kolibrios-gitea/drivers/video/drm/radeon/evergreen.c

/*
 * Copyright 2010 Advanced Micro Devices, Inc.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
 * OTHER DEALINGS IN THE SOFTWARE.
 *
 * Authors: Alex Deucher
 */
#include <linux/firmware.h>
//#include <linux/platform_device.h>
#include <linux/slab.h>
#include "drmP.h"
#include "radeon.h"
#include "radeon_asic.h"
#include "radeon_drm.h"
#include "evergreend.h"
#include "atom.h"
#include "avivod.h"
#include "evergreen_reg.h"
#include "evergreen_blit_shaders.h"

#define EVERGREEN_PFP_UCODE_SIZE 1120
#define EVERGREEN_PM4_UCODE_SIZE 1376

static void evergreen_gpu_init(struct radeon_device *rdev);
void evergreen_fini(struct radeon_device *rdev);
static void evergreen_pcie_gen2_enable(struct radeon_device *rdev);



u32 evergreen_page_flip(struct radeon_device *rdev, int crtc_id, u64 crtc_base)
{
	struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc_id];
	u32 tmp = RREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset);

	/* Lock the graphics update lock */
	tmp |= EVERGREEN_GRPH_UPDATE_LOCK;
	WREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset, tmp);

	/* update the scanout addresses */
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + radeon_crtc->crtc_offset,
	       upper_32_bits(crtc_base));
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
	       (u32)crtc_base);

	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + radeon_crtc->crtc_offset,
	       upper_32_bits(crtc_base));
	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset,
	       (u32)crtc_base);

	/* Wait for update_pending to go high. */
	while (!(RREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset) & EVERGREEN_GRPH_SURFACE_UPDATE_PENDING));
	DRM_DEBUG("Update pending now high. Unlocking vupdate_lock.\n");

	/* Unlock the lock, so double-buffering can take place inside vblank */
	tmp &= ~EVERGREEN_GRPH_UPDATE_LOCK;
	WREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset, tmp);

	/* Return current update_pending status: */
	return RREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset) & EVERGREEN_GRPH_SURFACE_UPDATE_PENDING;
}

/* get temperature in millidegrees */
int evergreen_get_temp(struct radeon_device *rdev)
{
	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;

		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)
{
	u32 temp = RREG32(CG_THERMAL_STATUS) & 0xff;
	int actual_temp = temp - 49;

	return actual_temp * 1000;
}

void evergreen_pm_misc(struct radeon_device *rdev)
{
	int req_ps_idx = rdev->pm.requested_power_state_index;
	int req_cm_idx = rdev->pm.requested_clock_mode_index;
	struct radeon_power_state *ps = &rdev->pm.power_state[req_ps_idx];
	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;
			DRM_DEBUG("Setting: vddci: %d\n", voltage->vddci);
		}
	}
}

void evergreen_pm_prepare(struct radeon_device *rdev)
{
	struct drm_device *ddev = rdev->ddev;
	struct drm_crtc *crtc;
	struct radeon_crtc *radeon_crtc;
	u32 tmp;

	/* disable any active CRTCs */
	list_for_each_entry(crtc, &ddev->mode_config.crtc_list, head) {
		radeon_crtc = to_radeon_crtc(crtc);
		if (radeon_crtc->enabled) {
			tmp = RREG32(EVERGREEN_CRTC_CONTROL + radeon_crtc->crtc_offset);
			tmp |= EVERGREEN_CRTC_DISP_READ_REQUEST_DISABLE;
			WREG32(EVERGREEN_CRTC_CONTROL + radeon_crtc->crtc_offset, tmp);
		}
	}
}

void evergreen_pm_finish(struct radeon_device *rdev)
{
	struct drm_device *ddev = rdev->ddev;
	struct drm_crtc *crtc;
	struct radeon_crtc *radeon_crtc;
	u32 tmp;

	/* enable any active CRTCs */
	list_for_each_entry(crtc, &ddev->mode_config.crtc_list, head) {
		radeon_crtc = to_radeon_crtc(crtc);
		if (radeon_crtc->enabled) {
			tmp = RREG32(EVERGREEN_CRTC_CONTROL + radeon_crtc->crtc_offset);
			tmp &= ~EVERGREEN_CRTC_DISP_READ_REQUEST_DISABLE;
			WREG32(EVERGREEN_CRTC_CONTROL + radeon_crtc->crtc_offset, tmp);
		}
	}
}

bool evergreen_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd)
{
	bool connected = false;

	switch (hpd) {
	case RADEON_HPD_1:
		if (RREG32(DC_HPD1_INT_STATUS) & DC_HPDx_SENSE)
			connected = true;
		break;
	case RADEON_HPD_2:
		if (RREG32(DC_HPD2_INT_STATUS) & DC_HPDx_SENSE)
			connected = true;
		break;
	case RADEON_HPD_3:
		if (RREG32(DC_HPD3_INT_STATUS) & DC_HPDx_SENSE)
			connected = true;
		break;
	case RADEON_HPD_4:
		if (RREG32(DC_HPD4_INT_STATUS) & DC_HPDx_SENSE)
			connected = true;
		break;
	case RADEON_HPD_5:
		if (RREG32(DC_HPD5_INT_STATUS) & DC_HPDx_SENSE)
			connected = true;
		break;
	case RADEON_HPD_6:
		if (RREG32(DC_HPD6_INT_STATUS) & DC_HPDx_SENSE)
			connected = true;
			break;
	default:
		break;
	}

	return connected;
}

void evergreen_hpd_set_polarity(struct radeon_device *rdev,
				enum radeon_hpd_id hpd)
{
	u32 tmp;
	bool connected = evergreen_hpd_sense(rdev, hpd);

	switch (hpd) {
	case RADEON_HPD_1:
		tmp = RREG32(DC_HPD1_INT_CONTROL);
		if (connected)
			tmp &= ~DC_HPDx_INT_POLARITY;
		else
			tmp |= DC_HPDx_INT_POLARITY;
		WREG32(DC_HPD1_INT_CONTROL, tmp);
		break;
	case RADEON_HPD_2:
		tmp = RREG32(DC_HPD2_INT_CONTROL);
		if (connected)
			tmp &= ~DC_HPDx_INT_POLARITY;
		else
			tmp |= DC_HPDx_INT_POLARITY;
		WREG32(DC_HPD2_INT_CONTROL, tmp);
		break;
	case RADEON_HPD_3:
		tmp = RREG32(DC_HPD3_INT_CONTROL);
		if (connected)
			tmp &= ~DC_HPDx_INT_POLARITY;
		else
			tmp |= DC_HPDx_INT_POLARITY;
		WREG32(DC_HPD3_INT_CONTROL, tmp);
		break;
	case RADEON_HPD_4:
		tmp = RREG32(DC_HPD4_INT_CONTROL);
		if (connected)
			tmp &= ~DC_HPDx_INT_POLARITY;
		else
			tmp |= DC_HPDx_INT_POLARITY;
		WREG32(DC_HPD4_INT_CONTROL, tmp);
		break;
	case RADEON_HPD_5:
		tmp = RREG32(DC_HPD5_INT_CONTROL);
		if (connected)
			tmp &= ~DC_HPDx_INT_POLARITY;
		else
			tmp |= DC_HPDx_INT_POLARITY;
		WREG32(DC_HPD5_INT_CONTROL, tmp);
			break;
	case RADEON_HPD_6:
		tmp = RREG32(DC_HPD6_INT_CONTROL);
		if (connected)
			tmp &= ~DC_HPDx_INT_POLARITY;
		else
			tmp |= DC_HPDx_INT_POLARITY;
		WREG32(DC_HPD6_INT_CONTROL, tmp);
		break;
	default:
		break;
	}
}

#if 0
void evergreen_hpd_init(struct radeon_device *rdev)
{
	struct drm_device *dev = rdev->ddev;
	struct drm_connector *connector;
	u32 tmp = DC_HPDx_CONNECTION_TIMER(0x9c4) |
		DC_HPDx_RX_INT_TIMER(0xfa) | DC_HPDx_EN;

	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
		struct radeon_connector *radeon_connector = to_radeon_connector(connector);
		switch (radeon_connector->hpd.hpd) {
		case RADEON_HPD_1:
			WREG32(DC_HPD1_CONTROL, tmp);
			rdev->irq.hpd[0] = true;
			break;
		case RADEON_HPD_2:
			WREG32(DC_HPD2_CONTROL, tmp);
			rdev->irq.hpd[1] = true;
			break;
		case RADEON_HPD_3:
			WREG32(DC_HPD3_CONTROL, tmp);
			rdev->irq.hpd[2] = true;
			break;
		case RADEON_HPD_4:
			WREG32(DC_HPD4_CONTROL, tmp);
			rdev->irq.hpd[3] = true;
			break;
		case RADEON_HPD_5:
			WREG32(DC_HPD5_CONTROL, tmp);
			rdev->irq.hpd[4] = true;
			break;
		case RADEON_HPD_6:
			WREG32(DC_HPD6_CONTROL, tmp);
			rdev->irq.hpd[5] = true;
			break;
		default:
			break;
		}
	}
	if (rdev->irq.installed)
		evergreen_irq_set(rdev);
}

void evergreen_hpd_fini(struct radeon_device *rdev)
{
	struct drm_device *dev = rdev->ddev;
	struct drm_connector *connector;

	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
		struct radeon_connector *radeon_connector = to_radeon_connector(connector);
		switch (radeon_connector->hpd.hpd) {
		case RADEON_HPD_1:
			WREG32(DC_HPD1_CONTROL, 0);
			rdev->irq.hpd[0] = false;
			break;
		case RADEON_HPD_2:
			WREG32(DC_HPD2_CONTROL, 0);
			rdev->irq.hpd[1] = false;
			break;
		case RADEON_HPD_3:
			WREG32(DC_HPD3_CONTROL, 0);
			rdev->irq.hpd[2] = false;
			break;
		case RADEON_HPD_4:
			WREG32(DC_HPD4_CONTROL, 0);
			rdev->irq.hpd[3] = false;
			break;
		case RADEON_HPD_5:
			WREG32(DC_HPD5_CONTROL, 0);
			rdev->irq.hpd[4] = false;
			break;
		case RADEON_HPD_6:
			WREG32(DC_HPD6_CONTROL, 0);
			rdev->irq.hpd[5] = false;
			break;
		default:
			break;
		}
	}
}

#endif
/* watermark setup */

static u32 evergreen_line_buffer_adjust(struct radeon_device *rdev,
					struct radeon_crtc *radeon_crtc,
					struct drm_display_mode *mode,
					struct drm_display_mode *other_mode)
{
	u32 tmp;
	/*
	 * Line Buffer Setup
	 * There are 3 line buffers, each one shared by 2 display controllers.
	 * DC_LB_MEMORY_SPLIT controls how that line buffer is shared between
	 * the display controllers.  The paritioning is done via one of four
	 * preset allocations specified in bits 2:0:
	 * first display controller
	 *  0 - first half of lb (3840 * 2)
	 *  1 - first 3/4 of lb (5760 * 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), other crtc must be disabled
	 *  7 - last 1/4 of lb (1920 * 2)
	 */
	/* 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
			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);

	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)
{
	u32 tmp = RREG32(MC_SHARED_CHMAP);

	switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) {
	case 0:
	default:
		return 1;
	case 1:
		return 2;
	case 2:
		return 4;
	case 3:
		return 8;
	}
}

struct evergreen_wm_params {
	u32 dram_channels; /* number of dram channels */
	u32 yclk;          /* bandwidth per dram data pin in kHz */
	u32 sclk;          /* engine clock in kHz */
	u32 disp_clk;      /* display clock in kHz */
	u32 src_width;     /* viewport width */
	u32 active_time;   /* active display time in ns */
	u32 blank_time;    /* blank time in ns */
	bool interlaced;    /* mode is interlaced */
	fixed20_12 vsc;    /* vertical scale ratio */
	u32 num_heads;     /* number of active crtcs */
	u32 bytes_per_pixel; /* bytes per pixel display + overlay */
	u32 lb_size;       /* line buffer allocated to pipe */
	u32 vtaps;         /* vertical scaler taps */
};

static u32 evergreen_dram_bandwidth(struct evergreen_wm_params *wm)
{
	/* Calculate DRAM Bandwidth and the part allocated to display. */
	fixed20_12 dram_efficiency; /* 0.7 */
	fixed20_12 yclk, dram_channels, bandwidth;
	fixed20_12 a;

	a.full = dfixed_const(1000);
	yclk.full = dfixed_const(wm->yclk);
	yclk.full = dfixed_div(yclk, a);
	dram_channels.full = dfixed_const(wm->dram_channels * 4);
	a.full = dfixed_const(10);
	dram_efficiency.full = dfixed_const(7);
	dram_efficiency.full = dfixed_div(dram_efficiency, a);
	bandwidth.full = dfixed_mul(dram_channels, yclk);
	bandwidth.full = dfixed_mul(bandwidth, dram_efficiency);

	return dfixed_trunc(bandwidth);
}

static u32 evergreen_dram_bandwidth_for_display(struct evergreen_wm_params *wm)
{
	/* Calculate DRAM Bandwidth and the part allocated to display. */
	fixed20_12 disp_dram_allocation; /* 0.3 to 0.7 */
	fixed20_12 yclk, dram_channels, bandwidth;
	fixed20_12 a;

	a.full = dfixed_const(1000);
	yclk.full = dfixed_const(wm->yclk);
	yclk.full = dfixed_div(yclk, a);
	dram_channels.full = dfixed_const(wm->dram_channels * 4);
	a.full = dfixed_const(10);
	disp_dram_allocation.full = dfixed_const(3); /* XXX worse case value 0.3 */
	disp_dram_allocation.full = dfixed_div(disp_dram_allocation, a);
	bandwidth.full = dfixed_mul(dram_channels, yclk);
	bandwidth.full = dfixed_mul(bandwidth, disp_dram_allocation);

	return dfixed_trunc(bandwidth);
}

static u32 evergreen_data_return_bandwidth(struct evergreen_wm_params *wm)
{
	/* Calculate the display Data return Bandwidth */
	fixed20_12 return_efficiency; /* 0.8 */
	fixed20_12 sclk, bandwidth;
	fixed20_12 a;

	a.full = dfixed_const(1000);
	sclk.full = dfixed_const(wm->sclk);
	sclk.full = dfixed_div(sclk, a);
	a.full = dfixed_const(10);
	return_efficiency.full = dfixed_const(8);
	return_efficiency.full = dfixed_div(return_efficiency, a);
	a.full = dfixed_const(32);
	bandwidth.full = dfixed_mul(a, sclk);
	bandwidth.full = dfixed_mul(bandwidth, return_efficiency);

	return dfixed_trunc(bandwidth);
}

static u32 evergreen_dmif_request_bandwidth(struct evergreen_wm_params *wm)
{
	/* Calculate the DMIF Request Bandwidth */
	fixed20_12 disp_clk_request_efficiency; /* 0.8 */
	fixed20_12 disp_clk, bandwidth;
	fixed20_12 a;

	a.full = dfixed_const(1000);
	disp_clk.full = dfixed_const(wm->disp_clk);
	disp_clk.full = dfixed_div(disp_clk, a);
	a.full = dfixed_const(10);
	disp_clk_request_efficiency.full = dfixed_const(8);
	disp_clk_request_efficiency.full = dfixed_div(disp_clk_request_efficiency, a);
	a.full = dfixed_const(32);
	bandwidth.full = dfixed_mul(a, disp_clk);
	bandwidth.full = dfixed_mul(bandwidth, disp_clk_request_efficiency);

	return dfixed_trunc(bandwidth);
}

static u32 evergreen_available_bandwidth(struct evergreen_wm_params *wm)
{
	/* Calculate the Available bandwidth. Display can use this temporarily but not in average. */
	u32 dram_bandwidth = evergreen_dram_bandwidth(wm);
	u32 data_return_bandwidth = evergreen_data_return_bandwidth(wm);
	u32 dmif_req_bandwidth = evergreen_dmif_request_bandwidth(wm);

	return min(dram_bandwidth, min(data_return_bandwidth, dmif_req_bandwidth));
}

static u32 evergreen_average_bandwidth(struct evergreen_wm_params *wm)
{
	/* Calculate the display mode Average Bandwidth
	 * DisplayMode should contain the source and destination dimensions,
	 * timing, etc.
	 */
	fixed20_12 bpp;
	fixed20_12 line_time;
	fixed20_12 src_width;
	fixed20_12 bandwidth;
	fixed20_12 a;

	a.full = dfixed_const(1000);
	line_time.full = dfixed_const(wm->active_time + wm->blank_time);
	line_time.full = dfixed_div(line_time, a);
	bpp.full = dfixed_const(wm->bytes_per_pixel);
	src_width.full = dfixed_const(wm->src_width);
	bandwidth.full = dfixed_mul(src_width, bpp);
	bandwidth.full = dfixed_mul(bandwidth, wm->vsc);
	bandwidth.full = dfixed_div(bandwidth, line_time);

	return dfixed_trunc(bandwidth);
}

static u32 evergreen_latency_watermark(struct evergreen_wm_params *wm)
{
	/* First calcualte the latency in ns */
	u32 mc_latency = 2000; /* 2000 ns. */
	u32 available_bandwidth = evergreen_available_bandwidth(wm);
	u32 worst_chunk_return_time = (512 * 8 * 1000) / available_bandwidth;
	u32 cursor_line_pair_return_time = (128 * 4 * 1000) / available_bandwidth;
	u32 dc_latency = 40000000 / wm->disp_clk; /* dc pipe latency */
	u32 other_heads_data_return_time = ((wm->num_heads + 1) * worst_chunk_return_time) +
		(wm->num_heads * cursor_line_pair_return_time);
	u32 latency = mc_latency + other_heads_data_return_time + dc_latency;
	u32 max_src_lines_per_dst_line, lb_fill_bw, line_fill_time;
	fixed20_12 a, b, c;

	if (wm->num_heads == 0)
		return 0;

	a.full = dfixed_const(2);
	b.full = dfixed_const(1);
	if ((wm->vsc.full > a.full) ||
	    ((wm->vsc.full > b.full) && (wm->vtaps >= 3)) ||
	    (wm->vtaps >= 5) ||
	    ((wm->vsc.full >= a.full) && wm->interlaced))
		max_src_lines_per_dst_line = 4;
	else
		max_src_lines_per_dst_line = 2;

	a.full = dfixed_const(available_bandwidth);
	b.full = dfixed_const(wm->num_heads);
	a.full = dfixed_div(a, b);

	b.full = dfixed_const(1000);
	c.full = dfixed_const(wm->disp_clk);
	b.full = dfixed_div(c, b);
	c.full = dfixed_const(wm->bytes_per_pixel);
	b.full = dfixed_mul(b, c);

	lb_fill_bw = min(dfixed_trunc(a), dfixed_trunc(b));

	a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel);
	b.full = dfixed_const(1000);
	c.full = dfixed_const(lb_fill_bw);
	b.full = dfixed_div(c, b);
	a.full = dfixed_div(a, b);
	line_fill_time = dfixed_trunc(a);

	if (line_fill_time < wm->active_time)
		return latency;
	else
		return latency + (line_fill_time - wm->active_time);

}

static bool evergreen_average_bandwidth_vs_dram_bandwidth_for_display(struct evergreen_wm_params *wm)
{
	if (evergreen_average_bandwidth(wm) <=
	    (evergreen_dram_bandwidth_for_display(wm) / wm->num_heads))
		return true;
	else
		return false;
};

static bool evergreen_average_bandwidth_vs_available_bandwidth(struct evergreen_wm_params *wm)
{
	if (evergreen_average_bandwidth(wm) <=
	    (evergreen_available_bandwidth(wm) / wm->num_heads))
		return true;
	else
		return false;
};

static bool evergreen_check_latency_hiding(struct evergreen_wm_params *wm)
{
	u32 lb_partitions = wm->lb_size / wm->src_width;
	u32 line_time = wm->active_time + wm->blank_time;
	u32 latency_tolerant_lines;
	u32 latency_hiding;
	fixed20_12 a;

	a.full = dfixed_const(1);
	if (wm->vsc.full > a.full)
		latency_tolerant_lines = 1;
	else {
		if (lb_partitions <= (wm->vtaps + 1))
			latency_tolerant_lines = 1;
		else
			latency_tolerant_lines = 2;
	}

	latency_hiding = (latency_tolerant_lines * line_time + wm->blank_time);

	if (evergreen_latency_watermark(wm) <= latency_hiding)
		return true;
	else
		return false;
}

static void evergreen_program_watermarks(struct radeon_device *rdev,
					 struct radeon_crtc *radeon_crtc,
					 u32 lb_size, u32 num_heads)
{
	struct drm_display_mode *mode = &radeon_crtc->base.mode;
	struct evergreen_wm_params wm;
	u32 pixel_period;
	u32 line_time = 0;
	u32 latency_watermark_a = 0, latency_watermark_b = 0;
	u32 priority_a_mark = 0, priority_b_mark = 0;
	u32 priority_a_cnt = PRIORITY_OFF;
	u32 priority_b_cnt = PRIORITY_OFF;
	u32 pipe_offset = radeon_crtc->crtc_id * 16;
	u32 tmp, arb_control3;
	fixed20_12 a, b, c;

	if (radeon_crtc->base.enabled && num_heads && mode) {
		pixel_period = 1000000 / (u32)mode->clock;
		line_time = min((u32)mode->crtc_htotal * pixel_period, (u32)65535);
		priority_a_cnt = 0;
		priority_b_cnt = 0;

		wm.yclk = rdev->pm.current_mclk * 10;
		wm.sclk = rdev->pm.current_sclk * 10;
		wm.disp_clk = mode->clock;
		wm.src_width = mode->crtc_hdisplay;
		wm.active_time = mode->crtc_hdisplay * pixel_period;
		wm.blank_time = line_time - wm.active_time;
		wm.interlaced = false;
		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
			wm.interlaced = true;
		wm.vsc = radeon_crtc->vsc;
		wm.vtaps = 1;
		if (radeon_crtc->rmx_type != RMX_OFF)
			wm.vtaps = 2;
		wm.bytes_per_pixel = 4; /* XXX: get this from fb config */
		wm.lb_size = lb_size;
		wm.dram_channels = evergreen_get_number_of_dram_channels(rdev);
		wm.num_heads = num_heads;

		/* set for high clocks */
		latency_watermark_a = min(evergreen_latency_watermark(&wm), (u32)65535);
		/* set for low clocks */
		/* wm.yclk = low clk; wm.sclk = low clk */
		latency_watermark_b = min(evergreen_latency_watermark(&wm), (u32)65535);

		/* possibly force display priority to high */
		/* should really do this at mode validation time... */
		if (!evergreen_average_bandwidth_vs_dram_bandwidth_for_display(&wm) ||
		    !evergreen_average_bandwidth_vs_available_bandwidth(&wm) ||
		    !evergreen_check_latency_hiding(&wm) ||
		    (rdev->disp_priority == 2)) {
			DRM_INFO("force priority to high\n");
			priority_a_cnt |= PRIORITY_ALWAYS_ON;
			priority_b_cnt |= PRIORITY_ALWAYS_ON;
		}

		a.full = dfixed_const(1000);
		b.full = dfixed_const(mode->clock);
		b.full = dfixed_div(b, a);
		c.full = dfixed_const(latency_watermark_a);
		c.full = dfixed_mul(c, b);
		c.full = dfixed_mul(c, radeon_crtc->hsc);
		c.full = dfixed_div(c, a);
		a.full = dfixed_const(16);
		c.full = dfixed_div(c, a);
		priority_a_mark = dfixed_trunc(c);
		priority_a_cnt |= priority_a_mark & PRIORITY_MARK_MASK;

		a.full = dfixed_const(1000);
		b.full = dfixed_const(mode->clock);
		b.full = dfixed_div(b, a);
		c.full = dfixed_const(latency_watermark_b);
		c.full = dfixed_mul(c, b);
		c.full = dfixed_mul(c, radeon_crtc->hsc);
		c.full = dfixed_div(c, a);
		a.full = dfixed_const(16);
		c.full = dfixed_div(c, a);
		priority_b_mark = dfixed_trunc(c);
		priority_b_cnt |= priority_b_mark & PRIORITY_MARK_MASK;
	}

	/* select wm A */
	arb_control3 = RREG32(PIPE0_ARBITRATION_CONTROL3 + pipe_offset);
	tmp = arb_control3;
	tmp &= ~LATENCY_WATERMARK_MASK(3);
	tmp |= LATENCY_WATERMARK_MASK(1);
	WREG32(PIPE0_ARBITRATION_CONTROL3 + pipe_offset, tmp);
	WREG32(PIPE0_LATENCY_CONTROL + pipe_offset,
	       (LATENCY_LOW_WATERMARK(latency_watermark_a) |
		LATENCY_HIGH_WATERMARK(line_time)));
	/* select wm B */
	tmp = RREG32(PIPE0_ARBITRATION_CONTROL3 + pipe_offset);
	tmp &= ~LATENCY_WATERMARK_MASK(3);
	tmp |= LATENCY_WATERMARK_MASK(2);
	WREG32(PIPE0_ARBITRATION_CONTROL3 + pipe_offset, tmp);
	WREG32(PIPE0_LATENCY_CONTROL + pipe_offset,
	       (LATENCY_LOW_WATERMARK(latency_watermark_b) |
		LATENCY_HIGH_WATERMARK(line_time)));
	/* restore original selection */
	WREG32(PIPE0_ARBITRATION_CONTROL3 + pipe_offset, arb_control3);

	/* write the priority marks */
	WREG32(PRIORITY_A_CNT + radeon_crtc->crtc_offset, priority_a_cnt);
	WREG32(PRIORITY_B_CNT + radeon_crtc->crtc_offset, priority_b_cnt);

}

void evergreen_bandwidth_update(struct radeon_device *rdev)
{
	struct drm_display_mode *mode0 = NULL;
	struct drm_display_mode *mode1 = NULL;
	u32 num_heads = 0, lb_size;
	int i;

	radeon_update_display_priority(rdev);

	for (i = 0; i < rdev->num_crtc; i++) {
		if (rdev->mode_info.crtcs[i]->base.enabled)
			num_heads++;
	}
	for (i = 0; i < rdev->num_crtc; i += 2) {
		mode0 = &rdev->mode_info.crtcs[i]->base.mode;
		mode1 = &rdev->mode_info.crtcs[i+1]->base.mode;
		lb_size = evergreen_line_buffer_adjust(rdev, rdev->mode_info.crtcs[i], mode0, mode1);
		evergreen_program_watermarks(rdev, rdev->mode_info.crtcs[i], lb_size, num_heads);
		lb_size = evergreen_line_buffer_adjust(rdev, rdev->mode_info.crtcs[i+1], mode1, mode0);
		evergreen_program_watermarks(rdev, rdev->mode_info.crtcs[i+1], lb_size, num_heads);
	}
}

int evergreen_mc_wait_for_idle(struct radeon_device *rdev)
{
	unsigned i;
	u32 tmp;

	for (i = 0; i < rdev->usec_timeout; i++) {
		/* read MC_STATUS */
		tmp = RREG32(SRBM_STATUS) & 0x1F00;
		if (!tmp)
			return 0;
		udelay(1);
	}
	return -1;
}

/*
 * GART
 */
void evergreen_pcie_gart_tlb_flush(struct radeon_device *rdev)
{
	unsigned i;
	u32 tmp;

	WREG32(HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1);

	WREG32(VM_CONTEXT0_REQUEST_RESPONSE, REQUEST_TYPE(1));
	for (i = 0; i < rdev->usec_timeout; i++) {
		/* read MC_STATUS */
		tmp = RREG32(VM_CONTEXT0_REQUEST_RESPONSE);
		tmp = (tmp & RESPONSE_TYPE_MASK) >> RESPONSE_TYPE_SHIFT;
		if (tmp == 2) {
			printk(KERN_WARNING "[drm] r600 flush TLB failed\n");
			return;
		}
		if (tmp) {
			return;
		}
		udelay(1);
	}
}

int evergreen_pcie_gart_enable(struct radeon_device *rdev)
{
	u32 tmp;
	int r;

	if (rdev->gart.table.vram.robj == NULL) {
		dev_err(rdev->dev, "No VRAM object for PCIE GART.\n");
		return -EINVAL;
	}
	r = radeon_gart_table_vram_pin(rdev);
	if (r)
		return r;
	radeon_gart_restore(rdev);
	/* Setup L2 cache */
	WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING |
				ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
				EFFECTIVE_L2_QUEUE_SIZE(7));
	WREG32(VM_L2_CNTL2, 0);
	WREG32(VM_L2_CNTL3, BANK_SELECT(0) | CACHE_UPDATE_MODE(2));
	/* Setup TLB control */
	tmp = ENABLE_L1_TLB | ENABLE_L1_FRAGMENT_PROCESSING |
		SYSTEM_ACCESS_MODE_NOT_IN_SYS |
		SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU |
		EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5);
	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);
	WREG32(MC_VM_MB_L1_TLB3_CNTL, tmp);
	WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR, rdev->mc.gtt_start >> 12);
	WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, rdev->mc.gtt_end >> 12);
	WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR, rdev->gart.table_addr >> 12);
	WREG32(VM_CONTEXT0_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) |
				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT);
	WREG32(VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR,
			(u32)(rdev->dummy_page.addr >> 12));
	WREG32(VM_CONTEXT1_CNTL, 0);

	evergreen_pcie_gart_tlb_flush(rdev);
	rdev->gart.ready = true;
	return 0;
}

void evergreen_pcie_gart_disable(struct radeon_device *rdev)
{
	u32 tmp;
	int r;

	/* Disable all tables */
	WREG32(VM_CONTEXT0_CNTL, 0);
	WREG32(VM_CONTEXT1_CNTL, 0);

	/* Setup L2 cache */
	WREG32(VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING |
				EFFECTIVE_L2_QUEUE_SIZE(7));
	WREG32(VM_L2_CNTL2, 0);
	WREG32(VM_L2_CNTL3, BANK_SELECT(0) | CACHE_UPDATE_MODE(2));
	/* Setup TLB control */
	tmp = 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);
	WREG32(MC_VM_MB_L1_TLB0_CNTL, tmp);
	WREG32(MC_VM_MB_L1_TLB1_CNTL, tmp);
	WREG32(MC_VM_MB_L1_TLB2_CNTL, tmp);
	WREG32(MC_VM_MB_L1_TLB3_CNTL, tmp);
	if (rdev->gart.table.vram.robj) {
		r = radeon_bo_reserve(rdev->gart.table.vram.robj, false);
		if (likely(r == 0)) {
			radeon_bo_kunmap(rdev->gart.table.vram.robj);
			radeon_bo_unpin(rdev->gart.table.vram.robj);
			radeon_bo_unreserve(rdev->gart.table.vram.robj);
		}
	}
}

void evergreen_pcie_gart_fini(struct radeon_device *rdev)
{
	evergreen_pcie_gart_disable(rdev);
	radeon_gart_table_vram_free(rdev);
	radeon_gart_fini(rdev);
}


void evergreen_agp_enable(struct radeon_device *rdev)
{
	u32 tmp;

	/* Setup L2 cache */
	WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING |
				ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
				EFFECTIVE_L2_QUEUE_SIZE(7));
	WREG32(VM_L2_CNTL2, 0);
	WREG32(VM_L2_CNTL3, BANK_SELECT(0) | CACHE_UPDATE_MODE(2));
	/* Setup TLB control */
	tmp = ENABLE_L1_TLB | ENABLE_L1_FRAGMENT_PROCESSING |
		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);
	WREG32(MC_VM_MB_L1_TLB0_CNTL, tmp);
	WREG32(MC_VM_MB_L1_TLB1_CNTL, tmp);
	WREG32(MC_VM_MB_L1_TLB2_CNTL, tmp);
	WREG32(MC_VM_MB_L1_TLB3_CNTL, tmp);
	WREG32(VM_CONTEXT0_CNTL, 0);
	WREG32(VM_CONTEXT1_CNTL, 0);
}

void evergreen_mc_stop(struct radeon_device *rdev, struct evergreen_mc_save *save)
{
	save->vga_control[0] = RREG32(D1VGA_CONTROL);
	save->vga_control[1] = RREG32(D2VGA_CONTROL);
	save->vga_control[2] = RREG32(EVERGREEN_D3VGA_CONTROL);
	save->vga_control[3] = RREG32(EVERGREEN_D4VGA_CONTROL);
	save->vga_control[4] = RREG32(EVERGREEN_D5VGA_CONTROL);
	save->vga_control[5] = RREG32(EVERGREEN_D6VGA_CONTROL);
	save->vga_render_control = RREG32(VGA_RENDER_CONTROL);
	save->vga_hdp_control = RREG32(VGA_HDP_CONTROL);
	save->crtc_control[0] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET);
	save->crtc_control[1] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET);
	if (!(rdev->flags & RADEON_IS_IGP)) {
	save->crtc_control[2] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET);
	save->crtc_control[3] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET);
	save->crtc_control[4] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET);
	save->crtc_control[5] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET);
	}

	/* Stop all video */
	WREG32(VGA_RENDER_CONTROL, 0);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 1);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 1);
	if (!(rdev->flags & RADEON_IS_IGP)) {
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 1);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 1);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 1);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 1);
	}
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
	if (!(rdev->flags & RADEON_IS_IGP)) {
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, 0);
	}
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
	if (!(rdev->flags & RADEON_IS_IGP)) {
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0);
	}

	WREG32(D1VGA_CONTROL, 0);
	WREG32(D2VGA_CONTROL, 0);
	WREG32(EVERGREEN_D3VGA_CONTROL, 0);
	WREG32(EVERGREEN_D4VGA_CONTROL, 0);
	WREG32(EVERGREEN_D5VGA_CONTROL, 0);
	WREG32(EVERGREEN_D6VGA_CONTROL, 0);
}

void evergreen_mc_resume(struct radeon_device *rdev, struct evergreen_mc_save *save)
{
	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC0_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC0_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC0_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC0_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);

	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC1_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC1_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC1_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC1_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);

	if (!(rdev->flags & RADEON_IS_IGP)) {
	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC2_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC2_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC2_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC2_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);

	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC3_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC3_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC3_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC3_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);

	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC4_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC4_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC4_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC4_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);

	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC5_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC5_REGISTER_OFFSET,
	       upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC5_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);
	WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC5_REGISTER_OFFSET,
	       (u32)rdev->mc.vram_start);
	}

	WREG32(EVERGREEN_VGA_MEMORY_BASE_ADDRESS_HIGH, upper_32_bits(rdev->mc.vram_start));
	WREG32(EVERGREEN_VGA_MEMORY_BASE_ADDRESS, (u32)rdev->mc.vram_start);
	/* Unlock host access */
	WREG32(VGA_HDP_CONTROL, save->vga_hdp_control);
	mdelay(1);
	/* Restore video state */
	WREG32(D1VGA_CONTROL, save->vga_control[0]);
	WREG32(D2VGA_CONTROL, save->vga_control[1]);
	WREG32(EVERGREEN_D3VGA_CONTROL, save->vga_control[2]);
	WREG32(EVERGREEN_D4VGA_CONTROL, save->vga_control[3]);
	WREG32(EVERGREEN_D5VGA_CONTROL, save->vga_control[4]);
	WREG32(EVERGREEN_D6VGA_CONTROL, save->vga_control[5]);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 1);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 1);
	if (!(rdev->flags & RADEON_IS_IGP)) {
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 1);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 1);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 1);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 1);
	}
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, save->crtc_control[0]);
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, save->crtc_control[1]);
	if (!(rdev->flags & RADEON_IS_IGP)) {
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, save->crtc_control[2]);
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, save->crtc_control[3]);
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, save->crtc_control[4]);
	WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, save->crtc_control[5]);
	}
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
	if (!(rdev->flags & RADEON_IS_IGP)) {
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0);
	WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0);
	}
	WREG32(VGA_RENDER_CONTROL, save->vga_render_control);
}

void evergreen_mc_program(struct radeon_device *rdev)
{
	struct evergreen_mc_save save;
	u32 tmp;
	int i, j;

	/* Initialize HDP */
	for (i = 0, j = 0; i < 32; i++, j += 0x18) {
		WREG32((0x2c14 + j), 0x00000000);
		WREG32((0x2c18 + j), 0x00000000);
		WREG32((0x2c1c + j), 0x00000000);
		WREG32((0x2c20 + j), 0x00000000);
		WREG32((0x2c24 + j), 0x00000000);
	}
	WREG32(HDP_REG_COHERENCY_FLUSH_CNTL, 0);

	evergreen_mc_stop(rdev, &save);
	if (evergreen_mc_wait_for_idle(rdev)) {
		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
	}
	/* Lockout access through VGA aperture*/
	WREG32(VGA_HDP_CONTROL, VGA_MEMORY_DISABLE);
	/* Update configuration */
	if (rdev->flags & RADEON_IS_AGP) {
		if (rdev->mc.vram_start < rdev->mc.gtt_start) {
			/* VRAM before AGP */
			WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
				rdev->mc.vram_start >> 12);
			WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
				rdev->mc.gtt_end >> 12);
		} else {
			/* VRAM after AGP */
			WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
				rdev->mc.gtt_start >> 12);
			WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
				rdev->mc.vram_end >> 12);
		}
	} else {
		WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
			rdev->mc.vram_start >> 12);
		WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
			rdev->mc.vram_end >> 12);
	}
	WREG32(MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, 0);
	if (rdev->flags & RADEON_IS_IGP) {
		tmp = RREG32(MC_FUS_VM_FB_OFFSET) & 0x000FFFFF;
		tmp |= ((rdev->mc.vram_end >> 20) & 0xF) << 24;
		tmp |= ((rdev->mc.vram_start >> 20) & 0xF) << 20;
		WREG32(MC_FUS_VM_FB_OFFSET, tmp);
	}
	tmp = ((rdev->mc.vram_end >> 24) & 0xFFFF) << 16;
	tmp |= ((rdev->mc.vram_start >> 24) & 0xFFFF);
	WREG32(MC_VM_FB_LOCATION, tmp);
	WREG32(HDP_NONSURFACE_BASE, (rdev->mc.vram_start >> 8));
	WREG32(HDP_NONSURFACE_INFO, (2 << 7) | (1 << 30));
	WREG32(HDP_NONSURFACE_SIZE, 0x3FFFFFFF);
	if (rdev->flags & RADEON_IS_AGP) {
		WREG32(MC_VM_AGP_TOP, rdev->mc.gtt_end >> 16);
		WREG32(MC_VM_AGP_BOT, rdev->mc.gtt_start >> 16);
		WREG32(MC_VM_AGP_BASE, rdev->mc.agp_base >> 22);
	} else {
		WREG32(MC_VM_AGP_BASE, 0);
		WREG32(MC_VM_AGP_TOP, 0x0FFFFFFF);
		WREG32(MC_VM_AGP_BOT, 0x0FFFFFFF);
	}
	if (evergreen_mc_wait_for_idle(rdev)) {
		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
	}
	evergreen_mc_resume(rdev, &save);
	/* we need to own VRAM, so turn off the VGA renderer here
	 * to stop it overwriting our objects */
	rv515_vga_render_disable(rdev);
}

/*
 * CP.
 */
void evergreen_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
{
	/* set to DX10/11 mode */
	radeon_ring_write(rdev, PACKET3(PACKET3_MODE_CONTROL, 0));
	radeon_ring_write(rdev, 1);
	/* FIXME: implement */
	radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
	radeon_ring_write(rdev,
#ifdef __BIG_ENDIAN
			  (2 << 0) |
#endif
			  (ib->gpu_addr & 0xFFFFFFFC));
	radeon_ring_write(rdev, upper_32_bits(ib->gpu_addr) & 0xFF);
	radeon_ring_write(rdev, ib->length_dw);
}


static int evergreen_cp_load_microcode(struct radeon_device *rdev)
{
	const __be32 *fw_data;
	int i;

	if (!rdev->me_fw || !rdev->pfp_fw)
		return -EINVAL;

	r700_cp_stop(rdev);
	WREG32(CP_RB_CNTL,
#ifdef __BIG_ENDIAN
	       BUF_SWAP_32BIT |
#endif
	       RB_NO_UPDATE | RB_BLKSZ(15) | RB_BUFSZ(3));

	fw_data = (const __be32 *)rdev->pfp_fw->data;
	WREG32(CP_PFP_UCODE_ADDR, 0);
	for (i = 0; i < EVERGREEN_PFP_UCODE_SIZE; i++)
		WREG32(CP_PFP_UCODE_DATA, be32_to_cpup(fw_data++));
	WREG32(CP_PFP_UCODE_ADDR, 0);

	fw_data = (const __be32 *)rdev->me_fw->data;
	WREG32(CP_ME_RAM_WADDR, 0);
	for (i = 0; i < EVERGREEN_PM4_UCODE_SIZE; i++)
		WREG32(CP_ME_RAM_DATA, be32_to_cpup(fw_data++));

	WREG32(CP_PFP_UCODE_ADDR, 0);
	WREG32(CP_ME_RAM_WADDR, 0);
	WREG32(CP_ME_RAM_RADDR, 0);
	return 0;
}

static int evergreen_cp_start(struct radeon_device *rdev)
{
	int r, i;
	uint32_t cp_me;

	r = radeon_ring_lock(rdev, 7);
	if (r) {
		DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
		return r;
	}
	radeon_ring_write(rdev, PACKET3(PACKET3_ME_INITIALIZE, 5));
	radeon_ring_write(rdev, 0x1);
	radeon_ring_write(rdev, 0x0);
	radeon_ring_write(rdev, rdev->config.evergreen.max_hw_contexts - 1);
	radeon_ring_write(rdev, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
	radeon_ring_write(rdev, 0);
	radeon_ring_write(rdev, 0);
	radeon_ring_unlock_commit(rdev);

	cp_me = 0xff;
	WREG32(CP_ME_CNTL, cp_me);

	r = radeon_ring_lock(rdev, evergreen_default_size + 19);
	if (r) {
		DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
		return r;
	}

	/* setup clear context state */
	radeon_ring_write(rdev, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
	radeon_ring_write(rdev, PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);

	for (i = 0; i < evergreen_default_size; i++)
		radeon_ring_write(rdev, evergreen_default_state[i]);

	radeon_ring_write(rdev, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
	radeon_ring_write(rdev, PACKET3_PREAMBLE_END_CLEAR_STATE);

	/* set clear context state */
	radeon_ring_write(rdev, PACKET3(PACKET3_CLEAR_STATE, 0));
	radeon_ring_write(rdev, 0);

	/* SQ_VTX_BASE_VTX_LOC */
	radeon_ring_write(rdev, 0xc0026f00);
	radeon_ring_write(rdev, 0x00000000);
	radeon_ring_write(rdev, 0x00000000);
	radeon_ring_write(rdev, 0x00000000);

	/* Clear consts */
	radeon_ring_write(rdev, 0xc0036f00);
	radeon_ring_write(rdev, 0x00000bc4);
	radeon_ring_write(rdev, 0xffffffff);
	radeon_ring_write(rdev, 0xffffffff);
	radeon_ring_write(rdev, 0xffffffff);

	radeon_ring_write(rdev, 0xc0026900);
	radeon_ring_write(rdev, 0x00000316);
	radeon_ring_write(rdev, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */
	radeon_ring_write(rdev, 0x00000010); /*  */

	radeon_ring_unlock_commit(rdev);

	return 0;
}

int evergreen_cp_resume(struct radeon_device *rdev)
{
	u32 tmp;
	u32 rb_bufsz;
	int r;

	/* Reset cp; if cp is reset, then PA, SH, VGT also need to be reset */
	WREG32(GRBM_SOFT_RESET, (SOFT_RESET_CP |
				 SOFT_RESET_PA |
				 SOFT_RESET_SH |
				 SOFT_RESET_VGT |
				 SOFT_RESET_SX));
	RREG32(GRBM_SOFT_RESET);
	mdelay(15);
	WREG32(GRBM_SOFT_RESET, 0);
	RREG32(GRBM_SOFT_RESET);

	/* Set ring buffer size */
	rb_bufsz = drm_order(rdev->cp.ring_size / 8);
	tmp = (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
#ifdef __BIG_ENDIAN
	tmp |= BUF_SWAP_32BIT;
#endif
	WREG32(CP_RB_CNTL, tmp);
	WREG32(CP_SEM_WAIT_TIMER, 0x4);

	/* Set the write pointer delay */
	WREG32(CP_RB_WPTR_DELAY, 0);

	/* Initialize the ring buffer's read and write pointers */
	WREG32(CP_RB_CNTL, tmp | RB_RPTR_WR_ENA);
	WREG32(CP_RB_RPTR_WR, 0);
	WREG32(CP_RB_WPTR, 0);

	/* set the wb address wether it's enabled or not */
	WREG32(CP_RB_RPTR_ADDR,
#ifdef __BIG_ENDIAN
	       RB_RPTR_SWAP(2) |
#endif
	       ((rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC));
	WREG32(CP_RB_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFF);
	WREG32(SCRATCH_ADDR, ((rdev->wb.gpu_addr + RADEON_WB_SCRATCH_OFFSET) >> 8) & 0xFFFFFFFF);

	if (rdev->wb.enabled)
		WREG32(SCRATCH_UMSK, 0xff);
	else {
		tmp |= RB_NO_UPDATE;
		WREG32(SCRATCH_UMSK, 0);
	}

	mdelay(1);
	WREG32(CP_RB_CNTL, tmp);

	WREG32(CP_RB_BASE, rdev->cp.gpu_addr >> 8);
	WREG32(CP_DEBUG, (1 << 27) | (1 << 28));

	rdev->cp.rptr = RREG32(CP_RB_RPTR);
	rdev->cp.wptr = RREG32(CP_RB_WPTR);

	evergreen_cp_start(rdev);
	rdev->cp.ready = true;
	r = radeon_ring_test(rdev);
	if (r) {
		rdev->cp.ready = false;
		return r;
	}
	return 0;
}

/*
 * Core functions
 */
static u32 evergreen_get_tile_pipe_to_backend_map(struct radeon_device *rdev,
						  u32 num_tile_pipes,
						  u32 num_backends,
						  u32 backend_disable_mask)
{
	u32 backend_map = 0;
	u32 enabled_backends_mask = 0;
	u32 enabled_backends_count = 0;
	u32 cur_pipe;
	u32 swizzle_pipe[EVERGREEN_MAX_PIPES];
	u32 cur_backend = 0;
	u32 i;
	bool force_no_swizzle;

	if (num_tile_pipes > EVERGREEN_MAX_PIPES)
		num_tile_pipes = EVERGREEN_MAX_PIPES;
	if (num_tile_pipes < 1)
		num_tile_pipes = 1;
	if (num_backends > EVERGREEN_MAX_BACKENDS)
		num_backends = EVERGREEN_MAX_BACKENDS;
	if (num_backends < 1)
		num_backends = 1;

	for (i = 0; i < EVERGREEN_MAX_BACKENDS; ++i) {
		if (((backend_disable_mask >> i) & 1) == 0) {
			enabled_backends_mask |= (1 << i);
			++enabled_backends_count;
		}
		if (enabled_backends_count == num_backends)
			break;
	}

	if (enabled_backends_count == 0) {
		enabled_backends_mask = 1;
		enabled_backends_count = 1;
	}

	if (enabled_backends_count != num_backends)
		num_backends = enabled_backends_count;

	memset((uint8_t *)&swizzle_pipe[0], 0, sizeof(u32) * EVERGREEN_MAX_PIPES);
	switch (rdev->family) {
	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;
		break;
	case CHIP_CYPRESS:
	case CHIP_HEMLOCK:
	case CHIP_JUNIPER:
	case CHIP_BARTS:
	default:
		force_no_swizzle = true;
		break;
	}
	if (force_no_swizzle) {
		bool last_backend_enabled = false;

		force_no_swizzle = false;
		for (i = 0; i < EVERGREEN_MAX_BACKENDS; ++i) {
			if (((enabled_backends_mask >> i) & 1) == 1) {
				if (last_backend_enabled)
					force_no_swizzle = true;
				last_backend_enabled = true;
			} else
				last_backend_enabled = false;
		}
	}

	switch (num_tile_pipes) {
	case 1:
	case 3:
	case 5:
	case 7:
		DRM_ERROR("odd number of pipes!\n");
		break;
	case 2:
		swizzle_pipe[0] = 0;
		swizzle_pipe[1] = 1;
		break;
	case 4:
		if (force_no_swizzle) {
			swizzle_pipe[0] = 0;
			swizzle_pipe[1] = 1;
			swizzle_pipe[2] = 2;
			swizzle_pipe[3] = 3;
		} else {
			swizzle_pipe[0] = 0;
			swizzle_pipe[1] = 2;
			swizzle_pipe[2] = 1;
			swizzle_pipe[3] = 3;
		}
		break;
	case 6:
		if (force_no_swizzle) {
			swizzle_pipe[0] = 0;
			swizzle_pipe[1] = 1;
			swizzle_pipe[2] = 2;
			swizzle_pipe[3] = 3;
			swizzle_pipe[4] = 4;
			swizzle_pipe[5] = 5;
		} else {
			swizzle_pipe[0] = 0;
			swizzle_pipe[1] = 2;
			swizzle_pipe[2] = 4;
			swizzle_pipe[3] = 1;
			swizzle_pipe[4] = 3;
			swizzle_pipe[5] = 5;
		}
		break;
	case 8:
		if (force_no_swizzle) {
			swizzle_pipe[0] = 0;
			swizzle_pipe[1] = 1;
			swizzle_pipe[2] = 2;
			swizzle_pipe[3] = 3;
			swizzle_pipe[4] = 4;
			swizzle_pipe[5] = 5;
			swizzle_pipe[6] = 6;
			swizzle_pipe[7] = 7;
		} else {
			swizzle_pipe[0] = 0;
			swizzle_pipe[1] = 2;
			swizzle_pipe[2] = 4;
			swizzle_pipe[3] = 6;
			swizzle_pipe[4] = 1;
			swizzle_pipe[5] = 3;
			swizzle_pipe[6] = 5;
			swizzle_pipe[7] = 7;
		}
		break;
	}

	for (cur_pipe = 0; cur_pipe < num_tile_pipes; ++cur_pipe) {
		while (((1 << cur_backend) & enabled_backends_mask) == 0)
			cur_backend = (cur_backend + 1) % EVERGREEN_MAX_BACKENDS;

		backend_map |= (((cur_backend & 0xf) << (swizzle_pipe[cur_pipe] * 4)));

		cur_backend = (cur_backend + 1) % EVERGREEN_MAX_BACKENDS;
	}

	return backend_map;
}

static void evergreen_program_channel_remap(struct radeon_device *rdev)
{
	u32 tcp_chan_steer_lo, tcp_chan_steer_hi, mc_shared_chremap, tmp;

	tmp = RREG32(MC_SHARED_CHMAP);
	switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) {
	case 0:
	case 1:
	case 2:
	case 3:
	default:
		/* default mapping */
		mc_shared_chremap = 0x00fac688;
		break;
	}

	switch (rdev->family) {
	case CHIP_HEMLOCK:
	case CHIP_CYPRESS:
	case CHIP_BARTS:
		tcp_chan_steer_lo = 0x54763210;
		tcp_chan_steer_hi = 0x0000ba98;
		break;
	case CHIP_JUNIPER:
	case CHIP_REDWOOD:
	case CHIP_CEDAR:
	case CHIP_PALM:
	case CHIP_SUMO:
	case CHIP_SUMO2:
	case CHIP_TURKS:
	case CHIP_CAICOS:
	default:
		tcp_chan_steer_lo = 0x76543210;
		tcp_chan_steer_hi = 0x0000ba98;
		break;
	}

	WREG32(TCP_CHAN_STEER_LO, tcp_chan_steer_lo);
	WREG32(TCP_CHAN_STEER_HI, tcp_chan_steer_hi);
	WREG32(MC_SHARED_CHREMAP, mc_shared_chremap);
}

static void evergreen_gpu_init(struct radeon_device *rdev)
{
	u32 cc_rb_backend_disable = 0;
	u32 cc_gc_shader_pipe_config;
	u32 gb_addr_config = 0;
	u32 mc_shared_chmap, mc_arb_ramcfg;
	u32 gb_backend_map;
	u32 grbm_gfx_index;
	u32 sx_debug_1;
	u32 smx_dc_ctl0;
	u32 sq_config;
	u32 sq_lds_resource_mgmt;
	u32 sq_gpr_resource_mgmt_1;
	u32 sq_gpr_resource_mgmt_2;
	u32 sq_gpr_resource_mgmt_3;
	u32 sq_thread_resource_mgmt;
	u32 sq_thread_resource_mgmt_2;
	u32 sq_stack_resource_mgmt_1;
	u32 sq_stack_resource_mgmt_2;
	u32 sq_stack_resource_mgmt_3;
	u32 vgt_cache_invalidation;
	u32 hdp_host_path_cntl, tmp;
	int i, j, num_shader_engines, ps_thread_count;

	switch (rdev->family) {
	case CHIP_CYPRESS:
	case CHIP_HEMLOCK:
		rdev->config.evergreen.num_ses = 2;
		rdev->config.evergreen.max_pipes = 4;
		rdev->config.evergreen.max_tile_pipes = 8;
		rdev->config.evergreen.max_simds = 10;
		rdev->config.evergreen.max_backends = 4 * 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 = 0x100;
		rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30;
		rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130;
		break;
	case CHIP_JUNIPER:
		rdev->config.evergreen.num_ses = 1;
		rdev->config.evergreen.max_pipes = 4;
		rdev->config.evergreen.max_tile_pipes = 4;
		rdev->config.evergreen.max_simds = 10;
		rdev->config.evergreen.max_backends = 4 * 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 = 0x100;
		rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30;
		rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130;
		break;
	case CHIP_REDWOOD:
		rdev->config.evergreen.num_ses = 1;
		rdev->config.evergreen.max_pipes = 4;
		rdev->config.evergreen.max_tile_pipes = 4;
		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 = 0x100;
		rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30;
		rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130;
		break;
	case CHIP_CEDAR:
	default:
		rdev->config.evergreen.num_ses = 1;
		rdev->config.evergreen.max_pipes = 2;
		rdev->config.evergreen.max_tile_pipes = 2;
		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 = 192;
		rdev->config.evergreen.max_gs_threads = 16;
		rdev->config.evergreen.max_stack_entries = 256;
		rdev->config.evergreen.sx_num_of_sets = 4;
		rdev->config.evergreen.sx_max_export_size = 128;
		rdev->config.evergreen.sx_max_export_pos_size = 32;
		rdev->config.evergreen.sx_max_export_smx_size = 96;
		rdev->config.evergreen.max_hw_contexts = 4;
		rdev->config.evergreen.sq_num_cf_insts = 1;

		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_PALM:
		rdev->config.evergreen.num_ses = 1;
		rdev->config.evergreen.max_pipes = 2;
		rdev->config.evergreen.max_tile_pipes = 2;
		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 = 192;
		rdev->config.evergreen.max_gs_threads = 16;
		rdev->config.evergreen.max_stack_entries = 256;
		rdev->config.evergreen.sx_num_of_sets = 4;
		rdev->config.evergreen.sx_max_export_size = 128;
		rdev->config.evergreen.sx_max_export_pos_size = 32;
		rdev->config.evergreen.sx_max_export_smx_size = 96;
		rdev->config.evergreen.max_hw_contexts = 4;
		rdev->config.evergreen.sq_num_cf_insts = 1;

		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_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;
		rdev->config.evergreen.max_tile_pipes = 8;
		rdev->config.evergreen.max_simds = 7;
		rdev->config.evergreen.max_backends = 4 * 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 = 0x100;
		rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30;
		rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130;
		break;
	case CHIP_TURKS:
		rdev->config.evergreen.num_ses = 1;
		rdev->config.evergreen.max_pipes = 4;
		rdev->config.evergreen.max_tile_pipes = 4;
		rdev->config.evergreen.max_simds = 6;
		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 = 0x100;
		rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30;
		rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130;
		break;
	case CHIP_CAICOS:
		rdev->config.evergreen.num_ses = 1;
		rdev->config.evergreen.max_pipes = 4;
		rdev->config.evergreen.max_tile_pipes = 2;
		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 = 192;
		rdev->config.evergreen.max_gs_threads = 16;
		rdev->config.evergreen.max_stack_entries = 256;
		rdev->config.evergreen.sx_num_of_sets = 4;
		rdev->config.evergreen.sx_max_export_size = 128;
		rdev->config.evergreen.sx_max_export_pos_size = 32;
		rdev->config.evergreen.sx_max_export_smx_size = 96;
		rdev->config.evergreen.max_hw_contexts = 4;
		rdev->config.evergreen.sq_num_cf_insts = 1;

		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;
	}

	/* Initialize HDP */
	for (i = 0, j = 0; i < 32; i++, j += 0x18) {
		WREG32((0x2c14 + j), 0x00000000);
		WREG32((0x2c18 + j), 0x00000000);
		WREG32((0x2c1c + j), 0x00000000);
		WREG32((0x2c20 + j), 0x00000000);
		WREG32((0x2c24 + j), 0x00000000);
	}

	WREG32(GRBM_CNTL, GRBM_READ_TIMEOUT(0xff));

	cc_gc_shader_pipe_config = RREG32(CC_GC_SHADER_PIPE_CONFIG) & ~2;

	cc_gc_shader_pipe_config |=
		INACTIVE_QD_PIPES((EVERGREEN_MAX_PIPES_MASK << rdev->config.evergreen.max_pipes)
				  & EVERGREEN_MAX_PIPES_MASK);
	cc_gc_shader_pipe_config |=
		INACTIVE_SIMDS((EVERGREEN_MAX_SIMDS_MASK << rdev->config.evergreen.max_simds)
			       & EVERGREEN_MAX_SIMDS_MASK);

	cc_rb_backend_disable =
		BACKEND_DISABLE((EVERGREEN_MAX_BACKENDS_MASK << rdev->config.evergreen.max_backends)
				& EVERGREEN_MAX_BACKENDS_MASK);


	mc_shared_chmap = RREG32(MC_SHARED_CHMAP);
	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:
	default:
		gb_addr_config |= NUM_PIPES(0);
		break;
	case 2:
		gb_addr_config |= NUM_PIPES(1);
		break;
	case 4:
		gb_addr_config |= NUM_PIPES(2);
		break;
	case 8:
		gb_addr_config |= NUM_PIPES(3);
		break;
	}

	gb_addr_config |= PIPE_INTERLEAVE_SIZE((mc_arb_ramcfg & BURSTLENGTH_MASK) >> BURSTLENGTH_SHIFT);
	gb_addr_config |= BANK_INTERLEAVE_SIZE(0);
	gb_addr_config |= NUM_SHADER_ENGINES(rdev->config.evergreen.num_ses - 1);
	gb_addr_config |= SHADER_ENGINE_TILE_SIZE(1);
	gb_addr_config |= NUM_GPUS(0); /* Hemlock? */
	gb_addr_config |= MULTI_GPU_TILE_SIZE(2);

	if (((mc_arb_ramcfg & NOOFCOLS_MASK) >> NOOFCOLS_SHIFT) > 2)
		gb_addr_config |= ROW_SIZE(2);
	else
		gb_addr_config |= ROW_SIZE((mc_arb_ramcfg & NOOFCOLS_MASK) >> NOOFCOLS_SHIFT);

	if (rdev->ddev->pdev->device == 0x689e) {
		u32 efuse_straps_4;
		u32 efuse_straps_3;
		u8 efuse_box_bit_131_124;

		WREG32(RCU_IND_INDEX, 0x204);
		efuse_straps_4 = RREG32(RCU_IND_DATA);
		WREG32(RCU_IND_INDEX, 0x203);
		efuse_straps_3 = RREG32(RCU_IND_DATA);
		efuse_box_bit_131_124 = (u8)(((efuse_straps_4 & 0xf) << 4) | ((efuse_straps_3 & 0xf0000000) >> 28));

		switch(efuse_box_bit_131_124) {
		case 0x00:
			gb_backend_map = 0x76543210;
			break;
		case 0x55:
			gb_backend_map = 0x77553311;
			break;
		case 0x56:
			gb_backend_map = 0x77553300;
			break;
		case 0x59:
			gb_backend_map = 0x77552211;
			break;
		case 0x66:
			gb_backend_map = 0x77443300;
			break;
		case 0x99:
			gb_backend_map = 0x66552211;
			break;
		case 0x5a:
			gb_backend_map = 0x77552200;
			break;
		case 0xaa:
			gb_backend_map = 0x66442200;
			break;
		case 0x95:
			gb_backend_map = 0x66553311;
			break;
		default:
			DRM_ERROR("bad backend map, using default\n");
			gb_backend_map =
				evergreen_get_tile_pipe_to_backend_map(rdev,
								       rdev->config.evergreen.max_tile_pipes,
								       rdev->config.evergreen.max_backends,
								       ((EVERGREEN_MAX_BACKENDS_MASK <<
								   rdev->config.evergreen.max_backends) &
									EVERGREEN_MAX_BACKENDS_MASK));
			break;
		}
	} else if (rdev->ddev->pdev->device == 0x68b9) {
		u32 efuse_straps_3;
		u8 efuse_box_bit_127_124;

		WREG32(RCU_IND_INDEX, 0x203);
		efuse_straps_3 = RREG32(RCU_IND_DATA);
		efuse_box_bit_127_124 = (u8)((efuse_straps_3 & 0xF0000000) >> 28);

		switch(efuse_box_bit_127_124) {
		case 0x0:
			gb_backend_map = 0x00003210;
			break;
		case 0x5:
		case 0x6:
		case 0x9:
		case 0xa:
			gb_backend_map = 0x00003311;
			break;
		default:
			DRM_ERROR("bad backend map, using default\n");
			gb_backend_map =
				evergreen_get_tile_pipe_to_backend_map(rdev,
								       rdev->config.evergreen.max_tile_pipes,
								       rdev->config.evergreen.max_backends,
								       ((EVERGREEN_MAX_BACKENDS_MASK <<
								   rdev->config.evergreen.max_backends) &
									EVERGREEN_MAX_BACKENDS_MASK));
			break;
		}
	} else {
		switch (rdev->family) {
		case CHIP_CYPRESS:
		case CHIP_HEMLOCK:
		case CHIP_BARTS:
			gb_backend_map = 0x66442200;
			break;
		case CHIP_JUNIPER:
			gb_backend_map = 0x00006420;
			break;
		default:
			gb_backend_map =
				evergreen_get_tile_pipe_to_backend_map(rdev,
								       rdev->config.evergreen.max_tile_pipes,
								       rdev->config.evergreen.max_backends,
								       ((EVERGREEN_MAX_BACKENDS_MASK <<
									 rdev->config.evergreen.max_backends) &
									EVERGREEN_MAX_BACKENDS_MASK));
		}
	}

	/* setup tiling info dword.  gb_addr_config is not adequate since it does
	 * not have bank info, so create a custom tiling dword.
	 * bits 3:0   num_pipes
	 * bits 7:4   num_banks
	 * bits 11:8  group_size
	 * bits 15:12 row_size
	 */
	rdev->config.evergreen.tile_config = 0;
	switch (rdev->config.evergreen.max_tile_pipes) {
	case 1:
	default:
		rdev->config.evergreen.tile_config |= (0 << 0);
		break;
	case 2:
		rdev->config.evergreen.tile_config |= (1 << 0);
		break;
	case 4:
		rdev->config.evergreen.tile_config |= (2 << 0);
		break;
	case 8:
		rdev->config.evergreen.tile_config |= (3 << 0);
		break;
	}
	/* num banks is 8 on all fusion asics. 0 = 4, 1 = 8, 2 = 16 */
	if (rdev->flags & RADEON_IS_IGP)
		rdev->config.evergreen.tile_config |= 1 << 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 |=
		((gb_addr_config & 0x30000000) >> 28) << 12;

	WREG32(GB_BACKEND_MAP, gb_backend_map);
	WREG32(GB_ADDR_CONFIG, gb_addr_config);
	WREG32(DMIF_ADDR_CONFIG, gb_addr_config);
	WREG32(HDP_ADDR_CONFIG, gb_addr_config);

	evergreen_program_channel_remap(rdev);

	num_shader_engines = ((RREG32(GB_ADDR_CONFIG) & NUM_SHADER_ENGINES(3)) >> 12) + 1;
	grbm_gfx_index = INSTANCE_BROADCAST_WRITES;

	for (i = 0; i < rdev->config.evergreen.num_ses; i++) {
		u32 rb = cc_rb_backend_disable | (0xf0 << 16);
		u32 sp = cc_gc_shader_pipe_config;
		u32 gfx = grbm_gfx_index | SE_INDEX(i);

		if (i == num_shader_engines) {
			rb |= BACKEND_DISABLE(EVERGREEN_MAX_BACKENDS_MASK);
			sp |= INACTIVE_SIMDS(EVERGREEN_MAX_SIMDS_MASK);
		}

		WREG32(GRBM_GFX_INDEX, gfx);
		WREG32(RLC_GFX_INDEX, gfx);

		WREG32(CC_RB_BACKEND_DISABLE, rb);
		WREG32(CC_SYS_RB_BACKEND_DISABLE, rb);
		WREG32(GC_USER_RB_BACKEND_DISABLE, rb);
		WREG32(CC_GC_SHADER_PIPE_CONFIG, sp);
        }

	grbm_gfx_index |= SE_BROADCAST_WRITES;
	WREG32(GRBM_GFX_INDEX, grbm_gfx_index);
	WREG32(RLC_GFX_INDEX, grbm_gfx_index);

	WREG32(CGTS_SYS_TCC_DISABLE, 0);
	WREG32(CGTS_TCC_DISABLE, 0);
	WREG32(CGTS_USER_SYS_TCC_DISABLE, 0);
	WREG32(CGTS_USER_TCC_DISABLE, 0);

	/* set HW defaults for 3D engine */
	WREG32(CP_QUEUE_THRESHOLDS, (ROQ_IB1_START(0x16) |
				     ROQ_IB2_START(0x2b)));

	WREG32(CP_MEQ_THRESHOLDS, STQ_SPLIT(0x30));

	WREG32(TA_CNTL_AUX, (DISABLE_CUBE_ANISO |
			     SYNC_GRADIENT |
			     SYNC_WALKER |
			     SYNC_ALIGNER));

	sx_debug_1 = RREG32(SX_DEBUG_1);
	sx_debug_1 |= ENABLE_NEW_SMX_ADDRESS;
	WREG32(SX_DEBUG_1, sx_debug_1);


	smx_dc_ctl0 = RREG32(SMX_DC_CTL0);
	smx_dc_ctl0 &= ~NUMBER_OF_SETS(0x1ff);
	smx_dc_ctl0 |= NUMBER_OF_SETS(rdev->config.evergreen.sx_num_of_sets);
	WREG32(SMX_DC_CTL0, smx_dc_ctl0);

	WREG32(SX_EXPORT_BUFFER_SIZES, (COLOR_BUFFER_SIZE((rdev->config.evergreen.sx_max_export_size / 4) - 1) |
					POSITION_BUFFER_SIZE((rdev->config.evergreen.sx_max_export_pos_size / 4) - 1) |
					SMX_BUFFER_SIZE((rdev->config.evergreen.sx_max_export_smx_size / 4) - 1)));

	WREG32(PA_SC_FIFO_SIZE, (SC_PRIM_FIFO_SIZE(rdev->config.evergreen.sc_prim_fifo_size) |
				 SC_HIZ_TILE_FIFO_SIZE(rdev->config.evergreen.sc_hiz_tile_fifo_size) |
				 SC_EARLYZ_TILE_FIFO_SIZE(rdev->config.evergreen.sc_earlyz_tile_fifo_size)));

	WREG32(VGT_NUM_INSTANCES, 1);
	WREG32(SPI_CONFIG_CNTL, 0);
	WREG32(SPI_CONFIG_CNTL_1, VTX_DONE_DELAY(4));
	WREG32(CP_PERFMON_CNTL, 0);

	WREG32(SQ_MS_FIFO_SIZES, (CACHE_FIFO_SIZE(16 * rdev->config.evergreen.sq_num_cf_insts) |
				  FETCH_FIFO_HIWATER(0x4) |
				  DONE_FIFO_HIWATER(0xe0) |
				  ALU_UPDATE_FIFO_HIWATER(0x8)));

	sq_config = RREG32(SQ_CONFIG);
	sq_config &= ~(PS_PRIO(3) |
		       VS_PRIO(3) |
		       GS_PRIO(3) |
		       ES_PRIO(3));
	sq_config |= (VC_ENABLE |
		      EXPORT_SRC_C |
		      PS_PRIO(0) |
		      VS_PRIO(1) |
		      GS_PRIO(2) |
		      ES_PRIO(3));

	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;
		break;
	default:
		break;
	}

	sq_lds_resource_mgmt = RREG32(SQ_LDS_RESOURCE_MGMT);

	sq_gpr_resource_mgmt_1 = NUM_PS_GPRS((rdev->config.evergreen.max_gprs - (4 * 2))* 12 / 32);
	sq_gpr_resource_mgmt_1 |= NUM_VS_GPRS((rdev->config.evergreen.max_gprs - (4 * 2)) * 6 / 32);
	sq_gpr_resource_mgmt_1 |= NUM_CLAUSE_TEMP_GPRS(4);
	sq_gpr_resource_mgmt_2 = NUM_GS_GPRS((rdev->config.evergreen.max_gprs - (4 * 2)) * 4 / 32);
	sq_gpr_resource_mgmt_2 |= NUM_ES_GPRS((rdev->config.evergreen.max_gprs - (4 * 2)) * 4 / 32);
	sq_gpr_resource_mgmt_3 = NUM_HS_GPRS((rdev->config.evergreen.max_gprs - (4 * 2)) * 3 / 32);
	sq_gpr_resource_mgmt_3 |= NUM_LS_GPRS((rdev->config.evergreen.max_gprs - (4 * 2)) * 3 / 32);

	switch (rdev->family) {
	case CHIP_CEDAR:
	case CHIP_PALM:
	case CHIP_SUMO:
	case CHIP_SUMO2:
		ps_thread_count = 96;
		break;
	default:
		ps_thread_count = 128;
		break;
	}

	sq_thread_resource_mgmt = NUM_PS_THREADS(ps_thread_count);
	sq_thread_resource_mgmt |= NUM_VS_THREADS((((rdev->config.evergreen.max_threads - ps_thread_count) / 6) / 8) * 8);
	sq_thread_resource_mgmt |= NUM_GS_THREADS((((rdev->config.evergreen.max_threads - ps_thread_count) / 6) / 8) * 8);
	sq_thread_resource_mgmt |= NUM_ES_THREADS((((rdev->config.evergreen.max_threads - ps_thread_count) / 6) / 8) * 8);
	sq_thread_resource_mgmt_2 = NUM_HS_THREADS((((rdev->config.evergreen.max_threads - ps_thread_count) / 6) / 8) * 8);
	sq_thread_resource_mgmt_2 |= NUM_LS_THREADS((((rdev->config.evergreen.max_threads - ps_thread_count) / 6) / 8) * 8);

	sq_stack_resource_mgmt_1 = NUM_PS_STACK_ENTRIES((rdev->config.evergreen.max_stack_entries * 1) / 6);
	sq_stack_resource_mgmt_1 |= NUM_VS_STACK_ENTRIES((rdev->config.evergreen.max_stack_entries * 1) / 6);
	sq_stack_resource_mgmt_2 = NUM_GS_STACK_ENTRIES((rdev->config.evergreen.max_stack_entries * 1) / 6);
	sq_stack_resource_mgmt_2 |= NUM_ES_STACK_ENTRIES((rdev->config.evergreen.max_stack_entries * 1) / 6);
	sq_stack_resource_mgmt_3 = NUM_HS_STACK_ENTRIES((rdev->config.evergreen.max_stack_entries * 1) / 6);
	sq_stack_resource_mgmt_3 |= NUM_LS_STACK_ENTRIES((rdev->config.evergreen.max_stack_entries * 1) / 6);

	WREG32(SQ_CONFIG, sq_config);
	WREG32(SQ_GPR_RESOURCE_MGMT_1, sq_gpr_resource_mgmt_1);
	WREG32(SQ_GPR_RESOURCE_MGMT_2, sq_gpr_resource_mgmt_2);
	WREG32(SQ_GPR_RESOURCE_MGMT_3, sq_gpr_resource_mgmt_3);
	WREG32(SQ_THREAD_RESOURCE_MGMT, sq_thread_resource_mgmt);
	WREG32(SQ_THREAD_RESOURCE_MGMT_2, sq_thread_resource_mgmt_2);
	WREG32(SQ_STACK_RESOURCE_MGMT_1, sq_stack_resource_mgmt_1);
	WREG32(SQ_STACK_RESOURCE_MGMT_2, sq_stack_resource_mgmt_2);
	WREG32(SQ_STACK_RESOURCE_MGMT_3, sq_stack_resource_mgmt_3);
	WREG32(SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0);
	WREG32(SQ_LDS_RESOURCE_MGMT, sq_lds_resource_mgmt);

	WREG32(PA_SC_FORCE_EOV_MAX_CNTS, (FORCE_EOV_MAX_CLK_CNT(4095) |
					  FORCE_EOV_MAX_REZ_CNT(255)));

	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;
	default:
		vgt_cache_invalidation = CACHE_INVALIDATION(VC_AND_TC);
		break;
	}
	vgt_cache_invalidation |= AUTO_INVLD_EN(ES_AND_GS_AUTO);
	WREG32(VGT_CACHE_INVALIDATION, vgt_cache_invalidation);

	WREG32(VGT_GS_VERTEX_REUSE, 16);
	WREG32(PA_SU_LINE_STIPPLE_VALUE, 0);
	WREG32(PA_SC_LINE_STIPPLE_STATE, 0);

	WREG32(VGT_VERTEX_REUSE_BLOCK_CNTL, 14);
	WREG32(VGT_OUT_DEALLOC_CNTL, 16);

	WREG32(CB_PERF_CTR0_SEL_0, 0);
	WREG32(CB_PERF_CTR0_SEL_1, 0);
	WREG32(CB_PERF_CTR1_SEL_0, 0);
	WREG32(CB_PERF_CTR1_SEL_1, 0);
	WREG32(CB_PERF_CTR2_SEL_0, 0);
	WREG32(CB_PERF_CTR2_SEL_1, 0);
	WREG32(CB_PERF_CTR3_SEL_0, 0);
	WREG32(CB_PERF_CTR3_SEL_1, 0);

	/* clear render buffer base addresses */
	WREG32(CB_COLOR0_BASE, 0);
	WREG32(CB_COLOR1_BASE, 0);
	WREG32(CB_COLOR2_BASE, 0);
	WREG32(CB_COLOR3_BASE, 0);
	WREG32(CB_COLOR4_BASE, 0);
	WREG32(CB_COLOR5_BASE, 0);
	WREG32(CB_COLOR6_BASE, 0);
	WREG32(CB_COLOR7_BASE, 0);
	WREG32(CB_COLOR8_BASE, 0);
	WREG32(CB_COLOR9_BASE, 0);
	WREG32(CB_COLOR10_BASE, 0);
	WREG32(CB_COLOR11_BASE, 0);

	/* set the shader const cache sizes to 0 */
	for (i = SQ_ALU_CONST_BUFFER_SIZE_PS_0; i < 0x28200; i += 4)
		WREG32(i, 0);
	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);

	WREG32(PA_CL_ENHANCE, CLIP_VTX_REORDER_ENA | NUM_CLIP_SEQ(3));

	udelay(50);

}

int evergreen_mc_init(struct radeon_device *rdev)
{
	u32 tmp;
	int chansize, numchan;

	/* Get VRAM informations */
	rdev->mc.vram_is_ddr = true;
	tmp = RREG32(MC_ARB_RAMCFG);
	if (tmp & CHANSIZE_OVERRIDE) {
		chansize = 16;
	} else if (tmp & CHANSIZE_MASK) {
		chansize = 64;
	} else {
		chansize = 32;
	}
	tmp = RREG32(MC_SHARED_CHMAP);
	switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) {
	case 0:
	default:
		numchan = 1;
		break;
	case 1:
		numchan = 2;
		break;
	case 2:
		numchan = 4;
		break;
	case 3:
		numchan = 8;
		break;
	}
	rdev->mc.vram_width = numchan * chansize;
	/* Could aper size report 0 ? */
	rdev->mc.aper_base = pci_resource_start(rdev->pdev, 0);
	rdev->mc.aper_size = pci_resource_len(rdev->pdev, 0);
	/* Setup GPU memory space */
	if (rdev->flags & RADEON_IS_IGP) {
		/* size in bytes on fusion */
		rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE);
		rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE);
	} else {
	/* size in MB on evergreen */
	rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE) * 1024 * 1024;
	rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE) * 1024 * 1024;
	}
	rdev->mc.visible_vram_size = rdev->mc.aper_size;
	r700_vram_gtt_location(rdev, &rdev->mc);
	radeon_update_bandwidth_info(rdev);

	return 0;
}

bool evergreen_gpu_is_lockup(struct radeon_device *rdev)
{
	u32 srbm_status;
	u32 grbm_status;
	u32 grbm_status_se0, grbm_status_se1;
	struct r100_gpu_lockup *lockup = &rdev->config.evergreen.lockup;
	int r;

	srbm_status = RREG32(SRBM_STATUS);
	grbm_status = RREG32(GRBM_STATUS);
	grbm_status_se0 = RREG32(GRBM_STATUS_SE0);
	grbm_status_se1 = RREG32(GRBM_STATUS_SE1);
	if (!(grbm_status & GUI_ACTIVE)) {
		r100_gpu_lockup_update(lockup, &rdev->cp);
	return false;
	}
	/* force CP activities */
	r = radeon_ring_lock(rdev, 2);
	if (!r) {
		/* PACKET2 NOP */
		radeon_ring_write(rdev, 0x80000000);
		radeon_ring_write(rdev, 0x80000000);
		radeon_ring_unlock_commit(rdev);
	}
	rdev->cp.rptr = RREG32(CP_RB_RPTR);
	return r100_gpu_cp_is_lockup(rdev, lockup, &rdev->cp);
}

static int evergreen_gpu_soft_reset(struct radeon_device *rdev)
{
	struct evergreen_mc_save save;
	u32 grbm_reset = 0;

	if (!(RREG32(GRBM_STATUS) & GUI_ACTIVE))
		return 0;

	dev_info(rdev->dev, "GPU softreset \n");
	dev_info(rdev->dev, "  GRBM_STATUS=0x%08X\n",
		RREG32(GRBM_STATUS));
	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 (evergreen_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);

	/* reset all the gfx blocks */
	grbm_reset = (SOFT_RESET_CP |
		      SOFT_RESET_CB |
		      SOFT_RESET_DB |
		      SOFT_RESET_PA |
		      SOFT_RESET_SC |
		      SOFT_RESET_SPI |
		      SOFT_RESET_SH |
		      SOFT_RESET_SX |
		      SOFT_RESET_TC |
		      SOFT_RESET_TA |
		      SOFT_RESET_VC |
		      SOFT_RESET_VGT);

	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);
	/* 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_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;
}

int evergreen_asic_reset(struct radeon_device *rdev)
{
	return evergreen_gpu_soft_reset(rdev);
}

/* Interrupts */

u32 evergreen_get_vblank_counter(struct radeon_device *rdev, int crtc)
{
	switch (crtc) {
	case 0:
		return RREG32(CRTC_STATUS_FRAME_COUNT + EVERGREEN_CRTC0_REGISTER_OFFSET);
	case 1:
		return RREG32(CRTC_STATUS_FRAME_COUNT + EVERGREEN_CRTC1_REGISTER_OFFSET);
	case 2:
		return RREG32(CRTC_STATUS_FRAME_COUNT + EVERGREEN_CRTC2_REGISTER_OFFSET);
	case 3:
		return RREG32(CRTC_STATUS_FRAME_COUNT + EVERGREEN_CRTC3_REGISTER_OFFSET);
	case 4:
		return RREG32(CRTC_STATUS_FRAME_COUNT + EVERGREEN_CRTC4_REGISTER_OFFSET);
	case 5:
		return RREG32(CRTC_STATUS_FRAME_COUNT + EVERGREEN_CRTC5_REGISTER_OFFSET);
	default:
	return 0;
	}
}

void evergreen_disable_interrupt_state(struct radeon_device *rdev)
{
	u32 tmp;

	WREG32(CP_INT_CNTL, CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE);
	WREG32(GRBM_INT_CNTL, 0);
	WREG32(INT_MASK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
	WREG32(INT_MASK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
	if (!(rdev->flags & RADEON_IS_IGP)) {
	WREG32(INT_MASK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0);
	WREG32(INT_MASK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0);
	WREG32(INT_MASK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0);
	WREG32(INT_MASK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0);
	}

	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
	if (!(rdev->flags & RADEON_IS_IGP)) {
	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, 0);
	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, 0);
	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, 0);
	WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, 0);
	}

	WREG32(DACA_AUTODETECT_INT_CONTROL, 0);
	WREG32(DACB_AUTODETECT_INT_CONTROL, 0);

	tmp = RREG32(DC_HPD1_INT_CONTROL) & DC_HPDx_INT_POLARITY;
	WREG32(DC_HPD1_INT_CONTROL, tmp);
	tmp = RREG32(DC_HPD2_INT_CONTROL) & DC_HPDx_INT_POLARITY;
	WREG32(DC_HPD2_INT_CONTROL, tmp);
	tmp = RREG32(DC_HPD3_INT_CONTROL) & DC_HPDx_INT_POLARITY;
	WREG32(DC_HPD3_INT_CONTROL, tmp);
	tmp = RREG32(DC_HPD4_INT_CONTROL) & DC_HPDx_INT_POLARITY;
	WREG32(DC_HPD4_INT_CONTROL, tmp);
	tmp = RREG32(DC_HPD5_INT_CONTROL) & DC_HPDx_INT_POLARITY;
	WREG32(DC_HPD5_INT_CONTROL, tmp);
	tmp = RREG32(DC_HPD6_INT_CONTROL) & DC_HPDx_INT_POLARITY;
	WREG32(DC_HPD6_INT_CONTROL, tmp);

}
static int evergreen_startup(struct radeon_device *rdev)
{
	int r;

	/* enable pcie gen2 link */
	if (!ASIC_IS_DCE5(rdev))
		evergreen_pcie_gen2_enable(rdev);

	if (ASIC_IS_DCE5(rdev)) {
		if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw || !rdev->mc_fw) {
			r = ni_init_microcode(rdev);
			if (r) {
				DRM_ERROR("Failed to load firmware!\n");
				return r;
			}
		}
		r = ni_mc_load_microcode(rdev);
		if (r) {
			DRM_ERROR("Failed to load MC firmware!\n");
			return r;
		}
	} else {
	if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) {
		r = r600_init_microcode(rdev);
		if (r) {
			DRM_ERROR("Failed to load firmware!\n");
			return r;
		}
	}
	}

	evergreen_mc_program(rdev);
	if (rdev->flags & RADEON_IS_AGP) {
		evergreen_agp_enable(rdev);
	} else {
		r = evergreen_pcie_gart_enable(rdev);
		if (r)
			return r;
	}
	evergreen_gpu_init(rdev);
#if 0
	r = evergreen_blit_init(rdev);
	if (r) {
		evergreen_blit_fini(rdev);
		rdev->asic->copy = NULL;
		dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r);
	}

	/* allocate wb buffer */
	r = radeon_wb_init(rdev);
	if (r)
		return r;
#endif

	/* Enable IRQ */

    r = radeon_ring_init(rdev, rdev->cp.ring_size);
	if (r)
		return r;
	r = evergreen_cp_load_microcode(rdev);
	if (r)
		return r;
	r = evergreen_cp_resume(rdev);
	if (r)
		return r;

	return 0;
}







/* Plan is to move initialization in that function and use
 * helper function so that radeon_device_init pretty much
 * do nothing more than calling asic specific function. This
 * should also allow to remove a bunch of callback function
 * like vram_info.
 */
int evergreen_init(struct radeon_device *rdev)
{
	int r;

	/* This don't do much */
	r = radeon_gem_init(rdev);
	if (r)
		return r;
	/* Read BIOS */
	if (!radeon_get_bios(rdev)) {
		if (ASIC_IS_AVIVO(rdev))
			return -EINVAL;
	}
	/* Must be an ATOMBIOS */
	if (!rdev->is_atom_bios) {
		dev_err(rdev->dev, "Expecting atombios for evergreen GPU\n");
		return -EINVAL;
	}
	r = radeon_atombios_init(rdev);
	if (r)
		return r;
	/* reset the asic, the gfx blocks are often in a bad state
	 * after the driver is unloaded or after a resume
	 */
	if (radeon_asic_reset(rdev))
		dev_warn(rdev->dev, "GPU reset failed !\n");
	/* Post card if necessary */
	if (!radeon_card_posted(rdev)) {
		if (!rdev->bios) {
			dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n");
			return -EINVAL;
		}
		DRM_INFO("GPU not posted. posting now...\n");
		atom_asic_init(rdev->mode_info.atom_context);
	}
	/* Initialize scratch registers */
	r600_scratch_init(rdev);
	/* Initialize surface registers */
	radeon_surface_init(rdev);
	/* Initialize clocks */
	radeon_get_clock_info(rdev->ddev);
	/* Fence driver */
    /* initialize AGP */
	if (rdev->flags & RADEON_IS_AGP) {
		r = radeon_agp_init(rdev);
		if (r)
			radeon_agp_disable(rdev);
	}
	/* initialize memory controller */
	r = evergreen_mc_init(rdev);
	if (r)
		return r;
	/* Memory manager */
	r = radeon_bo_init(rdev);
	if (r)
		return r;


	rdev->cp.ring_obj = NULL;
	r600_ring_init(rdev, 1024 * 1024);

//   rdev->ih.ring_obj = NULL;
//   r600_ih_ring_init(rdev, 64 * 1024);

	r = r600_pcie_gart_init(rdev);
	if (r)
		return r;

	rdev->accel_working = true;
	r = evergreen_startup(rdev);
	if (r) {
		dev_err(rdev->dev, "disabling GPU acceleration\n");
		rdev->accel_working = false;
	}
	if (rdev->accel_working) {
	}
	return 0;
}


static void evergreen_pcie_gen2_enable(struct radeon_device *rdev)
{
	u32 link_width_cntl, speed_cntl;

	if (radeon_pcie_gen2 == 0)
		return;

	if (rdev->flags & RADEON_IS_IGP)
		return;

	if (!(rdev->flags & RADEON_IS_PCIE))
		return;

	/* x2 cards have a special sequence */
	if (ASIC_IS_X2(rdev))
		return;

	speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL);
	if ((speed_cntl & LC_OTHER_SIDE_EVER_SENT_GEN2) ||
	    (speed_cntl & LC_OTHER_SIDE_SUPPORTS_GEN2)) {

		link_width_cntl = RREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL);
		link_width_cntl &= ~LC_UPCONFIGURE_DIS;
		WREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl);

		speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL);
		speed_cntl &= ~LC_TARGET_LINK_SPEED_OVERRIDE_EN;
		WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl);

		speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL);
		speed_cntl |= LC_CLR_FAILED_SPD_CHANGE_CNT;
		WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl);

		speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL);
		speed_cntl &= ~LC_CLR_FAILED_SPD_CHANGE_CNT;
		WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl);

		speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL);
		speed_cntl |= LC_GEN2_EN_STRAP;
		WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl);

	} else {
		link_width_cntl = RREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL);
		/* XXX: only disable it if gen1 bridge vendor == 0x111d or 0x1106 */
		if (1)
			link_width_cntl |= LC_UPCONFIGURE_DIS;
		else
			link_width_cntl &= ~LC_UPCONFIGURE_DIS;
		WREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl);
	}
}