib_fini(struct radeon_device *rdev);
-int r100_ib_init(struct radeon_device *rdev);
+int r100_ib_test(struct radeon_device *rdev);
 void r100_irq_disable(struct radeon_device *rdev);
 void r100_mc_stop(struct radeon_device *rdev, struct r100_mc_save *save);
 void r100_mc_resume(struct radeon_device *rdev, struct r100_mc_save *save);
@@ -154,7 +157,7 @@ extern int r300_init(struct radeon_device *rdev);
 extern void r300_fini(struct radeon_device *rdev);
 extern int r300_suspend(struct radeon_device *rdev);
 extern int r300_resume(struct radeon_device *rdev);
-extern bool r300_gpu_is_lockup(struct radeon_device *rdev);
+extern bool r300_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *cp);
 extern int r300_asic_reset(struct radeon_device *rdev);
 extern void r300_ring_start(struct radeon_device *rdev);
 extern void r300_fence_ring_emit(struct radeon_device *rdev,
@@ -293,22 +296,25 @@ int r600_resume(struct radeon_device *rdev);
 void r600_vga_set_state(struct radeon_device *rdev, bool state);
 int r600_wb_init(struct radeon_device *rdev);
 void r600_wb_fini(struct radeon_device *rdev);
-void r600_cp_commit(struct radeon_device *rdev);
 void r600_pcie_gart_tlb_flush(struct radeon_device *rdev);
 uint32_t r600_pciep_rreg(struct radeon_device *rdev, uint32_t reg);
 void r600_pciep_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
 int r600_cs_parse(struct radeon_cs_parser *p);
 void r600_fence_ring_emit(struct radeon_device *rdev,
 			  struct radeon_fence *fence);
-bool r600_gpu_is_lockup(struct radeon_device *rdev);
+void r600_semaphore_ring_emit(struct radeon_device *rdev,
+			      struct radeon_ring *cp,
+			      struct radeon_semaphore *semaphore,
+			      bool emit_wait);
+bool r600_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *cp);
 int r600_asic_reset(struct radeon_device *rdev);
 int r600_set_surface_reg(struct radeon_device *rdev, int reg,
 			 uint32_t tiling_flags, uint32_t pitch,
 			 uint32_t offset, uint32_t obj_size);
 void r600_clear_surface_reg(struct radeon_device *rdev, int reg);
-int r600_ib_test(struct radeon_device *rdev);
+int r600_ib_test(struct radeon_device *rdev, int ring);
 void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib);
-int r600_ring_test(struct radeon_device *rdev);
+int r600_ring_test(struct radeon_device *rdev, struct radeon_ring *cp);
 int r600_copy_blit(struct radeon_device *rdev,
 		   uint64_t src_offset, uint64_t dst_offset,
 		   unsigned num_gpu_pages, struct radeon_fence *fence);
@@ -328,7 +334,7 @@ extern int r600_get_pcie_lanes(struct radeon_device *rdev);
 bool r600_card_posted(struct radeon_device *rdev);
 void r600_cp_stop(struct radeon_device *rdev);
 int r600_cp_start(struct radeon_device *rdev);
-void r600_ring_init(struct radeon_device *rdev, unsigned ring_size);
+void r600_ring_init(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ring_size);
 int r600_cp_resume(struct radeon_device *rdev);
 void r600_cp_fini(struct radeon_device *rdev);
 int r600_count_pipe_bits(uint32_t val);
@@ -397,7 +403,7 @@ int evergreen_init(struct radeon_device *rdev);
 void evergreen_fini(struct radeon_device *rdev);
 int evergreen_suspend(struct radeon_device *rdev);
 int evergreen_resume(struct radeon_device *rdev);
-bool evergreen_gpu_is_lockup(struct radeon_device *rdev);
+bool evergreen_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *cp);
 int evergreen_asic_reset(struct radeon_device *rdev);
 void evergreen_bandwidth_update(struct radeon_device *rdev);
 void evergreen_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib);
@@ -423,12 +429,14 @@ int evergreen_blit_init(struct radeon_device *rdev);
 /*
  * cayman
  */
+void cayman_fence_ring_emit(struct radeon_device *rdev,
+			    struct radeon_fence *fence);
 void cayman_pcie_gart_tlb_flush(struct radeon_device *rdev);
 int cayman_init(struct radeon_device *rdev);
 void cayman_fini(struct radeon_device *rdev);
 int cayman_suspend(struct radeon_device *rdev);
 int cayman_resume(struct radeon_device *rdev);
-bool cayman_gpu_is_lockup(struct radeon_device *rdev);
+bool cayman_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *cp);
 int cayman_asic_reset(struct radeon_device *rdev);
 
 #endif

drivers/gpu/drm/radeon/radeon_benchmark.c

@@ -43,7 +43,7 @@ static int radeon_benchmark_do_move(struct radeon_device *rdev, unsigned size,
 
 	start_jiffies = jiffies;
 	for (i = 0; i < n; i++) {
-		r = radeon_fence_create(rdev, &fence);
+		r = radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
 		if (r)
 			return r;
 

drivers/gpu/drm/radeon/radeon_cs.c

@@ -58,7 +58,7 @@ int radeon_cs_parser_relocs(struct radeon_cs_parser *p)
 
 		duplicate = false;
 		r = (struct drm_radeon_cs_reloc *)&chunk->kdata[i*4];
-		for (j = 0; j < p->nrelocs; j++) {
+		for (j = 0; j < i; j++) {
 			if (r->handle == p->relocs[j].handle) {
 				p->relocs_ptr[i] = &p->relocs[j];
 				duplicate = true;
@@ -84,7 +84,8 @@ int radeon_cs_parser_relocs(struct radeon_cs_parser *p)
 			p->relocs[i].flags = r->flags;
 			radeon_bo_list_add_object(&p->relocs[i].lobj,
 						  &p->validated);
-		}
+		} else
+			p->relocs[i].handle = 0;
 	}
 	return radeon_bo_list_validate(&p->validated);
 }
@@ -245,7 +246,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		radeon_mutex_unlock(&rdev->cs_mutex);
 		return r;
 	}
-	r =  radeon_ib_get(rdev, &parser.ib);
+	r =  radeon_ib_get(rdev, RADEON_RING_TYPE_GFX_INDEX, &parser.ib);
 	if (r) {
 		DRM_ERROR("Failed to get ib !\n");
 		radeon_cs_parser_fini(&parser, r);

drivers/gpu/drm/radeon/radeon_device.c

@@ -718,17 +718,20 @@ int radeon_device_init(struct radeon_device *rdev,
 	 * can recall function without having locking issues */
 	radeon_mutex_init(&rdev->cs_mutex);
 	mutex_init(&rdev->ib_pool.mutex);
-	mutex_init(&rdev->cp.mutex);
+	for (i = 0; i < RADEON_NUM_RINGS; ++i)
+		mutex_init(&rdev->ring[i].mutex);
 	mutex_init(&rdev->dc_hw_i2c_mutex);
 	if (rdev->family >= CHIP_R600)
 		spin_lock_init(&rdev->ih.lock);
 	mutex_init(&rdev->gem.mutex);
 	mutex_init(&rdev->pm.mutex);
 	mutex_init(&rdev->vram_mutex);
-	rwlock_init(&rdev->fence_drv.lock);
+	rwlock_init(&rdev->fence_lock);
+	rwlock_init(&rdev->semaphore_drv.lock);
 	INIT_LIST_HEAD(&rdev->gem.objects);
 	init_waitqueue_head(&rdev->irq.vblank_queue);
 	init_waitqueue_head(&rdev->irq.idle_queue);
+	INIT_LIST_HEAD(&rdev->semaphore_drv.free);
 
 	/* Set asic functions */
 	r = radeon_asic_init(rdev);
@@ -820,15 +823,20 @@ int radeon_device_init(struct radeon_device *rdev,
 		if (r)
 			return r;
 	}
-	if (radeon_testing) {
+	if ((radeon_testing & 1)) {
 		radeon_test_moves(rdev);
 	}
+	if ((radeon_testing & 2)) {
+		radeon_test_syncing(rdev);
+	}
 	if (radeon_benchmarking) {
 		radeon_benchmark(rdev, radeon_benchmarking);
 	}
 	return 0;
 }
 
+static void radeon_debugfs_remove_files(struct radeon_device *rdev);
+
 void radeon_device_fini(struct radeon_device *rdev)
 {
 	DRM_INFO("radeon: finishing device.\n");
@@ -843,6 +851,7 @@ void radeon_device_fini(struct radeon_device *rdev)
 	rdev->rio_mem = NULL;
 	iounmap(rdev->rmmio);
 	rdev->rmmio = NULL;
+	radeon_debugfs_remove_files(rdev);
 }
 
 
@@ -854,7 +863,7 @@ int radeon_suspend_kms(struct drm_device *dev, pm_message_t state)
 	struct radeon_device *rdev;
 	struct drm_crtc *crtc;
 	struct drm_connector *connector;
-	int r;
+	int i, r;
 
 	if (dev == NULL || dev->dev_private == NULL) {
 		return -ENODEV;
@@ -893,7 +902,8 @@ int radeon_suspend_kms(struct drm_device *dev, pm_message_t state)
 	/* evict vram memory */
 	radeon_bo_evict_vram(rdev);
 	/* wait for gpu to finish processing current batch */
-	radeon_fence_wait_last(rdev);
+	for (i = 0; i < RADEON_NUM_RINGS; i++)
+		radeon_fence_wait_last(rdev, i);
 
 	radeon_save_bios_scratch_regs(rdev);
 
@@ -992,36 +1002,29 @@ int radeon_gpu_reset(struct radeon_device *rdev)
 /*
  * Debugfs
  */
-struct radeon_debugfs {
-	struct drm_info_list	*files;
-	unsigned		num_files;
-};
-static struct radeon_debugfs _radeon_debugfs[RADEON_DEBUGFS_MAX_COMPONENTS];
-static unsigned _radeon_debugfs_count = 0;
-
 int radeon_debugfs_add_files(struct radeon_device *rdev,
 			     struct drm_info_list *files,
 			     unsigned nfiles)
 {
 	unsigned i;
 
-	for (i = 0; i < _radeon_debugfs_count; i++) {
-		if (_radeon_debugfs[i].files == files) {
+	for (i = 0; i < rdev->debugfs_count; i++) {
+		if (rdev->debugfs[i].files == files) {
 			/* Already registered */
 			return 0;
 		}
 	}
 
-	i = _radeon_debugfs_count + 1;
+	i = rdev->debugfs_count + 1;
 	if (i > RADEON_DEBUGFS_MAX_COMPONENTS) {
 		DRM_ERROR("Reached maximum number of debugfs components.\n");
 		DRM_ERROR("Report so we increase "
 		          "RADEON_DEBUGFS_MAX_COMPONENTS.\n");
 		return -EINVAL;
 	}
-	_radeon_debugfs[_radeon_debugfs_count].files = files;
-	_radeon_debugfs[_radeon_debugfs_count].num_files = nfiles;
-	_radeon_debugfs_count = i;
+	rdev->debugfs[rdev->debugfs_count].files = files;
+	rdev->debugfs[rdev->debugfs_count].num_files = nfiles;
+	rdev->debugfs_count = i;
 #if defined(CONFIG_DEBUG_FS)
 	drm_debugfs_create_files(files, nfiles,
 				 rdev->ddev->control->debugfs_root,
@@ -1033,6 +1036,22 @@ int radeon_debugfs_add_files(struct radeon_device *rdev,
 	return 0;
 }
 
+static void radeon_debugfs_remove_files(struct radeon_device *rdev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	unsigned i;
+
+	for (i = 0; i < rdev->debugfs_count; i++) {
+		drm_debugfs_remove_files(rdev->debugfs[i].files,
+					 rdev->debugfs[i].num_files,
+					 rdev->ddev->control);
+		drm_debugfs_remove_files(rdev->debugfs[i].files,
+					 rdev->debugfs[i].num_files,
+					 rdev->ddev->primary);
+	}
+#endif
+}
+
 #if defined(CONFIG_DEBUG_FS)
 int radeon_debugfs_init(struct drm_minor *minor)
 {
@@ -1041,11 +1060,5 @@ int radeon_debugfs_init(struct drm_minor *minor)
 
 void radeon_debugfs_cleanup(struct drm_minor *minor)
 {
-	unsigned i;
-
-	for (i = 0; i < _radeon_debugfs_count; i++) {
-		drm_debugfs_remove_files(_radeon_debugfs[i].files,
-					 _radeon_debugfs[i].num_files, minor);
-	}
 }
 #endif

drivers/gpu/drm/radeon/radeon_gem.c

@@ -152,6 +152,7 @@ int radeon_gem_info_ioctl(struct drm_device *dev, void *data,
 	struct radeon_device *rdev = dev->dev_private;
 	struct drm_radeon_gem_info *args = data;
 	struct ttm_mem_type_manager *man;
+	unsigned i;
 
 	man = &rdev->mman.bdev.man[TTM_PL_VRAM];
 
@@ -160,8 +161,9 @@ int radeon_gem_info_ioctl(struct drm_device *dev, void *data,
 	if (rdev->stollen_vga_memory)
 		args->vram_visible -= radeon_bo_size(rdev->stollen_vga_memory);
 	args->vram_visible -= radeon_fbdev_total_size(rdev);
-	args->gart_size = rdev->mc.gtt_size - rdev->cp.ring_size - 4096 -
-		RADEON_IB_POOL_SIZE*64*1024;
+	args->gart_size = rdev->mc.gtt_size - 4096 - RADEON_IB_POOL_SIZE*64*1024;
+	for(i = 0; i < RADEON_NUM_RINGS; ++i)
+		args->gart_size -= rdev->ring[i].ring_size;
 	return 0;
 }
 

drivers/gpu/drm/radeon/radeon_irq_kms.c

@@ -65,7 +65,8 @@ void radeon_driver_irq_preinstall_kms(struct drm_device *dev)
 	unsigned i;
 
 	/* Disable *all* interrupts */
-	rdev->irq.sw_int = false;
+	for (i = 0; i < RADEON_NUM_RINGS; i++)
+		rdev->irq.sw_int[i] = false;
 	rdev->irq.gui_idle = false;
 	for (i = 0; i < RADEON_MAX_HPD_PINS; i++)
 		rdev->irq.hpd[i] = false;
@@ -81,9 +82,11 @@ void radeon_driver_irq_preinstall_kms(struct drm_device *dev)
 int radeon_driver_irq_postinstall_kms(struct drm_device *dev)
 {
 	struct radeon_device *rdev = dev->dev_private;
+	unsigned i;
 
 	dev->max_vblank_count = 0x001fffff;
-	rdev->irq.sw_int = true;
+	for (i = 0; i < RADEON_NUM_RINGS; i++)
+		rdev->irq.sw_int[i] = true;
 	radeon_irq_set(rdev);
 	return 0;
 }
@@ -97,7 +100,8 @@ void radeon_driver_irq_uninstall_kms(struct drm_device *dev)
 		return;
 	}
 	/* Disable *all* interrupts */
-	rdev->irq.sw_int = false;
+	for (i = 0; i < RADEON_NUM_RINGS; i++)
+		rdev->irq.sw_int[i] = false;
 	rdev->irq.gui_idle = false;
 	for (i = 0; i < RADEON_MAX_HPD_PINS; i++)
 		rdev->irq.hpd[i] = false;
@@ -194,26 +198,26 @@ void radeon_irq_kms_fini(struct radeon_device *rdev)
 	flush_work_sync(&rdev->hotplug_work);
 }
 
-void radeon_irq_kms_sw_irq_get(struct radeon_device *rdev)
+void radeon_irq_kms_sw_irq_get(struct radeon_device *rdev, int ring)
 {
 	unsigned long irqflags;
 
 	spin_lock_irqsave(&rdev->irq.sw_lock, irqflags);
-	if (rdev->ddev->irq_enabled && (++rdev->irq.sw_refcount == 1)) {
-		rdev->irq.sw_int = true;
+	if (rdev->ddev->irq_enabled && (++rdev->irq.sw_refcount[ring] == 1)) {
+		rdev->irq.sw_int[ring] = true;
 		radeon_irq_set(rdev);
 	}
 	spin_unlock_irqrestore(&rdev->irq.sw_lock, irqflags);
 }
 
-void radeon_irq_kms_sw_irq_put(struct radeon_device *rdev)
+void radeon_irq_kms_sw_irq_put(struct radeon_device *rdev, int ring)
 {
 	unsigned long irqflags;
 
 	spin_lock_irqsave(&rdev->irq.sw_lock, irqflags);
-	BUG_ON(rdev->ddev->irq_enabled && rdev->irq.sw_refcount <= 0);
-	if (rdev->ddev->irq_enabled && (--rdev->irq.sw_refcount == 0)) {
-		rdev->irq.sw_int = false;
+	BUG_ON(rdev->ddev->irq_enabled && rdev->irq.sw_refcount[ring] <= 0);
+	if (rdev->ddev->irq_enabled && (--rdev->irq.sw_refcount[ring] == 0)) {
+		rdev->irq.sw_int[ring] = false;
 		radeon_irq_set(rdev);
 	}
 	spin_unlock_irqrestore(&rdev->irq.sw_lock, irqflags);

drivers/gpu/drm/radeon/radeon_object.h

@@ -128,4 +128,24 @@ extern void radeon_bo_move_notify(struct ttm_buffer_object *bo,
 					struct ttm_mem_reg *mem);
 extern int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo);
 extern int radeon_bo_get_surface_reg(struct radeon_bo *bo);
+
+/*
+ * sub allocation
+ */
+extern int radeon_sa_bo_manager_init(struct radeon_device *rdev,
+				     struct radeon_sa_manager *sa_manager,
+				     unsigned size, u32 domain);
+extern void radeon_sa_bo_manager_fini(struct radeon_device *rdev,
+				      struct radeon_sa_manager *sa_manager);
+extern int radeon_sa_bo_manager_start(struct radeon_device *rdev,
+				      struct radeon_sa_manager *sa_manager);
+extern int radeon_sa_bo_manager_suspend(struct radeon_device *rdev,
+					struct radeon_sa_manager *sa_manager);
+extern int radeon_sa_bo_new(struct radeon_device *rdev,
+			    struct radeon_sa_manager *sa_manager,
+			    struct radeon_sa_bo *sa_bo,
+			    unsigned size, unsigned align);
+extern void radeon_sa_bo_free(struct radeon_device *rdev,
+			      struct radeon_sa_bo *sa_bo);
+
 #endif

drivers/gpu/drm/radeon/radeon_pm.c

@@ -252,7 +252,10 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev)
 
 	mutex_lock(&rdev->ddev->struct_mutex);
 	mutex_lock(&rdev->vram_mutex);
-	mutex_lock(&rdev->cp.mutex);
+	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
+		if (rdev->ring[i].ring_obj)
+			mutex_lock(&rdev->ring[i].mutex);
+	}
 
 	/* gui idle int has issues on older chips it seems */
 	if (rdev->family >= CHIP_R600) {
@@ -268,12 +271,13 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev)
 			radeon_irq_set(rdev);
 		}
 	} else {
-		if (rdev->cp.ready) {
+		struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
+		if (ring->ready) {
 			struct radeon_fence *fence;
-			radeon_ring_alloc(rdev, 64);
-			radeon_fence_create(rdev, &fence);
+			radeon_ring_alloc(rdev, ring, 64);
+			radeon_fence_create(rdev, &fence, radeon_ring_index(rdev, ring));
 			radeon_fence_emit(rdev, fence);
-			radeon_ring_commit(rdev);
+			radeon_ring_commit(rdev, ring);
 			radeon_fence_wait(fence, false);
 			radeon_fence_unref(&fence);
 		}
@@ -307,7 +311,10 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev)
 
 	rdev->pm.dynpm_planned_action = DYNPM_ACTION_NONE;
 
-	mutex_unlock(&rdev->cp.mutex);
+	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
+		if (rdev->ring[i].ring_obj)
+			mutex_unlock(&rdev->ring[i].mutex);
+	}
 	mutex_unlock(&rdev->vram_mutex);
 	mutex_unlock(&rdev->ddev->struct_mutex);
 }
@@ -795,19 +802,14 @@ static void radeon_dynpm_idle_work_handler(struct work_struct *work)
 	resched = ttm_bo_lock_delayed_workqueue(&rdev->mman.bdev);
 	mutex_lock(&rdev->pm.mutex);
 	if (rdev->pm.dynpm_state == DYNPM_STATE_ACTIVE) {
-		unsigned long irq_flags;
 		int not_processed = 0;
+		int i;
 
-		read_lock_irqsave(&rdev->fence_drv.lock, irq_flags);
-		if (!list_empty(&rdev->fence_drv.emited)) {
-			struct list_head *ptr;
-			list_for_each(ptr, &rdev->fence_drv.emited) {
-				/* count up to 3, that's enought info */
-				if (++not_processed >= 3)
-					break;
-			}
+		for (i = 0; i < RADEON_NUM_RINGS; ++i) {
+			not_processed += radeon_fence_count_emitted(rdev, i);
+			if (not_processed >= 3)
+				break;
 		}
-		read_unlock_irqrestore(&rdev->fence_drv.lock, irq_flags);
 
 		if (not_processed >= 3) { /* should upclock */
 			if (rdev->pm.dynpm_planned_action == DYNPM_ACTION_DOWNCLOCK) {

drivers/gpu/drm/radeon/radeon_ttm.c

@@ -188,7 +188,7 @@ static void radeon_evict_flags(struct ttm_buffer_object *bo,
 	rbo = container_of(bo, struct radeon_bo, tbo);
 	switch (bo->mem.mem_type) {
 	case TTM_PL_VRAM:
-		if (rbo->rdev->cp.ready == false)
+		if (rbo->rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready == false)
 			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_CPU);
 		else
 			radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_GTT);
@@ -226,7 +226,7 @@ static int radeon_move_blit(struct ttm_buffer_object *bo,
 	int r;
 
 	rdev = radeon_get_rdev(bo->bdev);
-	r = radeon_fence_create(rdev, &fence);
+	r = radeon_fence_create(rdev, &fence, RADEON_RING_TYPE_GFX_INDEX);
 	if (unlikely(r)) {
 		return r;
 	}
@@ -255,7 +255,7 @@ static int radeon_move_blit(struct ttm_buffer_object *bo,
 		DRM_ERROR("Unknown placement %d\n", old_mem->mem_type);
 		return -EINVAL;
 	}
-	if (!rdev->cp.ready) {
+	if (!rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready) {
 		DRM_ERROR("Trying to move memory with CP turned off.\n");
 		return -EINVAL;
 	}
@@ -380,7 +380,7 @@ static int radeon_bo_move(struct ttm_buffer_object *bo,
 		radeon_move_null(bo, new_mem);
 		return 0;
 	}
-	if (!rdev->cp.ready || rdev->asic->copy == NULL) {
+	if (!rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready || rdev->asic->copy == NULL) {
 		/* use memcpy */
 		goto memcpy;
 	}