Merge branch 'next' into for-linus

Prepare first set of updates for 3.10 merge window.

1 files changed, 704 insertions, 107 deletions

diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c
index b0db712060f..9128120da04 100644
--- a/drivers/gpu/drm/radeon/si.c
+++ b/drivers/gpu/drm/radeon/si.c
@@ -38,6 +38,7 @@
 #define SI_CE_UCODE_SIZE 2144
 #define SI_RLC_UCODE_SIZE 2048
 #define SI_MC_UCODE_SIZE 7769
+#define OLAND_MC_UCODE_SIZE 7863
 
 MODULE_FIRMWARE("radeon/TAHITI_pfp.bin");
 MODULE_FIRMWARE("radeon/TAHITI_me.bin");
@@ -54,6 +55,11 @@ MODULE_FIRMWARE("radeon/VERDE_me.bin");
 MODULE_FIRMWARE("radeon/VERDE_ce.bin");
 MODULE_FIRMWARE("radeon/VERDE_mc.bin");
 MODULE_FIRMWARE("radeon/VERDE_rlc.bin");
+MODULE_FIRMWARE("radeon/OLAND_pfp.bin");
+MODULE_FIRMWARE("radeon/OLAND_me.bin");
+MODULE_FIRMWARE("radeon/OLAND_ce.bin");
+MODULE_FIRMWARE("radeon/OLAND_mc.bin");
+MODULE_FIRMWARE("radeon/OLAND_rlc.bin");
 
 extern int r600_ih_ring_alloc(struct radeon_device *rdev);
 extern void r600_ih_ring_fini(struct radeon_device *rdev);
@@ -61,6 +67,35 @@ extern void evergreen_fix_pci_max_read_req_size(struct radeon_device *rdev);
 extern void evergreen_mc_stop(struct radeon_device *rdev, struct evergreen_mc_save *save);
 extern void evergreen_mc_resume(struct radeon_device *rdev, struct evergreen_mc_save *save);
 extern u32 evergreen_get_number_of_dram_channels(struct radeon_device *rdev);
+extern void evergreen_print_gpu_status_regs(struct radeon_device *rdev);
+extern bool evergreen_is_display_hung(struct radeon_device *rdev);
+
+#define PCIE_BUS_CLK                10000
+#define TCLK                        (PCIE_BUS_CLK / 10)
+
+/**
+ * si_get_xclk - get the xclk
+ *
+ * @rdev: radeon_device pointer
+ *
+ * Returns the reference clock used by the gfx engine
+ * (SI).
+ */
+u32 si_get_xclk(struct radeon_device *rdev)
+{
+        u32 reference_clock = rdev->clock.spll.reference_freq;
+	u32 tmp;
+
+	tmp = RREG32(CG_CLKPIN_CNTL_2);
+	if (tmp & MUX_TCLK_TO_XCLK)
+		return TCLK;
+
+	tmp = RREG32(CG_CLKPIN_CNTL);
+	if (tmp & XTALIN_DIVIDE)
+		return reference_clock / 4;
+
+	return reference_clock;
+}
 
 /* get temperature in millidegrees */
 int si_get_temp(struct radeon_device *rdev)
@@ -200,6 +235,45 @@ static const u32 verde_io_mc_regs[TAHITI_IO_MC_REGS_SIZE][2] = {
 	{0x0000009f, 0x00a37400}
 };
 
+static const u32 oland_io_mc_regs[TAHITI_IO_MC_REGS_SIZE][2] = {
+	{0x0000006f, 0x03044000},
+	{0x00000070, 0x0480c018},
+	{0x00000071, 0x00000040},
+	{0x00000072, 0x01000000},
+	{0x00000074, 0x000000ff},
+	{0x00000075, 0x00143400},
+	{0x00000076, 0x08ec0800},
+	{0x00000077, 0x040000cc},
+	{0x00000079, 0x00000000},
+	{0x0000007a, 0x21000409},
+	{0x0000007c, 0x00000000},
+	{0x0000007d, 0xe8000000},
+	{0x0000007e, 0x044408a8},
+	{0x0000007f, 0x00000003},
+	{0x00000080, 0x00000000},
+	{0x00000081, 0x01000000},
+	{0x00000082, 0x02000000},
+	{0x00000083, 0x00000000},
+	{0x00000084, 0xe3f3e4f4},
+	{0x00000085, 0x00052024},
+	{0x00000087, 0x00000000},
+	{0x00000088, 0x66036603},
+	{0x00000089, 0x01000000},
+	{0x0000008b, 0x1c0a0000},
+	{0x0000008c, 0xff010000},
+	{0x0000008e, 0xffffefff},
+	{0x0000008f, 0xfff3efff},
+	{0x00000090, 0xfff3efbf},
+	{0x00000094, 0x00101101},
+	{0x00000095, 0x00000fff},
+	{0x00000096, 0x00116fff},
+	{0x00000097, 0x60010000},
+	{0x00000098, 0x10010000},
+	{0x00000099, 0x00006000},
+	{0x0000009a, 0x00001000},
+	{0x0000009f, 0x00a17730}
+};
+
 /* ucode loading */
 static int si_mc_load_microcode(struct radeon_device *rdev)
 {
@@ -228,6 +302,11 @@ static int si_mc_load_microcode(struct radeon_device *rdev)
 		ucode_size = SI_MC_UCODE_SIZE;
 		regs_size = TAHITI_IO_MC_REGS_SIZE;
 		break;
+	case CHIP_OLAND:
+		io_mc_regs = (u32 *)&oland_io_mc_regs;
+		ucode_size = OLAND_MC_UCODE_SIZE;
+		regs_size = TAHITI_IO_MC_REGS_SIZE;
+		break;
 	}
 
 	running = RREG32(MC_SEQ_SUP_CNTL) & RUN_MASK;
@@ -322,6 +401,15 @@ static int si_init_microcode(struct radeon_device *rdev)
 		rlc_req_size = SI_RLC_UCODE_SIZE * 4;
 		mc_req_size = SI_MC_UCODE_SIZE * 4;
 		break;
+	case CHIP_OLAND:
+		chip_name = "OLAND";
+		rlc_chip_name = "OLAND";
+		pfp_req_size = SI_PFP_UCODE_SIZE * 4;
+		me_req_size = SI_PM4_UCODE_SIZE * 4;
+		ce_req_size = SI_CE_UCODE_SIZE * 4;
+		rlc_req_size = SI_RLC_UCODE_SIZE * 4;
+		mc_req_size = OLAND_MC_UCODE_SIZE * 4;
+		break;
 	default: BUG();
 	}
 
@@ -1125,7 +1213,8 @@ static void si_tiling_mode_table_init(struct radeon_device *rdev)
 			}
 			WREG32(GB_TILE_MODE0 + (reg_offset * 4), gb_tile_moden);
 		}
-	} else if (rdev->family == CHIP_VERDE) {
+	} else if ((rdev->family == CHIP_VERDE) ||
+		   (rdev->family == CHIP_OLAND)) {
 		for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++) {
 			switch (reg_offset) {
 			case 0:  /* non-AA compressed depth or any compressed stencil */
@@ -1570,6 +1659,23 @@ static void si_gpu_init(struct radeon_device *rdev)
 		rdev->config.si.sc_earlyz_tile_fifo_size = 0x130;
 		gb_addr_config = VERDE_GB_ADDR_CONFIG_GOLDEN;
 		break;
+	case CHIP_OLAND:
+		rdev->config.si.max_shader_engines = 1;
+		rdev->config.si.max_tile_pipes = 4;
+		rdev->config.si.max_cu_per_sh = 6;
+		rdev->config.si.max_sh_per_se = 1;
+		rdev->config.si.max_backends_per_se = 2;
+		rdev->config.si.max_texture_channel_caches = 4;
+		rdev->config.si.max_gprs = 256;
+		rdev->config.si.max_gs_threads = 16;
+		rdev->config.si.max_hw_contexts = 8;
+
+		rdev->config.si.sc_prim_fifo_size_frontend = 0x20;
+		rdev->config.si.sc_prim_fifo_size_backend = 0x40;
+		rdev->config.si.sc_hiz_tile_fifo_size = 0x30;
+		rdev->config.si.sc_earlyz_tile_fifo_size = 0x130;
+		gb_addr_config = VERDE_GB_ADDR_CONFIG_GOLDEN;
+		break;
 	}
 
 	/* Initialize HDP */
@@ -1660,6 +1766,8 @@ static void si_gpu_init(struct radeon_device *rdev)
 	WREG32(GB_ADDR_CONFIG, gb_addr_config);
 	WREG32(DMIF_ADDR_CONFIG, gb_addr_config);
 	WREG32(HDP_ADDR_CONFIG, gb_addr_config);
+	WREG32(DMA_TILING_CONFIG + DMA0_REGISTER_OFFSET, gb_addr_config);
+	WREG32(DMA_TILING_CONFIG + DMA1_REGISTER_OFFSET, gb_addr_config);
 
 	si_tiling_mode_table_init(rdev);
 
@@ -1836,6 +1944,9 @@ static void si_cp_enable(struct radeon_device *rdev, bool enable)
 		radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size);
 		WREG32(CP_ME_CNTL, (CP_ME_HALT | CP_PFP_HALT | CP_CE_HALT));
 		WREG32(SCRATCH_UMSK, 0);
+		rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = false;
+		rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX].ready = false;
+		rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX].ready = false;
 	}
 	udelay(50);
 }
@@ -2007,7 +2118,7 @@ static int si_cp_resume(struct radeon_device *rdev)
 	ring->wptr = 0;
 	WREG32(CP_RB0_WPTR, ring->wptr);
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	WREG32(CP_RB0_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC);
 	WREG32(CP_RB0_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFF);
 
@@ -2040,7 +2151,7 @@ static int si_cp_resume(struct radeon_device *rdev)
 	ring->wptr = 0;
 	WREG32(CP_RB1_WPTR, ring->wptr);
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	WREG32(CP_RB1_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP1_RPTR_OFFSET) & 0xFFFFFFFC);
 	WREG32(CP_RB1_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP1_RPTR_OFFSET) & 0xFF);
 
@@ -2066,7 +2177,7 @@ static int si_cp_resume(struct radeon_device *rdev)
 	ring->wptr = 0;
 	WREG32(CP_RB2_WPTR, ring->wptr);
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	WREG32(CP_RB2_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP2_RPTR_OFFSET) & 0xFFFFFFFC);
 	WREG32(CP_RB2_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP2_RPTR_OFFSET) & 0xFF);
 
@@ -2101,92 +2212,281 @@ static int si_cp_resume(struct radeon_device *rdev)
 	return 0;
 }
 
-bool si_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring)
+static u32 si_gpu_check_soft_reset(struct radeon_device *rdev)
 {
-	u32 srbm_status;
-	u32 grbm_status, grbm_status2;
-	u32 grbm_status_se0, grbm_status_se1;
+	u32 reset_mask = 0;
+	u32 tmp;
 
-	srbm_status = RREG32(SRBM_STATUS);
-	grbm_status = RREG32(GRBM_STATUS);
-	grbm_status2 = RREG32(GRBM_STATUS2);
-	grbm_status_se0 = RREG32(GRBM_STATUS_SE0);
-	grbm_status_se1 = RREG32(GRBM_STATUS_SE1);
-	if (!(grbm_status & GUI_ACTIVE)) {
-		radeon_ring_lockup_update(ring);
-		return false;
+	/* GRBM_STATUS */
+	tmp = RREG32(GRBM_STATUS);
+	if (tmp & (PA_BUSY | SC_BUSY |
+		   BCI_BUSY | SX_BUSY |
+		   TA_BUSY | VGT_BUSY |
+		   DB_BUSY | CB_BUSY |
+		   GDS_BUSY | SPI_BUSY |
+		   IA_BUSY | IA_BUSY_NO_DMA))
+		reset_mask |= RADEON_RESET_GFX;
+
+	if (tmp & (CF_RQ_PENDING | PF_RQ_PENDING |
+		   CP_BUSY | CP_COHERENCY_BUSY))
+		reset_mask |= RADEON_RESET_CP;
+
+	if (tmp & GRBM_EE_BUSY)
+		reset_mask |= RADEON_RESET_GRBM | RADEON_RESET_GFX | RADEON_RESET_CP;
+
+	/* GRBM_STATUS2 */
+	tmp = RREG32(GRBM_STATUS2);
+	if (tmp & (RLC_RQ_PENDING | RLC_BUSY))
+		reset_mask |= RADEON_RESET_RLC;
+
+	/* DMA_STATUS_REG 0 */
+	tmp = RREG32(DMA_STATUS_REG + DMA0_REGISTER_OFFSET);
+	if (!(tmp & DMA_IDLE))
+		reset_mask |= RADEON_RESET_DMA;
+
+	/* DMA_STATUS_REG 1 */
+	tmp = RREG32(DMA_STATUS_REG + DMA1_REGISTER_OFFSET);
+	if (!(tmp & DMA_IDLE))
+		reset_mask |= RADEON_RESET_DMA1;
+
+	/* SRBM_STATUS2 */
+	tmp = RREG32(SRBM_STATUS2);
+	if (tmp & DMA_BUSY)
+		reset_mask |= RADEON_RESET_DMA;
+
+	if (tmp & DMA1_BUSY)
+		reset_mask |= RADEON_RESET_DMA1;
+
+	/* SRBM_STATUS */
+	tmp = RREG32(SRBM_STATUS);
+
+	if (tmp & IH_BUSY)
+		reset_mask |= RADEON_RESET_IH;
+
+	if (tmp & SEM_BUSY)
+		reset_mask |= RADEON_RESET_SEM;
+
+	if (tmp & GRBM_RQ_PENDING)
+		reset_mask |= RADEON_RESET_GRBM;
+
+	if (tmp & VMC_BUSY)
+		reset_mask |= RADEON_RESET_VMC;
+
+	if (tmp & (MCB_BUSY | MCB_NON_DISPLAY_BUSY |
+		   MCC_BUSY | MCD_BUSY))
+		reset_mask |= RADEON_RESET_MC;
+
+	if (evergreen_is_display_hung(rdev))
+		reset_mask |= RADEON_RESET_DISPLAY;
+
+	/* VM_L2_STATUS */
+	tmp = RREG32(VM_L2_STATUS);
+	if (tmp & L2_BUSY)
+		reset_mask |= RADEON_RESET_VMC;
+
+	/* Skip MC reset as it's mostly likely not hung, just busy */
+	if (reset_mask & RADEON_RESET_MC) {
+		DRM_DEBUG("MC busy: 0x%08X, clearing.\n", reset_mask);
+		reset_mask &= ~RADEON_RESET_MC;
 	}
-	/* force CP activities */
-	radeon_ring_force_activity(rdev, ring);
-	return radeon_ring_test_lockup(rdev, ring);
+
+	return reset_mask;
 }
 
-static int si_gpu_soft_reset(struct radeon_device *rdev)
+static void si_gpu_soft_reset(struct radeon_device *rdev, u32 reset_mask)
 {
 	struct evergreen_mc_save save;
-	u32 grbm_reset = 0;
+	u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
+	u32 tmp;
 
-	if (!(RREG32(GRBM_STATUS) & GUI_ACTIVE))
-		return 0;
+	if (reset_mask == 0)
+		return;
+
+	dev_info(rdev->dev, "GPU softreset: 0x%08X\n", reset_mask);
+
+	evergreen_print_gpu_status_regs(rdev);
+	dev_info(rdev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
+		 RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR));
+	dev_info(rdev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
+		 RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS));
 
-	dev_info(rdev->dev, "GPU softreset \n");
-	dev_info(rdev->dev, "  GRBM_STATUS=0x%08X\n",
-		RREG32(GRBM_STATUS));
-	dev_info(rdev->dev, "  GRBM_STATUS2=0x%08X\n",
-		RREG32(GRBM_STATUS2));
-	dev_info(rdev->dev, "  GRBM_STATUS_SE0=0x%08X\n",
-		RREG32(GRBM_STATUS_SE0));
-	dev_info(rdev->dev, "  GRBM_STATUS_SE1=0x%08X\n",
-		RREG32(GRBM_STATUS_SE1));
-	dev_info(rdev->dev, "  SRBM_STATUS=0x%08X\n",
-		RREG32(SRBM_STATUS));
-	evergreen_mc_stop(rdev, &save);
-	if (radeon_mc_wait_for_idle(rdev)) {
-		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
-	}
 	/* Disable CP parsing/prefetching */
 	WREG32(CP_ME_CNTL, CP_ME_HALT | CP_PFP_HALT | CP_CE_HALT);
 
-	/* reset all the gfx blocks */
-	grbm_reset = (SOFT_RESET_CP |
-		      SOFT_RESET_CB |
-		      SOFT_RESET_DB |
-		      SOFT_RESET_GDS |
-		      SOFT_RESET_PA |
-		      SOFT_RESET_SC |
-		      SOFT_RESET_BCI |
-		      SOFT_RESET_SPI |
-		      SOFT_RESET_SX |
-		      SOFT_RESET_TC |
-		      SOFT_RESET_TA |
-		      SOFT_RESET_VGT |
-		      SOFT_RESET_IA);
+	if (reset_mask & RADEON_RESET_DMA) {
+		/* dma0 */
+		tmp = RREG32(DMA_RB_CNTL + DMA0_REGISTER_OFFSET);
+		tmp &= ~DMA_RB_ENABLE;
+		WREG32(DMA_RB_CNTL + DMA0_REGISTER_OFFSET, tmp);
+	}
+	if (reset_mask & RADEON_RESET_DMA1) {
+		/* dma1 */
+		tmp = RREG32(DMA_RB_CNTL + DMA1_REGISTER_OFFSET);
+		tmp &= ~DMA_RB_ENABLE;
+		WREG32(DMA_RB_CNTL + DMA1_REGISTER_OFFSET, tmp);
+	}
 
-	dev_info(rdev->dev, "  GRBM_SOFT_RESET=0x%08X\n", grbm_reset);
-	WREG32(GRBM_SOFT_RESET, grbm_reset);
-	(void)RREG32(GRBM_SOFT_RESET);
 	udelay(50);
-	WREG32(GRBM_SOFT_RESET, 0);
-	(void)RREG32(GRBM_SOFT_RESET);
+
+	evergreen_mc_stop(rdev, &save);
+	if (evergreen_mc_wait_for_idle(rdev)) {
+		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
+	}
+
+	if (reset_mask & (RADEON_RESET_GFX | RADEON_RESET_COMPUTE | RADEON_RESET_CP)) {
+		grbm_soft_reset = SOFT_RESET_CB |
+			SOFT_RESET_DB |
+			SOFT_RESET_GDS |
+			SOFT_RESET_PA |
+			SOFT_RESET_SC |
+			SOFT_RESET_BCI |
+			SOFT_RESET_SPI |
+			SOFT_RESET_SX |
+			SOFT_RESET_TC |
+			SOFT_RESET_TA |
+			SOFT_RESET_VGT |
+			SOFT_RESET_IA;
+	}
+
+	if (reset_mask & RADEON_RESET_CP) {
+		grbm_soft_reset |= SOFT_RESET_CP | SOFT_RESET_VGT;
+
+		srbm_soft_reset |= SOFT_RESET_GRBM;
+	}
+
+	if (reset_mask & RADEON_RESET_DMA)
+		srbm_soft_reset |= SOFT_RESET_DMA;
+
+	if (reset_mask & RADEON_RESET_DMA1)
+		srbm_soft_reset |= SOFT_RESET_DMA1;
+
+	if (reset_mask & RADEON_RESET_DISPLAY)
+		srbm_soft_reset |= SOFT_RESET_DC;
+
+	if (reset_mask & RADEON_RESET_RLC)
+		grbm_soft_reset |= SOFT_RESET_RLC;
+
+	if (reset_mask & RADEON_RESET_SEM)
+		srbm_soft_reset |= SOFT_RESET_SEM;
+
+	if (reset_mask & RADEON_RESET_IH)
+		srbm_soft_reset |= SOFT_RESET_IH;
+
+	if (reset_mask & RADEON_RESET_GRBM)
+		srbm_soft_reset |= SOFT_RESET_GRBM;
+
+	if (reset_mask & RADEON_RESET_VMC)
+		srbm_soft_reset |= SOFT_RESET_VMC;
+
+	if (reset_mask & RADEON_RESET_MC)
+		srbm_soft_reset |= SOFT_RESET_MC;
+
+	if (grbm_soft_reset) {
+		tmp = RREG32(GRBM_SOFT_RESET);
+		tmp |= grbm_soft_reset;
+		dev_info(rdev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(GRBM_SOFT_RESET, tmp);
+		tmp = RREG32(GRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~grbm_soft_reset;
+		WREG32(GRBM_SOFT_RESET, tmp);
+		tmp = RREG32(GRBM_SOFT_RESET);
+	}
+
+	if (srbm_soft_reset) {
+		tmp = RREG32(SRBM_SOFT_RESET);
+		tmp |= srbm_soft_reset;
+		dev_info(rdev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(SRBM_SOFT_RESET, tmp);
+		tmp = RREG32(SRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~srbm_soft_reset;
+		WREG32(SRBM_SOFT_RESET, tmp);
+		tmp = RREG32(SRBM_SOFT_RESET);
+	}
+
 	/* Wait a little for things to settle down */
 	udelay(50);
-	dev_info(rdev->dev, "  GRBM_STATUS=0x%08X\n",
-		RREG32(GRBM_STATUS));
-	dev_info(rdev->dev, "  GRBM_STATUS2=0x%08X\n",
-		RREG32(GRBM_STATUS2));
-	dev_info(rdev->dev, "  GRBM_STATUS_SE0=0x%08X\n",
-		RREG32(GRBM_STATUS_SE0));
-	dev_info(rdev->dev, "  GRBM_STATUS_SE1=0x%08X\n",
-		RREG32(GRBM_STATUS_SE1));
-	dev_info(rdev->dev, "  SRBM_STATUS=0x%08X\n",
-		RREG32(SRBM_STATUS));
+
 	evergreen_mc_resume(rdev, &save);
-	return 0;
+	udelay(50);
+
+	evergreen_print_gpu_status_regs(rdev);
 }
 
 int si_asic_reset(struct radeon_device *rdev)
 {
-	return si_gpu_soft_reset(rdev);
+	u32 reset_mask;
+
+	reset_mask = si_gpu_check_soft_reset(rdev);
+
+	if (reset_mask)
+		r600_set_bios_scratch_engine_hung(rdev, true);
+
+	si_gpu_soft_reset(rdev, reset_mask);
+
+	reset_mask = si_gpu_check_soft_reset(rdev);
+
+	if (!reset_mask)
+		r600_set_bios_scratch_engine_hung(rdev, false);
+
+	return 0;
+}
+
+/**
+ * si_gfx_is_lockup - Check if the GFX engine is locked up
+ *
+ * @rdev: radeon_device pointer
+ * @ring: radeon_ring structure holding ring information
+ *
+ * Check if the GFX engine is locked up.
+ * Returns true if the engine appears to be locked up, false if not.
+ */
+bool si_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring)
+{
+	u32 reset_mask = si_gpu_check_soft_reset(rdev);
+
+	if (!(reset_mask & (RADEON_RESET_GFX |
+			    RADEON_RESET_COMPUTE |
+			    RADEON_RESET_CP))) {
+		radeon_ring_lockup_update(ring);
+		return false;
+	}
+	/* force CP activities */
+	radeon_ring_force_activity(rdev, ring);
+	return radeon_ring_test_lockup(rdev, ring);
+}
+
+/**
+ * si_dma_is_lockup - Check if the DMA engine is locked up
+ *
+ * @rdev: radeon_device pointer
+ * @ring: radeon_ring structure holding ring information
+ *
+ * Check if the async DMA engine is locked up.
+ * Returns true if the engine appears to be locked up, false if not.
+ */
+bool si_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring)
+{
+	u32 reset_mask = si_gpu_check_soft_reset(rdev);
+	u32 mask;
+
+	if (ring->idx == R600_RING_TYPE_DMA_INDEX)
+		mask = RADEON_RESET_DMA;
+	else
+		mask = RADEON_RESET_DMA1;
+
+	if (!(reset_mask & mask)) {
+		radeon_ring_lockup_update(ring);
+		return false;
+	}
+	/* force ring activities */
+	radeon_ring_force_activity(rdev, ring);
+	return radeon_ring_test_lockup(rdev, ring);
 }
 
 /* MC */
@@ -2426,9 +2726,20 @@ static int si_pcie_gart_enable(struct radeon_device *rdev)
 	/* enable context1-15 */
 	WREG32(VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR,
 	       (u32)(rdev->dummy_page.addr >> 12));
-	WREG32(VM_CONTEXT1_CNTL2, 0);
+	WREG32(VM_CONTEXT1_CNTL2, 4);
 	WREG32(VM_CONTEXT1_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(1) |
-				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT);
+				RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT |
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT |
+				PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT |
+				VALID_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT |
+				READ_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT |
+				WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT);
 
 	si_pcie_gart_tlb_flush(rdev);
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
@@ -2474,6 +2785,7 @@ static bool si_vm_reg_valid(u32 reg)
 	/* check config regs */
 	switch (reg) {
 	case GRBM_GFX_INDEX:
+	case CP_STRMOUT_CNTL:
 	case VGT_VTX_VECT_EJECT_REG:
 	case VGT_CACHE_INVALIDATION:
 	case VGT_ESGS_RING_SIZE:
@@ -2533,6 +2845,7 @@ static int si_vm_packet3_gfx_check(struct radeon_device *rdev,
 	u32 idx = pkt->idx + 1;
 	u32 idx_value = ib[idx];
 	u32 start_reg, end_reg, reg, i;
+	u32 command, info;
 
 	switch (pkt->opcode) {
 	case PACKET3_NOP:
@@ -2632,6 +2945,52 @@ static int si_vm_packet3_gfx_check(struct radeon_device *rdev,
 				return -EINVAL;
 		}
 		break;
+	case PACKET3_CP_DMA:
+		command = ib[idx + 4];
+		info = ib[idx + 1];
+		if (command & PACKET3_CP_DMA_CMD_SAS) {
+			/* src address space is register */
+			if (((info & 0x60000000) >> 29) == 0) {
+				start_reg = idx_value << 2;
+				if (command & PACKET3_CP_DMA_CMD_SAIC) {
+					reg = start_reg;
+					if (!si_vm_reg_valid(reg)) {
+						DRM_ERROR("CP DMA Bad SRC register\n");
+						return -EINVAL;
+					}
+				} else {
+					for (i = 0; i < (command & 0x1fffff); i++) {
+						reg = start_reg + (4 * i);
+						if (!si_vm_reg_valid(reg)) {
+							DRM_ERROR("CP DMA Bad SRC register\n");
+							return -EINVAL;
+						}
+					}
+				}
+			}
+		}
+		if (command & PACKET3_CP_DMA_CMD_DAS) {
+			/* dst address space is register */
+			if (((info & 0x00300000) >> 20) == 0) {
+				start_reg = ib[idx + 2];
+				if (command & PACKET3_CP_DMA_CMD_DAIC) {
+					reg = start_reg;
+					if (!si_vm_reg_valid(reg)) {
+						DRM_ERROR("CP DMA Bad DST register\n");
+						return -EINVAL;
+					}
+				} else {
+					for (i = 0; i < (command & 0x1fffff); i++) {
+						reg = start_reg + (4 * i);
+						if (!si_vm_reg_valid(reg)) {
+							DRM_ERROR("CP DMA Bad DST register\n");
+							return -EINVAL;
+						}
+					}
+				}
+			}
+		}
+		break;
 	default:
 		DRM_ERROR("Invalid GFX packet3: 0x%x\n", pkt->opcode);
 		return -EINVAL;
@@ -2729,19 +3088,19 @@ int si_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib)
 
 	do {
 		pkt.idx = idx;
-		pkt.type = CP_PACKET_GET_TYPE(ib->ptr[idx]);
-		pkt.count = CP_PACKET_GET_COUNT(ib->ptr[idx]);
+		pkt.type = RADEON_CP_PACKET_GET_TYPE(ib->ptr[idx]);
+		pkt.count = RADEON_CP_PACKET_GET_COUNT(ib->ptr[idx]);
 		pkt.one_reg_wr = 0;
 		switch (pkt.type) {
-		case PACKET_TYPE0:
+		case RADEON_PACKET_TYPE0:
 			dev_err(rdev->dev, "Packet0 not allowed!\n");
 			ret = -EINVAL;
 			break;
-		case PACKET_TYPE2:
+		case RADEON_PACKET_TYPE2:
 			idx += 1;
 			break;
-		case PACKET_TYPE3:
-			pkt.opcode = CP_PACKET3_GET_OPCODE(ib->ptr[idx]);
+		case RADEON_PACKET_TYPE3:
+			pkt.opcode = RADEON_CP_PACKET3_GET_OPCODE(ib->ptr[idx]);
 			if (ib->is_const_ib)
 				ret = si_vm_packet3_ce_check(rdev, ib->ptr, &pkt);
 			else {
@@ -2794,45 +3153,105 @@ void si_vm_fini(struct radeon_device *rdev)
  * si_vm_set_page - update the page tables using the CP
  *
  * @rdev: radeon_device pointer
+ * @ib: indirect buffer to fill with commands
  * @pe: addr of the page entry
  * @addr: dst addr to write into pe
  * @count: number of page entries to update
  * @incr: increase next addr by incr bytes
  * @flags: access flags
  *
- * Update the page tables using the CP (cayman-si).
+ * Update the page tables using the CP (SI).
  */
-void si_vm_set_page(struct radeon_device *rdev, uint64_t pe,
+void si_vm_set_page(struct radeon_device *rdev,
+		    struct radeon_ib *ib,
+		    uint64_t pe,
 		    uint64_t addr, unsigned count,
 		    uint32_t incr, uint32_t flags)
 {
-	struct radeon_ring *ring = &rdev->ring[rdev->asic->vm.pt_ring_index];
 	uint32_t r600_flags = cayman_vm_page_flags(rdev, flags);
+	uint64_t value;
+	unsigned ndw;
 
-	while (count) {
-		unsigned ndw = 2 + count * 2;
-		if (ndw > 0x3FFE)
-			ndw = 0x3FFE;
+	if (rdev->asic->vm.pt_ring_index == RADEON_RING_TYPE_GFX_INDEX) {
+		while (count) {
+			ndw = 2 + count * 2;
+			if (ndw > 0x3FFE)
+				ndw = 0x3FFE;
 
-		radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, ndw));
-		radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
-					 WRITE_DATA_DST_SEL(1)));
-		radeon_ring_write(ring, pe);
-		radeon_ring_write(ring, upper_32_bits(pe));
-		for (; ndw > 2; ndw -= 2, --count, pe += 8) {
-			uint64_t value;
-			if (flags & RADEON_VM_PAGE_SYSTEM) {
-				value = radeon_vm_map_gart(rdev, addr);
-				value &= 0xFFFFFFFFFFFFF000ULL;
-			} else if (flags & RADEON_VM_PAGE_VALID)
-				value = addr;
-			else
-				value = 0;
-			addr += incr;
-			value |= r600_flags;
-			radeon_ring_write(ring, value);
-			radeon_ring_write(ring, upper_32_bits(value));
+			ib->ptr[ib->length_dw++] = PACKET3(PACKET3_WRITE_DATA, ndw);
+			ib->ptr[ib->length_dw++] = (WRITE_DATA_ENGINE_SEL(0) |
+					WRITE_DATA_DST_SEL(1));
+			ib->ptr[ib->length_dw++] = pe;
+			ib->ptr[ib->length_dw++] = upper_32_bits(pe);
+			for (; ndw > 2; ndw -= 2, --count, pe += 8) {
+				if (flags & RADEON_VM_PAGE_SYSTEM) {
+					value = radeon_vm_map_gart(rdev, addr);
+					value &= 0xFFFFFFFFFFFFF000ULL;
+				} else if (flags & RADEON_VM_PAGE_VALID) {
+					value = addr;
+				} else {
+					value = 0;
+				}
+				addr += incr;
+				value |= r600_flags;
+				ib->ptr[ib->length_dw++] = value;
+				ib->ptr[ib->length_dw++] = upper_32_bits(value);
+			}
 		}
+	} else {
+		/* DMA */
+		if (flags & RADEON_VM_PAGE_SYSTEM) {
+			while (count) {
+				ndw = count * 2;
+				if (ndw > 0xFFFFE)
+					ndw = 0xFFFFE;
+
+				/* for non-physically contiguous pages (system) */
+				ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 0, ndw);
+				ib->ptr[ib->length_dw++] = pe;
+				ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff;
+				for (; ndw > 0; ndw -= 2, --count, pe += 8) {
+					if (flags & RADEON_VM_PAGE_SYSTEM) {
+						value = radeon_vm_map_gart(rdev, addr);
+						value &= 0xFFFFFFFFFFFFF000ULL;
+					} else if (flags & RADEON_VM_PAGE_VALID) {
+						value = addr;
+					} else {
+						value = 0;
+					}
+					addr += incr;
+					value |= r600_flags;
+					ib->ptr[ib->length_dw++] = value;
+					ib->ptr[ib->length_dw++] = upper_32_bits(value);
+				}
+			}
+		} else {
+			while (count) {
+				ndw = count * 2;
+				if (ndw > 0xFFFFE)
+					ndw = 0xFFFFE;
+
+				if (flags & RADEON_VM_PAGE_VALID)
+					value = addr;
+				else
+					value = 0;
+				/* for physically contiguous pages (vram) */
+				ib->ptr[ib->length_dw++] = DMA_PTE_PDE_PACKET(ndw);
+				ib->ptr[ib->length_dw++] = pe; /* dst addr */
+				ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff;
+				ib->ptr[ib->length_dw++] = r600_flags; /* mask */
+				ib->ptr[ib->length_dw++] = 0;
+				ib->ptr[ib->length_dw++] = value; /* value */
+				ib->ptr[ib->length_dw++] = upper_32_bits(value);
+				ib->ptr[ib->length_dw++] = incr; /* increment size */
+				ib->ptr[ib->length_dw++] = 0;
+				pe += ndw * 4;
+				addr += (ndw / 2) * incr;
+				count -= ndw / 2;
+			}
+		}
+		while (ib->length_dw & 0x7)
+			ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0);
 	}
 }
 
@@ -2879,6 +3298,32 @@ void si_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm)
 	radeon_ring_write(ring, 0x0);
 }
 
+void si_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm)
+{
+	struct radeon_ring *ring = &rdev->ring[ridx];
+
+	if (vm == NULL)
+		return;
+
+	radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0));
+	if (vm->id < 8) {
+		radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2));
+	} else {
+		radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2));
+	}
+	radeon_ring_write(ring, vm->pd_gpu_addr >> 12);
+
+	/* flush hdp cache */
+	radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0));
+	radeon_ring_write(ring, (0xf << 16) | (HDP_MEM_COHERENCY_FLUSH_CNTL >> 2));
+	radeon_ring_write(ring, 1);
+
+	/* bits 0-7 are the VM contexts0-7 */
+	radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0));
+	radeon_ring_write(ring, (0xf << 16) | (VM_INVALIDATE_REQUEST >> 2));
+	radeon_ring_write(ring, 1 << vm->id);
+}
+
 /*
  * RLC
  */
@@ -3047,6 +3492,10 @@ static void si_disable_interrupt_state(struct radeon_device *rdev)
 	WREG32(CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE);
 	WREG32(CP_INT_CNTL_RING1, 0);
 	WREG32(CP_INT_CNTL_RING2, 0);
+	tmp = RREG32(DMA_CNTL + DMA0_REGISTER_OFFSET) & ~TRAP_ENABLE;
+	WREG32(DMA_CNTL + DMA0_REGISTER_OFFSET, tmp);
+	tmp = RREG32(DMA_CNTL + DMA1_REGISTER_OFFSET) & ~TRAP_ENABLE;
+	WREG32(DMA_CNTL + DMA1_REGISTER_OFFSET, tmp);
 	WREG32(GRBM_INT_CNTL, 0);
 	WREG32(INT_MASK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
 	WREG32(INT_MASK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
@@ -3166,6 +3615,7 @@ int si_irq_set(struct radeon_device *rdev)
 	u32 hpd1, hpd2, hpd3, hpd4, hpd5, hpd6;
 	u32 grbm_int_cntl = 0;
 	u32 grph1 = 0, grph2 = 0, grph3 = 0, grph4 = 0, grph5 = 0, grph6 = 0;
+	u32 dma_cntl, dma_cntl1;
 
 	if (!rdev->irq.installed) {
 		WARN(1, "Can't enable IRQ/MSI because no handler is installed\n");
@@ -3186,6 +3636,9 @@ int si_irq_set(struct radeon_device *rdev)
 	hpd5 = RREG32(DC_HPD5_INT_CONTROL) & ~DC_HPDx_INT_EN;
 	hpd6 = RREG32(DC_HPD6_INT_CONTROL) & ~DC_HPDx_INT_EN;
 
+	dma_cntl = RREG32(DMA_CNTL + DMA0_REGISTER_OFFSET) & ~TRAP_ENABLE;
+	dma_cntl1 = RREG32(DMA_CNTL + DMA1_REGISTER_OFFSET) & ~TRAP_ENABLE;
+
 	/* enable CP interrupts on all rings */
 	if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) {
 		DRM_DEBUG("si_irq_set: sw int gfx\n");
@@ -3199,6 +3652,15 @@ int si_irq_set(struct radeon_device *rdev)
 		DRM_DEBUG("si_irq_set: sw int cp2\n");
 		cp_int_cntl2 |= TIME_STAMP_INT_ENABLE;
 	}
+	if (atomic_read(&rdev->irq.ring_int[R600_RING_TYPE_DMA_INDEX])) {
+		DRM_DEBUG("si_irq_set: sw int dma\n");
+		dma_cntl |= TRAP_ENABLE;
+	}
+
+	if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_DMA1_INDEX])) {
+		DRM_DEBUG("si_irq_set: sw int dma1\n");
+		dma_cntl1 |= TRAP_ENABLE;
+	}
 	if (rdev->irq.crtc_vblank_int[0] ||
 	    atomic_read(&rdev->irq.pflip[0])) {
 		DRM_DEBUG("si_irq_set: vblank 0\n");
@@ -3258,6 +3720,9 @@ int si_irq_set(struct radeon_device *rdev)
 	WREG32(CP_INT_CNTL_RING1, cp_int_cntl1);
 	WREG32(CP_INT_CNTL_RING2, cp_int_cntl2);
 
+	WREG32(DMA_CNTL + DMA0_REGISTER_OFFSET, dma_cntl);
+	WREG32(DMA_CNTL + DMA1_REGISTER_OFFSET, dma_cntl1);
+
 	WREG32(GRBM_INT_CNTL, grbm_int_cntl);
 
 	WREG32(INT_MASK + EVERGREEN_CRTC0_REGISTER_OFFSET, crtc1);
@@ -3683,6 +4148,16 @@ restart_ih:
 				break;
 			}
 			break;
+		case 146:
+		case 147:
+			dev_err(rdev->dev, "GPU fault detected: %d 0x%08x\n", src_id, src_data);
+			dev_err(rdev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
+				RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR));
+			dev_err(rdev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
+				RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS));
+			/* reset addr and status */
+			WREG32_P(VM_CONTEXT1_CNTL2, 1, ~1);
+			break;
 		case 176: /* RINGID0 CP_INT */
 			radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX);
 			break;
@@ -3706,9 +4181,17 @@ restart_ih:
 				break;
 			}
 			break;
+		case 224: /* DMA trap event */
+			DRM_DEBUG("IH: DMA trap\n");
+			radeon_fence_process(rdev, R600_RING_TYPE_DMA_INDEX);
+			break;
 		case 233: /* GUI IDLE */
 			DRM_DEBUG("IH: GUI idle\n");
 			break;
+		case 244: /* DMA trap event */
+			DRM_DEBUG("IH: DMA1 trap\n");
+			radeon_fence_process(rdev, CAYMAN_RING_TYPE_DMA1_INDEX);
+			break;
 		default:
 			DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
 			break;
@@ -3732,6 +4215,80 @@ restart_ih:
 	return IRQ_HANDLED;
 }
 
+/**
+ * si_copy_dma - copy pages using the DMA engine
+ *
+ * @rdev: radeon_device pointer
+ * @src_offset: src GPU address
+ * @dst_offset: dst GPU address
+ * @num_gpu_pages: number of GPU pages to xfer
+ * @fence: radeon fence object
+ *
+ * Copy GPU paging using the DMA engine (SI).
+ * Used by the radeon ttm implementation to move pages if
+ * registered as the asic copy callback.
+ */
+int si_copy_dma(struct radeon_device *rdev,
+		uint64_t src_offset, uint64_t dst_offset,
+		unsigned num_gpu_pages,
+		struct radeon_fence **fence)
+{
+	struct radeon_semaphore *sem = NULL;
+	int ring_index = rdev->asic->copy.dma_ring_index;
+	struct radeon_ring *ring = &rdev->ring[ring_index];
+	u32 size_in_bytes, cur_size_in_bytes;
+	int i, num_loops;
+	int r = 0;
+
+	r = radeon_semaphore_create(rdev, &sem);
+	if (r) {
+		DRM_ERROR("radeon: moving bo (%d).\n", r);
+		return r;
+	}
+
+	size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT);
+	num_loops = DIV_ROUND_UP(size_in_bytes, 0xfffff);
+	r = radeon_ring_lock(rdev, ring, num_loops * 5 + 11);
+	if (r) {
+		DRM_ERROR("radeon: moving bo (%d).\n", r);
+		radeon_semaphore_free(rdev, &sem, NULL);
+		return r;
+	}
+
+	if (radeon_fence_need_sync(*fence, ring->idx)) {
+		radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring,
+					    ring->idx);
+		radeon_fence_note_sync(*fence, ring->idx);
+	} else {
+		radeon_semaphore_free(rdev, &sem, NULL);
+	}
+
+	for (i = 0; i < num_loops; i++) {
+		cur_size_in_bytes = size_in_bytes;
+		if (cur_size_in_bytes > 0xFFFFF)
+			cur_size_in_bytes = 0xFFFFF;
+		size_in_bytes -= cur_size_in_bytes;
+		radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 1, 0, 0, cur_size_in_bytes));
+		radeon_ring_write(ring, dst_offset & 0xffffffff);
+		radeon_ring_write(ring, src_offset & 0xffffffff);
+		radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xff);
+		radeon_ring_write(ring, upper_32_bits(src_offset) & 0xff);
+		src_offset += cur_size_in_bytes;
+		dst_offset += cur_size_in_bytes;
+	}
+
+	r = radeon_fence_emit(rdev, fence, ring->idx);
+	if (r) {
+		radeon_ring_unlock_undo(rdev, ring);
+		return r;
+	}
+
+	radeon_ring_unlock_commit(rdev, ring);
+	radeon_semaphore_free(rdev, &sem, *fence);
+
+	return r;
+}
+
 /*
  * startup/shutdown callbacks
  */
@@ -3803,6 +4360,18 @@ static int si_startup(struct radeon_device *rdev)
 		return r;
 	}
 
+	r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_DMA_INDEX);
+	if (r) {
+		dev_err(rdev->dev, "failed initializing DMA fences (%d).\n", r);
+		return r;
+	}
+
+	r = radeon_fence_driver_start_ring(rdev, CAYMAN_RING_TYPE_DMA1_INDEX);
+	if (r) {
+		dev_err(rdev->dev, "failed initializing DMA fences (%d).\n", r);
+		return r;
+	}
+
 	/* Enable IRQ */
 	r = si_irq_init(rdev);
 	if (r) {
@@ -3833,6 +4402,22 @@ static int si_startup(struct radeon_device *rdev)
 	if (r)
 		return r;
 
+	ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX];
+	r = radeon_ring_init(rdev, ring, ring->ring_size, R600_WB_DMA_RPTR_OFFSET,
+			     DMA_RB_RPTR + DMA0_REGISTER_OFFSET,
+			     DMA_RB_WPTR + DMA0_REGISTER_OFFSET,
+			     2, 0x3fffc, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0));
+	if (r)
+		return r;
+
+	ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX];
+	r = radeon_ring_init(rdev, ring, ring->ring_size, CAYMAN_WB_DMA1_RPTR_OFFSET,
+			     DMA_RB_RPTR + DMA1_REGISTER_OFFSET,
+			     DMA_RB_WPTR + DMA1_REGISTER_OFFSET,
+			     2, 0x3fffc, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0));
+	if (r)
+		return r;
+
 	r = si_cp_load_microcode(rdev);
 	if (r)
 		return r;
@@ -3840,6 +4425,10 @@ static int si_startup(struct radeon_device *rdev)
 	if (r)
 		return r;
 
+	r = cayman_dma_resume(rdev);
+	if (r)
+		return r;
+
 	r = radeon_ib_pool_init(rdev);
 	if (r) {
 		dev_err(rdev->dev, "IB initialization failed (%d).\n", r);
@@ -3881,9 +4470,7 @@ int si_resume(struct radeon_device *rdev)
 int si_suspend(struct radeon_device *rdev)
 {
 	si_cp_enable(rdev, false);
-	rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = false;
-	rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX].ready = false;
-	rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX].ready = false;
+	cayman_dma_stop(rdev);
 	si_irq_suspend(rdev);
 	radeon_wb_disable(rdev);
 	si_pcie_gart_disable(rdev);
@@ -3961,6 +4548,14 @@ int si_init(struct radeon_device *rdev)
 	ring->ring_obj = NULL;
 	r600_ring_init(rdev, ring, 1024 * 1024);
 
+	ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX];
+	ring->ring_obj = NULL;
+	r600_ring_init(rdev, ring, 64 * 1024);
+
+	ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX];
+	ring->ring_obj = NULL;
+	r600_ring_init(rdev, ring, 64 * 1024);
+
 	rdev->ih.ring_obj = NULL;
 	r600_ih_ring_init(rdev, 64 * 1024);
 
@@ -3973,6 +4568,7 @@ int si_init(struct radeon_device *rdev)
 	if (r) {
 		dev_err(rdev->dev, "disabling GPU acceleration\n");
 		si_cp_fini(rdev);
+		cayman_dma_fini(rdev);
 		si_irq_fini(rdev);
 		si_rlc_fini(rdev);
 		radeon_wb_fini(rdev);
@@ -4001,6 +4597,7 @@ void si_fini(struct radeon_device *rdev)
 	r600_blit_fini(rdev);
 #endif
 	si_cp_fini(rdev);
+	cayman_dma_fini(rdev);
 	si_irq_fini(rdev);
 	si_rlc_fini(rdev);
 	radeon_wb_fini(rdev);
@@ -4018,14 +4615,14 @@ void si_fini(struct radeon_device *rdev)
 }
 
 /**
- * si_get_gpu_clock - return GPU clock counter snapshot
+ * si_get_gpu_clock_counter - return GPU clock counter snapshot
  *
  * @rdev: radeon_device pointer
  *
  * Fetches a GPU clock counter snapshot (SI).
  * Returns the 64 bit clock counter snapshot.
  */
-uint64_t si_get_gpu_clock(struct radeon_device *rdev)
+uint64_t si_get_gpu_clock_counter(struct radeon_device *rdev)
 {
 	uint64_t clock;