1 files changed, 174 insertions, 83 deletions

diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c
index 3453910ee0f..445af798163 100644
--- a/drivers/gpu/drm/radeon/evergreen.c
+++ b/drivers/gpu/drm/radeon/evergreen.c
@@ -88,21 +88,40 @@ u32 evergreen_page_flip(struct radeon_device *rdev, int crtc_id, u64 crtc_base)
 /* get temperature in millidegrees */
 int evergreen_get_temp(struct radeon_device *rdev)
 {
-	u32 temp = (RREG32(CG_MULT_THERMAL_STATUS) & ASIC_T_MASK) >>
-		ASIC_T_SHIFT;
-	u32 actual_temp = 0;
-
-	if (temp & 0x400)
-		actual_temp = -256;
-	else if (temp & 0x200)
-		actual_temp = 255;
-	else if (temp & 0x100) {
-		actual_temp = temp & 0x1ff;
-		actual_temp |= ~0x1ff;
-	} else
-		actual_temp = temp & 0xff;
+	u32 temp, toffset;
+	int actual_temp = 0;
+
+	if (rdev->family == CHIP_JUNIPER) {
+		toffset = (RREG32(CG_THERMAL_CTRL) & TOFFSET_MASK) >>
+			TOFFSET_SHIFT;
+		temp = (RREG32(CG_TS0_STATUS) & TS0_ADC_DOUT_MASK) >>
+			TS0_ADC_DOUT_SHIFT;
+
+		if (toffset & 0x100)
+			actual_temp = temp / 2 - (0x200 - toffset);
+		else
+			actual_temp = temp / 2 + toffset;
 
-	return (actual_temp * 1000) / 2;
+		actual_temp = actual_temp * 1000;
+
+	} else {
+		temp = (RREG32(CG_MULT_THERMAL_STATUS) & ASIC_T_MASK) >>
+			ASIC_T_SHIFT;
+
+		if (temp & 0x400)
+			actual_temp = -256;
+		else if (temp & 0x200)
+			actual_temp = 255;
+		else if (temp & 0x100) {
+			actual_temp = temp & 0x1ff;
+			actual_temp |= ~0x1ff;
+		} else
+			actual_temp = temp & 0xff;
+
+		actual_temp = (actual_temp * 1000) / 2;
+	}
+
+	return actual_temp;
 }
 
 int sumo_get_temp(struct radeon_device *rdev)
@@ -121,11 +140,17 @@ void evergreen_pm_misc(struct radeon_device *rdev)
 	struct radeon_voltage *voltage = &ps->clock_info[req_cm_idx].voltage;
 
 	if (voltage->type == VOLTAGE_SW) {
+		/* 0xff01 is a flag rather then an actual voltage */
+		if (voltage->voltage == 0xff01)
+			return;
 		if (voltage->voltage && (voltage->voltage != rdev->pm.current_vddc)) {
 			radeon_atom_set_voltage(rdev, voltage->voltage, SET_VOLTAGE_TYPE_ASIC_VDDC);
 			rdev->pm.current_vddc = voltage->voltage;
 			DRM_DEBUG("Setting: vddc: %d\n", voltage->voltage);
 		}
+		/* 0xff01 is a flag rather then an actual voltage */
+		if (voltage->vddci == 0xff01)
+			return;
 		if (voltage->vddci && (voltage->vddci != rdev->pm.current_vddci)) {
 			radeon_atom_set_voltage(rdev, voltage->vddci, SET_VOLTAGE_TYPE_ASIC_VDDCI);
 			rdev->pm.current_vddci = voltage->vddci;
@@ -353,7 +378,7 @@ static u32 evergreen_line_buffer_adjust(struct radeon_device *rdev,
 					struct drm_display_mode *mode,
 					struct drm_display_mode *other_mode)
 {
-	u32 tmp = 0;
+	u32 tmp;
 	/*
 	 * Line Buffer Setup
 	 * There are 3 line buffers, each one shared by 2 display controllers.
@@ -363,64 +388,63 @@ static u32 evergreen_line_buffer_adjust(struct radeon_device *rdev,
 	 * first display controller
 	 *  0 - first half of lb (3840 * 2)
 	 *  1 - first 3/4 of lb (5760 * 2)
-	 *  2 - whole lb (7680 * 2)
+	 *  2 - whole lb (7680 * 2), other crtc must be disabled
 	 *  3 - first 1/4 of lb (1920 * 2)
 	 * second display controller
 	 *  4 - second half of lb (3840 * 2)
 	 *  5 - second 3/4 of lb (5760 * 2)
-	 *  6 - whole lb (7680 * 2)
+	 *  6 - whole lb (7680 * 2), other crtc must be disabled
 	 *  7 - last 1/4 of lb (1920 * 2)
 	 */
-	if (mode && other_mode) {
-		if (mode->hdisplay > other_mode->hdisplay) {
-			if (mode->hdisplay > 2560)
-				tmp = 1; /* 3/4 */
-			else
-				tmp = 0; /* 1/2 */
-		} else if (other_mode->hdisplay > mode->hdisplay) {
-			if (other_mode->hdisplay > 2560)
-				tmp = 3; /* 1/4 */
-			else
-				tmp = 0; /* 1/2 */
-		} else
+	/* this can get tricky if we have two large displays on a paired group
+	 * of crtcs.  Ideally for multiple large displays we'd assign them to
+	 * non-linked crtcs for maximum line buffer allocation.
+	 */
+	if (radeon_crtc->base.enabled && mode) {
+		if (other_mode)
 			tmp = 0; /* 1/2 */
-	} else if (mode)
-		tmp = 2; /* whole */
-	else if (other_mode)
-		tmp = 3; /* 1/4 */
+		else
+			tmp = 2; /* whole */
+	} else
+		tmp = 0;
 
 	/* second controller of the pair uses second half of the lb */
 	if (radeon_crtc->crtc_id % 2)
 		tmp += 4;
 	WREG32(DC_LB_MEMORY_SPLIT + radeon_crtc->crtc_offset, tmp);
 
-	switch (tmp) {
-	case 0:
-	case 4:
-	default:
-		if (ASIC_IS_DCE5(rdev))
-			return 4096 * 2;
-		else
-			return 3840 * 2;
-	case 1:
-	case 5:
-		if (ASIC_IS_DCE5(rdev))
-			return 6144 * 2;
-		else
-			return 5760 * 2;
-	case 2:
-	case 6:
-		if (ASIC_IS_DCE5(rdev))
-			return 8192 * 2;
-		else
-			return 7680 * 2;
-	case 3:
-	case 7:
-		if (ASIC_IS_DCE5(rdev))
-			return 2048 * 2;
-		else
-			return 1920 * 2;
+	if (radeon_crtc->base.enabled && mode) {
+		switch (tmp) {
+		case 0:
+		case 4:
+		default:
+			if (ASIC_IS_DCE5(rdev))
+				return 4096 * 2;
+			else
+				return 3840 * 2;
+		case 1:
+		case 5:
+			if (ASIC_IS_DCE5(rdev))
+				return 6144 * 2;
+			else
+				return 5760 * 2;
+		case 2:
+		case 6:
+			if (ASIC_IS_DCE5(rdev))
+				return 8192 * 2;
+			else
+				return 7680 * 2;
+		case 3:
+		case 7:
+			if (ASIC_IS_DCE5(rdev))
+				return 2048 * 2;
+			else
+				return 1920 * 2;
+		}
 	}
+
+	/* controller not enabled, so no lb used */
+	return 0;
 }
 
 static u32 evergreen_get_number_of_dram_channels(struct radeon_device *rdev)
@@ -863,9 +887,15 @@ int evergreen_pcie_gart_enable(struct radeon_device *rdev)
 		SYSTEM_ACCESS_MODE_NOT_IN_SYS |
 		SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU |
 		EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5);
-	WREG32(MC_VM_MD_L1_TLB0_CNTL, tmp);
-	WREG32(MC_VM_MD_L1_TLB1_CNTL, tmp);
-	WREG32(MC_VM_MD_L1_TLB2_CNTL, tmp);
+	if (rdev->flags & RADEON_IS_IGP) {
+		WREG32(FUS_MC_VM_MD_L1_TLB0_CNTL, tmp);
+		WREG32(FUS_MC_VM_MD_L1_TLB1_CNTL, tmp);
+		WREG32(FUS_MC_VM_MD_L1_TLB2_CNTL, tmp);
+	} else {
+		WREG32(MC_VM_MD_L1_TLB0_CNTL, tmp);
+		WREG32(MC_VM_MD_L1_TLB1_CNTL, tmp);
+		WREG32(MC_VM_MD_L1_TLB2_CNTL, tmp);
+	}
 	WREG32(MC_VM_MB_L1_TLB0_CNTL, tmp);
 	WREG32(MC_VM_MB_L1_TLB1_CNTL, tmp);
 	WREG32(MC_VM_MB_L1_TLB2_CNTL, tmp);
@@ -1410,6 +1440,8 @@ static u32 evergreen_get_tile_pipe_to_backend_map(struct radeon_device *rdev,
 	case CHIP_CEDAR:
 	case CHIP_REDWOOD:
 	case CHIP_PALM:
+	case CHIP_SUMO:
+	case CHIP_SUMO2:
 	case CHIP_TURKS:
 	case CHIP_CAICOS:
 		force_no_swizzle = false;
@@ -1539,6 +1571,8 @@ static void evergreen_program_channel_remap(struct radeon_device *rdev)
 	case CHIP_REDWOOD:
 	case CHIP_CEDAR:
 	case CHIP_PALM:
+	case CHIP_SUMO:
+	case CHIP_SUMO2:
 	case CHIP_TURKS:
 	case CHIP_CAICOS:
 	default:
@@ -1573,7 +1607,7 @@ static void evergreen_gpu_init(struct radeon_device *rdev)
 	u32 sq_stack_resource_mgmt_2;
 	u32 sq_stack_resource_mgmt_3;
 	u32 vgt_cache_invalidation;
-	u32 hdp_host_path_cntl;
+	u32 hdp_host_path_cntl, tmp;
 	int i, j, num_shader_engines, ps_thread_count;
 
 	switch (rdev->family) {
@@ -1684,6 +1718,54 @@ static void evergreen_gpu_init(struct radeon_device *rdev)
 		rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30;
 		rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130;
 		break;
+	case CHIP_SUMO:
+		rdev->config.evergreen.num_ses = 1;
+		rdev->config.evergreen.max_pipes = 4;
+		rdev->config.evergreen.max_tile_pipes = 2;
+		if (rdev->pdev->device == 0x9648)
+			rdev->config.evergreen.max_simds = 3;
+		else if ((rdev->pdev->device == 0x9647) ||
+			 (rdev->pdev->device == 0x964a))
+			rdev->config.evergreen.max_simds = 4;
+		else
+			rdev->config.evergreen.max_simds = 5;
+		rdev->config.evergreen.max_backends = 2 * rdev->config.evergreen.num_ses;
+		rdev->config.evergreen.max_gprs = 256;
+		rdev->config.evergreen.max_threads = 248;
+		rdev->config.evergreen.max_gs_threads = 32;
+		rdev->config.evergreen.max_stack_entries = 256;
+		rdev->config.evergreen.sx_num_of_sets = 4;
+		rdev->config.evergreen.sx_max_export_size = 256;
+		rdev->config.evergreen.sx_max_export_pos_size = 64;
+		rdev->config.evergreen.sx_max_export_smx_size = 192;
+		rdev->config.evergreen.max_hw_contexts = 8;
+		rdev->config.evergreen.sq_num_cf_insts = 2;
+
+		rdev->config.evergreen.sc_prim_fifo_size = 0x40;
+		rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30;
+		rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130;
+		break;
+	case CHIP_SUMO2:
+		rdev->config.evergreen.num_ses = 1;
+		rdev->config.evergreen.max_pipes = 4;
+		rdev->config.evergreen.max_tile_pipes = 4;
+		rdev->config.evergreen.max_simds = 2;
+		rdev->config.evergreen.max_backends = 1 * rdev->config.evergreen.num_ses;
+		rdev->config.evergreen.max_gprs = 256;
+		rdev->config.evergreen.max_threads = 248;
+		rdev->config.evergreen.max_gs_threads = 32;
+		rdev->config.evergreen.max_stack_entries = 512;
+		rdev->config.evergreen.sx_num_of_sets = 4;
+		rdev->config.evergreen.sx_max_export_size = 256;
+		rdev->config.evergreen.sx_max_export_pos_size = 64;
+		rdev->config.evergreen.sx_max_export_smx_size = 192;
+		rdev->config.evergreen.max_hw_contexts = 8;
+		rdev->config.evergreen.sq_num_cf_insts = 2;
+
+		rdev->config.evergreen.sc_prim_fifo_size = 0x40;
+		rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30;
+		rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130;
+		break;
 	case CHIP_BARTS:
 		rdev->config.evergreen.num_ses = 2;
 		rdev->config.evergreen.max_pipes = 4;
@@ -1775,7 +1857,10 @@ static void evergreen_gpu_init(struct radeon_device *rdev)
 
 
 	mc_shared_chmap = RREG32(MC_SHARED_CHMAP);
-	mc_arb_ramcfg = RREG32(MC_ARB_RAMCFG);
+	if (rdev->flags & RADEON_IS_IGP)
+		mc_arb_ramcfg = RREG32(FUS_MC_ARB_RAMCFG);
+	else
+		mc_arb_ramcfg = RREG32(MC_ARB_RAMCFG);
 
 	switch (rdev->config.evergreen.max_tile_pipes) {
 	case 1:
@@ -1928,8 +2013,12 @@ static void evergreen_gpu_init(struct radeon_device *rdev)
 		rdev->config.evergreen.tile_config |= (3 << 0);
 		break;
 	}
-	rdev->config.evergreen.tile_config |=
-		((mc_arb_ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT) << 4;
+	/* num banks is 8 on all fusion asics */
+	if (rdev->flags & RADEON_IS_IGP)
+		rdev->config.evergreen.tile_config |= 8 << 4;
+	else
+		rdev->config.evergreen.tile_config |=
+			((mc_arb_ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT) << 4;
 	rdev->config.evergreen.tile_config |=
 		((mc_arb_ramcfg & BURSTLENGTH_MASK) >> BURSTLENGTH_SHIFT) << 8;
 	rdev->config.evergreen.tile_config |=
@@ -2027,6 +2116,8 @@ static void evergreen_gpu_init(struct radeon_device *rdev)
 	switch (rdev->family) {
 	case CHIP_CEDAR:
 	case CHIP_PALM:
+	case CHIP_SUMO:
+	case CHIP_SUMO2:
 	case CHIP_CAICOS:
 		/* no vertex cache */
 		sq_config &= ~VC_ENABLE;
@@ -2048,6 +2139,8 @@ static void evergreen_gpu_init(struct radeon_device *rdev)
 	switch (rdev->family) {
 	case CHIP_CEDAR:
 	case CHIP_PALM:
+	case CHIP_SUMO:
+	case CHIP_SUMO2:
 		ps_thread_count = 96;
 		break;
 	default:
@@ -2087,6 +2180,8 @@ static void evergreen_gpu_init(struct radeon_device *rdev)
 	switch (rdev->family) {
 	case CHIP_CEDAR:
 	case CHIP_PALM:
+	case CHIP_SUMO:
+	case CHIP_SUMO2:
 	case CHIP_CAICOS:
 		vgt_cache_invalidation = CACHE_INVALIDATION(TC_ONLY);
 		break;
@@ -2133,6 +2228,10 @@ static void evergreen_gpu_init(struct radeon_device *rdev)
 	for (i = SQ_ALU_CONST_BUFFER_SIZE_HS_0; i < 0x29000; i += 4)
 		WREG32(i, 0);
 
+	tmp = RREG32(HDP_MISC_CNTL);
+	tmp |= HDP_FLUSH_INVALIDATE_CACHE;
+	WREG32(HDP_MISC_CNTL, tmp);
+
 	hdp_host_path_cntl = RREG32(HDP_HOST_PATH_CNTL);
 	WREG32(HDP_HOST_PATH_CNTL, hdp_host_path_cntl);
 
@@ -2581,7 +2680,7 @@ static inline u32 evergreen_get_ih_wptr(struct radeon_device *rdev)
 	u32 wptr, tmp;
 
 	if (rdev->wb.enabled)
-		wptr = rdev->wb.wb[R600_WB_IH_WPTR_OFFSET/4];
+		wptr = le32_to_cpu(rdev->wb.wb[R600_WB_IH_WPTR_OFFSET/4]);
 	else
 		wptr = RREG32(IH_RB_WPTR);
 
@@ -2602,28 +2701,25 @@ static inline u32 evergreen_get_ih_wptr(struct radeon_device *rdev)
 
 int evergreen_irq_process(struct radeon_device *rdev)
 {
-	u32 wptr = evergreen_get_ih_wptr(rdev);
-	u32 rptr = rdev->ih.rptr;
+	u32 wptr;
+	u32 rptr;
 	u32 src_id, src_data;
 	u32 ring_index;
 	unsigned long flags;
 	bool queue_hotplug = false;
 
-	DRM_DEBUG("r600_irq_process start: rptr %d, wptr %d\n", rptr, wptr);
-	if (!rdev->ih.enabled)
+	if (!rdev->ih.enabled || rdev->shutdown)
 		return IRQ_NONE;
 
-	spin_lock_irqsave(&rdev->ih.lock, flags);
+	wptr = evergreen_get_ih_wptr(rdev);
+	rptr = rdev->ih.rptr;
+	DRM_DEBUG("r600_irq_process start: rptr %d, wptr %d\n", rptr, wptr);
 
+	spin_lock_irqsave(&rdev->ih.lock, flags);
 	if (rptr == wptr) {
 		spin_unlock_irqrestore(&rdev->ih.lock, flags);
 		return IRQ_NONE;
 	}
-	if (rdev->shutdown) {
-		spin_unlock_irqrestore(&rdev->ih.lock, flags);
-		return IRQ_NONE;
-	}
-
 restart_ih:
 	/* display interrupts */
 	evergreen_irq_ack(rdev);
@@ -2852,7 +2948,7 @@ restart_ih:
 			radeon_fence_process(rdev);
 			break;
 		case 233: /* GUI IDLE */
-			DRM_DEBUG("IH: CP EOP\n");
+			DRM_DEBUG("IH: GUI idle\n");
 			rdev->pm.gui_idle = true;
 			wake_up(&rdev->irq.idle_queue);
 			break;
@@ -2924,11 +3020,6 @@ static int evergreen_startup(struct radeon_device *rdev)
 		rdev->asic->copy = NULL;
 		dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r);
 	}
-	/* XXX: ontario has problems blitting to gart at the moment */
-	if (rdev->family == CHIP_PALM) {
-		rdev->asic->copy = NULL;
-		radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size);
-	}
 
 	/* allocate wb buffer */
 	r = radeon_wb_init(rdev);