/*
 * Copyright 2008 Advanced Micro Devices, Inc.
 * Copyright 2008 Red Hat Inc.
 * Copyright 2009 Jerome Glisse.
 *
 * 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: Dave Airlie
 *          Alex Deucher
 *          Jerome Glisse
 */

#include <linux/firmware.h>
#include <linux/module.h>
#include <linux/pci.h>
#include <linux/slab.h>
#include <linux/seq_file.h>

#include <drm/drm_debugfs.h>
#include <drm/drm_device.h>
#include <drm/drm_vblank.h>
#include <drm/radeon_drm.h>

#include "atom.h"
#include "avivod.h"
#include "evergreen.h"
#include "r600.h"
#include "r600d.h"
#include "rv770.h"
#include "radeon.h"
#include "radeon_asic.h"
#include "radeon_audio.h"
#include "radeon_mode.h"
#include "radeon_ucode.h"

/* Firmware Names */
MODULE_FIRMWARE("radeon/R600_pfp.bin");
MODULE_FIRMWARE("radeon/R600_me.bin");
MODULE_FIRMWARE("radeon/RV610_pfp.bin");
MODULE_FIRMWARE("radeon/RV610_me.bin");
MODULE_FIRMWARE("radeon/RV630_pfp.bin");
MODULE_FIRMWARE("radeon/RV630_me.bin");
MODULE_FIRMWARE("radeon/RV620_pfp.bin");
MODULE_FIRMWARE("radeon/RV620_me.bin");
MODULE_FIRMWARE("radeon/RV635_pfp.bin");
MODULE_FIRMWARE("radeon/RV635_me.bin");
MODULE_FIRMWARE("radeon/RV670_pfp.bin");
MODULE_FIRMWARE("radeon/RV670_me.bin");
MODULE_FIRMWARE("radeon/RS780_pfp.bin");
MODULE_FIRMWARE("radeon/RS780_me.bin");
MODULE_FIRMWARE("radeon/RV770_pfp.bin");
MODULE_FIRMWARE("radeon/RV770_me.bin");
MODULE_FIRMWARE("radeon/RV770_smc.bin");
MODULE_FIRMWARE("radeon/RV730_pfp.bin");
MODULE_FIRMWARE("radeon/RV730_me.bin");
MODULE_FIRMWARE("radeon/RV730_smc.bin");
MODULE_FIRMWARE("radeon/RV740_smc.bin");
MODULE_FIRMWARE("radeon/RV710_pfp.bin");
MODULE_FIRMWARE("radeon/RV710_me.bin");
MODULE_FIRMWARE("radeon/RV710_smc.bin");
MODULE_FIRMWARE("radeon/R600_rlc.bin");
MODULE_FIRMWARE("radeon/R700_rlc.bin");
MODULE_FIRMWARE("radeon/CEDAR_pfp.bin");
MODULE_FIRMWARE("radeon/CEDAR_me.bin");
MODULE_FIRMWARE("radeon/CEDAR_rlc.bin");
MODULE_FIRMWARE("radeon/CEDAR_smc.bin");
MODULE_FIRMWARE("radeon/REDWOOD_pfp.bin");
MODULE_FIRMWARE("radeon/REDWOOD_me.bin");
MODULE_FIRMWARE("radeon/REDWOOD_rlc.bin");
MODULE_FIRMWARE("radeon/REDWOOD_smc.bin");
MODULE_FIRMWARE("radeon/JUNIPER_pfp.bin");
MODULE_FIRMWARE("radeon/JUNIPER_me.bin");
MODULE_FIRMWARE("radeon/JUNIPER_rlc.bin");
MODULE_FIRMWARE("radeon/JUNIPER_smc.bin");
MODULE_FIRMWARE("radeon/CYPRESS_pfp.bin");
MODULE_FIRMWARE("radeon/CYPRESS_me.bin");
MODULE_FIRMWARE("radeon/CYPRESS_rlc.bin");
MODULE_FIRMWARE("radeon/CYPRESS_smc.bin");
MODULE_FIRMWARE("radeon/PALM_pfp.bin");
MODULE_FIRMWARE("radeon/PALM_me.bin");
MODULE_FIRMWARE("radeon/SUMO_rlc.bin");
MODULE_FIRMWARE("radeon/SUMO_pfp.bin");
MODULE_FIRMWARE("radeon/SUMO_me.bin");
MODULE_FIRMWARE("radeon/SUMO2_pfp.bin");
MODULE_FIRMWARE("radeon/SUMO2_me.bin");

static const u32 crtc_offsets[2] =
{
        0,
        AVIVO_D2CRTC_H_TOTAL - AVIVO_D1CRTC_H_TOTAL
};

int r600_debugfs_mc_info_init(struct radeon_device *rdev);

/* r600,rv610,rv630,rv620,rv635,rv670 */
int r600_mc_wait_for_idle(struct radeon_device *rdev);
static void r600_gpu_init(struct radeon_device *rdev);
void r600_fini(struct radeon_device *rdev);
void r600_irq_disable(struct radeon_device *rdev);
static void r600_pcie_gen2_enable(struct radeon_device *rdev);

/*
 * Indirect registers accessor
 */
u32 r600_rcu_rreg(struct radeon_device *rdev, u32 reg)
{
        unsigned long flags;
        u32 r;

        spin_lock_irqsave(&rdev->rcu_idx_lock, flags);
        WREG32(R600_RCU_INDEX, ((reg) & 0x1fff));
        r = RREG32(R600_RCU_DATA);
        spin_unlock_irqrestore(&rdev->rcu_idx_lock, flags);
        return r;
}

void r600_rcu_wreg(struct radeon_device *rdev, u32 reg, u32 v)
{
        unsigned long flags;

        spin_lock_irqsave(&rdev->rcu_idx_lock, flags);
        WREG32(R600_RCU_INDEX, ((reg) & 0x1fff));
        WREG32(R600_RCU_DATA, (v));
        spin_unlock_irqrestore(&rdev->rcu_idx_lock, flags);
}

u32 r600_uvd_ctx_rreg(struct radeon_device *rdev, u32 reg)
{
        unsigned long flags;
        u32 r;

        spin_lock_irqsave(&rdev->uvd_idx_lock, flags);
        WREG32(R600_UVD_CTX_INDEX, ((reg) & 0x1ff));
        r = RREG32(R600_UVD_CTX_DATA);
        spin_unlock_irqrestore(&rdev->uvd_idx_lock, flags);
        return r;
}

void r600_uvd_ctx_wreg(struct radeon_device *rdev, u32 reg, u32 v)
{
        unsigned long flags;

        spin_lock_irqsave(&rdev->uvd_idx_lock, flags);
        WREG32(R600_UVD_CTX_INDEX, ((reg) & 0x1ff));
        WREG32(R600_UVD_CTX_DATA, (v));
        spin_unlock_irqrestore(&rdev->uvd_idx_lock, flags);
}

/**
 * r600_get_allowed_info_register - fetch the register for the info ioctl
 *
 * @rdev: radeon_device pointer
 * @reg: register offset in bytes
 * @val: register value
 *
 * Returns 0 for success or -EINVAL for an invalid register
 *
 */
int r600_get_allowed_info_register(struct radeon_device *rdev,
                                   u32 reg, u32 *val)
{
        switch (reg) {
        case GRBM_STATUS:
        case GRBM_STATUS2:
        case R_000E50_SRBM_STATUS:
        case DMA_STATUS_REG:
        case UVD_STATUS:
                *val = RREG32(reg);
                return 0;
        default:
                return -EINVAL;
        }
}

/**
 * r600_get_xclk - get the xclk
 *
 * @rdev: radeon_device pointer
 *
 * Returns the reference clock used by the gfx engine
 * (r6xx, IGPs, APUs).
 */
u32 r600_get_xclk(struct radeon_device *rdev)
{
        return rdev->clock.spll.reference_freq;
}

int r600_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk)
{
        unsigned fb_div = 0, ref_div, vclk_div = 0, dclk_div = 0;
        int r;

        /* bypass vclk and dclk with bclk */
        WREG32_P(CG_UPLL_FUNC_CNTL_2,
                 VCLK_SRC_SEL(1) | DCLK_SRC_SEL(1),
                 ~(VCLK_SRC_SEL_MASK | DCLK_SRC_SEL_MASK));

        /* assert BYPASS_EN, deassert UPLL_RESET, UPLL_SLEEP and UPLL_CTLREQ */
        WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_BYPASS_EN_MASK, ~(
                 UPLL_RESET_MASK | UPLL_SLEEP_MASK | UPLL_CTLREQ_MASK));

        if (rdev->family >= CHIP_RS780)
                WREG32_P(GFX_MACRO_BYPASS_CNTL, UPLL_BYPASS_CNTL,
                         ~UPLL_BYPASS_CNTL);

        if (!vclk || !dclk) {
                /* keep the Bypass mode, put PLL to sleep */
                WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_SLEEP_MASK, ~UPLL_SLEEP_MASK);
                return 0;
        }

        if (rdev->clock.spll.reference_freq == 10000)
                ref_div = 34;
        else
                ref_div = 4;

        r = radeon_uvd_calc_upll_dividers(rdev, vclk, dclk, 50000, 160000,
                                          ref_div + 1, 0xFFF, 2, 30, ~0,
                                          &fb_div, &vclk_div, &dclk_div);
        if (r)
                return r;

        if (rdev->family >= CHIP_RV670 && rdev->family < CHIP_RS780)
                fb_div >>= 1;
        else
                fb_div |= 1;

        r = radeon_uvd_send_upll_ctlreq(rdev, CG_UPLL_FUNC_CNTL);
        if (r)
                return r;

        /* assert PLL_RESET */
        WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_RESET_MASK, ~UPLL_RESET_MASK);

        /* For RS780 we have to choose ref clk */
        if (rdev->family >= CHIP_RS780)
                WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_REFCLK_SRC_SEL_MASK,
                         ~UPLL_REFCLK_SRC_SEL_MASK);

        /* set the required fb, ref and post divder values */
        WREG32_P(CG_UPLL_FUNC_CNTL,
                 UPLL_FB_DIV(fb_div) |
                 UPLL_REF_DIV(ref_div),
                 ~(UPLL_FB_DIV_MASK | UPLL_REF_DIV_MASK));
        WREG32_P(CG_UPLL_FUNC_CNTL_2,
                 UPLL_SW_HILEN(vclk_div >> 1) |
                 UPLL_SW_LOLEN((vclk_div >> 1) + (vclk_div & 1)) |
                 UPLL_SW_HILEN2(dclk_div >> 1) |
                 UPLL_SW_LOLEN2((dclk_div >> 1) + (dclk_div & 1)) |
                 UPLL_DIVEN_MASK | UPLL_DIVEN2_MASK,
                 ~UPLL_SW_MASK);

        /* give the PLL some time to settle */
        mdelay(15);

        /* deassert PLL_RESET */
        WREG32_P(CG_UPLL_FUNC_CNTL, 0, ~UPLL_RESET_MASK);

        mdelay(15);

        /* deassert BYPASS EN */
        WREG32_P(CG_UPLL_FUNC_CNTL, 0, ~UPLL_BYPASS_EN_MASK);

        if (rdev->family >= CHIP_RS780)
                WREG32_P(GFX_MACRO_BYPASS_CNTL, 0, ~UPLL_BYPASS_CNTL);

        r = radeon_uvd_send_upll_ctlreq(rdev, CG_UPLL_FUNC_CNTL);
        if (r)
                return r;

        /* switch VCLK and DCLK selection */
        WREG32_P(CG_UPLL_FUNC_CNTL_2,
                 VCLK_SRC_SEL(2) | DCLK_SRC_SEL(2),
                 ~(VCLK_SRC_SEL_MASK | DCLK_SRC_SEL_MASK));

        mdelay(100);

        return 0;
}

void dce3_program_fmt(struct drm_encoder *encoder)
{
        struct drm_device *dev = encoder->dev;
        struct radeon_device *rdev = dev->dev_private;
        struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder);
        struct radeon_crtc *radeon_crtc = to_radeon_crtc(encoder->crtc);
        struct drm_connector *connector = radeon_get_connector_for_encoder(encoder);
        int bpc = 0;
        u32 tmp = 0;
        enum radeon_connector_dither dither = RADEON_FMT_DITHER_DISABLE;

        if (connector) {
                struct radeon_connector *radeon_connector = to_radeon_connector(connector);
                bpc = radeon_get_monitor_bpc(connector);
                dither = radeon_connector->dither;
        }

        /* LVDS FMT is set up by atom */
        if (radeon_encoder->devices & ATOM_DEVICE_LCD_SUPPORT)
                return;

        /* not needed for analog */
        if ((radeon_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1) ||
            (radeon_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2))
                return;

        if (bpc == 0)
                return;

        switch (bpc) {
        case 6:
                if (dither == RADEON_FMT_DITHER_ENABLE)
                        /* XXX sort out optimal dither settings */
                        tmp |= FMT_SPATIAL_DITHER_EN;
                else
                        tmp |= FMT_TRUNCATE_EN;
                break;
        case 8:
                if (dither == RADEON_FMT_DITHER_ENABLE)
                        /* XXX sort out optimal dither settings */
                        tmp |= (FMT_SPATIAL_DITHER_EN | FMT_SPATIAL_DITHER_DEPTH);
                else
                        tmp |= (FMT_TRUNCATE_EN | FMT_TRUNCATE_DEPTH);
                break;
        case 10:
        default:
                /* not needed */
                break;
        }

        WREG32(FMT_BIT_DEPTH_CONTROL + radeon_crtc->crtc_offset, tmp);
}

/* get temperature in millidegrees */
int rv6xx_get_temp(struct radeon_device *rdev)
{
        u32 temp = (RREG32(CG_THERMAL_STATUS) & ASIC_T_MASK) >>
                ASIC_T_SHIFT;
        int actual_temp = temp & 0xff;

        if (temp & 0x100)
                actual_temp -= 256;

        return actual_temp * 1000;
}

void r600_pm_get_dynpm_state(struct radeon_device *rdev)
{
        int i;

        rdev->pm.dynpm_can_upclock = true;
        rdev->pm.dynpm_can_downclock = true;

        /* power state array is low to high, default is first */
        if ((rdev->flags & RADEON_IS_IGP) || (rdev->family == CHIP_R600)) {
                int min_power_state_index = 0;

                if (rdev->pm.num_power_states > 2)
                        min_power_state_index = 1;

                switch (rdev->pm.dynpm_planned_action) {
                case DYNPM_ACTION_MINIMUM:
                        rdev->pm.requested_power_state_index = min_power_state_index;
                        rdev->pm.requested_clock_mode_index = 0;
                        rdev->pm.dynpm_can_downclock = false;
                        break;
                case DYNPM_ACTION_DOWNCLOCK:
                        if (rdev->pm.current_power_state_index == min_power_state_index) {
                                rdev->pm.requested_power_state_index = rdev->pm.current_power_state_index;
                                rdev->pm.dynpm_can_downclock = false;
                        } else {
                                if (rdev->pm.active_crtc_count > 1) {
                                        for (i = 0; i < rdev->pm.num_power_states; i++) {
                                                if (rdev->pm.power_state[i].flags & RADEON_PM_STATE_SINGLE_DISPLAY_ONLY)
                                                        continue;
                                                else if (i >= rdev->pm.current_power_state_index) {
                                                        rdev->pm.requested_power_state_index =
                                                                rdev->pm.current_power_state_index;
                                                        break;
                                                } else {
                                                        rdev->pm.requested_power_state_index = i;
                                                        break;
                                                }
                                        }
                                } else {
                                        if (rdev->pm.current_power_state_index == 0)
                                                rdev->pm.requested_power_state_index =
                                                        rdev->pm.num_power_states - 1;
                                        else
                                                rdev->pm.requested_power_state_index =
                                                        rdev->pm.current_power_state_index - 1;
                                }
                        }
                        rdev->pm.requested_clock_mode_index = 0;
                        /* don't use the power state if crtcs are active and no display flag is set */
                        if ((rdev->pm.active_crtc_count > 0) &&
                            (rdev->pm.power_state[rdev->pm.requested_power_state_index].
                             clock_info[rdev->pm.requested_clock_mode_index].flags &
                             RADEON_PM_MODE_NO_DISPLAY)) {
                                rdev->pm.requested_power_state_index++;
                        }
                        break;
                case DYNPM_ACTION_UPCLOCK:
                        if (rdev->pm.current_power_state_index == (rdev->pm.num_power_states - 1)) {
                                rdev->pm.requested_power_state_index = rdev->pm.current_power_state_index;
                                rdev->pm.dynpm_can_upclock = false;
                        } else {
                                if (rdev->pm.active_crtc_count > 1) {
                                        for (i = (rdev->pm.num_power_states - 1); i >= 0; i--) {
                                                if (rdev->pm.power_state[i].flags & RADEON_PM_STATE_SINGLE_DISPLAY_ONLY)
                                                        continue;
                                                else if (i <= rdev->pm.current_power_state_index) {
                                                        rdev->pm.requested_power_state_index =
                                                                rdev->pm.current_power_state_index;
                                                        break;
                                                } else {
                                                        rdev->pm.requested_power_state_index = i;
                                                        break;
                                                }
                                        }
                                } else
                                        rdev->pm.requested_power_state_index =
                                                rdev->pm.current_power_state_index + 1;
                        }
                        rdev->pm.requested_clock_mode_index = 0;
                        break;
                case DYNPM_ACTION_DEFAULT:
                        rdev->pm.requested_power_state_index = rdev->pm.default_power_state_index;
                        rdev->pm.requested_clock_mode_index = 0;
                        rdev->pm.dynpm_can_upclock = false;
                        break;
                case DYNPM_ACTION_NONE:
                default:
                        DRM_ERROR("Requested mode for not defined action\n");
                        return;
                }
        } else {
                /* XXX select a power state based on AC/DC, single/dualhead, etc. */
                /* for now just select the first power state and switch between clock modes */
                /* power state array is low to high, default is first (0) */
                if (rdev->pm.active_crtc_count > 1) {
                        rdev->pm.requested_power_state_index = -1;
                        /* start at 1 as we don't want the default mode */
                        for (i = 1; i < rdev->pm.num_power_states; i++) {
                                if (rdev->pm.power_state[i].flags & RADEON_PM_STATE_SINGLE_DISPLAY_ONLY)
                                        continue;
                                else if ((rdev->pm.power_state[i].type == POWER_STATE_TYPE_PERFORMANCE) ||
                                         (rdev->pm.power_state[i].type == POWER_STATE_TYPE_BATTERY)) {
                                        rdev->pm.requested_power_state_index = i;
                                        break;
                                }
                        }
                        /* if nothing selected, grab the default state. */
                        if (rdev->pm.requested_power_state_index == -1)
                                rdev->pm.requested_power_state_index = 0;
                } else
                        rdev->pm.requested_power_state_index = 1;

                switch (rdev->pm.dynpm_planned_action) {
                case DYNPM_ACTION_MINIMUM:
                        rdev->pm.requested_clock_mode_index = 0;
                        rdev->pm.dynpm_can_downclock = false;
                        break;
                case DYNPM_ACTION_DOWNCLOCK:
                        if (rdev->pm.requested_power_state_index == rdev->pm.current_power_state_index) {
                                if (rdev->pm.current_clock_mode_index == 0) {
                                        rdev->pm.requested_clock_mode_index = 0;
                                        rdev->pm.dynpm_can_downclock = false;
                                } else
                                        rdev->pm.requested_clock_mode_index =
                                                rdev->pm.current_clock_mode_index - 1;
                        } else {
                                rdev->pm.requested_clock_mode_index = 0;
                                rdev->pm.dynpm_can_downclock = false;
                        }
                        /* don't use the power state if crtcs are active and no display flag is set */
                        if ((rdev->pm.active_crtc_count > 0) &&
                            (rdev->pm.power_state[rdev->pm.requested_power_state_index].
                             clock_info[rdev->pm.requested_clock_mode_index].flags &
                             RADEON_PM_MODE_NO_DISPLAY)) {
                                rdev->pm.requested_clock_mode_index++;
                        }
                        break;
                case DYNPM_ACTION_UPCLOCK:
                        if (rdev->pm.requested_power_state_index == rdev->pm.current_power_state_index) {
                                if (rdev->pm.current_clock_mode_index ==
                                    (rdev->pm.power_state[rdev->pm.requested_power_state_index].num_clock_modes - 1)) {
                                        rdev->pm.requested_clock_mode_index = rdev->pm.current_clock_mode_index;
                                        rdev->pm.dynpm_can_upclock = false;
                                } else
                                        rdev->pm.requested_clock_mode_index =
                                                rdev->pm.current_clock_mode_index + 1;
                        } else {
                                rdev->pm.requested_clock_mode_index =
                                        rdev->pm.power_state[rdev->pm.requested_power_state_index].num_clock_modes - 1;
                                rdev->pm.dynpm_can_upclock = false;
                        }
                        break;
                case DYNPM_ACTION_DEFAULT:
                        rdev->pm.requested_power_state_index = rdev->pm.default_power_state_index;
                        rdev->pm.requested_clock_mode_index = 0;
                        rdev->pm.dynpm_can_upclock = false;
                        break;
                case DYNPM_ACTION_NONE:
                default:
                        DRM_ERROR("Requested mode for not defined action\n");
                        return;
                }
        }

        DRM_DEBUG_DRIVER("Requested: e: %d m: %d p: %d\n",
                  rdev->pm.power_state[rdev->pm.requested_power_state_index].
                  clock_info[rdev->pm.requested_clock_mode_index].sclk,
                  rdev->pm.power_state[rdev->pm.requested_power_state_index].
                  clock_info[rdev->pm.requested_clock_mode_index].mclk,
                  rdev->pm.power_state[rdev->pm.requested_power_state_index].
                  pcie_lanes);
}

void rs780_pm_init_profile(struct radeon_device *rdev)
{
        if (rdev->pm.num_power_states == 2) {
                /* default */
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 0;
                /* low sh */
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
                /* mid sh */
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0;
                /* high sh */
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 0;
                /* low mh */
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
                /* mid mh */
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0;
                /* high mh */
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 0;
        } else if (rdev->pm.num_power_states == 3) {
                /* default */
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 0;
                /* low sh */
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
                /* mid sh */
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0;
                /* high sh */
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 2;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 0;
                /* low mh */
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
                /* mid mh */
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0;
                /* high mh */
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 1;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 2;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 0;
        } else {
                /* default */
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 0;
                /* low sh */
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = 2;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 2;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
                /* mid sh */
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 2;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 2;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0;
                /* high sh */
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 2;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 3;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 0;
                /* low mh */
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = 2;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
                /* mid mh */
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 2;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0;
                /* high mh */
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 2;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 3;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 0;
        }
}

void r600_pm_init_profile(struct radeon_device *rdev)
{
        int idx;

        if (rdev->family == CHIP_R600) {
                /* XXX */
                /* default */
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 0;
                /* low sh */
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
                /* mid sh */
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 0;
                /* high sh */
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 0;
                /* low mh */
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
                /* mid mh */
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 0;
                /* high mh */
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0;
                rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 0;
        } else {
                if (rdev->pm.num_power_states < 4) {
                        /* default */
                        rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                        rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                        rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 2;
                        /* low sh */
                        rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = 1;
                        rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = 1;
                        rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
                        /* mid sh */
                        rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = 1;
                        rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = 1;
                        rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 1;
                        /* high sh */
                        rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = 1;
                        rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = 1;
                        rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 2;
                        /* low mh */
                        rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = 2;
                        rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = 2;
                        rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
                        /* low mh */
                        rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = 2;
                        rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = 2;
                        rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 1;
                        /* high mh */
                        rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = 2;
                        rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = 2;
                        rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 2;
                } else {
                        /* default */
                        rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_ps_idx = rdev->pm.default_power_state_index;
                        rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_ps_idx = rdev->pm.default_power_state_index;
                        rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_DEFAULT_IDX].dpms_on_cm_idx = 2;
                        /* low sh */
                        if (rdev->flags & RADEON_IS_MOBILITY)
                                idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 0);
                        else
                                idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0);
                        rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_LOW_SH_IDX].dpms_on_cm_idx = 0;
                        /* mid sh */
                        rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_MID_SH_IDX].dpms_on_cm_idx = 1;
                        /* high sh */
                        idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 0);
                        rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_HIGH_SH_IDX].dpms_on_cm_idx = 2;
                        /* low mh */
                        if (rdev->flags & RADEON_IS_MOBILITY)
                                idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_BATTERY, 1);
                        else
                                idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1);
                        rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_LOW_MH_IDX].dpms_on_cm_idx = 0;
                        /* mid mh */
                        rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_MID_MH_IDX].dpms_on_cm_idx = 1;
                        /* high mh */
                        idx = radeon_pm_get_type_index(rdev, POWER_STATE_TYPE_PERFORMANCE, 1);
                        rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_ps_idx = idx;
                        rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_off_cm_idx = 0;
                        rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx = 2;
                }
        }
}

void r600_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) && voltage->voltage) {
                /* 0xff01 is a flag rather then an actual voltage */
                if (voltage->voltage == 0xff01)
                        return;
                if (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_DRIVER("Setting: v: %d\n", voltage->voltage);
                }
        }
}

bool r600_gui_idle(struct radeon_device *rdev)
{
        if (RREG32(GRBM_STATUS) & GUI_ACTIVE)
                return false;
        else
                return true;
}

/* hpd for digital panel detect/disconnect */
bool r600_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd)
{
        bool connected = false;

        if (ASIC_IS_DCE3(rdev)) {
                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;
                        /* DCE 3.2 */
                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;
                }
        } else {
                switch (hpd) {
                case RADEON_HPD_1:
                        if (RREG32(DC_HOT_PLUG_DETECT1_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE)
                                connected = true;
                        break;
                case RADEON_HPD_2:
                        if (RREG32(DC_HOT_PLUG_DETECT2_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE)
                                connected = true;
                        break;
                case RADEON_HPD_3:
                        if (RREG32(DC_HOT_PLUG_DETECT3_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE)
                                connected = true;
                        break;
                default:
                        break;
                }
        }
        return connected;
}

void r600_hpd_set_polarity(struct radeon_device *rdev,
                           enum radeon_hpd_id hpd)
{
        u32 tmp;
        bool connected = r600_hpd_sense(rdev, hpd);

        if (ASIC_IS_DCE3(rdev)) {
                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;
                        /* DCE 3.2 */
                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;
                }
        } else {
                switch (hpd) {
                case RADEON_HPD_1:
                        tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL);
                        if (connected)
                                tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY;
                        else
                                tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY;
                        WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp);
                        break;
                case RADEON_HPD_2:
                        tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL);
                        if (connected)
                                tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY;
                        else
                                tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY;
                        WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp);
                        break;
                case RADEON_HPD_3:
                        tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL);
                        if (connected)
                                tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY;
                        else
                                tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY;
                        WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp);
                        break;
                default:
                        break;
                }
        }
}

void r600_hpd_init(struct radeon_device *rdev)
{
        struct drm_device *dev = rdev->ddev;
        struct drm_connector *connector;
        unsigned enable = 0;

        list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
                struct radeon_connector *radeon_connector = to_radeon_connector(connector);

                if (connector->connector_type == DRM_MODE_CONNECTOR_eDP ||
                    connector->connector_type == DRM_MODE_CONNECTOR_LVDS) {
                        /* don't try to enable hpd on eDP or LVDS avoid breaking the
                         * aux dp channel on imac and help (but not completely fix)
                         * https://bugzilla.redhat.com/show_bug.cgi?id=726143
                         */
                        continue;
                }
                if (ASIC_IS_DCE3(rdev)) {
                        u32 tmp = DC_HPDx_CONNECTION_TIMER(0x9c4) | DC_HPDx_RX_INT_TIMER(0xfa);
                        if (ASIC_IS_DCE32(rdev))
                                tmp |= DC_HPDx_EN;

                        switch (radeon_connector->hpd.hpd) {
                        case RADEON_HPD_1:
                                WREG32(DC_HPD1_CONTROL, tmp);
                                break;
                        case RADEON_HPD_2:
                                WREG32(DC_HPD2_CONTROL, tmp);
                                break;
                        case RADEON_HPD_3:
                                WREG32(DC_HPD3_CONTROL, tmp);
                                break;
                        case RADEON_HPD_4:
                                WREG32(DC_HPD4_CONTROL, tmp);
                                break;
                                /* DCE 3.2 */
                        case RADEON_HPD_5:
                                WREG32(DC_HPD5_CONTROL, tmp);
                                break;
                        case RADEON_HPD_6:
                                WREG32(DC_HPD6_CONTROL, tmp);
                                break;
                        default:
                                break;
                        }
                } else {
                        switch (radeon_connector->hpd.hpd) {
                        case RADEON_HPD_1:
                                WREG32(DC_HOT_PLUG_DETECT1_CONTROL, DC_HOT_PLUG_DETECTx_EN);

                                break;
                        case RADEON_HPD_2:
                                WREG32(DC_HOT_PLUG_DETECT2_CONTROL, DC_HOT_PLUG_DETECTx_EN);
                                break;
                        case RADEON_HPD_3:
                                WREG32(DC_HOT_PLUG_DETECT3_CONTROL, DC_HOT_PLUG_DETECTx_EN);
                                break;
                        default:
                                break;
                        }
                }
                if (radeon_connector->hpd.hpd != RADEON_HPD_NONE)
                        enable |= 1 << radeon_connector->hpd.hpd;
                radeon_hpd_set_polarity(rdev, radeon_connector->hpd.hpd);
        }
        radeon_irq_kms_enable_hpd(rdev, enable);
}

void r600_hpd_fini(struct radeon_device *rdev)
{
        struct drm_device *dev = rdev->ddev;
        struct drm_connector *connector;
        unsigned disable = 0;

        list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
                struct radeon_connector *radeon_connector = to_radeon_connector(connector);
                if (ASIC_IS_DCE3(rdev)) {
                        switch (radeon_connector->hpd.hpd) {
                        case RADEON_HPD_1:
                                WREG32(DC_HPD1_CONTROL, 0);
                                break;
                        case RADEON_HPD_2:
                                WREG32(DC_HPD2_CONTROL, 0);
                                break;
                        case RADEON_HPD_3:
                                WREG32(DC_HPD3_CONTROL, 0);
                                break;
                        case RADEON_HPD_4:
                                WREG32(DC_HPD4_CONTROL, 0);
                                break;
                                /* DCE 3.2 */
                        case RADEON_HPD_5:
                                WREG32(DC_HPD5_CONTROL, 0);
                                break;
                        case RADEON_HPD_6:
                                WREG32(DC_HPD6_CONTROL, 0);
                                break;
                        default:
                                break;
                        }
                } else {
                        switch (radeon_connector->hpd.hpd) {
                        case RADEON_HPD_1:
                                WREG32(DC_HOT_PLUG_DETECT1_CONTROL, 0);
                                break;
                        case RADEON_HPD_2:
                                WREG32(DC_HOT_PLUG_DETECT2_CONTROL, 0);
                                break;
                        case RADEON_HPD_3:
                                WREG32(DC_HOT_PLUG_DETECT3_CONTROL, 0);
                                break;
                        default:
                                break;
                        }
                }
                if (radeon_connector->hpd.hpd != RADEON_HPD_NONE)
                        disable |= 1 << radeon_connector->hpd.hpd;
        }
        radeon_irq_kms_disable_hpd(rdev, disable);
}

/*
 * R600 PCIE GART
 */
void r600_pcie_gart_tlb_flush(struct radeon_device *rdev)
{
        unsigned i;
        u32 tmp;

        /* flush hdp cache so updates hit vram */
        if ((rdev->family >= CHIP_RV770) && (rdev->family <= CHIP_RV740) &&
            !(rdev->flags & RADEON_IS_AGP)) {
                void __iomem *ptr = (void *)rdev->gart.ptr;

                /* r7xx hw bug.  write to HDP_DEBUG1 followed by fb read
                 * rather than write to HDP_REG_COHERENCY_FLUSH_CNTL
                 * This seems to cause problems on some AGP cards. Just use the old
                 * method for them.
                 */
                WREG32(HDP_DEBUG1, 0);
                readl((void __iomem *)ptr);
        } else
                WREG32(R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1);

        WREG32(VM_CONTEXT0_INVALIDATION_LOW_ADDR, rdev->mc.gtt_start >> 12);
        WREG32(VM_CONTEXT0_INVALIDATION_HIGH_ADDR, (rdev->mc.gtt_end - 1) >> 12);
        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) {
                        pr_warn("[drm] r600 flush TLB failed\n");
                        return;
                }
                if (tmp) {
                        return;
                }
                udelay(1);
        }
}

int r600_pcie_gart_init(struct radeon_device *rdev)
{
        int r;

        if (rdev->gart.robj) {
                WARN(1, "R600 PCIE GART already initialized\n");
                return 0;
        }
        /* Initialize common gart structure */
        r = radeon_gart_init(rdev);
        if (r)
                return r;
        rdev->gart.table_size = rdev->gart.num_gpu_pages * 8;
        return radeon_gart_table_vram_alloc(rdev);
}

static int r600_pcie_gart_enable(struct radeon_device *rdev)
{
        u32 tmp;
        int r, i;

        if (rdev->gart.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;

        /* 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(0) | BANK_SELECT_1(1));
        /* Setup TLB control */
        tmp = ENABLE_L1_TLB | ENABLE_L1_FRAGMENT_PROCESSING |
                SYSTEM_ACCESS_MODE_NOT_IN_SYS |
                EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5) |
                ENABLE_WAIT_L2_QUERY;
        WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp | ENABLE_L1_STRICT_ORDERING);
        WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_UVD_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_UVD_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE);
        WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE);
        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));
        for (i = 1; i < 7; i++)
                WREG32(VM_CONTEXT0_CNTL + (i * 4), 0);

        r600_pcie_gart_tlb_flush(rdev);
        DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
                 (unsigned)(rdev->mc.gtt_size >> 20),
                 (unsigned long long)rdev->gart.table_addr);
        rdev->gart.ready = true;
        return 0;
}

static void r600_pcie_gart_disable(struct radeon_device *rdev)
{
        u32 tmp;
        int i;

        /* Disable all tables */
        for (i = 0; i < 7; i++)
                WREG32(VM_CONTEXT0_CNTL + (i * 4), 0);

        /* Disable L2 cache */
        WREG32(VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING |
                                EFFECTIVE_L2_QUEUE_SIZE(7));
        WREG32(VM_L2_CNTL3, BANK_SELECT_0(0) | BANK_SELECT_1(1));
        /* Setup L1 TLB control */
        tmp = EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5) |
                ENABLE_WAIT_L2_QUERY;
        WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_UVD_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_UVD_CNTL, tmp);
        radeon_gart_table_vram_unpin(rdev);
}

static void r600_pcie_gart_fini(struct radeon_device *rdev)
{
        radeon_gart_fini(rdev);
        r600_pcie_gart_disable(rdev);
        radeon_gart_table_vram_free(rdev);
}

static void r600_agp_enable(struct radeon_device *rdev)
{
        u32 tmp;
        int i;

        /* 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(0) | BANK_SELECT_1(1));
        /* Setup TLB control */
        tmp = ENABLE_L1_TLB | ENABLE_L1_FRAGMENT_PROCESSING |
                SYSTEM_ACCESS_MODE_NOT_IN_SYS |
                EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5) |
                ENABLE_WAIT_L2_QUERY;
        WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp | ENABLE_L1_STRICT_ORDERING);
        WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp);
        WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE);
        WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE);
        for (i = 0; i < 7; i++)
                WREG32(VM_CONTEXT0_CNTL + (i * 4), 0);
}

int r600_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(R_000E50_SRBM_STATUS) & 0x3F00;
                if (!tmp)
                        return 0;
                udelay(1);
        }
        return -1;
}

uint32_t rs780_mc_rreg(struct radeon_device *rdev, uint32_t reg)
{
        unsigned long flags;
        uint32_t r;

        spin_lock_irqsave(&rdev->mc_idx_lock, flags);
        WREG32(R_0028F8_MC_INDEX, S_0028F8_MC_IND_ADDR(reg));
        r = RREG32(R_0028FC_MC_DATA);
        WREG32(R_0028F8_MC_INDEX, ~C_0028F8_MC_IND_ADDR);
        spin_unlock_irqrestore(&rdev->mc_idx_lock, flags);
        return r;
}

void rs780_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v)
{
        unsigned long flags;

        spin_lock_irqsave(&rdev->mc_idx_lock, flags);
        WREG32(R_0028F8_MC_INDEX, S_0028F8_MC_IND_ADDR(reg) |
                S_0028F8_MC_IND_WR_EN(1));
        WREG32(R_0028FC_MC_DATA, v);
        WREG32(R_0028F8_MC_INDEX, 0x7F);
        spin_unlock_irqrestore(&rdev->mc_idx_lock, flags);
}

static void r600_mc_program(struct radeon_device *rdev)
{
        struct rv515_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);

        rv515_mc_stop(rdev, &save);
        if (r600_mc_wait_for_idle(rdev)) {
                dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
        }
        /* Lockout access through VGA aperture (doesn't exist before R600) */
        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, rdev->vram_scratch.gpu_addr >> 12);
        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));
        WREG32(HDP_NONSURFACE_SIZE, 0x3FFFFFFF);
        if (rdev->flags & RADEON_IS_AGP) {
                WREG32(MC_VM_AGP_TOP, rdev->mc.gtt_end >> 22);
                WREG32(MC_VM_AGP_BOT, rdev->mc.gtt_start >> 22);
                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 (r600_mc_wait_for_idle(rdev)) {
                dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
        }
        rv515_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);
}

/**
 * r600_vram_gtt_location - try to find VRAM & GTT location
 * @rdev: radeon device structure holding all necessary informations
 * @mc: memory controller structure holding memory informations
 *
 * Function will place try to place VRAM at same place as in CPU (PCI)
 * address space as some GPU seems to have issue when we reprogram at
 * different address space.
 *
 * If there is not enough space to fit the unvisible VRAM after the
 * aperture then we limit the VRAM size to the aperture.
 *
 * If we are using AGP then place VRAM adjacent to AGP aperture are we need
 * them to be in one from GPU point of view so that we can program GPU to
 * catch access outside them (weird GPU policy see ??).
 *
 * This function will never fails, worst case are limiting VRAM or GTT.
 *
 * Note: GTT start, end, size should be initialized before calling this
 * function on AGP platform.
 */
static void r600_vram_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc)
{
        u64 size_bf, size_af;

        if (mc->mc_vram_size > 0xE0000000) {
                /* leave room for at least 512M GTT */
                dev_warn(rdev->dev, "limiting VRAM\n");
                mc->real_vram_size = 0xE0000000;
                mc->mc_vram_size = 0xE0000000;
        }
        if (rdev->flags & RADEON_IS_AGP) {
                size_bf = mc->gtt_start;
                size_af = mc->mc_mask - mc->gtt_end;
                if (size_bf > size_af) {
                        if (mc->mc_vram_size > size_bf) {
                                dev_warn(rdev->dev, "limiting VRAM\n");
                                mc->real_vram_size = size_bf;
                                mc->mc_vram_size = size_bf;
                        }
                        mc->vram_start = mc->gtt_start - mc->mc_vram_size;
                } else {
                        if (mc->mc_vram_size > size_af) {
                                dev_warn(rdev->dev, "limiting VRAM\n");
                                mc->real_vram_size = size_af;
                                mc->mc_vram_size = size_af;
                        }
                        mc->vram_start = mc->gtt_end + 1;
                }
                mc->vram_end = mc->vram_start + mc->mc_vram_size - 1;
                dev_info(rdev->dev, "VRAM: %lluM 0x%08llX - 0x%08llX (%lluM used)\n",
                                mc->mc_vram_size >> 20, mc->vram_start,
                                mc->vram_end, mc->real_vram_size >> 20);
        } else {
                u64 base = 0;
                if (rdev->flags & RADEON_IS_IGP) {
                        base = RREG32(MC_VM_FB_LOCATION) & 0xFFFF;
                        base <<= 24;
                }
                radeon_vram_location(rdev, &rdev->mc, base);
                rdev->mc.gtt_base_align = 0;
                radeon_gtt_location(rdev, mc);
        }
}

static int r600_mc_init(struct radeon_device *rdev)
{
        u32 tmp;
        int chansize, numchan;
        uint32_t h_addr, l_addr;
        unsigned long long k8_addr;

        /* Get VRAM informations */
        rdev->mc.vram_is_ddr = true;
        tmp = RREG32(RAMCFG);
        if (tmp & CHANSIZE_OVERRIDE) {
                chansize = 16;
        } else if (tmp & CHANSIZE_MASK) {
                chansize = 64;
        } else {
                chansize = 32;
        }
        tmp = RREG32(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 */
        rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE);
        rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE);
        rdev->mc.visible_vram_size = rdev->mc.aper_size;
        r600_vram_gtt_location(rdev, &rdev->mc);

        if (rdev->flags & RADEON_IS_IGP) {
                rs690_pm_info(rdev);
                rdev->mc.igp_sideport_enabled = radeon_atombios_sideport_present(rdev);

                if (rdev->family == CHIP_RS780 || rdev->family == CHIP_RS880) {
                        /* Use K8 direct mapping for fast fb access. */
                        rdev->fastfb_working = false;
                        h_addr = G_000012_K8_ADDR_EXT(RREG32_MC(R_000012_MC_MISC_UMA_CNTL));
                        l_addr = RREG32_MC(R_000011_K8_FB_LOCATION);
                        k8_addr = ((unsigned long long)h_addr) << 32 | l_addr;
#if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
                        if (k8_addr + rdev->mc.visible_vram_size < 0x100000000ULL)
#endif
                        {
                                /* FastFB shall be used with UMA memory. Here it is simply disabled when sideport
                                * memory is present.
                                */
                                if (rdev->mc.igp_sideport_enabled == false && radeon_fastfb == 1) {
                                        DRM_INFO("Direct mapping: aper base at 0x%llx, replaced by direct mapping base 0x%llx.\n",
                                                (unsigned long long)rdev->mc.aper_base, k8_addr);
                                        rdev->mc.aper_base = (resource_size_t)k8_addr;
                                        rdev->fastfb_working = true;
                                }
                        }
                }
        }

        radeon_update_bandwidth_info(rdev);
        return 0;
}

int r600_vram_scratch_init(struct radeon_device *rdev)
{
        int r;

        if (rdev->vram_scratch.robj == NULL) {
                r = radeon_bo_create(rdev, RADEON_GPU_PAGE_SIZE,
                                     PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM,
                                     0, NULL, NULL, &rdev->vram_scratch.robj);
                if (r) {
                        return r;
                }
        }

        r = radeon_bo_reserve(rdev->vram_scratch.robj, false);
        if (unlikely(r != 0))
                return r;
        r = radeon_bo_pin(rdev->vram_scratch.robj,
                          RADEON_GEM_DOMAIN_VRAM, &rdev->vram_scratch.gpu_addr);
        if (r) {
                radeon_bo_unreserve(rdev->vram_scratch.robj);
                return r;
        }
        r = radeon_bo_kmap(rdev->vram_scratch.robj,
                                (void **)&rdev->vram_scratch.ptr);
        if (r)
                radeon_bo_unpin(rdev->vram_scratch.robj);
        radeon_bo_unreserve(rdev->vram_scratch.robj);

        return r;
}

void r600_vram_scratch_fini(struct radeon_device *rdev)
{
        int r;

        if (rdev->vram_scratch.robj == NULL) {
                return;
        }
        r = radeon_bo_reserve(rdev->vram_scratch.robj, false);
        if (likely(r == 0)) {
                radeon_bo_kunmap(rdev->vram_scratch.robj);
                radeon_bo_unpin(rdev->vram_scratch.robj);
                radeon_bo_unreserve(rdev->vram_scratch.robj);
        }
        radeon_bo_unref(&rdev->vram_scratch.robj);
}

void r600_set_bios_scratch_engine_hung(struct radeon_device *rdev, bool hung)
{
        u32 tmp = RREG32(R600_BIOS_3_SCRATCH);

        if (hung)
                tmp |= ATOM_S3_ASIC_GUI_ENGINE_HUNG;
        else
                tmp &= ~ATOM_S3_ASIC_GUI_ENGINE_HUNG;

        WREG32(R600_BIOS_3_SCRATCH, tmp);
}

static void r600_print_gpu_status_regs(struct radeon_device *rdev)
{
        dev_info(rdev->dev, "  R_008010_GRBM_STATUS      = 0x%08X\n",
                 RREG32(R_008010_GRBM_STATUS));
        dev_info(rdev->dev, "  R_008014_GRBM_STATUS2     = 0x%08X\n",
                 RREG32(R_008014_GRBM_STATUS2));
        dev_info(rdev->dev, "  R_000E50_SRBM_STATUS      = 0x%08X\n",
                 RREG32(R_000E50_SRBM_STATUS));
        dev_info(rdev->dev, "  R_008674_CP_STALLED_STAT1 = 0x%08X\n",
                 RREG32(CP_STALLED_STAT1));
        dev_info(rdev->dev, "  R_008678_CP_STALLED_STAT2 = 0x%08X\n",
                 RREG32(CP_STALLED_STAT2));
        dev_info(rdev->dev, "  R_00867C_CP_BUSY_STAT     = 0x%08X\n",
                 RREG32(CP_BUSY_STAT));
        dev_info(rdev->dev, "  R_008680_CP_STAT          = 0x%08X\n",
                 RREG32(CP_STAT));
        dev_info(rdev->dev, "  R_00D034_DMA_STATUS_REG   = 0x%08X\n",
                RREG32(DMA_STATUS_REG));
}

static bool r600_is_display_hung(struct radeon_device *rdev)
{
        u32 crtc_hung = 0;
        u32 crtc_status[2];
        u32 i, j, tmp;

        for (i = 0; i < rdev->num_crtc; i++) {
                if (RREG32(AVIVO_D1CRTC_CONTROL + crtc_offsets[i]) & AVIVO_CRTC_EN) {
                        crtc_status[i] = RREG32(AVIVO_D1CRTC_STATUS_HV_COUNT + crtc_offsets[i]);
                        crtc_hung |= (1 << i);
                }
        }

        for (j = 0; j < 10; j++) {
                for (i = 0; i < rdev->num_crtc; i++) {
                        if (crtc_hung & (1 << i)) {
                                tmp = RREG32(AVIVO_D1CRTC_STATUS_HV_COUNT + crtc_offsets[i]);
                                if (tmp != crtc_status[i])
                                        crtc_hung &= ~(1 << i);
                        }
                }
                if (crtc_hung == 0)
                        return false;
                udelay(100);
        }

        return true;
}

u32 r600_gpu_check_soft_reset(struct radeon_device *rdev)
{
        u32 reset_mask = 0;
        u32 tmp;

        /* GRBM_STATUS */
        tmp = RREG32(R_008010_GRBM_STATUS);
        if (rdev->family >= CHIP_RV770) {
                if (G_008010_PA_BUSY(tmp) | G_008010_SC_BUSY(tmp) |
                    G_008010_SH_BUSY(tmp) | G_008010_SX_BUSY(tmp) |
                    G_008010_TA_BUSY(tmp) | G_008010_VGT_BUSY(tmp) |
                    G_008010_DB03_BUSY(tmp) | G_008010_CB03_BUSY(tmp) |
                    G_008010_SPI03_BUSY(tmp) | G_008010_VGT_BUSY_NO_DMA(tmp))
                        reset_mask |= RADEON_RESET_GFX;
        } else {
                if (G_008010_PA_BUSY(tmp) | G_008010_SC_BUSY(tmp) |
                    G_008010_SH_BUSY(tmp) | G_008010_SX_BUSY(tmp) |
                    G_008010_TA03_BUSY(tmp) | G_008010_VGT_BUSY(tmp) |
                    G_008010_DB03_BUSY(tmp) | G_008010_CB03_BUSY(tmp) |
                    G_008010_SPI03_BUSY(tmp) | G_008010_VGT_BUSY_NO_DMA(tmp))
                        reset_mask |= RADEON_RESET_GFX;
        }

        if (G_008010_CF_RQ_PENDING(tmp) | G_008010_PF_RQ_PENDING(tmp) |
            G_008010_CP_BUSY(tmp) | G_008010_CP_COHERENCY_BUSY(tmp))
                reset_mask |= RADEON_RESET_CP;

        if (G_008010_GRBM_EE_BUSY(tmp))
                reset_mask |= RADEON_RESET_GRBM | RADEON_RESET_GFX | RADEON_RESET_CP;

        /* DMA_STATUS_REG */
        tmp = RREG32(DMA_STATUS_REG);
        if (!(tmp & DMA_IDLE))
                reset_mask |= RADEON_RESET_DMA;

        /* SRBM_STATUS */
        tmp = RREG32(R_000E50_SRBM_STATUS);
        if (G_000E50_RLC_RQ_PENDING(tmp) | G_000E50_RLC_BUSY(tmp))
                reset_mask |= RADEON_RESET_RLC;

        if (G_000E50_IH_BUSY(tmp))
                reset_mask |= RADEON_RESET_IH;

        if (G_000E50_SEM_BUSY(tmp))
                reset_mask |= RADEON_RESET_SEM;

        if (G_000E50_GRBM_RQ_PENDING(tmp))
                reset_mask |= RADEON_RESET_GRBM;

        if (G_000E50_VMC_BUSY(tmp))
                reset_mask |= RADEON_RESET_VMC;

        if (G_000E50_MCB_BUSY(tmp) | G_000E50_MCDZ_BUSY(tmp) |
            G_000E50_MCDY_BUSY(tmp) | G_000E50_MCDX_BUSY(tmp) |
            G_000E50_MCDW_BUSY(tmp))
                reset_mask |= RADEON_RESET_MC;

        if (r600_is_display_hung(rdev))
                reset_mask |= RADEON_RESET_DISPLAY;

        /* Skip MC reset as it's mostly likely not hung, just busy */
        if (reset_mask & RADEON_RESET_MC) {
                DRM_DEBUG("MC busy: 0x%08X, clearing.\n", reset_mask);
                reset_mask &= ~RADEON_RESET_MC;
        }

        return reset_mask;
}

static void r600_gpu_soft_reset(struct radeon_device *rdev, u32 reset_mask)
{
        struct rv515_mc_save save;
        u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
        u32 tmp;

        if (reset_mask == 0)
                return;

        dev_info(rdev->dev, "GPU softreset: 0x%08X\n", reset_mask);

        r600_print_gpu_status_regs(rdev);

        /* Disable CP parsing/prefetching */
        if (rdev->family >= CHIP_RV770)
                WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1) | S_0086D8_CP_PFP_HALT(1));
        else
                WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1));

        /* disable the RLC */
        WREG32(RLC_CNTL, 0);

        if (reset_mask & RADEON_RESET_DMA) {
                /* Disable DMA */
                tmp = RREG32(DMA_RB_CNTL);
                tmp &= ~DMA_RB_ENABLE;
                WREG32(DMA_RB_CNTL, tmp);
        }

        mdelay(50);

        rv515_mc_stop(rdev, &save);
        if (r600_mc_wait_for_idle(rdev)) {
                dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
        }

        if (reset_mask & (RADEON_RESET_GFX | RADEON_RESET_COMPUTE)) {
                if (rdev->family >= CHIP_RV770)
                        grbm_soft_reset |= S_008020_SOFT_RESET_DB(1) |
                                S_008020_SOFT_RESET_CB(1) |
                                S_008020_SOFT_RESET_PA(1) |
                                S_008020_SOFT_RESET_SC(1) |
                                S_008020_SOFT_RESET_SPI(1) |
                                S_008020_SOFT_RESET_SX(1) |
                                S_008020_SOFT_RESET_SH(1) |
                                S_008020_SOFT_RESET_TC(1) |
                                S_008020_SOFT_RESET_TA(1) |
                                S_008020_SOFT_RESET_VC(1) |
                                S_008020_SOFT_RESET_VGT(1);
                else
                        grbm_soft_reset |= S_008020_SOFT_RESET_CR(1) |
                                S_008020_SOFT_RESET_DB(1) |
                                S_008020_SOFT_RESET_CB(1) |
                                S_008020_SOFT_RESET_PA(1) |
                                S_008020_SOFT_RESET_SC(1) |
                                S_008020_SOFT_RESET_SMX(1) |
                                S_008020_SOFT_RESET_SPI(1) |
                                S_008020_SOFT_RESET_SX(1) |
                                S_008020_SOFT_RESET_SH(1) |
                                S_008020_SOFT_RESET_TC(1) |
                                S_008020_SOFT_RESET_TA(1) |
                                S_008020_SOFT_RESET_VC(1) |
                                S_008020_SOFT_RESET_VGT(1);
        }

        if (reset_mask & RADEON_RESET_CP) {
                grbm_soft_reset |= S_008020_SOFT_RESET_CP(1) |
                        S_008020_SOFT_RESET_VGT(1);

                srbm_soft_reset |= S_000E60_SOFT_RESET_GRBM(1);
        }

        if (reset_mask & RADEON_RESET_DMA) {
                if (rdev->family >= CHIP_RV770)
                        srbm_soft_reset |= RV770_SOFT_RESET_DMA;
                else
                        srbm_soft_reset |= SOFT_RESET_DMA;
        }

        if (reset_mask & RADEON_RESET_RLC)
                srbm_soft_reset |= S_000E60_SOFT_RESET_RLC(1);

        if (reset_mask & RADEON_RESET_SEM)
                srbm_soft_reset |= S_000E60_SOFT_RESET_SEM(1);

        if (reset_mask & RADEON_RESET_IH)
                srbm_soft_reset |= S_000E60_SOFT_RESET_IH(1);

        if (reset_mask & RADEON_RESET_GRBM)
                srbm_soft_reset |= S_000E60_SOFT_RESET_GRBM(1);

        if (!(rdev->flags & RADEON_IS_IGP)) {
                if (reset_mask & RADEON_RESET_MC)
                        srbm_soft_reset |= S_000E60_SOFT_RESET_MC(1);
        }

        if (reset_mask & RADEON_RESET_VMC)
                srbm_soft_reset |= S_000E60_SOFT_RESET_VMC(1);

        if (grbm_soft_reset) {
                tmp = RREG32(R_008020_GRBM_SOFT_RESET);
                tmp |= grbm_soft_reset;
                dev_info(rdev->dev, "R_008020_GRBM_SOFT_RESET=0x%08X\n", tmp);
                WREG32(R_008020_GRBM_SOFT_RESET, tmp);
                tmp = RREG32(R_008020_GRBM_SOFT_RESET);

                udelay(50);

                tmp &= ~grbm_soft_reset;
                WREG32(R_008020_GRBM_SOFT_RESET, tmp);
                tmp = RREG32(R_008020_GRBM_SOFT_RESET);
        }

        if (srbm_soft_reset) {
                tmp = RREG32(SRBM_SOFT_RESET);
                tmp |= srbm_soft_reset;
                dev_info(rdev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
                WREG32(SRBM_SOFT_RESET, tmp);
                tmp = RREG32(SRBM_SOFT_RESET);

                udelay(50);

                tmp &= ~srbm_soft_reset;
                WREG32(SRBM_SOFT_RESET, tmp);
                tmp = RREG32(SRBM_SOFT_RESET);
        }

        /* Wait a little for things to settle down */
        mdelay(1);

        rv515_mc_resume(rdev, &save);
        udelay(50);

        r600_print_gpu_status_regs(rdev);
}

static void r600_gpu_pci_config_reset(struct radeon_device *rdev)
{
        struct rv515_mc_save save;
        u32 tmp, i;

        dev_info(rdev->dev, "GPU pci config reset\n");

        /* disable dpm? */

        /* Disable CP parsing/prefetching */
        if (rdev->family >= CHIP_RV770)
                WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1) | S_0086D8_CP_PFP_HALT(1));
        else
                WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1));

        /* disable the RLC */
        WREG32(RLC_CNTL, 0);

        /* Disable DMA */
        tmp = RREG32(DMA_RB_CNTL);
        tmp &= ~DMA_RB_ENABLE;
        WREG32(DMA_RB_CNTL, tmp);

        mdelay(50);

        /* set mclk/sclk to bypass */
        if (rdev->family >= CHIP_RV770)
                rv770_set_clk_bypass_mode(rdev);
        /* disable BM */
        pci_clear_master(rdev->pdev);
        /* disable mem access */
        rv515_mc_stop(rdev, &save);
        if (r600_mc_wait_for_idle(rdev)) {
                dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
        }

        /* BIF reset workaround.  Not sure if this is needed on 6xx */
        tmp = RREG32(BUS_CNTL);
        tmp |= VGA_COHE_SPEC_TIMER_DIS;
        WREG32(BUS_CNTL, tmp);

        tmp = RREG32(BIF_SCRATCH0);

        /* reset */
        radeon_pci_config_reset(rdev);
        mdelay(1);

        /* BIF reset workaround.  Not sure if this is needed on 6xx */
        tmp = SOFT_RESET_BIF;
        WREG32(SRBM_SOFT_RESET, tmp);
        mdelay(1);
        WREG32(SRBM_SOFT_RESET, 0);

        /* wait for asic to come out of reset */
        for (i = 0; i < rdev->usec_timeout; i++) {
                if (RREG32(CONFIG_MEMSIZE) != 0xffffffff)
                        break;
                udelay(1);
        }
}

int r600_asic_reset(struct radeon_device *rdev, bool hard)
{
        u32 reset_mask;

        if (hard) {
                r600_gpu_pci_config_reset(rdev);
                return 0;
        }

        reset_mask = r600_gpu_check_soft_reset(rdev);

        if (reset_mask)
                r600_set_bios_scratch_engine_hung(rdev, true);

        /* try soft reset */
        r600_gpu_soft_reset(rdev, reset_mask);

        reset_mask = r600_gpu_check_soft_reset(rdev);

        /* try pci config reset */
        if (reset_mask && radeon_hard_reset)
                r600_gpu_pci_config_reset(rdev);

        reset_mask = r600_gpu_check_soft_reset(rdev);

        if (!reset_mask)
                r600_set_bios_scratch_engine_hung(rdev, false);

        return 0;
}

/**
 * r600_gfx_is_lockup - Check if the GFX engine is locked up
 *
 * @rdev: radeon_device pointer
 * @ring: radeon_ring structure holding ring information
 *
 * Check if the GFX engine is locked up.
 * Returns true if the engine appears to be locked up, false if not.
 */
bool r600_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring)
{
        u32 reset_mask = r600_gpu_check_soft_reset(rdev);

        if (!(reset_mask & (RADEON_RESET_GFX |
                            RADEON_RESET_COMPUTE |
                            RADEON_RESET_CP))) {
                radeon_ring_lockup_update(rdev, ring);
                return false;
        }
        return radeon_ring_test_lockup(rdev, ring);
}

u32 r6xx_remap_render_backend(struct radeon_device *rdev,
                              u32 tiling_pipe_num,
                              u32 max_rb_num,
                              u32 total_max_rb_num,
                              u32 disabled_rb_mask)
{
        u32 rendering_pipe_num, rb_num_width, req_rb_num;
        u32 pipe_rb_ratio, pipe_rb_remain, tmp;
        u32 data = 0, mask = 1 << (max_rb_num - 1);
        unsigned i, j;

        /* mask out the RBs that don't exist on that asic */
        tmp = disabled_rb_mask | ((0xff << max_rb_num) & 0xff);
        /* make sure at least one RB is available */
        if ((tmp & 0xff) != 0xff)
                disabled_rb_mask = tmp;

        rendering_pipe_num = 1 << tiling_pipe_num;
        req_rb_num = total_max_rb_num - r600_count_pipe_bits(disabled_rb_mask);
        BUG_ON(rendering_pipe_num < req_rb_num);

        pipe_rb_ratio = rendering_pipe_num / req_rb_num;
        pipe_rb_remain = rendering_pipe_num - pipe_rb_ratio * req_rb_num;

        if (rdev->family <= CHIP_RV740) {
                /* r6xx/r7xx */
                rb_num_width = 2;
        } else {
                /* eg+ */
                rb_num_width = 4;
        }

        for (i = 0; i < max_rb_num; i++) {
                if (!(mask & disabled_rb_mask)) {
                        for (j = 0; j < pipe_rb_ratio; j++) {
                                data <<= rb_num_width;
                                data |= max_rb_num - i - 1;
                        }
                        if (pipe_rb_remain) {
                                data <<= rb_num_width;
                                data |= max_rb_num - i - 1;
                                pipe_rb_remain--;
                        }
                }
                mask >>= 1;
        }

        return data;
}

int r600_count_pipe_bits(uint32_t val)
{
        return hweight32(val);
}

static void r600_gpu_init(struct radeon_device *rdev)
{
        u32 tiling_config;
        u32 ramcfg;
        u32 cc_gc_shader_pipe_config;
        u32 tmp;
        int i, j;
        u32 sq_config;
        u32 sq_gpr_resource_mgmt_1 = 0;
        u32 sq_gpr_resource_mgmt_2 = 0;
        u32 sq_thread_resource_mgmt = 0;

        u32 sq_stack_resource_mgmt_1 = 0;
        u32 sq_stack_resource_mgmt_2 = 0;
        u32 disabled_rb_mask;

        rdev->config.r600.tiling_group_size = 256;
        switch (rdev->family) {
        case CHIP_R600:
                rdev->config.r600.max_pipes = 4;
                rdev->config.r600.max_tile_pipes = 8;
                rdev->config.r600.max_simds = 4;
                rdev->config.r600.max_backends = 4;
                rdev->config.r600.max_gprs = 256;
                rdev->config.r600.max_threads = 192;
                rdev->config.r600.max_stack_entries = 256;
                rdev->config.r600.max_hw_contexts = 8;
                rdev->config.r600.max_gs_threads = 16;
                rdev->config.r600.sx_max_export_size = 128;
                rdev->config.r600.sx_max_export_pos_size = 16;
                rdev->config.r600.sx_max_export_smx_size = 128;
                rdev->config.r600.sq_num_cf_insts = 2;
                break;
        case CHIP_RV630:
        case CHIP_RV635:
                rdev->config.r600.max_pipes = 2;
                rdev->config.r600.max_tile_pipes = 2;
                rdev->config.r600.max_simds = 3;
                rdev->config.r600.max_backends = 1;
                rdev->config.r600.max_gprs = 128;
                rdev->config.r600.max_threads = 192;
                rdev->config.r600.max_stack_entries = 128;
                rdev->config.r600.max_hw_contexts = 8;
                rdev->config.r600.max_gs_threads = 4;
                rdev->config.r600.sx_max_export_size = 128;
                rdev->config.r600.sx_max_export_pos_size = 16;
                rdev->config.r600.sx_max_export_smx_size = 128;
                rdev->config.r600.sq_num_cf_insts = 2;
                break;
        case CHIP_RV610:
        case CHIP_RV620:
        case CHIP_RS780:
        case CHIP_RS880:
                rdev->config.r600.max_pipes = 1;
                rdev->config.r600.max_tile_pipes = 1;
                rdev->config.r600.max_simds = 2;
                rdev->config.r600.max_backends = 1;
                rdev->config.r600.max_gprs = 128;
                rdev->config.r600.max_threads = 192;
                rdev->config.r600.max_stack_entries = 128;
                rdev->config.r600.max_hw_contexts = 4;
                rdev->config.r600.max_gs_threads = 4;
                rdev->config.r600.sx_max_export_size = 128;
                rdev->config.r600.sx_max_export_pos_size = 16;
                rdev->config.r600.sx_max_export_smx_size = 128;
                rdev->config.r600.sq_num_cf_insts = 1;
                break;
        case CHIP_RV670:
                rdev->config.r600.max_pipes = 4;
                rdev->config.r600.max_tile_pipes = 4;
                rdev->config.r600.max_simds = 4;
                rdev->config.r600.max_backends = 4;
                rdev->config.r600.max_gprs = 192;
                rdev->config.r600.max_threads = 192;
                rdev->config.r600.max_stack_entries = 256;
                rdev->config.r600.max_hw_contexts = 8;
                rdev->config.r600.max_gs_threads = 16;
                rdev->config.r600.sx_max_export_size = 128;
                rdev->config.r600.sx_max_export_pos_size = 16;
                rdev->config.r600.sx_max_export_smx_size = 128;
                rdev->config.r600.sq_num_cf_insts = 2;
                break;
        default:
                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));

        /* Setup tiling */
        tiling_config = 0;
        ramcfg = RREG32(RAMCFG);
        switch (rdev->config.r600.max_tile_pipes) {
        case 1:
                tiling_config |= PIPE_TILING(0);
                break;
        case 2:
                tiling_config |= PIPE_TILING(1);
                break;
        case 4:
                tiling_config |= PIPE_TILING(2);
                break;
        case 8:
                tiling_config |= PIPE_TILING(3);
                break;
        default:
                break;
        }
        rdev->config.r600.tiling_npipes = rdev->config.r600.max_tile_pipes;
        rdev->config.r600.tiling_nbanks = 4 << ((ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT);
        tiling_config |= BANK_TILING((ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT);
        tiling_config |= GROUP_SIZE((ramcfg & BURSTLENGTH_MASK) >> BURSTLENGTH_SHIFT);

        tmp = (ramcfg & NOOFROWS_MASK) >> NOOFROWS_SHIFT;
        if (tmp > 3) {
                tiling_config |= ROW_TILING(3);
                tiling_config |= SAMPLE_SPLIT(3);
        } else {
                tiling_config |= ROW_TILING(tmp);
                tiling_config |= SAMPLE_SPLIT(tmp);
        }
        tiling_config |= BANK_SWAPS(1);

        cc_gc_shader_pipe_config = RREG32(CC_GC_SHADER_PIPE_CONFIG) & 0x00ffff00;
        tmp = rdev->config.r600.max_simds -
                r600_count_pipe_bits((cc_gc_shader_pipe_config >> 16) & R6XX_MAX_SIMDS_MASK);
        rdev->config.r600.active_simds = tmp;

        disabled_rb_mask = (RREG32(CC_RB_BACKEND_DISABLE) >> 16) & R6XX_MAX_BACKENDS_MASK;
        tmp = 0;
        for (i = 0; i < rdev->config.r600.max_backends; i++)
                tmp |= (1 << i);
        /* if all the backends are disabled, fix it up here */
        if ((disabled_rb_mask & tmp) == tmp) {
                for (i = 0; i < rdev->config.r600.max_backends; i++)
                        disabled_rb_mask &= ~(1 << i);
        }
        tmp = (tiling_config & PIPE_TILING__MASK) >> PIPE_TILING__SHIFT;
        tmp = r6xx_remap_render_backend(rdev, tmp, rdev->config.r600.max_backends,
                                        R6XX_MAX_BACKENDS, disabled_rb_mask);
        tiling_config |= tmp << 16;
        rdev->config.r600.backend_map = tmp;

        rdev->config.r600.tile_config = tiling_config;
        WREG32(GB_TILING_CONFIG, tiling_config);
        WREG32(DCP_TILING_CONFIG, tiling_config & 0xffff);
        WREG32(HDP_TILING_CONFIG, tiling_config & 0xffff);
        WREG32(DMA_TILING_CONFIG, tiling_config & 0xffff);

        tmp = R6XX_MAX_PIPES - r600_count_pipe_bits((cc_gc_shader_pipe_config & INACTIVE_QD_PIPES_MASK) >> 8);
        WREG32(VGT_OUT_DEALLOC_CNTL, (tmp * 4) & DEALLOC_DIST_MASK);
        WREG32(VGT_VERTEX_REUSE_BLOCK_CNTL, ((tmp * 4) - 2) & VTX_REUSE_DEPTH_MASK);

        /* Setup some CP states */
        WREG32(CP_QUEUE_THRESHOLDS, (ROQ_IB1_START(0x16) | ROQ_IB2_START(0x2b)));
        WREG32(CP_MEQ_THRESHOLDS, (MEQ_END(0x40) | ROQ_END(0x40)));

        WREG32(TA_CNTL_AUX, (DISABLE_CUBE_ANISO | SYNC_GRADIENT |
                             SYNC_WALKER | SYNC_ALIGNER));
        /* Setup various GPU states */
        if (rdev->family == CHIP_RV670)
                WREG32(ARB_GDEC_RD_CNTL, 0x00000021);

        tmp = RREG32(SX_DEBUG_1);
        tmp |= SMX_EVENT_RELEASE;
        if ((rdev->family > CHIP_R600))
                tmp |= ENABLE_NEW_SMX_ADDRESS;
        WREG32(SX_DEBUG_1, tmp);

        if (((rdev->family) == CHIP_R600) ||
            ((rdev->family) == CHIP_RV630) ||
            ((rdev->family) == CHIP_RV610) ||
            ((rdev->family) == CHIP_RV620) ||
            ((rdev->family) == CHIP_RS780) ||
            ((rdev->family) == CHIP_RS880)) {
                WREG32(DB_DEBUG, PREZ_MUST_WAIT_FOR_POSTZ_DONE);
        } else {
                WREG32(DB_DEBUG, 0);
        }
        WREG32(DB_WATERMARKS, (DEPTH_FREE(4) | DEPTH_CACHELINE_FREE(16) |
                               DEPTH_FLUSH(16) | DEPTH_PENDING_FREE(4)));

        WREG32(PA_SC_MULTI_CHIP_CNTL, 0);
        WREG32(VGT_NUM_INSTANCES, 0);

        WREG32(SPI_CONFIG_CNTL, GPR_WRITE_PRIORITY(0));
        WREG32(SPI_CONFIG_CNTL_1, VTX_DONE_DELAY(0));

        tmp = RREG32(SQ_MS_FIFO_SIZES);
        if (((rdev->family) == CHIP_RV610) ||
            ((rdev->family) == CHIP_RV620) ||
            ((rdev->family) == CHIP_RS780) ||
            ((rdev->family) == CHIP_RS880)) {
                tmp = (CACHE_FIFO_SIZE(0xa) |
                       FETCH_FIFO_HIWATER(0xa) |
                       DONE_FIFO_HIWATER(0xe0) |
                       ALU_UPDATE_FIFO_HIWATER(0x8));
        } else if (((rdev->family) == CHIP_R600) ||
                   ((rdev->family) == CHIP_RV630)) {
                tmp &= ~DONE_FIFO_HIWATER(0xff);
                tmp |= DONE_FIFO_HIWATER(0x4);
        }
        WREG32(SQ_MS_FIFO_SIZES, tmp);

        /* SQ_CONFIG, SQ_GPR_RESOURCE_MGMT, SQ_THREAD_RESOURCE_MGMT, SQ_STACK_RESOURCE_MGMT
         * should be adjusted as needed by the 2D/3D drivers.  This just sets default values
         */
        sq_config = RREG32(SQ_CONFIG);
        sq_config &= ~(PS_PRIO(3) |
                       VS_PRIO(3) |
                       GS_PRIO(3) |
                       ES_PRIO(3));
        sq_config |= (DX9_CONSTS |
                      VC_ENABLE |
                      PS_PRIO(0) |
                      VS_PRIO(1) |
                      GS_PRIO(2) |
                      ES_PRIO(3));

        if ((rdev->family) == CHIP_R600) {
                sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(124) |
                                          NUM_VS_GPRS(124) |
                                          NUM_CLAUSE_TEMP_GPRS(4));
                sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(0) |
                                          NUM_ES_GPRS(0));
                sq_thread_resource_mgmt = (NUM_PS_THREADS(136) |
                                           NUM_VS_THREADS(48) |
                                           NUM_GS_THREADS(4) |
                                           NUM_ES_THREADS(4));
                sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(128) |
                                            NUM_VS_STACK_ENTRIES(128));
                sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(0) |
                                            NUM_ES_STACK_ENTRIES(0));
        } else if (((rdev->family) == CHIP_RV610) ||
                   ((rdev->family) == CHIP_RV620) ||
                   ((rdev->family) == CHIP_RS780) ||
                   ((rdev->family) == CHIP_RS880)) {
                /* no vertex cache */
                sq_config &= ~VC_ENABLE;

                sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(44) |
                                          NUM_VS_GPRS(44) |
                                          NUM_CLAUSE_TEMP_GPRS(2));
                sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(17) |
                                          NUM_ES_GPRS(17));
                sq_thread_resource_mgmt = (NUM_PS_THREADS(79) |
                                           NUM_VS_THREADS(78) |
                                           NUM_GS_THREADS(4) |
                                           NUM_ES_THREADS(31));
                sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(40) |
                                            NUM_VS_STACK_ENTRIES(40));
                sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(32) |
                                            NUM_ES_STACK_ENTRIES(16));
        } else if (((rdev->family) == CHIP_RV630) ||
                   ((rdev->family) == CHIP_RV635)) {
                sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(44) |
                                          NUM_VS_GPRS(44) |
                                          NUM_CLAUSE_TEMP_GPRS(2));
                sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(18) |
                                          NUM_ES_GPRS(18));
                sq_thread_resource_mgmt = (NUM_PS_THREADS(79) |
                                           NUM_VS_THREADS(78) |
                                           NUM_GS_THREADS(4) |
                                           NUM_ES_THREADS(31));
                sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(40) |
                                            NUM_VS_STACK_ENTRIES(40));
                sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(32) |
                                            NUM_ES_STACK_ENTRIES(16));
        } else if ((rdev->family) == CHIP_RV670) {
                sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(44) |
                                          NUM_VS_GPRS(44) |
                                          NUM_CLAUSE_TEMP_GPRS(2));
                sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(17) |
                                          NUM_ES_GPRS(17));
                sq_thread_resource_mgmt = (NUM_PS_THREADS(79) |
                                           NUM_VS_THREADS(78) |
                                           NUM_GS_THREADS(4) |
                                           NUM_ES_THREADS(31));
                sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(64) |
                                            NUM_VS_STACK_ENTRIES(64));
                sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(64) |
                                            NUM_ES_STACK_ENTRIES(64));
        }

        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_THREAD_RESOURCE_MGMT, sq_thread_resource_mgmt);
        WREG32(SQ_STACK_RESOURCE_MGMT_1, sq_stack_resource_mgmt_1);
        WREG32(SQ_STACK_RESOURCE_MGMT_2, sq_stack_resource_mgmt_2);

        if (((rdev->family) == CHIP_RV610) ||
            ((rdev->family) == CHIP_RV620) ||
            ((rdev->family) == CHIP_RS780) ||
            ((rdev->family) == CHIP_RS880)) {
                WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(TC_ONLY));
        } else {
                WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(VC_AND_TC));
        }

        /* More default values. 2D/3D driver should adjust as needed */
        WREG32(PA_SC_AA_SAMPLE_LOCS_2S, (S0_X(0xc) | S0_Y(0x4) |
                                         S1_X(0x4) | S1_Y(0xc)));
        WREG32(PA_SC_AA_SAMPLE_LOCS_4S, (S0_X(0xe) | S0_Y(0xe) |
                                         S1_X(0x2) | S1_Y(0x2) |
                                         S2_X(0xa) | S2_Y(0x6) |
                                         S3_X(0x6) | S3_Y(0xa)));
        WREG32(PA_SC_AA_SAMPLE_LOCS_8S_WD0, (S0_X(0xe) | S0_Y(0xb) |
                                             S1_X(0x4) | S1_Y(0xc) |
                                             S2_X(0x1) | S2_Y(0x6) |
                                             S3_X(0xa) | S3_Y(0xe)));
        WREG32(PA_SC_AA_SAMPLE_LOCS_8S_WD1, (S4_X(0x6) | S4_Y(0x1) |
                                             S5_X(0x0) | S5_Y(0x0) |
                                             S6_X(0xb) | S6_Y(0x4) |
                                             S7_X(0x7) | S7_Y(0x8)));

        WREG32(VGT_STRMOUT_EN, 0);
        tmp = rdev->config.r600.max_pipes * 16;
        switch (rdev->family) {
        case CHIP_RV610:
        case CHIP_RV620:
        case CHIP_RS780:
        case CHIP_RS880:
                tmp += 32;
                break;
        case CHIP_RV670:
                tmp += 128;
                break;
        default:
                break;
        }
        if (tmp > 256) {
                tmp = 256;
        }
        WREG32(VGT_ES_PER_GS, 128);
        WREG32(VGT_GS_PER_ES, tmp);
        WREG32(VGT_GS_PER_VS, 2);
        WREG32(VGT_GS_VERTEX_REUSE, 16);

        /* more default values. 2D/3D driver should adjust as needed */
        WREG32(PA_SC_LINE_STIPPLE_STATE, 0);
        WREG32(VGT_STRMOUT_EN, 0);
        WREG32(SX_MISC, 0);
        WREG32(PA_SC_MODE_CNTL, 0);
        WREG32(PA_SC_AA_CONFIG, 0);
        WREG32(PA_SC_LINE_STIPPLE, 0);
        WREG32(SPI_INPUT_Z, 0);
        WREG32(SPI_PS_IN_CONTROL_0, NUM_INTERP(2));
        WREG32(CB_COLOR7_FRAG, 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_COLOR7_FRAG, 0);

        switch (rdev->family) {
        case CHIP_RV610:
        case CHIP_RV620:
        case CHIP_RS780:
        case CHIP_RS880:
                tmp = TC_L2_SIZE(8);
                break;
        case CHIP_RV630:
        case CHIP_RV635:
                tmp = TC_L2_SIZE(4);
                break;
        case CHIP_R600:
                tmp = TC_L2_SIZE(0) | L2_DISABLE_LATE_HIT;
                break;
        default:
                tmp = TC_L2_SIZE(0);
                break;
        }
        WREG32(TC_CNTL, tmp);

        tmp = RREG32(HDP_HOST_PATH_CNTL);
        WREG32(HDP_HOST_PATH_CNTL, tmp);

        tmp = RREG32(ARB_POP);
        tmp |= ENABLE_TC128;
        WREG32(ARB_POP, tmp);

        WREG32(PA_SC_MULTI_CHIP_CNTL, 0);
        WREG32(PA_CL_ENHANCE, (CLIP_VTX_REORDER_ENA |
                               NUM_CLIP_SEQ(3)));
        WREG32(PA_SC_ENHANCE, FORCE_EOV_MAX_CLK_CNT(4095));
        WREG32(VC_ENHANCE, 0);
}


/*
 * Indirect registers accessor
 */
u32 r600_pciep_rreg(struct radeon_device *rdev, u32 reg)
{
        unsigned long flags;
        u32 r;

        spin_lock_irqsave(&rdev->pciep_idx_lock, flags);
        WREG32(PCIE_PORT_INDEX, ((reg) & 0xff));
        (void)RREG32(PCIE_PORT_INDEX);
        r = RREG32(PCIE_PORT_DATA);
        spin_unlock_irqrestore(&rdev->pciep_idx_lock, flags);
        return r;
}

void r600_pciep_wreg(struct radeon_device *rdev, u32 reg, u32 v)
{
        unsigned long flags;

        spin_lock_irqsave(&rdev->pciep_idx_lock, flags);
        WREG32(PCIE_PORT_INDEX, ((reg) & 0xff));
        (void)RREG32(PCIE_PORT_INDEX);
        WREG32(PCIE_PORT_DATA, (v));
        (void)RREG32(PCIE_PORT_DATA);
        spin_unlock_irqrestore(&rdev->pciep_idx_lock, flags);
}

/*
 * CP & Ring
 */
void r600_cp_stop(struct radeon_device *rdev)
{
        if (rdev->asic->copy.copy_ring_index == RADEON_RING_TYPE_GFX_INDEX)
                radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size);
        WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1));
        WREG32(SCRATCH_UMSK, 0);
        rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = false;
}

int r600_init_microcode(struct radeon_device *rdev)
{
        const char *chip_name;
        const char *rlc_chip_name;
        const char *smc_chip_name = "RV770";
        size_t pfp_req_size, me_req_size, rlc_req_size, smc_req_size = 0;
        char fw_name[30];
        int err;

        DRM_DEBUG("\n");

        switch (rdev->family) {
        case CHIP_R600:
                chip_name = "R600";
                rlc_chip_name = "R600";
                break;
        case CHIP_RV610:
                chip_name = "RV610";
                rlc_chip_name = "R600";
                break;
        case CHIP_RV630:
                chip_name = "RV630";
                rlc_chip_name = "R600";
                break;
        case CHIP_RV620:
                chip_name = "RV620";
                rlc_chip_name = "R600";
                break;
        case CHIP_RV635:
                chip_name = "RV635";
                rlc_chip_name = "R600";
                break;
        case CHIP_RV670:
                chip_name = "RV670";
                rlc_chip_name = "R600";
                break;
        case CHIP_RS780:
        case CHIP_RS880:
                chip_name = "RS780";
                rlc_chip_name = "R600";
                break;
        case CHIP_RV770:
                chip_name = "RV770";
                rlc_chip_name = "R700";
                smc_chip_name = "RV770";
                smc_req_size = ALIGN(RV770_SMC_UCODE_SIZE, 4);
                break;
        case CHIP_RV730:
                chip_name = "RV730";
                rlc_chip_name = "R700";
                smc_chip_name = "RV730";
                smc_req_size = ALIGN(RV730_SMC_UCODE_SIZE, 4);
                break;
        case CHIP_RV710:
                chip_name = "RV710";
                rlc_chip_name = "R700";
                smc_chip_name = "RV710";
                smc_req_size = ALIGN(RV710_SMC_UCODE_SIZE, 4);
                break;
        case CHIP_RV740:
                chip_name = "RV730";
                rlc_chip_name = "R700";
                smc_chip_name = "RV740";
                smc_req_size = ALIGN(RV740_SMC_UCODE_SIZE, 4);
                break;
        case CHIP_CEDAR:
                chip_name = "CEDAR";
                rlc_chip_name = "CEDAR";
                smc_chip_name = "CEDAR";
                smc_req_size = ALIGN(CEDAR_SMC_UCODE_SIZE, 4);
                break;
        case CHIP_REDWOOD:
                chip_name = "REDWOOD";
                rlc_chip_name = "REDWOOD";
                smc_chip_name = "REDWOOD";
                smc_req_size = ALIGN(REDWOOD_SMC_UCODE_SIZE, 4);
                break;
        case CHIP_JUNIPER:
                chip_name = "JUNIPER";
                rlc_chip_name = "JUNIPER";
                smc_chip_name = "JUNIPER";
                smc_req_size = ALIGN(JUNIPER_SMC_UCODE_SIZE, 4);
                break;
        case CHIP_CYPRESS:
        case CHIP_HEMLOCK:
                chip_name = "CYPRESS";
                rlc_chip_name = "CYPRESS";
                smc_chip_name = "CYPRESS";
                smc_req_size = ALIGN(CYPRESS_SMC_UCODE_SIZE, 4);
                break;
        case CHIP_PALM:
                chip_name = "PALM";
                rlc_chip_name = "SUMO";
                break;
        case CHIP_SUMO:
                chip_name = "SUMO";
                rlc_chip_name = "SUMO";
                break;
        case CHIP_SUMO2:
                chip_name = "SUMO2";
                rlc_chip_name = "SUMO";
                break;
        default: BUG();
        }

        if (rdev->family >= CHIP_CEDAR) {
                pfp_req_size = EVERGREEN_PFP_UCODE_SIZE * 4;
                me_req_size = EVERGREEN_PM4_UCODE_SIZE * 4;
                rlc_req_size = EVERGREEN_RLC_UCODE_SIZE * 4;
        } else if (rdev->family >= CHIP_RV770) {
                pfp_req_size = R700_PFP_UCODE_SIZE * 4;
                me_req_size = R700_PM4_UCODE_SIZE * 4;
                rlc_req_size = R700_RLC_UCODE_SIZE * 4;
        } else {
                pfp_req_size = R600_PFP_UCODE_SIZE * 4;
                me_req_size = R600_PM4_UCODE_SIZE * 12;
                rlc_req_size = R600_RLC_UCODE_SIZE * 4;
        }

        DRM_INFO("Loading %s Microcode\n", chip_name);

        snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", chip_name);
        err = request_firmware(&rdev->pfp_fw, fw_name, rdev->dev);
        if (err)
                goto out;
        if (rdev->pfp_fw->size != pfp_req_size) {
                pr_err("r600_cp: Bogus length %zu in firmware \"%s\"\n",
                       rdev->pfp_fw->size, fw_name);
                err = -EINVAL;
                goto out;
        }

        snprintf(fw_name, sizeof(fw_name), "radeon/%s_me.bin", chip_name);
        err = request_firmware(&rdev->me_fw, fw_name, rdev->dev);
        if (err)
                goto out;
        if (rdev->me_fw->size != me_req_size) {
                pr_err("r600_cp: Bogus length %zu in firmware \"%s\"\n",
                       rdev->me_fw->size, fw_name);
                err = -EINVAL;
        }

        snprintf(fw_name, sizeof(fw_name), "radeon/%s_rlc.bin", rlc_chip_name);
        err = request_firmware(&rdev->rlc_fw, fw_name, rdev->dev);
        if (err)
                goto out;
        if (rdev->rlc_fw->size != rlc_req_size) {
                pr_err("r600_rlc: Bogus length %zu in firmware \"%s\"\n",
                       rdev->rlc_fw->size, fw_name);
                err = -EINVAL;
        }

        if ((rdev->family >= CHIP_RV770) && (rdev->family <= CHIP_HEMLOCK)) {
                snprintf(fw_name, sizeof(fw_name), "radeon/%s_smc.bin", smc_chip_name);
                err = request_firmware(&rdev->smc_fw, fw_name, rdev->dev);
                if (err) {
                        pr_err("smc: error loading firmware \"%s\"\n", fw_name);
                        release_firmware(rdev->smc_fw);
                        rdev->smc_fw = NULL;
                        err = 0;
                } else if (rdev->smc_fw->size != smc_req_size) {
                        pr_err("smc: Bogus length %zu in firmware \"%s\"\n",
                               rdev->smc_fw->size, fw_name);
                        err = -EINVAL;
                }
        }

out:
        if (err) {
                if (err != -EINVAL)
                        pr_err("r600_cp: Failed to load firmware \"%s\"\n",
                               fw_name);
                release_firmware(rdev->pfp_fw);
                rdev->pfp_fw = NULL;
                release_firmware(rdev->me_fw);
                rdev->me_fw = NULL;
                release_firmware(rdev->rlc_fw);
                rdev->rlc_fw = NULL;
                release_firmware(rdev->smc_fw);
                rdev->smc_fw = NULL;
        }
        return err;
}

u32 r600_gfx_get_rptr(struct radeon_device *rdev,
                      struct radeon_ring *ring)
{
        u32 rptr;

        if (rdev->wb.enabled)
                rptr = rdev->wb.wb[ring->rptr_offs/4];
        else
                rptr = RREG32(R600_CP_RB_RPTR);

        return rptr;
}

u32 r600_gfx_get_wptr(struct radeon_device *rdev,
                      struct radeon_ring *ring)
{
        return RREG32(R600_CP_RB_WPTR);
}

void r600_gfx_set_wptr(struct radeon_device *rdev,
                       struct radeon_ring *ring)
{
        WREG32(R600_CP_RB_WPTR, ring->wptr);
        (void)RREG32(R600_CP_RB_WPTR);
}

static int r600_cp_load_microcode(struct radeon_device *rdev)
{
        const __be32 *fw_data;
        int i;

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

        r600_cp_stop(rdev);

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

        /* Reset cp */
        WREG32(GRBM_SOFT_RESET, SOFT_RESET_CP);
        RREG32(GRBM_SOFT_RESET);
        mdelay(15);
        WREG32(GRBM_SOFT_RESET, 0);

        WREG32(CP_ME_RAM_WADDR, 0);

        fw_data = (const __be32 *)rdev->me_fw->data;
        WREG32(CP_ME_RAM_WADDR, 0);
        for (i = 0; i < R600_PM4_UCODE_SIZE * 3; i++)
                WREG32(CP_ME_RAM_DATA,
                       be32_to_cpup(fw_data++));

        fw_data = (const __be32 *)rdev->pfp_fw->data;
        WREG32(CP_PFP_UCODE_ADDR, 0);
        for (i = 0; i < R600_PFP_UCODE_SIZE; i++)
                WREG32(CP_PFP_UCODE_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;
}

int r600_cp_start(struct radeon_device *rdev)
{
        struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
        int r;
        uint32_t cp_me;

        r = radeon_ring_lock(rdev, ring, 7);
        if (r) {
                DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
                return r;
        }
        radeon_ring_write(ring, PACKET3(PACKET3_ME_INITIALIZE, 5));
        radeon_ring_write(ring, 0x1);
        if (rdev->family >= CHIP_RV770) {
                radeon_ring_write(ring, 0x0);
                radeon_ring_write(ring, rdev->config.rv770.max_hw_contexts - 1);
        } else {
                radeon_ring_write(ring, 0x3);
                radeon_ring_write(ring, rdev->config.r600.max_hw_contexts - 1);
        }
        radeon_ring_write(ring, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
        radeon_ring_write(ring, 0);
        radeon_ring_write(ring, 0);
        radeon_ring_unlock_commit(rdev, ring, false);

        cp_me = 0xff;
        WREG32(R_0086D8_CP_ME_CNTL, cp_me);
        return 0;
}

int r600_cp_resume(struct radeon_device *rdev)
{
        struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
        u32 tmp;
        u32 rb_bufsz;
        int r;

        /* Reset cp */
        WREG32(GRBM_SOFT_RESET, SOFT_RESET_CP);
        RREG32(GRBM_SOFT_RESET);
        mdelay(15);
        WREG32(GRBM_SOFT_RESET, 0);

        /* Set ring buffer size */
        rb_bufsz = order_base_2(ring->ring_size / 8);
        tmp = (order_base_2(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, 0x0);

        /* 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);
        ring->wptr = 0;
        WREG32(CP_RB_WPTR, ring->wptr);

        /* set the wb address whether it's enabled or not */
        WREG32(CP_RB_RPTR_ADDR,
               ((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, ring->gpu_addr >> 8);
        WREG32(CP_DEBUG, (1 << 27) | (1 << 28));

        r600_cp_start(rdev);
        ring->ready = true;
        r = radeon_ring_test(rdev, RADEON_RING_TYPE_GFX_INDEX, ring);
        if (r) {
                ring->ready = false;
                return r;
        }

        if (rdev->asic->copy.copy_ring_index == RADEON_RING_TYPE_GFX_INDEX)
                radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size);

        return 0;
}

void r600_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsigned ring_size)
{
        u32 rb_bufsz;
        int r;

        /* Align ring size */
        rb_bufsz = order_base_2(ring_size / 8);
        ring_size = (1 << (rb_bufsz + 1)) * 4;
        ring->ring_size = ring_size;
        ring->align_mask = 16 - 1;

        if (radeon_ring_supports_scratch_reg(rdev, ring)) {
                r = radeon_scratch_get(rdev, &ring->rptr_save_reg);
                if (r) {
                        DRM_ERROR("failed to get scratch reg for rptr save (%d).\n", r);
                        ring->rptr_save_reg = 0;
                }
        }
}

void r600_cp_fini(struct radeon_device *rdev)
{
        struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
        r600_cp_stop(rdev);
        radeon_ring_fini(rdev, ring);
        radeon_scratch_free(rdev, ring->rptr_save_reg);
}

/*
 * GPU scratch registers helpers function.
 */
void r600_scratch_init(struct radeon_device *rdev)
{
        int i;

        rdev->scratch.num_reg = 7;
        rdev->scratch.reg_base = SCRATCH_REG0;
        for (i = 0; i < rdev->scratch.num_reg; i++) {
                rdev->scratch.free[i] = true;
                rdev->scratch.reg[i] = rdev->scratch.reg_base + (i * 4);
        }
}

int r600_ring_test(struct radeon_device *rdev, struct radeon_ring *ring)
{
        uint32_t scratch;
        uint32_t tmp = 0;
        unsigned i;
        int r;

        r = radeon_scratch_get(rdev, &scratch);
        if (r) {
                DRM_ERROR("radeon: cp failed to get scratch reg (%d).\n", r);
                return r;
        }
        WREG32(scratch, 0xCAFEDEAD);
        r = radeon_ring_lock(rdev, ring, 3);
        if (r) {
                DRM_ERROR("radeon: cp failed to lock ring %d (%d).\n", ring->idx, r);
                radeon_scratch_free(rdev, scratch);
                return r;
        }
        radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
        radeon_ring_write(ring, ((scratch - PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
        radeon_ring_write(ring, 0xDEADBEEF);
        radeon_ring_unlock_commit(rdev, ring, false);
        for (i = 0; i < rdev->usec_timeout; i++) {
                tmp = RREG32(scratch);
                if (tmp == 0xDEADBEEF)
                        break;
                udelay(1);
        }
        if (i < rdev->usec_timeout) {
                DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i);
        } else {
                DRM_ERROR("radeon: ring %d test failed (scratch(0x%04X)=0x%08X)\n",
                          ring->idx, scratch, tmp);
                r = -EINVAL;
        }
        radeon_scratch_free(rdev, scratch);
        return r;
}

/*
 * CP fences/semaphores
 */

void r600_fence_ring_emit(struct radeon_device *rdev,
                          struct radeon_fence *fence)
{
        struct radeon_ring *ring = &rdev->ring[fence->ring];
        u32 cp_coher_cntl = PACKET3_TC_ACTION_ENA | PACKET3_VC_ACTION_ENA |
                PACKET3_SH_ACTION_ENA;

        if (rdev->family >= CHIP_RV770)
                cp_coher_cntl |= PACKET3_FULL_CACHE_ENA;

        if (rdev->wb.use_event) {
                u64 addr = rdev->fence_drv[fence->ring].gpu_addr;
                /* flush read cache over gart */
                radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3));
                radeon_ring_write(ring, cp_coher_cntl);
                radeon_ring_write(ring, 0xFFFFFFFF);
                radeon_ring_write(ring, 0);
                radeon_ring_write(ring, 10); /* poll interval */
                /* EVENT_WRITE_EOP - flush caches, send int */
                radeon_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4));
                radeon_ring_write(ring, EVENT_TYPE(CACHE_FLUSH_AND_INV_EVENT_TS) | EVENT_INDEX(5));
                radeon_ring_write(ring, lower_32_bits(addr));
                radeon_ring_write(ring, (upper_32_bits(addr) & 0xff) | DATA_SEL(1) | INT_SEL(2));
                radeon_ring_write(ring, fence->seq);
                radeon_ring_write(ring, 0);
        } else {
                /* flush read cache over gart */
                radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3));
                radeon_ring_write(ring, cp_coher_cntl);
                radeon_ring_write(ring, 0xFFFFFFFF);
                radeon_ring_write(ring, 0);
                radeon_ring_write(ring, 10); /* poll interval */
                radeon_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE, 0));
                radeon_ring_write(ring, EVENT_TYPE(CACHE_FLUSH_AND_INV_EVENT) | EVENT_INDEX(0));
                /* wait for 3D idle clean */
                radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
                radeon_ring_write(ring, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
                radeon_ring_write(ring, WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit);
                /* Emit fence sequence & fire IRQ */
                radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
                radeon_ring_write(ring, ((rdev->fence_drv[fence->ring].scratch_reg - PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
                radeon_ring_write(ring, fence->seq);
                /* CP_INTERRUPT packet 3 no longer exists, use packet 0 */
                radeon_ring_write(ring, PACKET0(CP_INT_STATUS, 0));
                radeon_ring_write(ring, RB_INT_STAT);
        }
}

/**
 * r600_semaphore_ring_emit - emit a semaphore on the CP ring
 *
 * @rdev: radeon_device pointer
 * @ring: radeon ring buffer object
 * @semaphore: radeon semaphore object
 * @emit_wait: Is this a sempahore wait?
 *
 * Emits a semaphore signal/wait packet to the CP ring and prevents the PFP
 * from running ahead of semaphore waits.
 */
bool r600_semaphore_ring_emit(struct radeon_device *rdev,
                              struct radeon_ring *ring,
                              struct radeon_semaphore *semaphore,
                              bool emit_wait)
{
        uint64_t addr = semaphore->gpu_addr;
        unsigned sel = emit_wait ? PACKET3_SEM_SEL_WAIT : PACKET3_SEM_SEL_SIGNAL;

        if (rdev->family < CHIP_CAYMAN)
                sel |= PACKET3_SEM_WAIT_ON_SIGNAL;

        radeon_ring_write(ring, PACKET3(PACKET3_MEM_SEMAPHORE, 1));
        radeon_ring_write(ring, lower_32_bits(addr));
        radeon_ring_write(ring, (upper_32_bits(addr) & 0xff) | sel);

        /* PFP_SYNC_ME packet only exists on 7xx+, only enable it on eg+ */
        if (emit_wait && (rdev->family >= CHIP_CEDAR)) {
                /* Prevent the PFP from running ahead of the semaphore wait */
                radeon_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
                radeon_ring_write(ring, 0x0);
        }

        return true;
}

/**
 * r600_copy_cpdma - copy pages using the CP DMA engine
 *
 * @rdev: radeon_device pointer
 * @src_offset: src GPU address
 * @dst_offset: dst GPU address
 * @num_gpu_pages: number of GPU pages to xfer
 * @resv: DMA reservation object to manage fences
 *
 * Copy GPU paging using the CP DMA engine (r6xx+).
 * Used by the radeon ttm implementation to move pages if
 * registered as the asic copy callback.
 */
struct radeon_fence *r600_copy_cpdma(struct radeon_device *rdev,
                                     uint64_t src_offset, uint64_t dst_offset,
                                     unsigned num_gpu_pages,
                                     struct dma_resv *resv)
{
        struct radeon_fence *fence;
        struct radeon_sync sync;
        int ring_index = rdev->asic->copy.blit_ring_index;
        struct radeon_ring *ring = &rdev->ring[ring_index];
        u32 size_in_bytes, cur_size_in_bytes, tmp;
        int i, num_loops;
        int r = 0;

        radeon_sync_create(&sync);

        size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT);
        num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff);
        r = radeon_ring_lock(rdev, ring, num_loops * 6 + 24);
        if (r) {
                DRM_ERROR("radeon: moving bo (%d).\n", r);
                radeon_sync_free(rdev, &sync, NULL);
                return ERR_PTR(r);
        }

        radeon_sync_resv(rdev, &sync, resv, false);
        radeon_sync_rings(rdev, &sync, ring->idx);

        radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
        radeon_ring_write(ring, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
        radeon_ring_write(ring, WAIT_3D_IDLE_bit);
        for (i = 0; i < num_loops; i++) {
                cur_size_in_bytes = size_in_bytes;
                if (cur_size_in_bytes > 0x1fffff)
                        cur_size_in_bytes = 0x1fffff;
                size_in_bytes -= cur_size_in_bytes;
                tmp = upper_32_bits(src_offset) & 0xff;
                if (size_in_bytes == 0)
                        tmp |= PACKET3_CP_DMA_CP_SYNC;