/*
 * Copyright © 2012 Intel Corporation
 *
 * 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 (including the next
 * paragraph) 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 AUTHORS OR COPYRIGHT HOLDERS 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:
 *    Eugeni Dodonov <eugeni.dodonov@intel.com>
 *
 */

#include <linux/cpufreq.h>
#include "i915_drv.h"
#include "intel_drv.h"
#include "../../../platform/x86/intel_ips.h"
#include <linux/module.h>

/**
 * RC6 is a special power stage which allows the GPU to enter an very
 * low-voltage mode when idle, using down to 0V while at this stage.  This
 * stage is entered automatically when the GPU is idle when RC6 support is
 * enabled, and as soon as new workload arises GPU wakes up automatically as well.
 *
 * There are different RC6 modes available in Intel GPU, which differentiate
 * among each other with the latency required to enter and leave RC6 and
 * voltage consumed by the GPU in different states.
 *
 * The combination of the following flags define which states GPU is allowed
 * to enter, while RC6 is the normal RC6 state, RC6p is the deep RC6, and
 * RC6pp is deepest RC6. Their support by hardware varies according to the
 * GPU, BIOS, chipset and platform. RC6 is usually the safest one and the one
 * which brings the most power savings; deeper states save more power, but
 * require higher latency to switch to and wake up.
 */
#define INTEL_RC6_ENABLE                        (1<<0)
#define INTEL_RC6p_ENABLE                       (1<<1)
#define INTEL_RC6pp_ENABLE                      (1<<2)

static void bxt_init_clock_gating(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;

        /* WaDisableSDEUnitClockGating:bxt */
        I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) |
                   GEN8_SDEUNIT_CLOCK_GATE_DISABLE);

        /*
         * FIXME:
         * GEN8_HDCUNIT_CLOCK_GATE_DISABLE_HDCREQ applies on 3x6 GT SKUs only.
         */
        I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) |
                   GEN8_HDCUNIT_CLOCK_GATE_DISABLE_HDCREQ);
}

static void i915_pineview_get_mem_freq(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        u32 tmp;

        tmp = I915_READ(CLKCFG);

        switch (tmp & CLKCFG_FSB_MASK) {
        case CLKCFG_FSB_533:
                dev_priv->fsb_freq = 533; /* 133*4 */
                break;
        case CLKCFG_FSB_800:
                dev_priv->fsb_freq = 800; /* 200*4 */
                break;
        case CLKCFG_FSB_667:
                dev_priv->fsb_freq =  667; /* 167*4 */
                break;
        case CLKCFG_FSB_400:
                dev_priv->fsb_freq = 400; /* 100*4 */
                break;
        }

        switch (tmp & CLKCFG_MEM_MASK) {
        case CLKCFG_MEM_533:
                dev_priv->mem_freq = 533;
                break;
        case CLKCFG_MEM_667:
                dev_priv->mem_freq = 667;
                break;
        case CLKCFG_MEM_800:
                dev_priv->mem_freq = 800;
                break;
        }

        /* detect pineview DDR3 setting */
        tmp = I915_READ(CSHRDDR3CTL);
        dev_priv->is_ddr3 = (tmp & CSHRDDR3CTL_DDR3) ? 1 : 0;
}

static void i915_ironlake_get_mem_freq(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        u16 ddrpll, csipll;

        ddrpll = I915_READ16(DDRMPLL1);
        csipll = I915_READ16(CSIPLL0);

        switch (ddrpll & 0xff) {
        case 0xc:
                dev_priv->mem_freq = 800;
                break;
        case 0x10:
                dev_priv->mem_freq = 1066;
                break;
        case 0x14:
                dev_priv->mem_freq = 1333;
                break;
        case 0x18:
                dev_priv->mem_freq = 1600;
                break;
        default:
                DRM_DEBUG_DRIVER("unknown memory frequency 0x%02x\n",
                                 ddrpll & 0xff);
                dev_priv->mem_freq = 0;
                break;
        }

        dev_priv->ips.r_t = dev_priv->mem_freq;

        switch (csipll & 0x3ff) {
        case 0x00c:
                dev_priv->fsb_freq = 3200;
                break;
        case 0x00e:
                dev_priv->fsb_freq = 3733;
                break;
        case 0x010:
                dev_priv->fsb_freq = 4266;
                break;
        case 0x012:
                dev_priv->fsb_freq = 4800;
                break;
        case 0x014:
                dev_priv->fsb_freq = 5333;
                break;
        case 0x016:
                dev_priv->fsb_freq = 5866;
                break;
        case 0x018:
                dev_priv->fsb_freq = 6400;
                break;
        default:
                DRM_DEBUG_DRIVER("unknown fsb frequency 0x%04x\n",
                                 csipll & 0x3ff);
                dev_priv->fsb_freq = 0;
                break;
        }

        if (dev_priv->fsb_freq == 3200) {
                dev_priv->ips.c_m = 0;
        } else if (dev_priv->fsb_freq > 3200 && dev_priv->fsb_freq <= 4800) {
                dev_priv->ips.c_m = 1;
        } else {
                dev_priv->ips.c_m = 2;
        }
}

static const struct cxsr_latency cxsr_latency_table[] = {
        {1, 0, 800, 400, 3382, 33382, 3983, 33983},    /* DDR2-400 SC */
        {1, 0, 800, 667, 3354, 33354, 3807, 33807},    /* DDR2-667 SC */
        {1, 0, 800, 800, 3347, 33347, 3763, 33763},    /* DDR2-800 SC */
        {1, 1, 800, 667, 6420, 36420, 6873, 36873},    /* DDR3-667 SC */
        {1, 1, 800, 800, 5902, 35902, 6318, 36318},    /* DDR3-800 SC */

        {1, 0, 667, 400, 3400, 33400, 4021, 34021},    /* DDR2-400 SC */
        {1, 0, 667, 667, 3372, 33372, 3845, 33845},    /* DDR2-667 SC */
        {1, 0, 667, 800, 3386, 33386, 3822, 33822},    /* DDR2-800 SC */
        {1, 1, 667, 667, 6438, 36438, 6911, 36911},    /* DDR3-667 SC */
        {1, 1, 667, 800, 5941, 35941, 6377, 36377},    /* DDR3-800 SC */

        {1, 0, 400, 400, 3472, 33472, 4173, 34173},    /* DDR2-400 SC */
        {1, 0, 400, 667, 3443, 33443, 3996, 33996},    /* DDR2-667 SC */
        {1, 0, 400, 800, 3430, 33430, 3946, 33946},    /* DDR2-800 SC */
        {1, 1, 400, 667, 6509, 36509, 7062, 37062},    /* DDR3-667 SC */
        {1, 1, 400, 800, 5985, 35985, 6501, 36501},    /* DDR3-800 SC */

        {0, 0, 800, 400, 3438, 33438, 4065, 34065},    /* DDR2-400 SC */
        {0, 0, 800, 667, 3410, 33410, 3889, 33889},    /* DDR2-667 SC */
        {0, 0, 800, 800, 3403, 33403, 3845, 33845},    /* DDR2-800 SC */
        {0, 1, 800, 667, 6476, 36476, 6955, 36955},    /* DDR3-667 SC */
        {0, 1, 800, 800, 5958, 35958, 6400, 36400},    /* DDR3-800 SC */

        {0, 0, 667, 400, 3456, 33456, 4103, 34106},    /* DDR2-400 SC */
        {0, 0, 667, 667, 3428, 33428, 3927, 33927},    /* DDR2-667 SC */
        {0, 0, 667, 800, 3443, 33443, 3905, 33905},    /* DDR2-800 SC */
        {0, 1, 667, 667, 6494, 36494, 6993, 36993},    /* DDR3-667 SC */
        {0, 1, 667, 800, 5998, 35998, 6460, 36460},    /* DDR3-800 SC */

        {0, 0, 400, 400, 3528, 33528, 4255, 34255},    /* DDR2-400 SC */
        {0, 0, 400, 667, 3500, 33500, 4079, 34079},    /* DDR2-667 SC */
        {0, 0, 400, 800, 3487, 33487, 4029, 34029},    /* DDR2-800 SC */
        {0, 1, 400, 667, 6566, 36566, 7145, 37145},    /* DDR3-667 SC */
        {0, 1, 400, 800, 6042, 36042, 6584, 36584},    /* DDR3-800 SC */
};

static const struct cxsr_latency *intel_get_cxsr_latency(int is_desktop,
                                                         int is_ddr3,
                                                         int fsb,
                                                         int mem)
{
        const struct cxsr_latency *latency;
        int i;

        if (fsb == 0 || mem == 0)
                return NULL;

        for (i = 0; i < ARRAY_SIZE(cxsr_latency_table); i++) {
                latency = &cxsr_latency_table[i];
                if (is_desktop == latency->is_desktop &&
                    is_ddr3 == latency->is_ddr3 &&
                    fsb == latency->fsb_freq && mem == latency->mem_freq)
                        return latency;
        }

        DRM_DEBUG_KMS("Unknown FSB/MEM found, disable CxSR\n");

        return NULL;
}

static void chv_set_memory_dvfs(struct drm_i915_private *dev_priv, bool enable)
{
        u32 val;

        mutex_lock(&dev_priv->rps.hw_lock);

        val = vlv_punit_read(dev_priv, PUNIT_REG_DDR_SETUP2);
        if (enable)
                val &= ~FORCE_DDR_HIGH_FREQ;
        else
                val |= FORCE_DDR_HIGH_FREQ;
        val &= ~FORCE_DDR_LOW_FREQ;
        val |= FORCE_DDR_FREQ_REQ_ACK;
        vlv_punit_write(dev_priv, PUNIT_REG_DDR_SETUP2, val);

        if (wait_for((vlv_punit_read(dev_priv, PUNIT_REG_DDR_SETUP2) &
                      FORCE_DDR_FREQ_REQ_ACK) == 0, 3))
                DRM_ERROR("timed out waiting for Punit DDR DVFS request\n");

        mutex_unlock(&dev_priv->rps.hw_lock);
}

static void chv_set_memory_pm5(struct drm_i915_private *dev_priv, bool enable)
{
        u32 val;

        mutex_lock(&dev_priv->rps.hw_lock);

        val = vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ);
        if (enable)
                val |= DSP_MAXFIFO_PM5_ENABLE;
        else
                val &= ~DSP_MAXFIFO_PM5_ENABLE;
        vlv_punit_write(dev_priv, PUNIT_REG_DSPFREQ, val);

        mutex_unlock(&dev_priv->rps.hw_lock);
}

#define FW_WM(value, plane) \
        (((value) << DSPFW_ ## plane ## _SHIFT) & DSPFW_ ## plane ## _MASK)

void intel_set_memory_cxsr(struct drm_i915_private *dev_priv, bool enable)
{
        struct drm_device *dev = dev_priv->dev;
        u32 val;

        if (IS_VALLEYVIEW(dev)) {
                I915_WRITE(FW_BLC_SELF_VLV, enable ? FW_CSPWRDWNEN : 0);
                POSTING_READ(FW_BLC_SELF_VLV);
                dev_priv->wm.vlv.cxsr = enable;
        } else if (IS_G4X(dev) || IS_CRESTLINE(dev)) {
                I915_WRITE(FW_BLC_SELF, enable ? FW_BLC_SELF_EN : 0);
                POSTING_READ(FW_BLC_SELF);
        } else if (IS_PINEVIEW(dev)) {
                val = I915_READ(DSPFW3) & ~PINEVIEW_SELF_REFRESH_EN;
                val |= enable ? PINEVIEW_SELF_REFRESH_EN : 0;
                I915_WRITE(DSPFW3, val);
                POSTING_READ(DSPFW3);
        } else if (IS_I945G(dev) || IS_I945GM(dev)) {
                val = enable ? _MASKED_BIT_ENABLE(FW_BLC_SELF_EN) :
                               _MASKED_BIT_DISABLE(FW_BLC_SELF_EN);
                I915_WRITE(FW_BLC_SELF, val);
                POSTING_READ(FW_BLC_SELF);
        } else if (IS_I915GM(dev)) {
                val = enable ? _MASKED_BIT_ENABLE(INSTPM_SELF_EN) :
                               _MASKED_BIT_DISABLE(INSTPM_SELF_EN);
                I915_WRITE(INSTPM, val);
                POSTING_READ(INSTPM);
        } else {
                return;
        }

        DRM_DEBUG_KMS("memory self-refresh is %s\n",
                      enable ? "enabled" : "disabled");
}


/*
 * Latency for FIFO fetches is dependent on several factors:
 *   - memory configuration (speed, channels)
 *   - chipset
 *   - current MCH state
 * It can be fairly high in some situations, so here we assume a fairly
 * pessimal value.  It's a tradeoff between extra memory fetches (if we
 * set this value too high, the FIFO will fetch frequently to stay full)
 * and power consumption (set it too low to save power and we might see
 * FIFO underruns and display "flicker").
 *
 * A value of 5us seems to be a good balance; safe for very low end
 * platforms but not overly aggressive on lower latency configs.
 */
static const int pessimal_latency_ns = 5000;

#define VLV_FIFO_START(dsparb, dsparb2, lo_shift, hi_shift) \
        ((((dsparb) >> (lo_shift)) & 0xff) | ((((dsparb2) >> (hi_shift)) & 0x1) << 8))

static int vlv_get_fifo_size(struct drm_device *dev,
                              enum pipe pipe, int plane)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        int sprite0_start, sprite1_start, size;

        switch (pipe) {
                uint32_t dsparb, dsparb2, dsparb3;
        case PIPE_A:
                dsparb = I915_READ(DSPARB);
                dsparb2 = I915_READ(DSPARB2);
                sprite0_start = VLV_FIFO_START(dsparb, dsparb2, 0, 0);
                sprite1_start = VLV_FIFO_START(dsparb, dsparb2, 8, 4);
                break;
        case PIPE_B:
                dsparb = I915_READ(DSPARB);
                dsparb2 = I915_READ(DSPARB2);
                sprite0_start = VLV_FIFO_START(dsparb, dsparb2, 16, 8);
                sprite1_start = VLV_FIFO_START(dsparb, dsparb2, 24, 12);
                break;
        case PIPE_C:
                dsparb2 = I915_READ(DSPARB2);
                dsparb3 = I915_READ(DSPARB3);
                sprite0_start = VLV_FIFO_START(dsparb3, dsparb2, 0, 16);
                sprite1_start = VLV_FIFO_START(dsparb3, dsparb2, 8, 20);
                break;
        default:
                return 0;
        }

        switch (plane) {
        case 0:
                size = sprite0_start;
                break;
        case 1:
                size = sprite1_start - sprite0_start;
                break;
        case 2:
                size = 512 - 1 - sprite1_start;
                break;
        default:
                return 0;
        }

        DRM_DEBUG_KMS("Pipe %c %s %c FIFO size: %d\n",
                      pipe_name(pipe), plane == 0 ? "primary" : "sprite",
                      plane == 0 ? plane_name(pipe) : sprite_name(pipe, plane - 1),
                      size);

        return size;
}

static int i9xx_get_fifo_size(struct drm_device *dev, int plane)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        uint32_t dsparb = I915_READ(DSPARB);
        int size;

        size = dsparb & 0x7f;
        if (plane)
                size = ((dsparb >> DSPARB_CSTART_SHIFT) & 0x7f) - size;

        DRM_DEBUG_KMS("FIFO size - (0x%08x) %s: %d\n", dsparb,
                      plane ? "B" : "A", size);

        return size;
}

static int i830_get_fifo_size(struct drm_device *dev, int plane)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        uint32_t dsparb = I915_READ(DSPARB);
        int size;

        size = dsparb & 0x1ff;
        if (plane)
                size = ((dsparb >> DSPARB_BEND_SHIFT) & 0x1ff) - size;
        size >>= 1; /* Convert to cachelines */

        DRM_DEBUG_KMS("FIFO size - (0x%08x) %s: %d\n", dsparb,
                      plane ? "B" : "A", size);

        return size;
}

static int i845_get_fifo_size(struct drm_device *dev, int plane)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        uint32_t dsparb = I915_READ(DSPARB);
        int size;

        size = dsparb & 0x7f;
        size >>= 2; /* Convert to cachelines */

        DRM_DEBUG_KMS("FIFO size - (0x%08x) %s: %d\n", dsparb,
                      plane ? "B" : "A",
                      size);

        return size;
}

/* Pineview has different values for various configs */
static const struct intel_watermark_params pineview_display_wm = {
        .fifo_size = PINEVIEW_DISPLAY_FIFO,
        .max_wm = PINEVIEW_MAX_WM,
        .default_wm = PINEVIEW_DFT_WM,
        .guard_size = PINEVIEW_GUARD_WM,
        .cacheline_size = PINEVIEW_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params pineview_display_hplloff_wm = {
        .fifo_size = PINEVIEW_DISPLAY_FIFO,
        .max_wm = PINEVIEW_MAX_WM,
        .default_wm = PINEVIEW_DFT_HPLLOFF_WM,
        .guard_size = PINEVIEW_GUARD_WM,
        .cacheline_size = PINEVIEW_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params pineview_cursor_wm = {
        .fifo_size = PINEVIEW_CURSOR_FIFO,
        .max_wm = PINEVIEW_CURSOR_MAX_WM,
        .default_wm = PINEVIEW_CURSOR_DFT_WM,
        .guard_size = PINEVIEW_CURSOR_GUARD_WM,
        .cacheline_size = PINEVIEW_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params pineview_cursor_hplloff_wm = {
        .fifo_size = PINEVIEW_CURSOR_FIFO,
        .max_wm = PINEVIEW_CURSOR_MAX_WM,
        .default_wm = PINEVIEW_CURSOR_DFT_WM,
        .guard_size = PINEVIEW_CURSOR_GUARD_WM,
        .cacheline_size = PINEVIEW_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params g4x_wm_info = {
        .fifo_size = G4X_FIFO_SIZE,
        .max_wm = G4X_MAX_WM,
        .default_wm = G4X_MAX_WM,
        .guard_size = 2,
        .cacheline_size = G4X_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params g4x_cursor_wm_info = {
        .fifo_size = I965_CURSOR_FIFO,
        .max_wm = I965_CURSOR_MAX_WM,
        .default_wm = I965_CURSOR_DFT_WM,
        .guard_size = 2,
        .cacheline_size = G4X_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params valleyview_wm_info = {
        .fifo_size = VALLEYVIEW_FIFO_SIZE,
        .max_wm = VALLEYVIEW_MAX_WM,
        .default_wm = VALLEYVIEW_MAX_WM,
        .guard_size = 2,
        .cacheline_size = G4X_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params valleyview_cursor_wm_info = {
        .fifo_size = I965_CURSOR_FIFO,
        .max_wm = VALLEYVIEW_CURSOR_MAX_WM,
        .default_wm = I965_CURSOR_DFT_WM,
        .guard_size = 2,
        .cacheline_size = G4X_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params i965_cursor_wm_info = {
        .fifo_size = I965_CURSOR_FIFO,
        .max_wm = I965_CURSOR_MAX_WM,
        .default_wm = I965_CURSOR_DFT_WM,
        .guard_size = 2,
        .cacheline_size = I915_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params i945_wm_info = {
        .fifo_size = I945_FIFO_SIZE,
        .max_wm = I915_MAX_WM,
        .default_wm = 1,
        .guard_size = 2,
        .cacheline_size = I915_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params i915_wm_info = {
        .fifo_size = I915_FIFO_SIZE,
        .max_wm = I915_MAX_WM,
        .default_wm = 1,
        .guard_size = 2,
        .cacheline_size = I915_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params i830_a_wm_info = {
        .fifo_size = I855GM_FIFO_SIZE,
        .max_wm = I915_MAX_WM,
        .default_wm = 1,
        .guard_size = 2,
        .cacheline_size = I830_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params i830_bc_wm_info = {
        .fifo_size = I855GM_FIFO_SIZE,
        .max_wm = I915_MAX_WM/2,
        .default_wm = 1,
        .guard_size = 2,
        .cacheline_size = I830_FIFO_LINE_SIZE,
};
static const struct intel_watermark_params i845_wm_info = {
        .fifo_size = I830_FIFO_SIZE,
        .max_wm = I915_MAX_WM,
        .default_wm = 1,
        .guard_size = 2,
        .cacheline_size = I830_FIFO_LINE_SIZE,
};

/**
 * intel_calculate_wm - calculate watermark level
 * @clock_in_khz: pixel clock
 * @wm: chip FIFO params
 * @pixel_size: display pixel size
 * @latency_ns: memory latency for the platform
 *
 * Calculate the watermark level (the level at which the display plane will
 * start fetching from memory again).  Each chip has a different display
 * FIFO size and allocation, so the caller needs to figure that out and pass
 * in the correct intel_watermark_params structure.
 *
 * As the pixel clock runs, the FIFO will be drained at a rate that depends
 * on the pixel size.  When it reaches the watermark level, it'll start
 * fetching FIFO line sized based chunks from memory until the FIFO fills
 * past the watermark point.  If the FIFO drains completely, a FIFO underrun
 * will occur, and a display engine hang could result.
 */
static unsigned long intel_calculate_wm(unsigned long clock_in_khz,
                                        const struct intel_watermark_params *wm,
                                        int fifo_size,
                                        int pixel_size,
                                        unsigned long latency_ns)
{
        long entries_required, wm_size;

        /*
         * Note: we need to make sure we don't overflow for various clock &
         * latency values.
         * clocks go from a few thousand to several hundred thousand.
         * latency is usually a few thousand
         */
        entries_required = ((clock_in_khz / 1000) * pixel_size * latency_ns) /
                1000;
        entries_required = DIV_ROUND_UP(entries_required, wm->cacheline_size);

        DRM_DEBUG_KMS("FIFO entries required for mode: %ld\n", entries_required);

        wm_size = fifo_size - (entries_required + wm->guard_size);

        DRM_DEBUG_KMS("FIFO watermark level: %ld\n", wm_size);

        /* Don't promote wm_size to unsigned... */
        if (wm_size > (long)wm->max_wm)
                wm_size = wm->max_wm;
        if (wm_size <= 0)
                wm_size = wm->default_wm;

        /*
         * Bspec seems to indicate that the value shouldn't be lower than
         * 'burst size + 1'. Certainly 830 is quite unhappy with low values.
         * Lets go for 8 which is the burst size since certain platforms
         * already use a hardcoded 8 (which is what the spec says should be
         * done).
         */
        if (wm_size <= 8)
                wm_size = 8;

        return wm_size;
}

static struct drm_crtc *single_enabled_crtc(struct drm_device *dev)
{
        struct drm_crtc *crtc, *enabled = NULL;

        for_each_crtc(dev, crtc) {
                if (intel_crtc_active(crtc)) {
                        if (enabled)
                                return NULL;
                        enabled = crtc;
                }
        }

        return enabled;
}

static void pineview_update_wm(struct drm_crtc *unused_crtc)
{
        struct drm_device *dev = unused_crtc->dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_crtc *crtc;
        const struct cxsr_latency *latency;
        u32 reg;
        unsigned long wm;

        latency = intel_get_cxsr_latency(IS_PINEVIEW_G(dev), dev_priv->is_ddr3,
                                         dev_priv->fsb_freq, dev_priv->mem_freq);
        if (!latency) {
                DRM_DEBUG_KMS("Unknown FSB/MEM found, disable CxSR\n");
                intel_set_memory_cxsr(dev_priv, false);
                return;
        }

        crtc = single_enabled_crtc(dev);
        if (crtc) {
                const struct drm_display_mode *adjusted_mode = &to_intel_crtc(crtc)->config->base.adjusted_mode;
                int pixel_size = crtc->primary->state->fb->bits_per_pixel / 8;
                int clock = adjusted_mode->crtc_clock;

                /* Display SR */
                wm = intel_calculate_wm(clock, &pineview_display_wm,
                                        pineview_display_wm.fifo_size,
                                        pixel_size, latency->display_sr);
                reg = I915_READ(DSPFW1);
                reg &= ~DSPFW_SR_MASK;
                reg |= FW_WM(wm, SR);
                I915_WRITE(DSPFW1, reg);
                DRM_DEBUG_KMS("DSPFW1 register is %x\n", reg);

                /* cursor SR */
                wm = intel_calculate_wm(clock, &pineview_cursor_wm,
                                        pineview_display_wm.fifo_size,
                                        pixel_size, latency->cursor_sr);
                reg = I915_READ(DSPFW3);
                reg &= ~DSPFW_CURSOR_SR_MASK;
                reg |= FW_WM(wm, CURSOR_SR);
                I915_WRITE(DSPFW3, reg);

                /* Display HPLL off SR */
                wm = intel_calculate_wm(clock, &pineview_display_hplloff_wm,
                                        pineview_display_hplloff_wm.fifo_size,
                                        pixel_size, latency->display_hpll_disable);
                reg = I915_READ(DSPFW3);
                reg &= ~DSPFW_HPLL_SR_MASK;
                reg |= FW_WM(wm, HPLL_SR);
                I915_WRITE(DSPFW3, reg);

                /* cursor HPLL off SR */
                wm = intel_calculate_wm(clock, &pineview_cursor_hplloff_wm,
                                        pineview_display_hplloff_wm.fifo_size,
                                        pixel_size, latency->cursor_hpll_disable);
                reg = I915_READ(DSPFW3);
                reg &= ~DSPFW_HPLL_CURSOR_MASK;
                reg |= FW_WM(wm, HPLL_CURSOR);
                I915_WRITE(DSPFW3, reg);
                DRM_DEBUG_KMS("DSPFW3 register is %x\n", reg);

                intel_set_memory_cxsr(dev_priv, true);
        } else {
                intel_set_memory_cxsr(dev_priv, false);
        }
}

static bool g4x_compute_wm0(struct drm_device *dev,
                            int plane,
                            const struct intel_watermark_params *display,
                            int display_latency_ns,
                            const struct intel_watermark_params *cursor,
                            int cursor_latency_ns,
                            int *plane_wm,
                            int *cursor_wm)
{
        struct drm_crtc *crtc;
        const struct drm_display_mode *adjusted_mode;
        int htotal, hdisplay, clock, pixel_size;
        int line_time_us, line_count;
        int entries, tlb_miss;

        crtc = intel_get_crtc_for_plane(dev, plane);
        if (!intel_crtc_active(crtc)) {
                *cursor_wm = cursor->guard_size;
                *plane_wm = display->guard_size;
                return false;
        }

        adjusted_mode = &to_intel_crtc(crtc)->config->base.adjusted_mode;
        clock = adjusted_mode->crtc_clock;
        htotal = adjusted_mode->crtc_htotal;
        hdisplay = to_intel_crtc(crtc)->config->pipe_src_w;
        pixel_size = crtc->primary->state->fb->bits_per_pixel / 8;

        /* Use the small buffer method to calculate plane watermark */
        entries = ((clock * pixel_size / 1000) * display_latency_ns) / 1000;
        tlb_miss = display->fifo_size*display->cacheline_size - hdisplay * 8;
        if (tlb_miss > 0)
                entries += tlb_miss;
        entries = DIV_ROUND_UP(entries, display->cacheline_size);
        *plane_wm = entries + display->guard_size;
        if (*plane_wm > (int)display->max_wm)
                *plane_wm = display->max_wm;

        /* Use the large buffer method to calculate cursor watermark */
        line_time_us = max(htotal * 1000 / clock, 1);
        line_count = (cursor_latency_ns / line_time_us + 1000) / 1000;
        entries = line_count * crtc->cursor->state->crtc_w * pixel_size;
        tlb_miss = cursor->fifo_size*cursor->cacheline_size - hdisplay * 8;
        if (tlb_miss > 0)
                entries += tlb_miss;
        entries = DIV_ROUND_UP(entries, cursor->cacheline_size);
        *cursor_wm = entries + cursor->guard_size;
        if (*cursor_wm > (int)cursor->max_wm)
                *cursor_wm = (int)cursor->max_wm;

        return true;
}

/*
 * Check the wm result.
 *
 * If any calculated watermark values is larger than the maximum value that
 * can be programmed into the associated watermark register, that watermark
 * must be disabled.
 */
static bool g4x_check_srwm(struct drm_device *dev,
                           int display_wm, int cursor_wm,
                           const struct intel_watermark_params *display,
                           const struct intel_watermark_params *cursor)
{
        DRM_DEBUG_KMS("SR watermark: display plane %d, cursor %d\n",
                      display_wm, cursor_wm);

        if (display_wm > display->max_wm) {
                DRM_DEBUG_KMS("display watermark is too large(%d/%ld), disabling\n",
                              display_wm, display->max_wm);
                return false;
        }

        if (cursor_wm > cursor->max_wm) {
                DRM_DEBUG_KMS("cursor watermark is too large(%d/%ld), disabling\n",
                              cursor_wm, cursor->max_wm);
                return false;
        }

        if (!(display_wm || cursor_wm)) {
                DRM_DEBUG_KMS("SR latency is 0, disabling\n");
                return false;
        }

        return true;
}

static bool g4x_compute_srwm(struct drm_device *dev,
                             int plane,
                             int latency_ns,
                             const struct intel_watermark_params *display,
                             const struct intel_watermark_params *cursor,
                             int *display_wm, int *cursor_wm)
{
        struct drm_crtc *crtc;
        const struct drm_display_mode *adjusted_mode;
        int hdisplay, htotal, pixel_size, clock;
        unsigned long line_time_us;
        int line_count, line_size;
        int small, large;
        int entries;

        if (!latency_ns) {
                *display_wm = *cursor_wm = 0;
                return false;
        }

        crtc = intel_get_crtc_for_plane(dev, plane);
        adjusted_mode = &to_intel_crtc(crtc)->config->base.adjusted_mode;
        clock = adjusted_mode->crtc_clock;
        htotal = adjusted_mode->crtc_htotal;
        hdisplay = to_intel_crtc(crtc)->config->pipe_src_w;
        pixel_size = crtc->primary->state->fb->bits_per_pixel / 8;

        line_time_us = max(htotal * 1000 / clock, 1);
        line_count = (latency_ns / line_time_us + 1000) / 1000;
        line_size = hdisplay * pixel_size;

        /* Use the minimum of the small and large buffer method for primary */
        small = ((clock * pixel_size / 1000) * latency_ns) / 1000;
        large = line_count * line_size;

        entries = DIV_ROUND_UP(min(small, large), display->cacheline_size);
        *display_wm = entries + display->guard_size;

        /* calculate the self-refresh watermark for display cursor */
        entries = line_count * pixel_size * crtc->cursor->state->crtc_w;
        entries = DIV_ROUND_UP(entries, cursor->cacheline_size);
        *cursor_wm = entries + cursor->guard_size;

        return g4x_check_srwm(dev,
                              *display_wm, *cursor_wm,
                              display, cursor);
}

#define FW_WM_VLV(value, plane) \
        (((value) << DSPFW_ ## plane ## _SHIFT) & DSPFW_ ## plane ## _MASK_VLV)

static void vlv_write_wm_values(struct intel_crtc *crtc,
                                const struct vlv_wm_values *wm)
{
        struct drm_i915_private *dev_priv = to_i915(crtc->base.dev);
        enum pipe pipe = crtc->pipe;

        I915_WRITE(VLV_DDL(pipe),
                   (wm->ddl[pipe].cursor << DDL_CURSOR_SHIFT) |
                   (wm->ddl[pipe].sprite[1] << DDL_SPRITE_SHIFT(1)) |
                   (wm->ddl[pipe].sprite[0] << DDL_SPRITE_SHIFT(0)) |
                   (wm->ddl[pipe].primary << DDL_PLANE_SHIFT));

        I915_WRITE(DSPFW1,
                   FW_WM(wm->sr.plane, SR) |
                   FW_WM(wm->pipe[PIPE_B].cursor, CURSORB) |
                   FW_WM_VLV(wm->pipe[PIPE_B].primary, PLANEB) |
                   FW_WM_VLV(wm->pipe[PIPE_A].primary, PLANEA));
        I915_WRITE(DSPFW2,
                   FW_WM_VLV(wm->pipe[PIPE_A].sprite[1], SPRITEB) |
                   FW_WM(wm->pipe[PIPE_A].cursor, CURSORA) |
                   FW_WM_VLV(wm->pipe[PIPE_A].sprite[0], SPRITEA));
        I915_WRITE(DSPFW3,
                   FW_WM(wm->sr.cursor, CURSOR_SR));

        if (IS_CHERRYVIEW(dev_priv)) {
                I915_WRITE(DSPFW7_CHV,
                           FW_WM_VLV(wm->pipe[PIPE_B].sprite[1], SPRITED) |
                           FW_WM_VLV(wm->pipe[PIPE_B].sprite[0], SPRITEC));
                I915_WRITE(DSPFW8_CHV,
                           FW_WM_VLV(wm->pipe[PIPE_C].sprite[1], SPRITEF) |
                           FW_WM_VLV(wm->pipe[PIPE_C].sprite[0], SPRITEE));
                I915_WRITE(DSPFW9_CHV,
                           FW_WM_VLV(wm->pipe[PIPE_C].primary, PLANEC) |
                           FW_WM(wm->pipe[PIPE_C].cursor, CURSORC));
                I915_WRITE(DSPHOWM,
                           FW_WM(wm->sr.plane >> 9, SR_HI) |
                           FW_WM(wm->pipe[PIPE_C].sprite[1] >> 8, SPRITEF_HI) |
                           FW_WM(wm->pipe[PIPE_C].sprite[0] >> 8, SPRITEE_HI) |
                           FW_WM(wm->pipe[PIPE_C].primary >> 8, PLANEC_HI) |
                           FW_WM(wm->pipe[PIPE_B].sprite[1] >> 8, SPRITED_HI) |
                           FW_WM(wm->pipe[PIPE_B].sprite[0] >> 8, SPRITEC_HI) |
                           FW_WM(wm->pipe[PIPE_B].primary >> 8, PLANEB_HI) |
                           FW_WM(wm->pipe[PIPE_A].sprite[1] >> 8, SPRITEB_HI) |
                           FW_WM(wm->pipe[PIPE_A].sprite[0] >> 8, SPRITEA_HI) |
                           FW_WM(wm->pipe[PIPE_A].primary >> 8, PLANEA_HI));
        } else {
                I915_WRITE(DSPFW7,
                           FW_WM_VLV(wm->pipe[PIPE_B].sprite[1], SPRITED) |
                           FW_WM_VLV(wm->pipe[PIPE_B].sprite[0], SPRITEC));
                I915_WRITE(DSPHOWM,
                           FW_WM(wm->sr.plane >> 9, SR_HI) |
                           FW_WM(wm->pipe[PIPE_B].sprite[1] >> 8, SPRITED_HI) |
                           FW_WM(wm->pipe[PIPE_B].sprite[0] >> 8, SPRITEC_HI) |
                           FW_WM(wm->pipe[PIPE_B].primary >> 8, PLANEB_HI) |
                           FW_WM(wm->pipe[PIPE_A].sprite[1] >> 8, SPRITEB_HI) |
                           FW_WM(wm->pipe[PIPE_A].sprite[0] >> 8, SPRITEA_HI) |
                           FW_WM(wm->pipe[PIPE_A].primary >> 8, PLANEA_HI));
        }

        /* zero (unused) WM1 watermarks */
        I915_WRITE(DSPFW4, 0);
        I915_WRITE(DSPFW5, 0);
        I915_WRITE(DSPFW6, 0);
        I915_WRITE(DSPHOWM1, 0);

        POSTING_READ(DSPFW1);
}

#undef FW_WM_VLV

enum vlv_wm_level {
        VLV_WM_LEVEL_PM2,
        VLV_WM_LEVEL_PM5,
        VLV_WM_LEVEL_DDR_DVFS,
};

/* latency must be in 0.1us units. */
static unsigned int vlv_wm_method2(unsigned int pixel_rate,
                                   unsigned int pipe_htotal,
                                   unsigned int horiz_pixels,
                                   unsigned int bytes_per_pixel,
                                   unsigned int latency)
{
        unsigned int ret;

        ret = (latency * pixel_rate) / (pipe_htotal * 10000);
        ret = (ret + 1) * horiz_pixels * bytes_per_pixel;
        ret = DIV_ROUND_UP(ret, 64);

        return ret;
}

static void vlv_setup_wm_latency(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;

        /* all latencies in usec */
        dev_priv->wm.pri_latency[VLV_WM_LEVEL_PM2] = 3;

        dev_priv->wm.max_level = VLV_WM_LEVEL_PM2;

        if (IS_CHERRYVIEW(dev_priv)) {
                dev_priv->wm.pri_latency[VLV_WM_LEVEL_PM5] = 12;
                dev_priv->wm.pri_latency[VLV_WM_LEVEL_DDR_DVFS] = 33;

                dev_priv->wm.max_level = VLV_WM_LEVEL_DDR_DVFS;
        }
}

static uint16_t vlv_compute_wm_level(struct intel_plane *plane,
                                     struct intel_crtc *crtc,
                                     const struct intel_plane_state *state,
                                     int level)
{
        struct drm_i915_private *dev_priv = to_i915(plane->base.dev);
        int clock, htotal, pixel_size, width, wm;

        if (dev_priv->wm.pri_latency[level] == 0)
                return USHRT_MAX;

        if (!state->visible)
                return 0;

        pixel_size = drm_format_plane_cpp(state->base.fb->pixel_format, 0);
        clock = crtc->config->base.adjusted_mode.crtc_clock;
        htotal = crtc->config->base.adjusted_mode.crtc_htotal;
        width = crtc->config->pipe_src_w;
        if (WARN_ON(htotal == 0))
                htotal = 1;

        if (plane->base.type == DRM_PLANE_TYPE_CURSOR) {
                /*
                 * FIXME the formula gives values that are
                 * too big for the cursor FIFO, and hence we
                 * would never be able to use cursors. For
                 * now just hardcode the watermark.
                 */
                wm = 63;
        } else {
                wm = vlv_wm_method2(clock, htotal, width, pixel_size,
                                    dev_priv->wm.pri_latency[level] * 10);
        }

        return min_t(int, wm, USHRT_MAX);
}

static void vlv_compute_fifo(struct intel_crtc *crtc)
{
        struct drm_device *dev = crtc->base.dev;
        struct vlv_wm_state *wm_state = &crtc->wm_state;
        struct intel_plane *plane;
        unsigned int total_rate = 0;
        const int fifo_size = 512 - 1;
        int fifo_extra, fifo_left = fifo_size;

        for_each_intel_plane_on_crtc(dev, crtc, plane) {
                struct intel_plane_state *state =
                        to_intel_plane_state(plane->base.state);

                if (plane->base.type == DRM_PLANE_TYPE_CURSOR)
                        continue;

                if (state->visible) {
                        wm_state->num_active_planes++;
                        total_rate += drm_format_plane_cpp(state->base.fb->pixel_format, 0);
                }
        }

        for_each_intel_plane_on_crtc(dev, crtc, plane) {
                struct intel_plane_state *state =
                        to_intel_plane_state(plane->base.state);
                unsigned int rate;

                if (plane->base.type == DRM_PLANE_TYPE_CURSOR) {
                        plane->wm.fifo_size = 63;
                        continue;
                }

                if (!state->visible) {
                        plane->wm.fifo_size = 0;
                        continue;
                }

                rate = drm_format_plane_cpp(state->base.fb->pixel_format, 0);
                plane->wm.fifo_size = fifo_size * rate / total_rate;
                fifo_left -= plane->wm.fifo_size;
        }

        fifo_extra = DIV_ROUND_UP(fifo_left, wm_state->num_active_planes ?: 1);

        /* spread the remainder evenly */
        for_each_intel_plane_on_crtc(dev, crtc, plane) {
                int plane_extra;

                if (fifo_left == 0)
                        break;

                if (plane->base.type == DRM_PLANE_TYPE_CURSOR)
                        continue;

                /* give it all to the first plane if none are active */
                if (plane->wm.fifo_size == 0 &&
                    wm_state->num_active_planes)
                        continue;

                plane_extra = min(fifo_extra, fifo_left);
                plane->wm.fifo_size += plane_extra;
                fifo_left -= plane_extra;
        }

        WARN_ON(fifo_left != 0);
}

static void vlv_invert_wms(struct intel_crtc *crtc)
{
        struct vlv_wm_state *wm_state = &crtc->wm_state;
        int level;

        for (level = 0; level < wm_state->num_levels; level++) {
                struct drm_device *dev = crtc->base.dev;
                const int sr_fifo_size = INTEL_INFO(dev)->num_pipes * 512 - 1;
                struct intel_plane *plane;

                wm_state->sr[level].plane = sr_fifo_size - wm_state->sr[level].plane;
                wm_state->sr[level].cursor = 63 - wm_state->sr[level].cursor;

                for_each_intel_plane_on_crtc(dev, crtc, plane) {
                        switch (plane->base.type) {
                                int sprite;
                        case DRM_PLANE_TYPE_CURSOR:
                                wm_state->wm[level].cursor = plane->wm.fifo_size -
                                        wm_state->wm[level].cursor;
                                break;
                        case DRM_PLANE_TYPE_PRIMARY:
                                wm_state->wm[level].primary = plane->wm.fifo_size -
                                        wm_state->wm[level].primary;
                                break;
                        case DRM_PLANE_TYPE_OVERLAY:
                                sprite = plane->plane;
                                wm_state->wm[level].sprite[sprite] = plane->wm.fifo_size -
                                        wm_state->wm[level].sprite[sprite];
                                break;
                        }
                }
        }
}

static void vlv_compute_wm(struct intel_crtc *crtc)
{
        struct drm_device *dev = crtc->base.dev;
        struct vlv_wm_state *wm_state = &crtc->wm_state;
        struct intel_plane *plane;
        int sr_fifo_size = INTEL_INFO(dev)->num_pipes * 512 - 1;
        int level;

        memset(wm_state, 0, sizeof(*wm_state));

        wm_state->cxsr = crtc->pipe != PIPE_C && crtc->wm.cxsr_allowed;
        wm_state->num_levels = to_i915(dev)->wm.max_level + 1;

        wm_state->num_active_planes = 0;

        vlv_compute_fifo(crtc);

        if (wm_state->num_active_planes != 1)
                wm_state->cxsr = false;

        if (wm_state->cxsr) {
                for (level = 0; level < wm_state->num_levels; level++) {
                        wm_state->sr[level].plane = sr_fifo_size;
                        wm_state->sr[level].cursor = 63;
                }
        }

        for_each_intel_plane_on_crtc(dev, crtc, plane) {
                struct intel_plane_state *state =
                        to_intel_plane_state(plane->base.state);

                if (!state->visible)
                        continue;

                /* normal watermarks */
                for (level = 0; level < wm_state->num_levels; level++) {
                        int wm = vlv_compute_wm_level(plane, crtc, state, level);
                        int max_wm = plane->base.type == DRM_PLANE_TYPE_CURSOR ? 63 : 511;

                        /* hack */
                        if (WARN_ON(level == 0 && wm > max_wm))
                                wm = max_wm;

                        if (wm > plane->wm.fifo_size)
                                break;

                        switch (plane->base.type) {
                                int sprite;
                        case DRM_PLANE_TYPE_CURSOR:
                                wm_state->wm[level].cursor = wm;
                                break;
                        case DRM_PLANE_TYPE_PRIMARY:
                                wm_state->wm[level].primary = wm;
                                break;
                        case DRM_PLANE_TYPE_OVERLAY:
                                sprite = plane->plane;
                                wm_state->wm[level].sprite[sprite] = wm;
                                break;
                        }
                }

                wm_state->num_levels = level;

                if (!wm_state->cxsr)
                        continue;

                /* maxfifo watermarks */
                switch (plane->base.type) {
                        int sprite, level;
                case DRM_PLANE_TYPE_CURSOR:
                        for (level = 0; level < wm_state->num_levels; level++)
                                wm_state->sr[level].cursor =
                                        wm_state->wm[level].cursor;
                        break;
                case DRM_PLANE_TYPE_PRIMARY:
                        for (level = 0; level < wm_state->num_levels; level++)
                                wm_state->sr[level].plane =
                                        min(wm_state->sr[level].plane,
                                            wm_state->wm[level].primary);
                        break;
                case DRM_PLANE_TYPE_OVERLAY:
                        sprite = plane->plane;
                        for (level = 0; level < wm_state->num_levels; level++)
                                wm_state->sr[level].plane =
                                        min(wm_state->sr[level].plane,
                                            wm_state->wm[level].sprite[sprite]);
                        break;
                }
        }

        /* clear any (partially) filled invalid levels */
        for (level = wm_state->num_levels; level < to_i915(dev)->wm.max_level + 1; level++) {
                memset(&wm_state->wm[level], 0, sizeof(wm_state->wm[level]));
                memset(&wm_state->sr[level], 0, sizeof(wm_state->sr[level]));
        }

        vlv_invert_wms(crtc);
}

#define VLV_FIFO(plane, value) \
        (((value) << DSPARB_ ## plane ## _SHIFT_VLV) & DSPARB_ ## plane ## _MASK_VLV)

static void vlv_pipe_set_fifo_size(struct intel_crtc *crtc)
{
        struct drm_device *dev = crtc->base.dev;
        struct drm_i915_private *dev_priv = to_i915(dev);
        struct intel_plane *plane;
        int sprite0_start = 0, sprite1_start = 0, fifo_size = 0;

        for_each_intel_plane_on_crtc(dev, crtc, plane) {
                if (plane->base.type == DRM_PLANE_TYPE_CURSOR) {
                        WARN_ON(plane->wm.fifo_size != 63);
                        continue;
                }

                if (plane->base.type == DRM_PLANE_TYPE_PRIMARY)
                        sprite0_start = plane->wm.fifo_size;
                else if (plane->plane == 0)
                        sprite1_start = sprite0_start + plane->wm.fifo_size;
                else
                        fifo_size = sprite1_start + plane->wm.fifo_size;
        }

        WARN_ON(fifo_size != 512 - 1);

        DRM_DEBUG_KMS("Pipe %c FIFO split %d / %d / %d\n",
                      pipe_name(crtc->pipe), sprite0_start,
                      sprite1_start, fifo_size);

        switch (crtc->pipe) {
                uint32_t dsparb, dsparb2, dsparb3;
        case PIPE_A:
                dsparb = I915_READ(DSPARB);
                dsparb2 = I915_READ(DSPARB2);

                dsparb &= ~(VLV_FIFO(SPRITEA, 0xff) |
                            VLV_FIFO(SPRITEB, 0xff));
                dsparb |= (VLV_FIFO(SPRITEA, sprite0_start) |
                           VLV_FIFO(SPRITEB, sprite1_start));

                dsparb2 &= ~(VLV_FIFO(SPRITEA_HI, 0x1) |
                             VLV_FIFO(SPRITEB_HI, 0x1));
                dsparb2 |= (VLV_FIFO(SPRITEA_HI, sprite0_start >> 8) |
                           VLV_FIFO(SPRITEB_HI, sprite1_start >> 8));

                I915_WRITE(DSPARB, dsparb);
                I915_WRITE(DSPARB2, dsparb2);
                break;
        case PIPE_B:
                dsparb = I915_READ(DSPARB);
                dsparb2 = I915_READ(DSPARB2);

                dsparb &= ~(VLV_FIFO(SPRITEC, 0xff) |
                            VLV_FIFO(SPRITED, 0xff));
                dsparb |= (VLV_FIFO(SPRITEC, sprite0_start) |
                           VLV_FIFO(SPRITED, sprite1_start));

                dsparb2 &= ~(VLV_FIFO(SPRITEC_HI, 0xff) |
                             VLV_FIFO(SPRITED_HI, 0xff));
                dsparb2 |= (VLV_FIFO(SPRITEC_HI, sprite0_start >> 8) |
                           VLV_FIFO(SPRITED_HI, sprite1_start >> 8));

                I915_WRITE(DSPARB, dsparb);
                I915_WRITE(DSPARB2, dsparb2);
                break;
        case PIPE_C:
                dsparb3 = I915_READ(DSPARB3);
                dsparb2 = I915_READ(DSPARB2);

                dsparb3 &= ~(VLV_FIFO(SPRITEE, 0xff) |
                             VLV_FIFO(SPRITEF, 0xff));
                dsparb3 |= (VLV_FIFO(SPRITEE, sprite0_start) |
                            VLV_FIFO(SPRITEF, sprite1_start));

                dsparb2 &= ~(VLV_FIFO(SPRITEE_HI, 0xff) |
                             VLV_FIFO(SPRITEF_HI, 0xff));
                dsparb2 |= (VLV_FIFO(SPRITEE_HI, sprite0_start >> 8) |
                           VLV_FIFO(SPRITEF_HI, sprite1_start >> 8));

                I915_WRITE(DSPARB3, dsparb3);
                I915_WRITE(DSPARB2, dsparb2);
                break;
        default:
                break;
        }
}

#undef VLV_FIFO

static void vlv_merge_wm(struct drm_device *dev,
                         struct vlv_wm_values *wm)
{
        struct intel_crtc *crtc;
        int num_active_crtcs = 0;

        wm->level = to_i915(dev)->wm.max_level;
        wm->cxsr = true;

        for_each_intel_crtc(dev, crtc) {
                const struct vlv_wm_state *wm_state = &crtc->wm_state;

                if (!crtc->active)
                        continue;

                if (!wm_state->cxsr)
                        wm->cxsr = false;

                num_active_crtcs++;
                wm->level = min_t(int, wm->level, wm_state->num_levels - 1);
        }

        if (num_active_crtcs != 1)
                wm->cxsr = false;

        if (num_active_crtcs > 1)
                wm->level = VLV_WM_LEVEL_PM2;

        for_each_intel_crtc(dev, crtc) {
                struct vlv_wm_state *wm_state = &crtc->wm_state;
                enum pipe pipe = crtc->pipe;

                if (!crtc->active)
                        continue;

                wm->pipe[pipe] = wm_state->wm[wm->level];
                if (wm->cxsr)
                        wm->sr = wm_state->sr[wm->level];

                wm->ddl[pipe].primary = DDL_PRECISION_HIGH | 2;
                wm->ddl[pipe].sprite[0] = DDL_PRECISION_HIGH | 2;
                wm->ddl[pipe].sprite[1] = DDL_PRECISION_HIGH | 2;
                wm->ddl[pipe].cursor = DDL_PRECISION_HIGH | 2;
        }
}

static void vlv_update_wm(struct drm_crtc *crtc)
{
        struct drm_device *dev = crtc->dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
        enum pipe pipe = intel_crtc->pipe;
        struct vlv_wm_values wm = {};

        vlv_compute_wm(intel_crtc);
        vlv_merge_wm(dev, &wm);

        if (memcmp(&dev_priv->wm.vlv, &wm, sizeof(wm)) == 0) {
                /* FIXME should be part of crtc atomic commit */
                vlv_pipe_set_fifo_size(intel_crtc);
                return;
        }

        if (wm.level < VLV_WM_LEVEL_DDR_DVFS &&
            dev_priv->wm.vlv.level >= VLV_WM_LEVEL_DDR_DVFS)
                chv_set_memory_dvfs(dev_priv, false);

        if (wm.level < VLV_WM_LEVEL_PM5 &&
            dev_priv->wm.vlv.level >= VLV_WM_LEVEL_PM5)
                chv_set_memory_pm5(dev_priv, false);

        if (!wm.cxsr && dev_priv->wm.vlv.cxsr)
                intel_set_memory_cxsr(dev_priv, false);

        /* FIXME should be part of crtc atomic commit */
        vlv_pipe_set_fifo_size(intel_crtc);

        vlv_write_wm_values(intel_crtc, &wm);

        DRM_DEBUG_KMS("Setting FIFO watermarks - %c: plane=%d, cursor=%d, "
                      "sprite0=%d, sprite1=%d, SR: plane=%d, cursor=%d level=%d cxsr=%d\n",
                      pipe_name(pipe), wm.pipe[pipe].primary, wm.pipe[pipe].cursor,
                      wm.pipe[pipe].sprite[0], wm.pipe[pipe].sprite[1],
                      wm.sr.plane, wm.sr.cursor, wm.level, wm.cxsr);

        if (wm.cxsr && !dev_priv->wm.vlv.cxsr)
                intel_set_memory_cxsr(dev_priv, true);

        if (wm.level >= VLV_WM_LEVEL_PM5 &&
            dev_priv->wm.vlv.level < VLV_WM_LEVEL_PM5)
                chv_set_memory_pm5(dev_priv, true);

        if (wm.level >= VLV_WM_LEVEL_DDR_DVFS &&
            dev_priv->wm.vlv.level < VLV_WM_LEVEL_DDR_DVFS)
                chv_set_memory_dvfs(dev_priv, true);

        dev_priv->wm.vlv = wm;
}

#define single_plane_enabled(mask) is_power_of_2(mask)

static void g4x_update_wm(struct drm_crtc *crtc)
{
        struct drm_device *dev = crtc->dev;
        static const int sr_latency_ns = 12000;
        struct drm_i915_private *dev_priv = dev->dev_private;
        int planea_wm, planeb_wm, cursora_wm, cursorb_wm;
        int plane_sr, cursor_sr;
        unsigned int enabled = 0;
        bool cxsr_enabled;

        if (g4x_compute_wm0(dev, PIPE_A,
                            &g4x_wm_info, pessimal_latency_ns,
                            &g4x_cursor_wm_info, pessimal_latency_ns,
                            &planea_wm, &cursora_wm))
                enabled |= 1 << PIPE_A;

        if (g4x_compute_wm0(dev, PIPE_B,
                            &g4x_wm_info, pessimal_latency_ns,
                            &g4x_cursor_wm_info, pessimal_latency_ns,
                            &planeb_wm, &cursorb_wm))
                enabled |= 1 << PIPE_B;

        if (single_plane_enabled(enabled) &&
            g4x_compute_srwm(dev, ffs(enabled) - 1,
                             sr_latency_ns,
                             &g4x_wm_info,
                             &g4x_cursor_wm_info,
                             &plane_sr, &cursor_sr)) {
                cxsr_enabled = true;
        } else {
                cxsr_enabled = false;
                intel_set_memory_cxsr(dev_priv, false);
                plane_sr = cursor_sr = 0;
        }

        DRM_DEBUG_KMS("Setting FIFO watermarks - A: plane=%d, cursor=%d, "
                      "B: plane=%d, cursor=%d, SR: plane=%d, cursor=%d\n",
                      planea_wm, cursora_wm,
                      planeb_wm, cursorb_wm,
                      plane_sr, cursor_sr);

        I915_WRITE(DSPFW1,
                   FW_WM(plane_sr, SR) |
                   FW_WM(cursorb_wm, CURSORB) |
                   FW_WM(planeb_wm, PLANEB) |
                   FW_WM(planea_wm, PLANEA));
        I915_WRITE(DSPFW2,
                   (I915_READ(DSPFW2) & ~DSPFW_CURSORA_MASK) |
                   FW_WM(cursora_wm, CURSORA));
        /* HPLL off in SR has some issues on G4x... disable it */
        I915_WRITE(DSPFW3,
                   (I915_READ(DSPFW3) & ~(DSPFW_HPLL_SR_EN | DSPFW_CURSOR_SR_MASK)) |
                   FW_WM(cursor_sr, CURSOR_SR));

        if (cxsr_enabled)
                intel_set_memory_cxsr(dev_priv, true);
}

static void i965_update_wm(struct drm_crtc *unused_crtc)
{
        struct drm_device *dev = unused_crtc->dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_crtc *crtc;
        int srwm = 1;
        int cursor_sr = 16;
        bool cxsr_enabled;

        /* Calc sr entries for one plane configs */
        crtc = single_enabled_crtc(dev);
        if (crtc) {
                /* self-refresh has much higher latency */
                static const int sr_latency_ns = 12000;
                const struct drm_display_mode *adjusted_mode = &to_intel_crtc(crtc)->config->base.adjusted_mode;
                int clock = adjusted_mode->crtc_clock;
                int htotal = adjusted_mode->crtc_htotal;
                int hdisplay = to_intel_crtc(crtc)->config->pipe_src_w;
                int pixel_size = crtc->primary->state->fb->bits_per_pixel / 8;
                unsigned long line_time_us;
                int entries;

                line_time_us = max(htotal * 1000 / clock, 1);

                /* Use ns/us then divide to preserve precision */
                entries = (((sr_latency_ns / line_time_us) + 1000) / 1000) *
                        pixel_size * hdisplay;
                entries = DIV_ROUND_UP(entries, I915_FIFO_LINE_SIZE);
                srwm = I965_FIFO_SIZE - entries;
                if (srwm < 0)
                        srwm = 1;
                srwm &= 0x1ff;
                DRM_DEBUG_KMS("self-refresh entries: %d, wm: %d\n",
                              entries, srwm);

                entries = (((sr_latency_ns / line_time_us) + 1000) / 1000) *
                        pixel_size * crtc->cursor->state->crtc_w;
                entries = DIV_ROUND_UP(entries,
                                          i965_cursor_wm_info.cacheline_size);
                cursor_sr = i965_cursor_wm_info.fifo_size -
                        (entries + i965_cursor_wm_info.guard_size);

                if (cursor_sr > i965_cursor_wm_info.max_wm)
                        cursor_sr = i965_cursor_wm_info.max_wm;

                DRM_DEBUG_KMS("self-refresh watermark: display plane %d "
                              "cursor %d\n", srwm, cursor_sr);

                cxsr_enabled = true;
        } else {
                cxsr_enabled = false;
                /* Turn off self refresh if both pipes are enabled */
                intel_set_memory_cxsr(dev_priv, false);
        }

        DRM_DEBUG_KMS("Setting FIFO watermarks - A: 8, B: 8, C: 8, SR %d\n",
                      srwm);

        /* 965 has limitations... */
        I915_WRITE(DSPFW1, FW_WM(srwm, SR) |
                   FW_WM(8, CURSORB) |
                   FW_WM(8, PLANEB) |
                   FW_WM(8, PLANEA));
        I915_WRITE(DSPFW2, FW_WM(8, CURSORA) |
                   FW_WM(8, PLANEC_OLD));
        /* update cursor SR watermark */
        I915_WRITE(DSPFW3, FW_WM(cursor_sr, CURSOR_SR));

        if (cxsr_enabled)
                intel_set_memory_cxsr(dev_priv, true);
}

#undef FW_WM

static void i9xx_update_wm(struct drm_crtc *unused_crtc)
{
        struct drm_device *dev = unused_crtc->dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        const struct intel_watermark_params *wm_info;
        uint32_t fwater_lo;
        uint32_t fwater_hi;
        int cwm, srwm = 1;
        int fifo_size;
        int planea_wm, planeb_wm;
        struct drm_crtc *crtc, *enabled = NULL;

        if (IS_I945GM(dev))
                wm_info = &i945_wm_info;
        else if (!IS_GEN2(dev))
                wm_info = &i915_wm_info;
        else
                wm_info = &i830_a_wm_info;

        fifo_size = dev_priv->display.get_fifo_size(dev, 0);
        crtc = intel_get_crtc_for_plane(dev, 0);
        if (intel_crtc_active(crtc)) {
                const struct drm_display_mode *adjusted_mode;
                int cpp = crtc->primary->state->fb->bits_per_pixel / 8;
                if (IS_GEN2(dev))
                        cpp = 4;

                adjusted_mode = &to_intel_crtc(crtc)->config->base.adjusted_mode;
                planea_wm = intel_calculate_wm(adjusted_mode->crtc_clock,
                                               wm_info, fifo_size, cpp,
                                               pessimal_latency_ns);
                enabled = crtc;
        } else {
                planea_wm = fifo_size - wm_info->guard_size;
                if (planea_wm > (long)wm_info->max_wm)
                        planea_wm = wm_info->max_wm;
        }

        if (IS_GEN2(dev))
                wm_info = &i830_bc_wm_info;

        fifo_size = dev_priv->display.get_fifo_size(dev, 1);
        crtc = intel_get_crtc_for_plane(dev, 1);
        if (intel_crtc_active(crtc)) {
                const struct drm_display_mode *adjusted_mode;
                int cpp = crtc->primary->state->fb->bits_per_pixel / 8;
                if (IS_GEN2(dev))
                        cpp = 4;

                adjusted_mode = &to_intel_crtc(crtc)->config->base.adjusted_mode;
                planeb_wm = intel_calculate_wm(adjusted_mode->crtc_clock,
                                               wm_info, fifo_size, cpp,
                                               pessimal_latency_ns);
                if (enabled == NULL)
                        enabled = crtc;
                else
                        enabled = NULL;
        } else {
                planeb_wm = fifo_size - wm_info->guard_size;
                if (planeb_wm > (long)wm_info->max_wm)
                        planeb_wm = wm_info->max_wm;
        }

        DRM_DEBUG_KMS("FIFO watermarks - A: %d, B: %d\n", planea_wm, planeb_wm);

        if (IS_I915GM(dev) && enabled) {
                struct drm_i915_gem_object *obj;

                obj = intel_fb_obj(enabled->primary->state->fb);

                /* self-refresh seems busted with untiled */
                if (obj->tiling_mode == I915_TILING_NONE)
                        enabled = NULL;
        }

        /*
         * Overlay gets an aggressive default since video jitter is bad.
         */
        cwm = 2;

        /* Play safe and disable self-refresh before adjusting watermarks. */
        intel_set_memory_cxsr(dev_priv, false);

        /* Calc sr entries for one plane configs */
        if (HAS_FW_BLC(dev) && enabled) {
                /* self-refresh has much higher latency */
                static const int sr_latency_ns = 6000;
                const struct drm_display_mode *adjusted_mode = &to_intel_crtc(enabled)->config->base.adjusted_mode;
                int clock = adjusted_mode->crtc_clock;
                int htotal = adjusted_mode->crtc_htotal;
                int hdisplay = to_intel_crtc(enabled)->config->pipe_src_w;
                int pixel_size = enabled->primary->state->fb->bits_per_pixel / 8;
                unsigned long line_time_us;
                int entries;

                line_time_us = max(htotal * 1000 / clock, 1);

                /* Use ns/us then divide to preserve precision */
                entries = (((sr_latency_ns / line_time_us) + 1000) / 1000) *
                        pixel_size * hdisplay;
                entries = DIV_ROUND_UP(entries, wm_info->cacheline_size);
                DRM_DEBUG_KMS("self-refresh entries: %d\n", entries);
                srwm = wm_info->fifo_size - entries;
                if (srwm < 0)
                        srwm = 1;

                if (IS_I945G(dev) || IS_I945GM(dev))
                        I915_WRITE(FW_BLC_SELF,
                                   FW_BLC_SELF_FIFO_MASK | (srwm & 0xff));
                else if (IS_I915GM(dev))
                        I915_WRITE(FW_BLC_SELF, srwm & 0x3f);
        }

        DRM_DEBUG_KMS("Setting FIFO watermarks - A: %d, B: %d, C: %d, SR %d\n",
                      planea_wm, planeb_wm, cwm, srwm);

        fwater_lo = ((planeb_wm & 0x3f) << 16) | (planea_wm & 0x3f);
        fwater_hi = (cwm & 0x1f);

        /* Set request length to 8 cachelines per fetch */
        fwater_lo = fwater_lo | (1 << 24) | (1 << 8);
        fwater_hi = fwater_hi | (1 << 8);

        I915_WRITE(FW_BLC, fwater_lo);
        I915_WRITE(FW_BLC2, fwater_hi);

        if (enabled)
                intel_set_memory_cxsr(dev_priv, true);
}

static void i845_update_wm(struct drm_crtc *unused_crtc)
{
        struct drm_device *dev = unused_crtc->dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_crtc *crtc;
        const struct drm_display_mode *adjusted_mode;
        uint32_t fwater_lo;
        int planea_wm;

        crtc = single_enabled_crtc(dev);
        if (crtc == NULL)
                return;

        adjusted_mode = &to_intel_crtc(crtc)->config->base.adjusted_mode;
        planea_wm = intel_calculate_wm(adjusted_mode->crtc_clock,
                                       &i845_wm_info,
                                       dev_priv->display.get_fifo_size(dev, 0),
                                       4, pessimal_latency_ns);
        fwater_lo = I915_READ(FW_BLC) & ~0xfff;
        fwater_lo |= (3<<8) | planea_wm;

        DRM_DEBUG_KMS("Setting FIFO watermarks - A: %d\n", planea_wm);

        I915_WRITE(FW_BLC, fwater_lo);
}

uint32_t ilk_pipe_pixel_rate(const struct intel_crtc_state *pipe_config)
{
        uint32_t pixel_rate;

        pixel_rate = pipe_config->base.adjusted_mode.crtc_clock;

        /* We only use IF-ID interlacing. If we ever use PF-ID we'll need to
         * adjust the pixel_rate here. */

        if (pipe_config->pch_pfit.enabled) {
                uint64_t pipe_w, pipe_h, pfit_w, pfit_h;
                uint32_t pfit_size = pipe_config->pch_pfit.size;

                pipe_w = pipe_config->pipe_src_w;
                pipe_h = pipe_config->pipe_src_h;

                pfit_w = (pfit_size >> 16) & 0xFFFF;
                pfit_h = pfit_size & 0xFFFF;
                if (pipe_w < pfit_w)
                        pipe_w = pfit_w;
                if (pipe_h < pfit_h)
                        pipe_h = pfit_h;

                pixel_rate = div_u64((uint64_t) pixel_rate * pipe_w * pipe_h,
                                     pfit_w * pfit_h);
        }

        return pixel_rate;
}

/* latency must be in 0.1us units. */
static uint32_t ilk_wm_method1(uint32_t pixel_rate, uint8_t bytes_per_pixel,
                               uint32_t latency)
{
        uint64_t ret;

        if (WARN(latency == 0, "Latency value missing\n"))
                return UINT_MAX;

        ret = (uint64_t) pixel_rate * bytes_per_pixel * latency;
        ret = DIV_ROUND_UP_ULL(ret, 64 * 10000) + 2;

        return ret;
}

/* latency must be in 0.1us units. */
static uint32_t ilk_wm_method2(uint32_t pixel_rate, uint32_t pipe_htotal,
                               uint32_t horiz_pixels, uint8_t bytes_per_pixel,
                               uint32_t latency)
{
        uint32_t ret;

        if (WARN(latency == 0, "Latency value missing\n"))
                return UINT_MAX;

        ret = (latency * pixel_rate) / (pipe_htotal * 10000);
        ret = (ret + 1) * horiz_pixels * bytes_per_pixel;
        ret = DIV_ROUND_UP(ret, 64) + 2;
        return ret;
}

static uint32_t ilk_wm_fbc(uint32_t pri_val, uint32_t horiz_pixels,
                           uint8_t bytes_per_pixel)
{
        return DIV_ROUND_UP(pri_val * 64, horiz_pixels * bytes_per_pixel) + 2;
}

struct skl_pipe_wm_parameters {
        bool active;
        uint32_t pipe_htotal;
        uint32_t pixel_rate; /* in KHz */
        struct intel_plane_wm_parameters plane[I915_MAX_PLANES];
};

struct ilk_wm_maximums {
        uint16_t pri;
        uint16_t spr;
        uint16_t cur;
        uint16_t fbc;
};

/* used in computing the new watermarks state */
struct intel_wm_config {
        unsigned int num_pipes_active;
        bool sprites_enabled;
        bool sprites_scaled;
};

/*
 * For both WM_PIPE and WM_LP.
 * mem_value must be in 0.1us units.
 */
static uint32_t ilk_compute_pri_wm(const struct intel_crtc_state *cstate,
                                   const struct intel_plane_state *pstate,
                                   uint32_t mem_value,
                                   bool is_lp)
{
        int bpp = pstate->base.fb ? pstate->base.fb->bits_per_pixel / 8 : 0;
        uint32_t method1, method2;

        if (!cstate->base.active || !pstate->visible)
                return 0;

        method1 = ilk_wm_method1(ilk_pipe_pixel_rate(cstate), bpp, mem_value);

        if (!is_lp)
                return method1;

        method2 = ilk_wm_method2(ilk_pipe_pixel_rate(cstate),
                                 cstate->base.adjusted_mode.crtc_htotal,
                                 drm_rect_width(&pstate->dst),
                                 bpp,
                                 mem_value);

        return min(method1, method2);
}

/*
 * For both WM_PIPE and WM_LP.
 * mem_value must be in 0.1us units.
 */
static uint32_t ilk_compute_spr_wm(const struct intel_crtc_state *cstate,
                                   const struct intel_plane_state *pstate,
                                   uint32_t mem_value)
{
        int bpp = pstate->base.fb ? pstate->base.fb->bits_per_pixel / 8 : 0;
        uint32_t method1, method2;

        if (!cstate->base.active || !pstate->visible)
                return 0;

        method1 = ilk_wm_method1(ilk_pipe_pixel_rate(cstate), bpp, mem_value);
        method2 = ilk_wm_method2(ilk_pipe_pixel_rate(cstate),
                                 cstate->base.adjusted_mode.crtc_htotal,
                                 drm_rect_width(&pstate->dst),
                                 bpp,
                                 mem_value);
        return min(method1, method2);
}

/*
 * For both WM_PIPE and WM_LP.
 * mem_value must be in 0.1us units.
 */
static uint32_t ilk_compute_cur_wm(const struct intel_crtc_state *cstate,
                                   const struct intel_plane_state *pstate,
                                   uint32_t mem_value)
{
        int bpp = pstate->base.fb ? pstate->base.fb->bits_per_pixel / 8 : 0;

        if (!cstate->base.active || !pstate->visible)
                return 0;

        return ilk_wm_method2(ilk_pipe_pixel_rate(cstate),
                              cstate->base.adjusted_mode.crtc_htotal,
                              drm_rect_width(&pstate->dst),
                              bpp,
                              mem_value);
}

/* Only for WM_LP. */
static uint32_t ilk_compute_fbc_wm(const struct intel_crtc_state *cstate,
                                   const struct intel_plane_state *pstate,
                                   uint32_t pri_val)
{
        int bpp = pstate->base.fb ? pstate->base.fb->bits_per_pixel / 8 : 0;

        if (!cstate->base.active || !pstate->visible)
                return 0;

        return ilk_wm_fbc(pri_val, drm_rect_width(&pstate->dst), bpp);
}

static unsigned int ilk_display_fifo_size(const struct drm_device *dev)
{
        if (INTEL_INFO(dev)->gen >= 8)
                return 3072;
        else if (INTEL_INFO(dev)->gen >= 7)
                return 768;
        else
                return 512;
}

static unsigned int ilk_plane_wm_reg_max(const struct drm_device *dev,
                                         int level, bool is_sprite)
{
        if (INTEL_INFO(dev)->gen >= 8)
                /* BDW primary/sprite plane watermarks */
                return level == 0 ? 255 : 2047;
        else if (INTEL_INFO(dev)->gen >= 7)
                /* IVB/HSW primary/sprite plane watermarks */
                return level == 0 ? 127 : 1023;
        else if (!is_sprite)
                /* ILK/SNB primary plane watermarks */
                return level == 0 ? 127 : 511;
        else
                /* ILK/SNB sprite plane watermarks */
                return level == 0 ? 63 : 255;
}

static unsigned int ilk_cursor_wm_reg_max(const struct drm_device *dev,
                                          int level)
{
        if (INTEL_INFO(dev)->gen >= 7)
                return level == 0 ? 63 : 255;
        else
                return level == 0 ? 31 : 63;
}

static unsigned int ilk_fbc_wm_reg_max(const struct drm_device *dev)
{
        if (INTEL_INFO(dev)->gen >= 8)
                return 31;
        else
                return 15;
}

/* Calculate the maximum primary/sprite plane watermark */
static unsigned int ilk_plane_wm_max(const struct drm_device *dev,
                                     int level,
                                     const struct intel_wm_config *config,
                                     enum intel_ddb_partitioning ddb_partitioning,
                                     bool is_sprite)
{
        unsigned int fifo_size = ilk_display_fifo_size(dev);

        /* if sprites aren't enabled, sprites get nothing */
        if (is_sprite && !config->sprites_enabled)
                return 0;

        /* HSW allows LP1+ watermarks even with multiple pipes */
        if (level == 0 || config->num_pipes_active > 1) {
                fifo_size /= INTEL_INFO(dev)->num_pipes;

                /*
                 * For some reason the non self refresh
                 * FIFO size is only half of the self
                 * refresh FIFO size on ILK/SNB.
                 */
                if (INTEL_INFO(dev)->gen <= 6)
                        fifo_size /= 2;
        }

        if (config->sprites_enabled) {
                /* level 0 is always calculated with 1:1 split */
                if (level > 0 && ddb_partitioning == INTEL_DDB_PART_5_6) {
                        if (is_sprite)
                                fifo_size *= 5;
                        fifo_size /= 6;
                } else {
                        fifo_size /= 2;
                }
        }

        /* clamp to max that the registers can hold */
        return min(fifo_size, ilk_plane_wm_reg_max(dev, level, is_sprite));
}

/* Calculate the maximum cursor plane watermark */
static unsigned int ilk_cursor_wm_max(const struct drm_device *dev,
                                      int level,
                                      const struct intel_wm_config *config)
{
        /* HSW LP1+ watermarks w/ multiple pipes */
        if (level > 0 && config->num_pipes_active > 1)
                return 64;

        /* otherwise just report max that registers can hold */
        return ilk_cursor_wm_reg_max(dev, level);
}

static void ilk_compute_wm_maximums(const struct drm_device *dev,
                                    int level,
                                    const struct intel_wm_config *config,
                                    enum intel_ddb_partitioning ddb_partitioning,
                                    struct ilk_wm_maximums *max)
{
        max->pri = ilk_plane_wm_max(dev, level, config, ddb_partitioning, false);
        max->spr = ilk_plane_wm_max(dev, level, config, ddb_partitioning, true);
        max->cur = ilk_cursor_wm_max(dev, level, config);
        max->fbc = ilk_fbc_wm_reg_max(dev);
}

static void ilk_compute_wm_reg_maximums(struct drm_device *dev,
                                        int level,
                                        struct ilk_wm_maximums *max)
{
        max->pri = ilk_plane_wm_reg_max(dev, level, false);
        max->spr = ilk_plane_wm_reg_max(dev, level, true);
        max->cur = ilk_cursor_wm_reg_max(dev, level);
        max->fbc = ilk_fbc_wm_reg_max(dev);
}

static bool ilk_validate_wm_level(int level,
                                  const struct ilk_wm_maximums *max,
                                  struct intel_wm_level *result)
{
        bool ret;

        /* already determined to be invalid? */
        if (!result->enable)
                return false;

        result->enable = result->pri_val <= max->pri &&
                         result->spr_val <= max->spr &&
                         result->cur_val <= max->cur;

        ret = result->enable;

        /*
         * HACK until we can pre-compute everything,
         * and thus fail gracefully if LP0 watermarks
         * are exceeded...
         */
        if (level == 0 && !result->enable) {
                if (result->pri_val > max->pri)
                        DRM_DEBUG_KMS("Primary WM%d too large %u (max %u)\n",
                                      level, result->pri_val, max->pri);
                if (result->spr_val > max->spr)
                        DRM_DEBUG_KMS("Sprite WM%d too large %u (max %u)\n",
                                      level, result->spr_val, max->spr);
                if (result->cur_val > max->cur)
                        DRM_DEBUG_KMS("Cursor WM%d too large %u (max %u)\n",
                                      level, result->cur_val, max->cur);

                result->pri_val = min_t(uint32_t, result->pri_val, max->pri);
                result->spr_val = min_t(uint32_t, result->spr_val, max->spr);
                result->cur_val = min_t(uint32_t, result->cur_val, max->cur);
                result->enable = true;
        }

        return ret;
}

static void ilk_compute_wm_level(const struct drm_i915_private *dev_priv,
                                 const struct intel_crtc *intel_crtc,
                                 int level,
                                 struct intel_crtc_state *cstate,
                                 struct intel_wm_level *result)
{
        struct intel_plane *intel_plane;
        uint16_t pri_latency = dev_priv->wm.pri_latency[level];
        uint16_t spr_latency = dev_priv->wm.spr_latency[level];
        uint16_t cur_latency = dev_priv->wm.cur_latency[level];

        /* WM1+ latency values stored in 0.5us units */
        if (level > 0) {
                pri_latency *= 5;
                spr_latency *= 5;
                cur_latency *= 5;
        }

        for_each_intel_plane_on_crtc(dev_priv->dev, intel_crtc, intel_plane) {
                struct intel_plane_state *pstate =
                        to_intel_plane_state(intel_plane->base.state);

                switch (intel_plane->base.type) {

                case DRM_PLANE_TYPE_PRIMARY:
                        result->pri_val = ilk_compute_pri_wm(cstate, pstate,
                                                             pri_latency,
                                                             level);
                        result->fbc_val = ilk_compute_fbc_wm(cstate, pstate,
                                                             result->pri_val);
                        break;
                case DRM_PLANE_TYPE_OVERLAY:
                        result->spr_val = ilk_compute_spr_wm(cstate, pstate,
                                                             spr_latency);
                        break;
                case DRM_PLANE_TYPE_CURSOR:
                        result->cur_val = ilk_compute_cur_wm(cstate, pstate,
                                                             cur_latency);
                        break;
                }
        }

        result->enable = true;
}

static uint32_t
hsw_compute_linetime_wm(struct drm_device *dev, struct drm_crtc *crtc)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
        const struct drm_display_mode *adjusted_mode = &intel_crtc->config->base.adjusted_mode;
        u32 linetime, ips_linetime;

        if (!intel_crtc->active)
                return 0;

        /* The WM are computed with base on how long it takes to fill a single
         * row at the given clock rate, multiplied by 8.
         * */
        linetime = DIV_ROUND_CLOSEST(adjusted_mode->crtc_htotal * 1000 * 8,
                                     adjusted_mode->crtc_clock);
        ips_linetime = DIV_ROUND_CLOSEST(adjusted_mode->crtc_htotal * 1000 * 8,
                                         dev_priv->cdclk_freq);

        return PIPE_WM_LINETIME_IPS_LINETIME(ips_linetime) |
               PIPE_WM_LINETIME_TIME(linetime);
}

static void intel_read_wm_latency(struct drm_device *dev, uint16_t wm[8])
{
        struct drm_i915_private *dev_priv = dev->dev_private;

        if (IS_GEN9(dev)) {
                uint32_t val;
                int ret, i;
                int level, max_level = ilk_wm_max_level(dev);

                /* read the first set of memory latencies[0:3] */
                val = 0; /* data0 to be programmed to 0 for first set */
                mutex_lock(&dev_priv->rps.hw_lock);
                ret = sandybridge_pcode_read(dev_priv,
                                             GEN9_PCODE_READ_MEM_LATENCY,
                                             &val);
                mutex_unlock(&dev_priv->rps.hw_lock);

                if (ret) {
                        DRM_ERROR("SKL Mailbox read error = %d\n", ret);
                        return;
                }

                wm[0] = val & GEN9_MEM_LATENCY_LEVEL_MASK;
                wm[1] = (val >> GEN9_MEM_LATENCY_LEVEL_1_5_SHIFT) &
                                GEN9_MEM_LATENCY_LEVEL_MASK;
                wm[2] = (val >> GEN9_MEM_LATENCY_LEVEL_2_6_SHIFT) &
                                GEN9_MEM_LATENCY_LEVEL_MASK;
                wm[3] = (val >> GEN9_MEM_LATENCY_LEVEL_3_7_SHIFT) &
                                GEN9_MEM_LATENCY_LEVEL_MASK;

                /* read the second set of memory latencies[4:7] */
                val = 1; /* data0 to be programmed to 1 for second set */
                mutex_lock(&dev_priv->rps.hw_lock);
                ret = sandybridge_pcode_read(dev_priv,
                                             GEN9_PCODE_READ_MEM_LATENCY,
                                             &val);
                mutex_unlock(&dev_priv->rps.hw_lock);
                if (ret) {
                        DRM_ERROR("SKL Mailbox read error = %d\n", ret);
                        return;
                }

                wm[4] = val & GEN9_MEM_LATENCY_LEVEL_MASK;
                wm[5] = (val >> GEN9_MEM_LATENCY_LEVEL_1_5_SHIFT) &
                                GEN9_MEM_LATENCY_LEVEL_MASK;
                wm[6] = (val >> GEN9_MEM_LATENCY_LEVEL_2_6_SHIFT) &
                                GEN9_MEM_LATENCY_LEVEL_MASK;
                wm[7] = (val >> GEN9_MEM_LATENCY_LEVEL_3_7_SHIFT) &
                                GEN9_MEM_LATENCY_LEVEL_MASK;

                /*
                 * WaWmMemoryReadLatency:skl
                 *
                 * punit doesn't take into account the read latency so we need
                 * to add 2us to the various latency levels we retrieve from
                 * the punit.
                 *   - W0 is a bit special in that it's the only level that
                 *   can't be disabled if we want to have display working, so
                 *   we always add 2us there.
                 *   - For levels >=1, punit returns 0us latency when they are
                 *   disabled, so we respect that and don't add 2us then
                 *
                 * Additionally, if a level n (n > 1) has a 0us latency, all
                 * levels m (m >= n) need to be disabled. We make sure to
                 * sanitize the values out of the punit to satisfy this
                 * requirement.
                 */
                wm[0] += 2;
                for (level = 1; level <= max_level; level++)
                        if (wm[level] != 0)
                                wm[level] += 2;
                        else {
                                for (i = level + 1; i <= max_level; i++)
                                        wm[i] = 0;

                                break;
                        }
        } else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) {
                uint64_t sskpd = I915_READ64(MCH_SSKPD);

                wm[0] = (sskpd >> 56) & 0xFF;
                if (wm[0] == 0)
                        wm[0] = sskpd & 0xF;
                wm[1] = (sskpd >> 4) & 0xFF;
                wm[2] = (sskpd >> 12) & 0xFF;
                wm[3] = (sskpd >> 20) & 0x1FF;
                wm[4] = (sskpd >> 32) & 0x1FF;
        } else if (INTEL_INFO(dev)->gen >= 6) {
                uint32_t sskpd = I915_READ(MCH_SSKPD);

                wm[0] = (sskpd >> SSKPD_WM0_SHIFT) & SSKPD_WM_MASK;
                wm[1] = (sskpd >> SSKPD_WM1_SHIFT) & SSKPD_WM_MASK;
                wm[2] = (sskpd >> SSKPD_WM2_SHIFT) & SSKPD_WM_MASK;
                wm[3] = (sskpd >> SSKPD_WM3_SHIFT) & SSKPD_WM_MASK;
        } else if (INTEL_INFO(dev)->gen >= 5) {
                uint32_t mltr = I915_READ(MLTR_ILK);

                /* ILK primary LP0 latency is 700 ns */
                wm[0] = 7;
                wm[1] = (mltr >> MLTR_WM1_SHIFT) & ILK_SRLT_MASK;
                wm[2] = (mltr >> MLTR_WM2_SHIFT) & ILK_SRLT_MASK;
        }
}

static void intel_fixup_spr_wm_latency(struct drm_device *dev, uint16_t wm[5])
{
        /* ILK sprite LP0 latency is 1300 ns */
        if (INTEL_INFO(dev)->gen == 5)
                wm[0] = 13;
}

static void intel_fixup_cur_wm_latency(struct drm_device *dev, uint16_t wm[5])
{
        /* ILK cursor LP0 latency is 1300 ns */
        if (INTEL_INFO(dev)->gen == 5)
                wm[0] = 13;

        /* WaDoubleCursorLP3Latency:ivb */
        if (IS_IVYBRIDGE(dev))
                wm[3] *= 2;
}

int ilk_wm_max_level(const struct drm_device *dev)
{
        /* how many WM levels are we expecting */
        if (INTEL_INFO(dev)->gen >= 9)
                return 7;
        else if (IS_HASWELL(dev) || IS_BROADWELL(dev))
                return 4;
        else if (INTEL_INFO(dev)->gen >= 6)
                return 3;
        else
                return 2;
}

static void intel_print_wm_latency(struct drm_device *dev,
                                   const char *name,
                                   const uint16_t wm[8])
{
        int level, max_level = ilk_wm_max_level(dev);

        for (level = 0; level <= max_level; level++) {
                unsigned int latency = wm[level];

                if (latency == 0) {
                        DRM_ERROR("%s WM%d latency not provided\n",
                                  name, level);
                        continue;
                }

                /*
                 * - latencies are in us on gen9.
                 * - before then, WM1+ latency values are in 0.5us units
                 */
                if (IS_GEN9(dev))
                        latency *= 10;
                else if (level > 0)
                        latency *= 5;

                DRM_DEBUG_KMS("%s WM%d latency %u (%u.%u usec)\n",
                              name, level, wm[level],
                              latency / 10, latency % 10);
        }
}

static bool ilk_increase_wm_latency(struct drm_i915_private *dev_priv,
                                    uint16_t wm[5], uint16_t min)
{
        int level, max_level = ilk_wm_max_level(dev_priv->dev);

        if (wm[0] >= min)
                return false;

        wm[0] = max(wm[0], min);
        for (level = 1; level <= max_level; level++)
                wm[level] = max_t(uint16_t, wm[level], DIV_ROUND_UP(min, 5));

        return true;
}

static void snb_wm_latency_quirk(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        bool changed;

        /*
         * The BIOS provided WM memory latency values are often
         * inadequate for high resolution displays. Adjust them.
         */
        changed = ilk_increase_wm_latency(dev_priv, dev_priv->wm.pri_latency, 12) |
                ilk_increase_wm_latency(dev_priv, dev_priv->wm.spr_latency, 12) |
                ilk_increase_wm_latency(dev_priv, dev_priv->wm.cur_latency, 12);

        if (!changed)
                return;

        DRM_DEBUG_KMS("WM latency values increased to avoid potential underruns\n");
        intel_print_wm_latency(dev, "Primary", dev_priv->wm.pri_latency);
        intel_print_wm_latency(dev, "Sprite", dev_priv->wm.spr_latency);
        intel_print_wm_latency(dev, "Cursor", dev_priv->wm.cur_latency);
}

static void ilk_setup_wm_latency(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;

        intel_read_wm_latency(dev, dev_priv->wm.pri_latency);

        memcpy(dev_priv->wm.spr_latency, dev_priv->wm.pri_latency,
               sizeof(dev_priv->wm.pri_latency));
        memcpy(dev_priv->wm.cur_latency, dev_priv->wm.pri_latency,
               sizeof(dev_priv->wm.pri_latency));

        intel_fixup_spr_wm_latency(dev, dev_priv->wm.spr_latency);
        intel_fixup_cur_wm_latency(dev, dev_priv->wm.cur_latency);

        intel_print_wm_latency(dev, "Primary", dev_priv->wm.pri_latency);
        intel_print_wm_latency(dev, "Sprite", dev_priv->wm.spr_latency);
        intel_print_wm_latency(dev, "Cursor", dev_priv->wm.cur_latency);

        if (IS_GEN6(dev))
                snb_wm_latency_quirk(dev);
}

static void skl_setup_wm_latency(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;

        intel_read_wm_latency(dev, dev_priv->wm.skl_latency);
        intel_print_wm_latency(dev, "Gen9 Plane", dev_priv->wm.skl_latency);
}

static void ilk_compute_wm_config(struct drm_device *dev,
                                  struct intel_wm_config *config)
{
        struct intel_crtc *intel_crtc;

        /* Compute the currently _active_ config */
        for_each_intel_crtc(dev, intel_crtc) {
                const struct intel_pipe_wm *wm = &intel_crtc->wm.active;

                if (!wm->pipe_enabled)
                        continue;

                config->sprites_enabled |= wm->sprites_enabled;
                config->sprites_scaled |= wm->sprites_scaled;
                config->num_pipes_active++;
        }
}

/* Compute new watermarks for the pipe */
static bool intel_compute_pipe_wm(struct intel_crtc_state *cstate,
                                  struct intel_pipe_wm *pipe_wm)
{
        struct drm_crtc *crtc = cstate->base.crtc;
        struct drm_device *dev = crtc->dev;
        const struct drm_i915_private *dev_priv = dev->dev_private;
        struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
        struct intel_plane *intel_plane;
        struct intel_plane_state *sprstate = NULL;
        int level, max_level = ilk_wm_max_level(dev);
        /* LP0 watermark maximums depend on this pipe alone */
        struct intel_wm_config config = {
                .num_pipes_active = 1,
        };
        struct ilk_wm_maximums max;

        for_each_intel_plane_on_crtc(dev, intel_crtc, intel_plane) {
                if (intel_plane->base.type == DRM_PLANE_TYPE_OVERLAY) {
                        sprstate = to_intel_plane_state(intel_plane->base.state);
                        break;
                }
        }

        config.sprites_enabled = sprstate->visible;
        config.sprites_scaled = sprstate->visible &&
                (drm_rect_width(&sprstate->dst) != drm_rect_width(&sprstate->src) >> 16 ||
                drm_rect_height(&sprstate->dst) != drm_rect_height(&sprstate->src) >> 16);

        pipe_wm->pipe_enabled = cstate->base.active;
        pipe_wm->sprites_enabled = sprstate->visible;
        pipe_wm->sprites_scaled = config.sprites_scaled;

        /* ILK/SNB: LP2+ watermarks only w/o sprites */
        if (INTEL_INFO(dev)->gen <= 6 && sprstate->visible)
                max_level = 1;

        /* ILK/SNB/IVB: LP1+ watermarks only w/o scaling */
        if (config.sprites_scaled)
                max_level = 0;

        ilk_compute_wm_level(dev_priv, intel_crtc, 0, cstate, &pipe_wm->wm[0]);

        if (IS_HASWELL(dev) || IS_BROADWELL(dev))
                pipe_wm->linetime = hsw_compute_linetime_wm(dev, crtc);

        /* LP0 watermarks always use 1/2 DDB partitioning */
        ilk_compute_wm_maximums(dev, 0, &config, INTEL_DDB_PART_1_2, &max);

        /* At least LP0 must be valid */
        if (!ilk_validate_wm_level(0, &max, &pipe_wm->wm[0]))
                return false;

        ilk_compute_wm_reg_maximums(dev, 1, &max);

        for (level = 1; level <= max_level; level++) {
                struct intel_wm_level wm = {};

                ilk_compute_wm_level(dev_priv, intel_crtc, level, cstate, &wm);

                /*
                 * Disable any watermark level that exceeds the
                 * register maximums since such watermarks are
                 * always invalid.
                 */
                if (!ilk_validate_wm_level(level, &max, &wm))
                        break;

                pipe_wm->wm[level] = wm;
        }

        return true;
}

/*
 * Merge the watermarks from all active pipes for a specific level.
 */
static void ilk_merge_wm_level(struct drm_device *dev,
                               int level,
                               struct intel_wm_level *ret_wm)
{
        const struct intel_crtc *intel_crtc;

        ret_wm->enable = true;

        for_each_intel_crtc(dev, intel_crtc) {
                const struct intel_pipe_wm *active = &intel_crtc->wm.active;
                const struct intel_wm_level *wm = &active->wm[level];

                if (!active->pipe_enabled)
                        continue;

                /*
                 * The watermark values may have been used in the past,
                 * so we must maintain them in the registers for some
                 * time even if the level is now disabled.
                 */
                if (!wm->enable)
                        ret_wm->enable = false;

                ret_wm->pri_val = max(ret_wm->pri_val, wm->pri_val);
                ret_wm->spr_val = max(ret_wm->spr_val, wm->spr_val);
                ret_wm->cur_val = max(ret_wm->cur_val, wm->cur_val);
                ret_wm->fbc_val = max(ret_wm->fbc_val, wm->fbc_val);
        }
}

/*
 * Merge all low power watermarks for all active pipes.
 */
static void ilk_wm_merge(struct drm_device *dev,
                         const struct intel_wm_config *config,
                         const struct ilk_wm_maximums *max,
                         struct intel_pipe_wm *merged)
{
        struct drm_i915_private *dev_priv = dev->dev_private;
        int level, max_level = ilk_wm_max_level(dev);
        int last_enabled_level = max_level;

        /* ILK/SNB/IVB: LP1+ watermarks only w/ single pipe */
        if ((INTEL_INFO(dev)->gen <= 6 || IS_IVYBRIDGE(dev)) &&
            config->num_pipes_active > 1)
                return;

        /* ILK: FBC WM must be disabled always */
        merged->fbc_wm_enabled = INTEL_INFO(dev)->gen >= 6;

        /* merge each WM1+ level */
        for (level = 1; level <= max_level; level++) {
                struct intel_wm_level *wm = &merged->wm[level];

                ilk_merge_wm_level(dev, level, wm);

                if (level > last_enabled_level)
                        wm->enable = false;
                else if (!ilk_validate_wm_level(level, max, wm))
                        /* make sure all following levels get disabled */
                        last_enabled_level = level - 1;

                /*
                 * The spec says it is preferred to disable
                 * FBC WMs instead of disabling a WM level.
                 */
                if (wm->fbc_val > max->fbc) {
                        if (wm->enable)
                                merged->fbc_wm_enabled = false;
                        wm->fbc_val = 0;
                }
        }

        /* ILK: LP2+ must be disabled when FBC WM is disabled but FBC enabled */
        /*
         * FIXME this is racy. FBC might get enabled later.
         * What we should check here is whether FBC can be
         * enabled sometime later.
         */
        if (IS_GEN5(dev) && !merged->fbc_wm_enabled &&
            intel_fbc_enabled(dev_priv)) {
                for (level = 2; level <= max_level; level++) {
                        struct intel_wm_level *wm = &merged->wm[level];

                        wm->enable = false;
                }
        }
}

static int ilk_wm_lp_to_level(int wm_lp, const struct intel_pipe_wm *pipe_wm)
{
        /* LP1,LP2,LP3 levels are either 1,2,3 or 1,3,4 */
        return wm_lp + (wm_lp >= 2 && pipe_wm->wm[4].enable);
}

/* The value we need to program into the WM_LPx latency field */
static unsigned int ilk_wm_lp_latency(struct drm_device *dev, int level)
{
        struct drm_i915_private *dev_priv = dev->dev_private;

        if (IS_HASWELL(dev) || IS_BROADWELL(dev))
                return 2 * level;
        else
                return dev_priv->wm.pri_latency[level];
}

static void ilk_compute_wm_results(struct drm_device *dev,
                                   const struct intel_pipe_wm *merged,
                                   enum intel_ddb_partitioning partitioning,
                                   struct ilk_wm_values *results)
{
        struct intel_crtc *intel_crtc;
        int level, wm_lp;

        results->enable_fbc_wm = merged->fbc_wm_enabled;
        results->partitioning = partitioning;

        /* LP1+ register values */
        for (wm_lp = 1; wm_lp <= 3; wm_lp++) {
                const struct intel_wm_level *r;

                level = ilk_wm_lp_to_level(wm_lp, merged);

                r = &merged->wm[level];

                /*
                 * Maintain the watermark values even if the level is
                 * disabled. Doing otherwise could cause underruns.
                 */
                results->wm_lp[wm_lp - 1] =
                        (ilk_wm_lp_latency(dev, level) << WM1_LP_LATENCY_SHIFT) |
                        (r->pri_val << WM1_LP_SR_SHIFT) |
                        r->cur_val;

                if (r->enable)
                        results->wm_lp[wm_lp - 1] |= WM1_LP_SR_EN;

                if (INTEL_INFO(dev)->gen >= 8)
                        results->wm_lp[wm_lp - 1] |=
                                r->fbc_val << WM1_LP_FBC_SHIFT_BDW;
                else
                        results->wm_lp[wm_lp - 1] |=
                                r->fbc_val << WM1_LP_FBC_SHIFT;

                /*
                 * Always set WM1S_LP_EN when spr_val != 0, even if the
                 * level is disabled. Doing otherwise could cause underruns.
                 */
                if (INTEL_INFO(dev)->gen <= 6 && r->spr_val) {
                        WARN_ON(wm_lp != 1);
                        results->wm_lp_spr[wm_lp - 1] = WM1S_LP_EN | r->spr_val;
                } else
                        results->wm_lp_spr[wm_lp - 1] = r->spr_val;
        }

        /* LP0 register values */
        for_each_intel_crtc(dev, intel_crtc) {
                enum pipe pipe = intel_crtc->pipe;
                const struct intel_wm_level *r =
                        &intel_crtc->wm.active.wm[0];

                if (WARN_ON(!r->enable))
                        continue;

                results->wm_linetime[pipe] = intel_crtc->wm.active.linetime;

                results->wm_pipe[pipe] =
                        (r->pri_val << WM0_PIPE_PLANE_SHIFT) |
                        (r->spr_val << WM0_PIPE_SPRITE_SHIFT) |
                        r->cur_val;
        }
}

/* Find the result with the highest level enabled. Check for enable_fbc_wm in
 * case both are at the same level. Prefer r1 in case they're the same. */
static struct intel_pipe_wm *ilk_find_best_result(struct drm_device *dev,
                                                  struct intel_pipe_wm *r1,
                                                  struct intel_pipe_wm *r2)
{
        int level, max_level = ilk_wm_max_level(dev);
        int level1 = 0, level2 = 0;

        for (level = 1; level <= max_level; level++) {
                if (r1->wm[level].enable)
                        level1 = level;
                if (r2->wm[level].enable)
                        level2 = level;
        }

        if (level1 == level2) {
                if (r2->fbc_wm_enabled && !r1->fbc_wm_enabled)
                        return r2;
                else
                        return r1;
        } else if (level1 > level2) {
                return r1;
        } else {
                return r2;
        }
}

/* dirty bits used to track which watermarks need changes */
#define WM_DIRTY_PIPE(pipe) (1 << (pipe))
#define WM_DIRTY_LINETIME(pipe) (1 << (8 + (pipe)))
#define WM_DIRTY_LP(wm_lp) (1 << (15 + (wm_lp)))
#define WM_DIRTY_LP_ALL (WM_DIRTY_LP(1) | WM_DIRTY_LP(2) | WM_DIRTY_LP(3))
#define WM_DIRTY_FBC (1 << 24)
#define WM_DIRTY_DDB (1 << 25)

static unsigned int ilk_compute_wm_dirty(struct drm_i915_private *dev_priv,
                                         const struct ilk_wm_values *old,
                                         const struct ilk_wm_values *new)
{
        unsigned int dirty = 0;
        enum pipe pipe;
        int wm_lp;

        for_each_pipe(dev_priv, pipe) {
                if (old->wm_linetime[pipe] != new->wm_linetime[pipe]) {
                        dirty |= WM_DIRTY_LINETIME(pipe);
                        /* Must disable LP1+ watermarks too */
                        dirty |= WM_DIRTY_LP_ALL;
                }

                if (old->wm_pipe[pipe] != new->wm_pipe[pipe]) {
                        dirty |= WM_DIRTY_PIPE(pipe);
                        /* Must disable LP1+ watermarks too */
                        dirty |= WM_DIRTY_LP_ALL;
                }
        }

        if (old->enable_fbc_wm != new->enable_fbc_wm) {
                dirty |= WM_DIRTY_FBC;
                /* Must disable LP1+ watermarks too */
                dirty |= WM_DIRTY_LP_ALL;
        }

        if (old->partitioning != new->partitioning) {
                dirty |= WM_DIRTY_DDB;
                /* Must disable LP1+ watermarks too */
                dirty |= WM_DIRTY_LP_ALL;
        }

        /* LP1+ watermarks already deemed dirty, no need to continue */
        if (dirty & WM_DIRTY_LP_ALL)
                return dirty;

        /* Find the lowest numbered LP1+ watermark in need of an update... */
        for (wm_lp = 1; wm_lp <= 3; wm_lp++) {
                if (old->wm_lp[wm_lp - 1] != new->wm_lp[wm_lp - 1] ||
                    old->wm_lp_spr[wm_lp - 1] != new->wm_lp_spr[wm_lp - 1])
                        break;
        }

        /* ...and mark it and all higher numbered LP1+ watermarks as dirty */
        for (; wm_lp <= 3; wm_lp++)
                dirty |= WM_DIRTY_LP(wm_lp);

        return dirty;
}

static bool _ilk_disable_lp_wm(struct drm_i915_private *dev_priv,
                               unsigned int dirty)
{
        struct ilk_wm_values *previous = &dev_priv->wm.hw;
        bool changed = false;

        if (dirty & WM_DIRTY_LP(3) && previous->wm_lp[2] & WM1_LP_SR_EN) {
                previous->wm_lp[2] &= ~WM1_LP_SR_EN;
                I915_WRITE(WM3_LP_ILK, previous->wm_lp[2]);
                changed = true;
        }
        if (dirty & WM_DIRTY_LP(2) && previous->wm_lp[1] & WM1_LP_SR_EN) {
                previous->wm_lp[1] &= ~WM1_LP_SR_EN;
                I915_WRITE(WM2_LP_ILK, previous->wm_lp[1]);
                changed = true;
        }
        if (dirty & WM_DIRTY_LP(1) && previous->wm_lp[0] & WM1_LP_SR_EN) {
                previous->wm_lp[0] &= ~WM1_LP_SR_EN;
                I915_WRITE(WM1_LP_ILK, previous->wm_lp[0]);
                changed = true;
        }

        /*
         * Don't touch WM1S_LP_EN here.
         * Doing so could cause underruns.
         */

        return changed;
}

/*
 * The spec says we shouldn't write when we don't need, because every write
 * causes WMs to be re-evaluated, expending some power.
 */
static void ilk_write_wm_values(struct drm_i915_private *dev_priv,
                                struct ilk_wm_values *results)
{
        struct drm_device *dev = dev_priv->dev;
        struct ilk_wm_values *previous = &dev_priv->wm.hw;
        unsigned int dirty;
        uint32_t val;

        dirty = ilk_compute_wm_dirty(dev_priv, previous, results);
        if (!dirty)
                return;

        _ilk_disable_lp_wm(dev_priv, dirty);

        if (dirty & WM_DIRTY_PIPE(PIPE_A))
                I915_WRITE(WM0_PIPEA_ILK, results->wm_pipe[0]);
        if (dirty & WM_DIRTY_PIPE(PIPE_B))
                I915_WRITE(WM0_PIPEB_ILK, results->wm_pipe[1]);
        if (dirty & WM_DIRTY_PIPE(PIPE_C))
                I915_WRITE(WM0_PIPEC_IVB, results->wm_pipe[2]);

        if (dirty & WM_DIRTY_LINETIME(PIPE_A))
                I915_WRITE(PIPE_WM_LINETIME(PIPE_A), results->wm_linetime[0]);
        if (dirty & WM_DIRTY_LINETIME(PIPE_B))
                I915_WRITE(PIPE_WM_LINETIME(PIPE_B), results->wm_linetime[1]);
        if (dirty & WM_DIRTY_LINETIME(PIPE_C))
                I915_WRITE(PIPE_WM_LINETIME(PIPE_C), results->wm_linetime[2]);

        if (dirty & WM_DIRTY_DDB) {
                if (IS_HASWELL(dev) || IS_BROADWELL(dev)) {
                        val = I915_READ(WM_MISC);
                        if (results->partitioning == INTEL_DDB_PART_1_2)
                                val &= ~WM_MISC_DATA_PARTITION_5_6;
                        else
                                val |= WM_MISC_DATA_PARTITION_5_6;
                        I915_WRITE(WM_MISC, val);
                } else {
                        val = I915_READ(DISP_ARB_CTL2);
                        if (results->partitioning == INTEL_DDB_PART_1_2)
                                val &= ~DISP_DATA_PARTITION_5_6;
                        else
                                val |= DISP_DATA_PARTITION_5_6;
                        I915_WRITE(DISP_ARB_CTL2, val);
                }
        }

        if (dirty & WM_DIRTY_FBC) {
                val = I915_READ(DISP_ARB_CTL);
                if (results->enable_fbc_wm)
                        val &= ~DISP_FBC_WM_DIS;
                else
                        val |= DISP_FBC_WM_DIS;
                I915_WRITE(DISP_ARB_CTL, val);
        }

        if (dirty & WM_DIRTY_LP(1) &&
            previous->wm_lp_spr[0] != results->wm_lp_spr[0])
                I915_WRITE(WM1S_LP_ILK, results->wm_lp_spr[0]);

        if (INTEL_INFO(dev)->gen >= 7) {
                if (dirty & WM_DIRTY_LP(2) && previous->wm_lp_spr[1] != results->wm_lp_spr[1])
                        I915_WRITE(WM2S_LP_IVB, results->wm_lp_spr[1]);
                if (dirty & WM_DIRTY_LP(3) && previous->wm_lp_spr[2] != results->wm_lp_spr[2])
                        I915_WRITE(WM3S_LP_IVB, results->wm_lp_spr[2]);
        }

        if (dirty & WM_DIRTY_LP(1) && previous->wm_lp[0] != results->wm_lp[0])
                I915_WRITE(WM1_LP_ILK, results->wm_lp[0]);
        if (dirty & WM_DIRTY_LP(2) && previous->wm_lp[1] != results->wm_lp[1])
                I915_WRITE(WM2_LP_ILK, results->wm_lp[1]);
        if (dirty & WM_DIRTY_LP(3) && previous->wm_lp[2] != results->wm_lp[2])
                I915_WRITE(WM3_LP_ILK, results->wm_lp[2]);

        dev_priv->wm.hw = *results;
}

static bool ilk_disable_lp_wm(struct drm_device *dev)
{
        struct drm_i915_private *dev_priv = dev->dev_private;

        return _ilk_disable_lp_wm(dev_priv, WM_DIRTY_LP_ALL);
}

/*
 * On gen9, we need to allocate Display Data Buffer (DDB) portions to the
 * different active planes.
 */

#define SKL_DDB_SIZE            896     /* in blocks */
#define BXT_DDB_SIZE            512

static void
skl_ddb_get_pipe_allocation_limits(struct drm_device *dev,
                                   struct drm_crtc *for_crtc,
                                   const struct intel_wm_config *config,
                                   const struct skl_pipe_wm_parameters *params,
                                   struct skl_ddb_entry *alloc /* out */)
{
        struct drm_crtc *crtc;
        unsigned int pipe_size, ddb_size;
        int nth_active_pipe;

        if (!params->active) {
                alloc->start = 0;
                alloc->end = 0;
                return;
        }

        if (IS_BROXTON(dev))
                ddb_size = BXT_DDB_SIZE;
        else
                ddb_size = SKL_DDB_SIZE;

        ddb_size -= 4; /* 4 blocks for bypass path allocation */

        nth_active_pipe = 0;
        for_each_crtc(dev, crtc) {
                if (!to_intel_crtc(crtc)->active)
                        continue;

                if (crtc == for_crtc)
                        break;

                nth_active_pipe++;
        }

        pipe_size = ddb_size / config->num_pipes_active;
        alloc->start = nth_active_pipe * ddb_size / config->num_pipes_active;
        alloc->end = alloc->start + pipe_size;
}

static unsigned int skl_cursor_allocation(const struct intel_wm_config *config)
{
        if (config->num_pipes_active == 1)
                return 32;

        return 8;
}

static void skl_ddb_entry_init_from_hw(struct skl_ddb_entry *entry, u32 reg)
{
        entry->start = reg & 0x3ff;
        entry->end = (reg >> 16) & 0x3ff;
        if (entry->end)
                entry->end += 1;
}

void skl_ddb_get_hw_state(struct drm_i915_private *dev_priv,
                          struct skl_ddb_allocation *ddb /* out */)
{
        enum pipe pipe;
        int plane;
        u32 val;

        memset(ddb, 0, sizeof(*ddb));

        for_each_pipe(dev_priv, pipe) {
                if (!intel_display_power_is_enabled(dev_priv, POWER_DOMAIN_PIPE(pipe)))
                        continue;

                for_each_plane(dev_priv, pipe, plane) {
                        val = I915_READ(PLANE_BUF_CFG(pipe, plane));
                        skl_ddb_entry_init_from_hw(&ddb->plane[pipe][plane],
                                                   val);
                }

                val = I915_READ(CUR_BUF_CFG(pipe));
                skl_ddb_entry_init_from_hw(&ddb->plane[pipe][PLANE_CURSOR],
                                           val);
        }
}

static unsigned int
skl_plane_relative_data_rate(const struct intel_plane_wm_parameters *p, int y)
{

        /* for planar format */
        if (p->y_bytes_per_pixel) {
                if (y)  /* y-plane data rate */
                        return p->horiz_pixels * p->vert_pixels * p->y_bytes_per_pixel;
                else    /* uv-plane data rate */
                        return (p->horiz_pixels/2) * (p->vert_pixels/2) * p->bytes_per_pixel;
        }

        /* for packed formats */
        return p->horiz_pixels * p->vert_pixels * p->bytes_per_pixel;
}

/*
 * We don't overflow 32 bits. Worst case is 3 planes enabled, each fetching
 * a 8192x4096@32bpp framebuffer:
 *   3 * 4096 * 8192  * 4 < 2^32
 */
static unsigned int
skl_get_total_relative_data_rate(struct intel_crtc *intel_crtc,
                                 const struct skl_pipe_wm_parameters *params)
{
        unsigned int total_data_rate = 0;
        int plane;

        for (plane = 0; plane < intel_num_planes(intel_crtc); plane++) {
                const struct intel_plane_wm_parameters *p;

                p = &params->plane[plane];
                if (!p->enabled)
                        continue;

                total_data_rate += skl_plane_relative_data_rate(p, 0); /* packed/uv */
                if (p->y_bytes_per_pixel) {
                        total_data_rate += skl_plane_relative_data_rate(p, 1); /* y-plane */
                }
        }

        return total_data_rate;
}

static void
skl_allocate_pipe_ddb(struct drm_crtc *crtc,
                      const struct intel_wm_config *config,
                      const struct skl_pipe_wm_parameters *params,
                      struct skl_ddb_allocation *ddb /* out */)
{
        struct drm_device *dev = crtc->dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
        enum pipe pipe = intel_crtc->pipe;
        struct skl_ddb_entry *alloc = &ddb->pipe[pipe];
        uint16_t alloc_size, start, cursor_blocks;
        uint16_t minimum[I915_MAX_PLANES];
        uint16_t y_minimum[I915_MAX_PLANES];
        unsigned int total_data_rate;
        int plane;

        skl_ddb_get_pipe_allocation_limits(dev, crtc, config, params, alloc);
        alloc_size = skl_ddb_entry_size(alloc);
        if (alloc_size == 0) {
                memset(ddb->plane[pipe], 0, sizeof(ddb->plane[pipe]));
                memset(&ddb->plane[pipe][PLANE_CURSOR], 0,
                       sizeof(ddb->plane[pipe][PLANE_CURSOR]));
                return;
        }

        cursor_blocks = skl_cursor_allocation(config);
        ddb->plane[pipe][PLANE_CURSOR].start = alloc->end - cursor_blocks;
        ddb->plane[pipe][PLANE_CURSOR].end = alloc->end;

        alloc_size -= cursor_blocks;
        alloc->end -= cursor_blocks;

        /* 1. Allocate the mininum required blocks for each active plane */
        for_each_plane(dev_priv, pipe, plane) {
                const struct intel_plane_wm_parameters *p;

                p = &params->plane[plane];
                if (!p->enabled)
                        continue;

                minimum[plane] = 8;
                alloc_size -= minimum[plane];
                y_minimum[plane] = p->y_bytes_per_pixel ? 8 : 0;
                alloc_size -= y_minimum[plane];
        }

        /*
         * 2. Distribute the remaining space in proportion to the amount of
         * data each plane needs to fetch from memory.
         *
         * FIXME: we may not allocate every single block here.
         */
        total_data_rate = skl_get_total_relative_data_rate(intel_crtc, params);

        start = alloc->start;
        for (plane = 0; plane < intel_num_planes(intel_crtc); plane++) {
                const struct intel_plane_wm_parameters *p;
                unsigned int data_rate, y_data_rate;
                uint16_t plane_blocks, y_plane_blocks = 0;

                p = &params->plane[plane];
                if (!p->enabled)
                        continue;

                data_rate = skl_plane_relative_data_rate(p, 0);

                /*
                 * allocation for (packed formats) or (uv-plane part of planar format):
                 * promote the expression to 64 bits to avoid overflowing, the
                 * result is < available as data_rate / total_data_rate < 1
                 */
                plane_blocks = minimum[plane];
                plane_blocks += div_u64((uint64_t)alloc_size * data_rate,
                                        total_data_rate);

                ddb->plane[pipe][plane].start = start;
                ddb->plane[pipe][plane].end = start + plane_blocks;

                start += plane_blocks;

                /*
                 * allocation for y_plane part of planar format:
                 */
                if (p->y_bytes_per_pixel) {
                        y_data_rate = skl_plane_relative_data_rate(p, 1);
                        y_plane_blocks = y_minimum[plane];
                        y_plane_blocks += div_u64((uint64_t)alloc_size * y_data_rate,
                                                total_data_rate);

                        ddb->y_plane[pipe][plane].start = start;
                        ddb->y_plane[pipe][plane].end = start + y_plane_blocks;

                        start += y_plane_blocks;
                }

        }

}

static uint32_t skl_pipe_pixel_rate(const struct intel_crtc_state *config)
{
        /* TODO: Take into account the scalers once we support them */
        return config->base.adjusted_mode.crtc_clock;
}

/*
 * The max latency should be 257 (max the punit can code is 255 and we add 2us
 * for the read latency) and bytes_per_pixel should always be <= 8, so that
 * should allow pixel_rate up to ~2 GHz which seems sufficient since max
 * 2xcdclk is 1350 MHz and the pixel rate should never exceed that.
*/
static uint32_t skl_wm_method1(uint32_t pixel_rate, uint8_t bytes_per_pixel,
                               uint32_t latency)
{
        uint32_t wm_intermediate_val, ret;

        if (latency == 0)