]> git.kernelconcepts.de Git - karo-tx-linux.git/commitdiff
drm/radeon/kms: add dpm support for KB/KV
authorAlex Deucher <alexander.deucher@amd.com>
Wed, 14 Aug 2013 05:01:40 +0000 (01:01 -0400)
committerAlex Deucher <alexander.deucher@amd.com>
Fri, 30 Aug 2013 20:30:28 +0000 (16:30 -0400)
This adds dpm support for KB/KV asics.  This includes:
- dynamic engine clock scaling
- dynamic voltage scaling
- power containment
- shader power scaling

Set radeon.dpm=1 to enable.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
12 files changed:
drivers/gpu/drm/radeon/Makefile
drivers/gpu/drm/radeon/cik.c
drivers/gpu/drm/radeon/cikd.h
drivers/gpu/drm/radeon/kv_dpm.c [new file with mode: 0644]
drivers/gpu/drm/radeon/kv_dpm.h [new file with mode: 0644]
drivers/gpu/drm/radeon/kv_smc.c [new file with mode: 0644]
drivers/gpu/drm/radeon/ppsmc.h
drivers/gpu/drm/radeon/radeon_asic.c
drivers/gpu/drm/radeon/radeon_asic.h
drivers/gpu/drm/radeon/radeon_pm.c
drivers/gpu/drm/radeon/smu7.h [new file with mode: 0644]
drivers/gpu/drm/radeon/smu7_fusion.h [new file with mode: 0644]

index bfabd69b4e3980c145fc8614a3730db74cb65080..d3265b5d466139ee1e4442fc482231e5692765fa 100644 (file)
@@ -79,7 +79,7 @@ radeon-y += radeon_device.o radeon_asic.o radeon_kms.o \
        si_blit_shaders.o radeon_prime.o radeon_uvd.o cik.o cik_blit_shaders.o \
        r600_dpm.o rs780_dpm.o rv6xx_dpm.o rv770_dpm.o rv730_dpm.o rv740_dpm.o \
        rv770_smc.o cypress_dpm.o btc_dpm.o sumo_dpm.o sumo_smc.o trinity_dpm.o \
-       trinity_smc.o ni_dpm.o si_smc.o si_dpm.o
+       trinity_smc.o ni_dpm.o si_smc.o si_dpm.o kv_smc.o kv_dpm.o
 
 radeon-$(CONFIG_COMPAT) += radeon_ioc32.o
 radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o
index d0804f79efed538c2bc51ab91a1521e9ff467ccc..87e5aeed6e888b66c2edefd21b600f3e6e7b087f 100644 (file)
@@ -6593,6 +6593,7 @@ int cik_irq_set(struct radeon_device *rdev)
        u32 hpd1, hpd2, hpd3, hpd4, hpd5, hpd6;
        u32 grbm_int_cntl = 0;
        u32 dma_cntl, dma_cntl1;
+       u32 thermal_int;
 
        if (!rdev->irq.installed) {
                WARN(1, "Can't enable IRQ/MSI because no handler is installed\n");
@@ -6625,6 +6626,9 @@ int cik_irq_set(struct radeon_device *rdev)
        cp_m2p2 = RREG32(CP_ME2_PIPE2_INT_CNTL) & ~TIME_STAMP_INT_ENABLE;
        cp_m2p3 = RREG32(CP_ME2_PIPE3_INT_CNTL) & ~TIME_STAMP_INT_ENABLE;
 
+       thermal_int = RREG32_SMC(CG_THERMAL_INT_CTRL) &
+               ~(THERM_INTH_MASK | THERM_INTL_MASK);
+
        /* enable CP interrupts on all rings */
        if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) {
                DRM_DEBUG("cik_irq_set: sw int gfx\n");
@@ -6782,6 +6786,11 @@ int cik_irq_set(struct radeon_device *rdev)
                hpd6 |= DC_HPDx_INT_EN;
        }
 
+       if (rdev->irq.dpm_thermal) {
+               DRM_DEBUG("dpm thermal\n");
+               thermal_int |= THERM_INTH_MASK | THERM_INTL_MASK;
+       }
+
        WREG32(CP_INT_CNTL_RING0, cp_int_cntl);
 
        WREG32(SDMA0_CNTL + SDMA0_REGISTER_OFFSET, dma_cntl);
@@ -6816,6 +6825,8 @@ int cik_irq_set(struct radeon_device *rdev)
        WREG32(DC_HPD5_INT_CONTROL, hpd5);
        WREG32(DC_HPD6_INT_CONTROL, hpd6);
 
+       WREG32_SMC(CG_THERMAL_INT_CTRL, thermal_int);
+
        return 0;
 }
 
@@ -7027,6 +7038,7 @@ int cik_irq_process(struct radeon_device *rdev)
        bool queue_hotplug = false;
        bool queue_reset = false;
        u32 addr, status, mc_client;
+       bool queue_thermal = false;
 
        if (!rdev->ih.enabled || rdev->shutdown)
                return IRQ_NONE;
@@ -7377,6 +7389,19 @@ restart_ih:
                                break;
                        }
                        break;
+               case 230: /* thermal low to high */
+                       DRM_DEBUG("IH: thermal low to high\n");
+                       rdev->pm.dpm.thermal.high_to_low = false;
+                       queue_thermal = true;
+                       break;
+               case 231: /* thermal high to low */
+                       DRM_DEBUG("IH: thermal high to low\n");
+                       rdev->pm.dpm.thermal.high_to_low = true;
+                       queue_thermal = true;
+                       break;
+               case 233: /* GUI IDLE */
+                       DRM_DEBUG("IH: GUI idle\n");
+                       break;
                case 241: /* SDMA Privileged inst */
                case 247: /* SDMA Privileged inst */
                        DRM_ERROR("Illegal instruction in SDMA command stream\n");
@@ -7416,9 +7441,6 @@ restart_ih:
                                break;
                        }
                        break;
-               case 233: /* GUI IDLE */
-                       DRM_DEBUG("IH: GUI idle\n");
-                       break;
                default:
                        DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
                        break;
@@ -7432,6 +7454,8 @@ restart_ih:
                schedule_work(&rdev->hotplug_work);
        if (queue_reset)
                schedule_work(&rdev->reset_work);
+       if (queue_thermal)
+               schedule_work(&rdev->pm.dpm.thermal.work);
        rdev->ih.rptr = rptr;
        WREG32(IH_RB_RPTR, rdev->ih.rptr);
        atomic_set(&rdev->ih.lock, 0);
index 65886caaf75639d2c2b233d5b4b39855036d4afb..179ca3625ae4d75c16aa1c99e8e81f7bf4697176 100644 (file)
 
 #define CIK_RB_BITMAP_WIDTH_PER_SH  2
 
+/* DIDT IND registers */
+#define DIDT_SQ_CTRL0                                     0x0
+#       define DIDT_CTRL_EN                               (1 << 0)
+#define DIDT_DB_CTRL0                                     0x20
+#define DIDT_TD_CTRL0                                     0x40
+#define DIDT_TCP_CTRL0                                    0x60
+
 /* SMC IND registers */
+#define NB_DPM_CONFIG_1                                   0x3F9E8
+#       define Dpm0PgNbPsLo(x)                            ((x) << 0)
+#       define Dpm0PgNbPsLo_MASK                          0x000000ff
+#       define Dpm0PgNbPsLo_SHIFT                         0
+#       define Dpm0PgNbPsHi(x)                            ((x) << 8)
+#       define Dpm0PgNbPsHi_MASK                          0x0000ff00
+#       define Dpm0PgNbPsHi_SHIFT                         8
+#       define DpmXNbPsLo(x)                              ((x) << 16)
+#       define DpmXNbPsLo_MASK                            0x00ff0000
+#       define DpmXNbPsLo_SHIFT                           16
+#       define DpmXNbPsHi(x)                              ((x) << 24)
+#       define DpmXNbPsHi_MASK                            0xff000000
+#       define DpmXNbPsHi_SHIFT                           24
+
+#define SMC_SYSCON_MSG_ARG_0                              0x80000068
+
 #define GENERAL_PWRMGT                                    0xC0200000
+#       define GLOBAL_PWRMGT_EN                           (1 << 0)
 #       define GPU_COUNTER_CLK                            (1 << 15)
 
+#define SCLK_PWRMGT_CNTL                                  0xC0200008
+#       define RESET_BUSY_CNT                             (1 << 4)
+#       define RESET_SCLK_CNT                             (1 << 5)
+#       define DYNAMIC_PM_EN                              (1 << 21)
+
+#define CG_FTV_0                                          0xC02001A8
+
+#define LCAC_SX0_OVR_SEL                                  0xC0400D04
+#define LCAC_SX0_OVR_VAL                                  0xC0400D08
+
+#define LCAC_MC0_OVR_SEL                                  0xC0400D34
+#define LCAC_MC0_OVR_VAL                                  0xC0400D38
+
+#define LCAC_MC1_OVR_SEL                                  0xC0400D40
+#define LCAC_MC1_OVR_VAL                                  0xC0400D44
+
+#define LCAC_MC2_OVR_SEL                                  0xC0400D4C
+#define LCAC_MC2_OVR_VAL                                  0xC0400D50
+
+#define LCAC_MC3_OVR_SEL                                  0xC0400D58
+#define LCAC_MC3_OVR_VAL                                  0xC0400D5C
+
+#define LCAC_CPL_OVR_SEL                                  0xC0400D84
+#define LCAC_CPL_OVR_VAL                                  0xC0400D88
+
 #define        CG_MULT_THERMAL_STATUS                          0xC0300014
 #define                ASIC_MAX_TEMP(x)                        ((x) << 0)
 #define                ASIC_MAX_TEMP_MASK                      0x000001ff
 #      define ZCLK_SEL(x)                              ((x) << 8)
 #      define ZCLK_SEL_MASK                            0xFF00
 
+#define        CG_THERMAL_INT_CTRL                             0xC2100028
+#define                DIG_THERM_INTH(x)                       ((x) << 0)
+#define                DIG_THERM_INTH_MASK                     0x000000FF
+#define                DIG_THERM_INTH_SHIFT                    0
+#define                DIG_THERM_INTL(x)                       ((x) << 8)
+#define                DIG_THERM_INTL_MASK                     0x0000FF00
+#define                DIG_THERM_INTL_SHIFT                    8
+#define        THERM_INTH_MASK                         (1 << 24)
+#define        THERM_INTL_MASK                         (1 << 25)
+
 /* PCIE registers idx/data 0x38/0x3c */
 #define PB0_PIF_PWRDOWN_0                                 0x1100012 /* PCIE */
 #       define PLL_POWER_STATE_IN_TXS2_0(x)               ((x) << 7)
 #define PCIE_INDEX                                     0x38
 #define PCIE_DATA                                      0x3C
 
+#define SMC_IND_INDEX_0                                0x200
+#define SMC_IND_DATA_0                                 0x204
+
+#define SMC_IND_ACCESS_CNTL                            0x240
+#define                AUTO_INCREMENT_IND_0                    (1 << 0)
+
+#define SMC_MESSAGE_0                                          0x250
+#define                SMC_MSG_MASK                            0xffff
+#define SMC_RESP_0                                     0x254
+#define                SMC_RESP_MASK                           0xffff
+
+#define SMC_MSG_ARG_0                                          0x290
+
 #define VGA_HDP_CONTROL                                0x328
 #define                VGA_MEMORY_DISABLE                              (1 << 4)
 
diff --git a/drivers/gpu/drm/radeon/kv_dpm.c b/drivers/gpu/drm/radeon/kv_dpm.c
new file mode 100644 (file)
index 0000000..2e40163
--- /dev/null
@@ -0,0 +1,2536 @@
+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "radeon.h"
+#include "cikd.h"
+#include "r600_dpm.h"
+#include "kv_dpm.h"
+
+#define KV_MAX_DEEPSLEEP_DIVIDER_ID     5
+#define KV_MINIMUM_ENGINE_CLOCK         800
+#define SMC_RAM_END                     0x40000
+
+static void kv_init_graphics_levels(struct radeon_device *rdev);
+static int kv_calculate_ds_divider(struct radeon_device *rdev);
+static int kv_calculate_nbps_level_settings(struct radeon_device *rdev);
+static int kv_calculate_dpm_settings(struct radeon_device *rdev);
+static void kv_enable_new_levels(struct radeon_device *rdev);
+static void kv_program_nbps_index_settings(struct radeon_device *rdev,
+                                          struct radeon_ps *new_rps);
+static int kv_set_enabled_levels(struct radeon_device *rdev);
+static int kv_force_dpm_lowest(struct radeon_device *rdev);
+static void kv_apply_state_adjust_rules(struct radeon_device *rdev,
+                                       struct radeon_ps *new_rps,
+                                       struct radeon_ps *old_rps);
+static int kv_set_thermal_temperature_range(struct radeon_device *rdev,
+                                           int min_temp, int max_temp);
+static int kv_init_fps_limits(struct radeon_device *rdev);
+
+static void kv_dpm_powergate_uvd(struct radeon_device *rdev, bool gate);
+static void kv_dpm_powergate_vce(struct radeon_device *rdev, bool gate);
+static void kv_dpm_powergate_samu(struct radeon_device *rdev, bool gate);
+static void kv_dpm_powergate_acp(struct radeon_device *rdev, bool gate);
+
+extern void cik_enter_rlc_safe_mode(struct radeon_device *rdev);
+extern void cik_exit_rlc_safe_mode(struct radeon_device *rdev);
+extern void cik_update_cg(struct radeon_device *rdev,
+                         u32 block, bool enable);
+
+static const struct kv_lcac_config_values sx_local_cac_cfg_kv[] =
+{
+       {  0,       4,        1    },
+       {  1,       4,        1    },
+       {  2,       5,        1    },
+       {  3,       4,        2    },
+       {  4,       1,        1    },
+       {  5,       5,        2    },
+       {  6,       6,        1    },
+       {  7,       9,        2    },
+       { 0xffffffff }
+};
+
+static const struct kv_lcac_config_values mc0_local_cac_cfg_kv[] =
+{
+       {  0,       4,        1    },
+       { 0xffffffff }
+};
+
+static const struct kv_lcac_config_values mc1_local_cac_cfg_kv[] =
+{
+       {  0,       4,        1    },
+       { 0xffffffff }
+};
+
+static const struct kv_lcac_config_values mc2_local_cac_cfg_kv[] =
+{
+       {  0,       4,        1    },
+       { 0xffffffff }
+};
+
+static const struct kv_lcac_config_values mc3_local_cac_cfg_kv[] =
+{
+       {  0,       4,        1    },
+       { 0xffffffff }
+};
+
+static const struct kv_lcac_config_values cpl_local_cac_cfg_kv[] =
+{
+       {  0,       4,        1    },
+       {  1,       4,        1    },
+       {  2,       5,        1    },
+       {  3,       4,        1    },
+       {  4,       1,        1    },
+       {  5,       5,        1    },
+       {  6,       6,        1    },
+       {  7,       9,        1    },
+       {  8,       4,        1    },
+       {  9,       2,        1    },
+       {  10,      3,        1    },
+       {  11,      6,        1    },
+       {  12,      8,        2    },
+       {  13,      1,        1    },
+       {  14,      2,        1    },
+       {  15,      3,        1    },
+       {  16,      1,        1    },
+       {  17,      4,        1    },
+       {  18,      3,        1    },
+       {  19,      1,        1    },
+       {  20,      8,        1    },
+       {  21,      5,        1    },
+       {  22,      1,        1    },
+       {  23,      1,        1    },
+       {  24,      4,        1    },
+       {  27,      6,        1    },
+       {  28,      1,        1    },
+       { 0xffffffff }
+};
+
+static const struct kv_lcac_config_reg sx0_cac_config_reg[] =
+{
+       { 0xc0400d00, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_lcac_config_reg mc0_cac_config_reg[] =
+{
+       { 0xc0400d30, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_lcac_config_reg mc1_cac_config_reg[] =
+{
+       { 0xc0400d3c, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_lcac_config_reg mc2_cac_config_reg[] =
+{
+       { 0xc0400d48, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_lcac_config_reg mc3_cac_config_reg[] =
+{
+       { 0xc0400d54, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_lcac_config_reg cpl_cac_config_reg[] =
+{
+       { 0xc0400d80, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_pt_config_reg didt_config_kv[] =
+{
+       { 0x10, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x10, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x10, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x10, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x11, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x11, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x11, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x11, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x12, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x12, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x12, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x12, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x2, 0x00003fff, 0, 0x4, KV_CONFIGREG_DIDT_IND },
+       { 0x2, 0x03ff0000, 16, 0x80, KV_CONFIGREG_DIDT_IND },
+       { 0x2, 0x78000000, 27, 0x3, KV_CONFIGREG_DIDT_IND },
+       { 0x1, 0x0000ffff, 0, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+       { 0x1, 0xffff0000, 16, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+       { 0x0, 0x00000001, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x30, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x30, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x30, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x30, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x31, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x31, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x31, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x31, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x32, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x32, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x32, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x32, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x22, 0x00003fff, 0, 0x4, KV_CONFIGREG_DIDT_IND },
+       { 0x22, 0x03ff0000, 16, 0x80, KV_CONFIGREG_DIDT_IND },
+       { 0x22, 0x78000000, 27, 0x3, KV_CONFIGREG_DIDT_IND },
+       { 0x21, 0x0000ffff, 0, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+       { 0x21, 0xffff0000, 16, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+       { 0x20, 0x00000001, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x50, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x50, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x50, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x50, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x51, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x51, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x51, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x51, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x52, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x52, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x52, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x52, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x42, 0x00003fff, 0, 0x4, KV_CONFIGREG_DIDT_IND },
+       { 0x42, 0x03ff0000, 16, 0x80, KV_CONFIGREG_DIDT_IND },
+       { 0x42, 0x78000000, 27, 0x3, KV_CONFIGREG_DIDT_IND },
+       { 0x41, 0x0000ffff, 0, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+       { 0x41, 0xffff0000, 16, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+       { 0x40, 0x00000001, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x70, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x70, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x70, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x70, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x71, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x71, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x71, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x71, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x72, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x72, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x72, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x72, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0x62, 0x00003fff, 0, 0x4, KV_CONFIGREG_DIDT_IND },
+       { 0x62, 0x03ff0000, 16, 0x80, KV_CONFIGREG_DIDT_IND },
+       { 0x62, 0x78000000, 27, 0x3, KV_CONFIGREG_DIDT_IND },
+       { 0x61, 0x0000ffff, 0, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+       { 0x61, 0xffff0000, 16, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+       { 0x60, 0x00000001, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+       { 0xFFFFFFFF }
+};
+
+static struct kv_ps *kv_get_ps(struct radeon_ps *rps)
+{
+       struct kv_ps *ps = rps->ps_priv;
+
+       return ps;
+}
+
+static struct kv_power_info *kv_get_pi(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = rdev->pm.dpm.priv;
+
+       return pi;
+}
+
+#if 0
+static void kv_program_local_cac_table(struct radeon_device *rdev,
+                                      const struct kv_lcac_config_values *local_cac_table,
+                                      const struct kv_lcac_config_reg *local_cac_reg)
+{
+       u32 i, count, data;
+       const struct kv_lcac_config_values *values = local_cac_table;
+
+       while (values->block_id != 0xffffffff) {
+               count = values->signal_id;
+               for (i = 0; i < count; i++) {
+                       data = ((values->block_id << local_cac_reg->block_shift) &
+                               local_cac_reg->block_mask);
+                       data |= ((i << local_cac_reg->signal_shift) &
+                                local_cac_reg->signal_mask);
+                       data |= ((values->t << local_cac_reg->t_shift) &
+                                local_cac_reg->t_mask);
+                       data |= ((1 << local_cac_reg->enable_shift) &
+                                local_cac_reg->enable_mask);
+                       WREG32_SMC(local_cac_reg->cntl, data);
+               }
+               values++;
+       }
+}
+#endif
+
+static int kv_program_pt_config_registers(struct radeon_device *rdev,
+                                         const struct kv_pt_config_reg *cac_config_regs)
+{
+       const struct kv_pt_config_reg *config_regs = cac_config_regs;
+       u32 data;
+       u32 cache = 0;
+
+       if (config_regs == NULL)
+               return -EINVAL;
+
+       while (config_regs->offset != 0xFFFFFFFF) {
+               if (config_regs->type == KV_CONFIGREG_CACHE) {
+                       cache |= ((config_regs->value << config_regs->shift) & config_regs->mask);
+               } else {
+                       switch (config_regs->type) {
+                       case KV_CONFIGREG_SMC_IND:
+                               data = RREG32_SMC(config_regs->offset);
+                               break;
+                       case KV_CONFIGREG_DIDT_IND:
+                               data = RREG32_DIDT(config_regs->offset);
+                               break;
+                       default:
+                               data = RREG32(config_regs->offset << 2);
+                               break;
+                       }
+
+                       data &= ~config_regs->mask;
+                       data |= ((config_regs->value << config_regs->shift) & config_regs->mask);
+                       data |= cache;
+                       cache = 0;
+
+                       switch (config_regs->type) {
+                       case KV_CONFIGREG_SMC_IND:
+                               WREG32_SMC(config_regs->offset, data);
+                               break;
+                       case KV_CONFIGREG_DIDT_IND:
+                               WREG32_DIDT(config_regs->offset, data);
+                               break;
+                       default:
+                               WREG32(config_regs->offset << 2, data);
+                               break;
+                       }
+               }
+               config_regs++;
+       }
+
+       return 0;
+}
+
+static void kv_do_enable_didt(struct radeon_device *rdev, bool enable)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 data;
+
+       if (pi->caps_sq_ramping) {
+               data = RREG32_DIDT(DIDT_SQ_CTRL0);
+               if (enable)
+                       data |= DIDT_CTRL_EN;
+               else
+                       data &= ~DIDT_CTRL_EN;
+               WREG32_DIDT(DIDT_SQ_CTRL0, data);
+       }
+
+       if (pi->caps_db_ramping) {
+               data = RREG32_DIDT(DIDT_DB_CTRL0);
+               if (enable)
+                       data |= DIDT_CTRL_EN;
+               else
+                       data &= ~DIDT_CTRL_EN;
+               WREG32_DIDT(DIDT_DB_CTRL0, data);
+       }
+
+       if (pi->caps_td_ramping) {
+               data = RREG32_DIDT(DIDT_TD_CTRL0);
+               if (enable)
+                       data |= DIDT_CTRL_EN;
+               else
+                       data &= ~DIDT_CTRL_EN;
+               WREG32_DIDT(DIDT_TD_CTRL0, data);
+       }
+
+       if (pi->caps_tcp_ramping) {
+               data = RREG32_DIDT(DIDT_TCP_CTRL0);
+               if (enable)
+                       data |= DIDT_CTRL_EN;
+               else
+                       data &= ~DIDT_CTRL_EN;
+               WREG32_DIDT(DIDT_TCP_CTRL0, data);
+       }
+}
+
+static int kv_enable_didt(struct radeon_device *rdev, bool enable)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret;
+
+       if (pi->caps_sq_ramping ||
+           pi->caps_db_ramping ||
+           pi->caps_td_ramping ||
+           pi->caps_tcp_ramping) {
+               cik_enter_rlc_safe_mode(rdev);
+
+               if (enable) {
+                       ret = kv_program_pt_config_registers(rdev, didt_config_kv);
+                       if (ret) {
+                               cik_exit_rlc_safe_mode(rdev);
+                               return ret;
+                       }
+               }
+
+               kv_do_enable_didt(rdev, enable);
+
+               cik_exit_rlc_safe_mode(rdev);
+       }
+
+       return 0;
+}
+
+#if 0
+static void kv_initialize_hardware_cac_manager(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       if (pi->caps_cac) {
+               WREG32_SMC(LCAC_SX0_OVR_SEL, 0);
+               WREG32_SMC(LCAC_SX0_OVR_VAL, 0);
+               kv_program_local_cac_table(rdev, sx_local_cac_cfg_kv, sx0_cac_config_reg);
+
+               WREG32_SMC(LCAC_MC0_OVR_SEL, 0);
+               WREG32_SMC(LCAC_MC0_OVR_VAL, 0);
+               kv_program_local_cac_table(rdev, mc0_local_cac_cfg_kv, mc0_cac_config_reg);
+
+               WREG32_SMC(LCAC_MC1_OVR_SEL, 0);
+               WREG32_SMC(LCAC_MC1_OVR_VAL, 0);
+               kv_program_local_cac_table(rdev, mc1_local_cac_cfg_kv, mc1_cac_config_reg);
+
+               WREG32_SMC(LCAC_MC2_OVR_SEL, 0);
+               WREG32_SMC(LCAC_MC2_OVR_VAL, 0);
+               kv_program_local_cac_table(rdev, mc2_local_cac_cfg_kv, mc2_cac_config_reg);
+
+               WREG32_SMC(LCAC_MC3_OVR_SEL, 0);
+               WREG32_SMC(LCAC_MC3_OVR_VAL, 0);
+               kv_program_local_cac_table(rdev, mc3_local_cac_cfg_kv, mc3_cac_config_reg);
+
+               WREG32_SMC(LCAC_CPL_OVR_SEL, 0);
+               WREG32_SMC(LCAC_CPL_OVR_VAL, 0);
+               kv_program_local_cac_table(rdev, cpl_local_cac_cfg_kv, cpl_cac_config_reg);
+       }
+}
+#endif
+
+static int kv_enable_smc_cac(struct radeon_device *rdev, bool enable)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret = 0;
+
+       if (pi->caps_cac) {
+               if (enable) {
+                       ret = kv_notify_message_to_smu(rdev, PPSMC_MSG_EnableCac);
+                       if (ret)
+                               pi->cac_enabled = false;
+                       else
+                               pi->cac_enabled = true;
+               } else if (pi->cac_enabled) {
+                       kv_notify_message_to_smu(rdev, PPSMC_MSG_DisableCac);
+                       pi->cac_enabled = false;
+               }
+       }
+
+       return ret;
+}
+
+static int kv_process_firmware_header(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 tmp;
+       int ret;
+
+       ret = kv_read_smc_sram_dword(rdev, SMU7_FIRMWARE_HEADER_LOCATION +
+                                    offsetof(SMU7_Firmware_Header, DpmTable),
+                                    &tmp, pi->sram_end);
+
+       if (ret == 0)
+               pi->dpm_table_start = tmp;
+
+       ret = kv_read_smc_sram_dword(rdev, SMU7_FIRMWARE_HEADER_LOCATION +
+                                    offsetof(SMU7_Firmware_Header, SoftRegisters),
+                                    &tmp, pi->sram_end);
+
+       if (ret == 0)
+               pi->soft_regs_start = tmp;
+
+       return ret;
+}
+
+static int kv_enable_dpm_voltage_scaling(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret;
+
+       pi->graphics_voltage_change_enable = 1;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, GraphicsVoltageChangeEnable),
+                                  &pi->graphics_voltage_change_enable,
+                                  sizeof(u8), pi->sram_end);
+
+       return ret;
+}
+
+static int kv_set_dpm_interval(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret;
+
+       pi->graphics_interval = 1;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, GraphicsInterval),
+                                  &pi->graphics_interval,
+                                  sizeof(u8), pi->sram_end);
+
+       return ret;
+}
+
+static int kv_set_dpm_boot_state(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, GraphicsBootLevel),
+                                  &pi->graphics_boot_level,
+                                  sizeof(u8), pi->sram_end);
+
+       return ret;
+}
+
+static void kv_program_vc(struct radeon_device *rdev)
+{
+       WREG32_SMC(CG_FTV_0, 0x3FFFC000);
+}
+
+static void kv_clear_vc(struct radeon_device *rdev)
+{
+       WREG32_SMC(CG_FTV_0, 0);
+}
+
+static int kv_set_divider_value(struct radeon_device *rdev,
+                               u32 index, u32 sclk)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct atom_clock_dividers dividers;
+       int ret;
+
+       ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM,
+                                            sclk, false, &dividers);
+       if (ret)
+               return ret;
+
+       pi->graphics_level[index].SclkDid = (u8)dividers.post_div;
+       pi->graphics_level[index].SclkFrequency = cpu_to_be32(sclk);
+
+       return 0;
+}
+
+static u16 kv_convert_8bit_index_to_voltage(struct radeon_device *rdev,
+                                           u16 voltage)
+{
+       return 6200 - (voltage * 25);
+}
+
+static u16 kv_convert_2bit_index_to_voltage(struct radeon_device *rdev,
+                                           u32 vid_2bit)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 vid_8bit = sumo_convert_vid2_to_vid7(rdev,
+                                                &pi->sys_info.vid_mapping_table,
+                                                vid_2bit);
+
+       return kv_convert_8bit_index_to_voltage(rdev, (u16)vid_8bit);
+}
+
+
+static int kv_set_vid(struct radeon_device *rdev, u32 index, u32 vid)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       pi->graphics_level[index].VoltageDownH = (u8)pi->voltage_drop_t;
+       pi->graphics_level[index].MinVddNb =
+               cpu_to_be32(kv_convert_2bit_index_to_voltage(rdev, vid));
+
+       return 0;
+}
+
+static int kv_set_at(struct radeon_device *rdev, u32 index, u32 at)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       pi->graphics_level[index].AT = cpu_to_be16((u16)at);
+
+       return 0;
+}
+
+static void kv_dpm_power_level_enable(struct radeon_device *rdev,
+                                     u32 index, bool enable)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       pi->graphics_level[index].EnabledForActivity = enable ? 1 : 0;
+}
+
+static void kv_start_dpm(struct radeon_device *rdev)
+{
+       u32 tmp = RREG32_SMC(GENERAL_PWRMGT);
+
+       tmp |= GLOBAL_PWRMGT_EN;
+       WREG32_SMC(GENERAL_PWRMGT, tmp);
+
+       kv_smc_dpm_enable(rdev, true);
+}
+
+static void kv_stop_dpm(struct radeon_device *rdev)
+{
+       kv_smc_dpm_enable(rdev, false);
+}
+
+static void kv_start_am(struct radeon_device *rdev)
+{
+       u32 sclk_pwrmgt_cntl = RREG32_SMC(SCLK_PWRMGT_CNTL);
+
+       sclk_pwrmgt_cntl &= ~(RESET_SCLK_CNT | RESET_BUSY_CNT);
+       sclk_pwrmgt_cntl |= DYNAMIC_PM_EN;
+
+       WREG32_SMC(SCLK_PWRMGT_CNTL, sclk_pwrmgt_cntl);
+}
+
+static void kv_reset_am(struct radeon_device *rdev)
+{
+       u32 sclk_pwrmgt_cntl = RREG32_SMC(SCLK_PWRMGT_CNTL);
+
+       sclk_pwrmgt_cntl |= (RESET_SCLK_CNT | RESET_BUSY_CNT);
+
+       WREG32_SMC(SCLK_PWRMGT_CNTL, sclk_pwrmgt_cntl);
+}
+
+static int kv_freeze_sclk_dpm(struct radeon_device *rdev, bool freeze)
+{
+       return kv_notify_message_to_smu(rdev, freeze ?
+                                       PPSMC_MSG_SCLKDPM_FreezeLevel : PPSMC_MSG_SCLKDPM_UnfreezeLevel);
+}
+
+static int kv_force_lowest_valid(struct radeon_device *rdev)
+{
+       return kv_force_dpm_lowest(rdev);
+}
+
+static int kv_unforce_levels(struct radeon_device *rdev)
+{
+       return kv_notify_message_to_smu(rdev, PPSMC_MSG_NoForcedLevel);
+}
+
+static int kv_update_sclk_t(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 low_sclk_interrupt_t = 0;
+       int ret = 0;
+
+       if (pi->caps_sclk_throttle_low_notification) {
+               low_sclk_interrupt_t = cpu_to_be32(pi->low_sclk_interrupt_t);
+
+               ret = kv_copy_bytes_to_smc(rdev,
+                                          pi->dpm_table_start +
+                                          offsetof(SMU7_Fusion_DpmTable, LowSclkInterruptT),
+                                          (u8 *)&low_sclk_interrupt_t,
+                                          sizeof(u32), pi->sram_end);
+       }
+       return ret;
+}
+
+static int kv_program_bootup_state(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 i;
+       struct radeon_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+
+       if (table && table->count) {
+               for (i = pi->graphics_dpm_level_count - 1; i >= 0; i--) {
+                       if ((table->entries[i].clk == pi->boot_pl.sclk) ||
+                           (i == 0))
+                               break;
+               }
+
+               pi->graphics_boot_level = (u8)i;
+               kv_dpm_power_level_enable(rdev, i, true);
+       } else {
+               struct sumo_sclk_voltage_mapping_table *table =
+                       &pi->sys_info.sclk_voltage_mapping_table;
+
+               if (table->num_max_dpm_entries == 0)
+                       return -EINVAL;
+
+               for (i = pi->graphics_dpm_level_count - 1; i >= 0; i--) {
+                       if ((table->entries[i].sclk_frequency == pi->boot_pl.sclk) ||
+                           (i == 0))
+                               break;
+               }
+
+               pi->graphics_boot_level = (u8)i;
+               kv_dpm_power_level_enable(rdev, i, true);
+       }
+       return 0;
+}
+
+static int kv_enable_auto_thermal_throttling(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret;
+
+       pi->graphics_therm_throttle_enable = 1;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, GraphicsThermThrottleEnable),
+                                  &pi->graphics_therm_throttle_enable,
+                                  sizeof(u8), pi->sram_end);
+
+       return ret;
+}
+
+static int kv_upload_dpm_settings(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, GraphicsLevel),
+                                  (u8 *)&pi->graphics_level,
+                                  sizeof(SMU7_Fusion_GraphicsLevel) * SMU7_MAX_LEVELS_GRAPHICS,
+                                  pi->sram_end);
+
+       if (ret)
+               return ret;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, GraphicsDpmLevelCount),
+                                  &pi->graphics_dpm_level_count,
+                                  sizeof(u8), pi->sram_end);
+
+       return ret;
+}
+
+static u32 kv_get_clock_difference(u32 a, u32 b)
+{
+       return (a >= b) ? a - b : b - a;
+}
+
+static u32 kv_get_clk_bypass(struct radeon_device *rdev, u32 clk)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 value;
+
+       if (pi->caps_enable_dfs_bypass) {
+               if (kv_get_clock_difference(clk, 40000) < 200)
+                       value = 3;
+               else if (kv_get_clock_difference(clk, 30000) < 200)
+                       value = 2;
+               else if (kv_get_clock_difference(clk, 20000) < 200)
+                       value = 7;
+               else if (kv_get_clock_difference(clk, 15000) < 200)
+                       value = 6;
+               else if (kv_get_clock_difference(clk, 10000) < 200)
+                       value = 8;
+               else
+                       value = 0;
+       } else {
+               value = 0;
+       }
+
+       return value;
+}
+
+static int kv_populate_uvd_table(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_uvd_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table;
+       struct atom_clock_dividers dividers;
+       int ret;
+       u32 i;
+
+       if (table == NULL || table->count == 0)
+               return 0;
+
+       pi->uvd_level_count = 0;
+       for (i = 0; i < table->count; i++) {
+               if (pi->high_voltage_t &&
+                   (pi->high_voltage_t < table->entries[i].v))
+                       break;
+
+               pi->uvd_level[i].VclkFrequency = cpu_to_be32(table->entries[i].vclk);
+               pi->uvd_level[i].DclkFrequency = cpu_to_be32(table->entries[i].dclk);
+               pi->uvd_level[i].MinVddNb = cpu_to_be16(table->entries[i].v);
+
+               pi->uvd_level[i].VClkBypassCntl =
+                       (u8)kv_get_clk_bypass(rdev, table->entries[i].vclk);
+               pi->uvd_level[i].DClkBypassCntl =
+                       (u8)kv_get_clk_bypass(rdev, table->entries[i].dclk);
+
+               ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM,
+                                                    table->entries[i].vclk, false, &dividers);
+               if (ret)
+                       return ret;
+               pi->uvd_level[i].VclkDivider = (u8)dividers.post_div;
+
+               ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM,
+                                                    table->entries[i].dclk, false, &dividers);
+               if (ret)
+                       return ret;
+               pi->uvd_level[i].DclkDivider = (u8)dividers.post_div;
+
+               pi->uvd_level_count++;
+       }
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, UvdLevelCount),
+                                  (u8 *)&pi->uvd_level_count,
+                                  sizeof(u8), pi->sram_end);
+       if (ret)
+               return ret;
+
+       pi->uvd_interval = 1;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, UVDInterval),
+                                  &pi->uvd_interval,
+                                  sizeof(u8), pi->sram_end);
+       if (ret)
+               return ret;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, UvdLevel),
+                                  (u8 *)&pi->uvd_level,
+                                  sizeof(SMU7_Fusion_UvdLevel) * SMU7_MAX_LEVELS_UVD,
+                                  pi->sram_end);
+
+       return ret;
+
+}
+
+static int kv_populate_vce_table(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret;
+       u32 i;
+       struct radeon_vce_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table;
+       struct atom_clock_dividers dividers;
+
+       if (table == NULL || table->count == 0)
+               return 0;
+
+       pi->vce_level_count = 0;
+       for (i = 0; i < table->count; i++) {
+               if (pi->high_voltage_t &&
+                   pi->high_voltage_t < table->entries[i].v)
+                       break;
+
+               pi->vce_level[i].Frequency = cpu_to_be32(table->entries[i].evclk);
+               pi->vce_level[i].MinVoltage = cpu_to_be16(table->entries[i].v);
+
+               pi->vce_level[i].ClkBypassCntl =
+                       (u8)kv_get_clk_bypass(rdev, table->entries[i].evclk);
+
+               ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM,
+                                                    table->entries[i].evclk, false, &dividers);
+               if (ret)
+                       return ret;
+               pi->vce_level[i].Divider = (u8)dividers.post_div;
+
+               pi->vce_level_count++;
+       }
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, VceLevelCount),
+                                  (u8 *)&pi->vce_level_count,
+                                  sizeof(u8),
+                                  pi->sram_end);
+       if (ret)
+               return ret;
+
+       pi->vce_interval = 1;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, VCEInterval),
+                                  (u8 *)&pi->vce_interval,
+                                  sizeof(u8),
+                                  pi->sram_end);
+       if (ret)
+               return ret;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, VceLevel),
+                                  (u8 *)&pi->vce_level,
+                                  sizeof(SMU7_Fusion_ExtClkLevel) * SMU7_MAX_LEVELS_VCE,
+                                  pi->sram_end);
+
+       return ret;
+}
+
+static int kv_populate_samu_table(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table;
+       struct atom_clock_dividers dividers;
+       int ret;
+       u32 i;
+
+       if (table == NULL || table->count == 0)
+               return 0;
+
+       pi->samu_level_count = 0;
+       for (i = 0; i < table->count; i++) {
+               if (pi->high_voltage_t &&
+                   pi->high_voltage_t < table->entries[i].v)
+                       break;
+
+               pi->samu_level[i].Frequency = cpu_to_be32(table->entries[i].clk);
+               pi->samu_level[i].MinVoltage = cpu_to_be16(table->entries[i].v);
+
+               pi->samu_level[i].ClkBypassCntl =
+                       (u8)kv_get_clk_bypass(rdev, table->entries[i].clk);
+
+               ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM,
+                                                    table->entries[i].clk, false, &dividers);
+               if (ret)
+                       return ret;
+               pi->samu_level[i].Divider = (u8)dividers.post_div;
+
+               pi->samu_level_count++;
+       }
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, SamuLevelCount),
+                                  (u8 *)&pi->samu_level_count,
+                                  sizeof(u8),
+                                  pi->sram_end);
+       if (ret)
+               return ret;
+
+       pi->samu_interval = 1;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, SAMUInterval),
+                                  (u8 *)&pi->samu_interval,
+                                  sizeof(u8),
+                                  pi->sram_end);
+       if (ret)
+               return ret;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, SamuLevel),
+                                  (u8 *)&pi->samu_level,
+                                  sizeof(SMU7_Fusion_ExtClkLevel) * SMU7_MAX_LEVELS_SAMU,
+                                  pi->sram_end);
+       if (ret)
+               return ret;
+
+       return ret;
+}
+
+
+static int kv_populate_acp_table(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table;
+       struct atom_clock_dividers dividers;
+       int ret;
+       u32 i;
+
+       if (table == NULL || table->count == 0)
+               return 0;
+
+       pi->acp_level_count = 0;
+       for (i = 0; i < table->count; i++) {
+               pi->acp_level[i].Frequency = cpu_to_be32(table->entries[i].clk);
+               pi->acp_level[i].MinVoltage = cpu_to_be16(table->entries[i].v);
+
+               ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM,
+                                                    table->entries[i].clk, false, &dividers);
+               if (ret)
+                       return ret;
+               pi->acp_level[i].Divider = (u8)dividers.post_div;
+
+               pi->acp_level_count++;
+       }
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, AcpLevelCount),
+                                  (u8 *)&pi->acp_level_count,
+                                  sizeof(u8),
+                                  pi->sram_end);
+       if (ret)
+               return ret;
+
+       pi->acp_interval = 1;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, ACPInterval),
+                                  (u8 *)&pi->acp_interval,
+                                  sizeof(u8),
+                                  pi->sram_end);
+       if (ret)
+               return ret;
+
+       ret = kv_copy_bytes_to_smc(rdev,
+                                  pi->dpm_table_start +
+                                  offsetof(SMU7_Fusion_DpmTable, AcpLevel),
+                                  (u8 *)&pi->acp_level,
+                                  sizeof(SMU7_Fusion_ExtClkLevel) * SMU7_MAX_LEVELS_ACP,
+                                  pi->sram_end);
+       if (ret)
+               return ret;
+
+       return ret;
+}
+
+static void kv_calculate_dfs_bypass_settings(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 i;
+       struct radeon_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+
+       if (table && table->count) {
+               for (i = 0; i < pi->graphics_dpm_level_count; i++) {
+                       if (pi->caps_enable_dfs_bypass) {
+                               if (kv_get_clock_difference(table->entries[i].clk, 40000) < 200)
+                                       pi->graphics_level[i].ClkBypassCntl = 3;
+                               else if (kv_get_clock_difference(table->entries[i].clk, 30000) < 200)
+                                       pi->graphics_level[i].ClkBypassCntl = 2;
+                               else if (kv_get_clock_difference(table->entries[i].clk, 26600) < 200)
+                                       pi->graphics_level[i].ClkBypassCntl = 7;
+                               else if (kv_get_clock_difference(table->entries[i].clk , 20000) < 200)
+                                       pi->graphics_level[i].ClkBypassCntl = 6;
+                               else if (kv_get_clock_difference(table->entries[i].clk , 10000) < 200)
+                                       pi->graphics_level[i].ClkBypassCntl = 8;
+                               else
+                                       pi->graphics_level[i].ClkBypassCntl = 0;
+                       } else {
+                               pi->graphics_level[i].ClkBypassCntl = 0;
+                       }
+               }
+       } else {
+               struct sumo_sclk_voltage_mapping_table *table =
+                       &pi->sys_info.sclk_voltage_mapping_table;
+               for (i = 0; i < pi->graphics_dpm_level_count; i++) {
+                       if (pi->caps_enable_dfs_bypass) {
+                               if (kv_get_clock_difference(table->entries[i].sclk_frequency, 40000) < 200)
+                                       pi->graphics_level[i].ClkBypassCntl = 3;
+                               else if (kv_get_clock_difference(table->entries[i].sclk_frequency, 30000) < 200)
+                                       pi->graphics_level[i].ClkBypassCntl = 2;
+                               else if (kv_get_clock_difference(table->entries[i].sclk_frequency, 26600) < 200)
+                                       pi->graphics_level[i].ClkBypassCntl = 7;
+                               else if (kv_get_clock_difference(table->entries[i].sclk_frequency, 20000) < 200)
+                                       pi->graphics_level[i].ClkBypassCntl = 6;
+                               else if (kv_get_clock_difference(table->entries[i].sclk_frequency, 10000) < 200)
+                                       pi->graphics_level[i].ClkBypassCntl = 8;
+                               else
+                                       pi->graphics_level[i].ClkBypassCntl = 0;
+                       } else {
+                               pi->graphics_level[i].ClkBypassCntl = 0;
+                       }
+               }
+       }
+}
+
+static int kv_enable_ulv(struct radeon_device *rdev, bool enable)
+{
+       return kv_notify_message_to_smu(rdev, enable ?
+                                       PPSMC_MSG_EnableULV : PPSMC_MSG_DisableULV);
+}
+
+static void kv_update_current_ps(struct radeon_device *rdev,
+                                struct radeon_ps *rps)
+{
+       struct kv_ps *new_ps = kv_get_ps(rps);
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       pi->current_rps = *rps;
+       pi->current_ps = *new_ps;
+       pi->current_rps.ps_priv = &pi->current_ps;
+}
+
+static void kv_update_requested_ps(struct radeon_device *rdev,
+                                  struct radeon_ps *rps)
+{
+       struct kv_ps *new_ps = kv_get_ps(rps);
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       pi->requested_rps = *rps;
+       pi->requested_ps = *new_ps;
+       pi->requested_rps.ps_priv = &pi->requested_ps;
+}
+
+int kv_dpm_enable(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret;
+
+       ret = kv_process_firmware_header(rdev);
+       if (ret) {
+               DRM_ERROR("kv_process_firmware_header failed\n");
+               return ret;
+       }
+       kv_init_fps_limits(rdev);
+       kv_init_graphics_levels(rdev);
+       ret = kv_program_bootup_state(rdev);
+       if (ret) {
+               DRM_ERROR("kv_program_bootup_state failed\n");
+               return ret;
+       }
+       kv_calculate_dfs_bypass_settings(rdev);
+       ret = kv_upload_dpm_settings(rdev);
+       if (ret) {
+               DRM_ERROR("kv_upload_dpm_settings failed\n");
+               return ret;
+       }
+       ret = kv_populate_uvd_table(rdev);
+       if (ret) {
+               DRM_ERROR("kv_populate_uvd_table failed\n");
+               return ret;
+       }
+       ret = kv_populate_vce_table(rdev);
+       if (ret) {
+               DRM_ERROR("kv_populate_vce_table failed\n");
+               return ret;
+       }
+       ret = kv_populate_samu_table(rdev);
+       if (ret) {
+               DRM_ERROR("kv_populate_samu_table failed\n");
+               return ret;
+       }
+       ret = kv_populate_acp_table(rdev);
+       if (ret) {
+               DRM_ERROR("kv_populate_acp_table failed\n");
+               return ret;
+       }
+       kv_program_vc(rdev);
+#if 0
+       kv_initialize_hardware_cac_manager(rdev);
+#endif
+       kv_start_am(rdev);
+       if (pi->enable_auto_thermal_throttling) {
+               ret = kv_enable_auto_thermal_throttling(rdev);
+               if (ret) {
+                       DRM_ERROR("kv_enable_auto_thermal_throttling failed\n");
+                       return ret;
+               }
+       }
+       ret = kv_enable_dpm_voltage_scaling(rdev);
+       if (ret) {
+               DRM_ERROR("kv_enable_dpm_voltage_scaling failed\n");
+               return ret;
+       }
+       ret = kv_set_dpm_interval(rdev);
+       if (ret) {
+               DRM_ERROR("kv_set_dpm_interval failed\n");
+               return ret;
+       }
+       ret = kv_set_dpm_boot_state(rdev);
+       if (ret) {
+               DRM_ERROR("kv_set_dpm_boot_state failed\n");
+               return ret;
+       }
+       ret = kv_enable_ulv(rdev, true);
+       if (ret) {
+               DRM_ERROR("kv_enable_ulv failed\n");
+               return ret;
+       }
+       kv_start_dpm(rdev);
+       ret = kv_enable_didt(rdev, true);
+       if (ret) {
+               DRM_ERROR("kv_enable_didt failed\n");
+               return ret;
+       }
+       ret = kv_enable_smc_cac(rdev, true);
+       if (ret) {
+               DRM_ERROR("kv_enable_smc_cac failed\n");
+               return ret;
+       }
+
+       if (rdev->irq.installed &&
+           r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) {
+               ret = kv_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX);
+               if (ret) {
+                       DRM_ERROR("kv_set_thermal_temperature_range failed\n");
+                       return ret;
+               }
+               rdev->irq.dpm_thermal = true;
+               radeon_irq_set(rdev);
+       }
+
+       /* powerdown unused blocks for now */
+       kv_dpm_powergate_acp(rdev, true);
+       kv_dpm_powergate_samu(rdev, true);
+       kv_dpm_powergate_vce(rdev, true);
+
+       kv_update_current_ps(rdev, rdev->pm.dpm.boot_ps);
+
+       return ret;
+}
+
+void kv_dpm_disable(struct radeon_device *rdev)
+{
+       kv_enable_smc_cac(rdev, false);
+       kv_enable_didt(rdev, false);
+       kv_clear_vc(rdev);
+       kv_stop_dpm(rdev);
+       kv_enable_ulv(rdev, false);
+       kv_reset_am(rdev);
+
+       kv_update_current_ps(rdev, rdev->pm.dpm.boot_ps);
+}
+
+#if 0
+static int kv_write_smc_soft_register(struct radeon_device *rdev,
+                                     u16 reg_offset, u32 value)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       return kv_copy_bytes_to_smc(rdev, pi->soft_regs_start + reg_offset,
+                                   (u8 *)&value, sizeof(u16), pi->sram_end);
+}
+
+static int kv_read_smc_soft_register(struct radeon_device *rdev,
+                                    u16 reg_offset, u32 *value)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       return kv_read_smc_sram_dword(rdev, pi->soft_regs_start + reg_offset,
+                                     value, pi->sram_end);
+}
+#endif
+
+static void kv_init_sclk_t(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       pi->low_sclk_interrupt_t = 0;
+}
+
+static int kv_init_fps_limits(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret = 0;
+
+       if (pi->caps_fps) {
+               u16 tmp;
+
+               tmp = 45;
+               pi->fps_high_t = cpu_to_be16(tmp);
+               ret = kv_copy_bytes_to_smc(rdev,
+                                          pi->dpm_table_start +
+                                          offsetof(SMU7_Fusion_DpmTable, FpsHighT),
+                                          (u8 *)&pi->fps_high_t,
+                                          sizeof(u16), pi->sram_end);
+
+               tmp = 30;
+               pi->fps_low_t = cpu_to_be16(tmp);
+
+               ret = kv_copy_bytes_to_smc(rdev,
+                                          pi->dpm_table_start +
+                                          offsetof(SMU7_Fusion_DpmTable, FpsLowT),
+                                          (u8 *)&pi->fps_low_t,
+                                          sizeof(u16), pi->sram_end);
+
+       }
+       return ret;
+}
+
+static void kv_init_powergate_state(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       pi->uvd_power_gated = false;
+       pi->vce_power_gated = false;
+       pi->samu_power_gated = false;
+       pi->acp_power_gated = false;
+
+}
+
+static int kv_enable_uvd_dpm(struct radeon_device *rdev, bool enable)
+{
+       return kv_notify_message_to_smu(rdev, enable ?
+                                       PPSMC_MSG_UVDDPM_Enable : PPSMC_MSG_UVDDPM_Disable);
+}
+
+#if 0
+static int kv_enable_vce_dpm(struct radeon_device *rdev, bool enable)
+{
+       return kv_notify_message_to_smu(rdev, enable ?
+                                       PPSMC_MSG_VCEDPM_Enable : PPSMC_MSG_VCEDPM_Disable);
+}
+#endif
+
+static int kv_enable_samu_dpm(struct radeon_device *rdev, bool enable)
+{
+       return kv_notify_message_to_smu(rdev, enable ?
+                                       PPSMC_MSG_SAMUDPM_Enable : PPSMC_MSG_SAMUDPM_Disable);
+}
+
+static int kv_enable_acp_dpm(struct radeon_device *rdev, bool enable)
+{
+       return kv_notify_message_to_smu(rdev, enable ?
+                                       PPSMC_MSG_ACPDPM_Enable : PPSMC_MSG_ACPDPM_Disable);
+}
+
+static int kv_update_uvd_dpm(struct radeon_device *rdev, bool gate)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_uvd_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table;
+       int ret;
+
+       if (!gate) {
+               if (!pi->caps_uvd_dpm || table->count || pi->caps_stable_p_state)
+                       pi->uvd_boot_level = table->count - 1;
+               else
+                       pi->uvd_boot_level = 0;
+
+               ret = kv_copy_bytes_to_smc(rdev,
+                                          pi->dpm_table_start +
+                                          offsetof(SMU7_Fusion_DpmTable, UvdBootLevel),
+                                          (uint8_t *)&pi->uvd_boot_level,
+                                          sizeof(u8), pi->sram_end);
+               if (ret)
+                       return ret;
+
+               if (!pi->caps_uvd_dpm ||
+                   pi->caps_stable_p_state)
+                       kv_send_msg_to_smc_with_parameter(rdev,
+                                                         PPSMC_MSG_UVDDPM_SetEnabledMask,
+                                                         (1 << pi->uvd_boot_level));
+       }
+
+       return kv_enable_uvd_dpm(rdev, !gate);
+}
+
+#if 0
+static u8 kv_get_vce_boot_level(struct radeon_device *rdev)
+{
+       u8 i;
+       struct radeon_vce_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table;
+
+       for (i = 0; i < table->count; i++) {
+               if (table->entries[i].evclk >= 0) /* XXX */
+                       break;
+       }
+
+       return i;
+}
+
+static int kv_update_vce_dpm(struct radeon_device *rdev,
+                            struct radeon_ps *radeon_new_state,
+                            struct radeon_ps *radeon_current_state)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_vce_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table;
+       int ret;
+
+       if (radeon_new_state->evclk > 0 && radeon_current_state->evclk == 0) {
+               if (pi->caps_stable_p_state)
+                       pi->vce_boot_level = table->count - 1;
+               else
+                       pi->vce_boot_level = kv_get_vce_boot_level(rdev);
+
+               ret = kv_copy_bytes_to_smc(rdev,
+                                          pi->dpm_table_start +
+                                          offsetof(SMU7_Fusion_DpmTable, VceBootLevel),
+                                          (u8 *)&pi->vce_boot_level,
+                                          sizeof(u8),
+                                          pi->sram_end);
+               if (ret)
+                       return ret;
+
+               if (pi->caps_stable_p_state)
+                       kv_send_msg_to_smc_with_parameter(rdev,
+                                                         PPSMC_MSG_VCEDPM_SetEnabledMask,
+                                                         (1 << pi->vce_boot_level));
+
+               kv_enable_vce_dpm(rdev, true);
+       } else if (radeon_new_state->evclk == 0 && radeon_current_state->evclk > 0) {
+               kv_enable_vce_dpm(rdev, false);
+       }
+
+       return 0;
+}
+#endif
+
+static int kv_update_samu_dpm(struct radeon_device *rdev, bool gate)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table;
+       int ret;
+
+       if (!gate) {
+               if (pi->caps_stable_p_state)
+                       pi->samu_boot_level = table->count - 1;
+               else
+                       pi->samu_boot_level = 0;
+
+               ret = kv_copy_bytes_to_smc(rdev,
+                                          pi->dpm_table_start +
+                                          offsetof(SMU7_Fusion_DpmTable, SamuBootLevel),
+                                          (u8 *)&pi->samu_boot_level,
+                                          sizeof(u8),
+                                          pi->sram_end);
+               if (ret)
+                       return ret;
+
+               if (pi->caps_stable_p_state)
+                       kv_send_msg_to_smc_with_parameter(rdev,
+                                                         PPSMC_MSG_SAMUDPM_SetEnabledMask,
+                                                         (1 << pi->samu_boot_level));
+       }
+
+       return kv_enable_samu_dpm(rdev, !gate);
+}
+
+static int kv_update_acp_dpm(struct radeon_device *rdev, bool gate)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table;
+       int ret;
+
+       if (!gate) {
+               if (pi->caps_stable_p_state)
+                       pi->acp_boot_level = table->count - 1;
+               else
+                       pi->acp_boot_level = 0;
+
+               ret = kv_copy_bytes_to_smc(rdev,
+                                          pi->dpm_table_start +
+                                          offsetof(SMU7_Fusion_DpmTable, AcpBootLevel),
+                                          (u8 *)&pi->acp_boot_level,
+                                          sizeof(u8),
+                                          pi->sram_end);
+               if (ret)
+                       return ret;
+
+               if (pi->caps_stable_p_state)
+                       kv_send_msg_to_smc_with_parameter(rdev,
+                                                         PPSMC_MSG_ACPDPM_SetEnabledMask,
+                                                         (1 << pi->acp_boot_level));
+       }
+
+       return kv_enable_acp_dpm(rdev, !gate);
+}
+
+static void kv_dpm_powergate_uvd(struct radeon_device *rdev, bool gate)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       if (pi->uvd_power_gated == gate)
+               return;
+
+       pi->uvd_power_gated = gate;
+
+       if (gate) {
+               kv_update_uvd_dpm(rdev, true);
+               if (pi->caps_uvd_pg)
+                       kv_notify_message_to_smu(rdev, PPSMC_MSG_UVDPowerOFF);
+       } else {
+               if (pi->caps_uvd_pg)
+                       kv_notify_message_to_smu(rdev, PPSMC_MSG_UVDPowerON);
+               kv_update_uvd_dpm(rdev, false);
+       }
+}
+
+static void kv_dpm_powergate_vce(struct radeon_device *rdev, bool gate)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       if (pi->vce_power_gated == gate)
+               return;
+
+       pi->vce_power_gated = gate;
+
+       if (gate) {
+               if (pi->caps_vce_pg)
+                       kv_notify_message_to_smu(rdev, PPSMC_MSG_VCEPowerOFF);
+       } else {
+               if (pi->caps_vce_pg)
+                       kv_notify_message_to_smu(rdev, PPSMC_MSG_VCEPowerON);
+       }
+}
+
+static void kv_dpm_powergate_samu(struct radeon_device *rdev, bool gate)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       if (pi->samu_power_gated == gate)
+               return;
+
+       pi->samu_power_gated = gate;
+
+       if (gate) {
+               kv_update_samu_dpm(rdev, true);
+               if (pi->caps_samu_pg)
+                       kv_notify_message_to_smu(rdev, PPSMC_MSG_SAMPowerOFF);
+       } else {
+               if (pi->caps_samu_pg)
+                       kv_notify_message_to_smu(rdev, PPSMC_MSG_SAMPowerON);
+               kv_update_samu_dpm(rdev, false);
+       }
+}
+
+static void kv_dpm_powergate_acp(struct radeon_device *rdev, bool gate)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       if (pi->acp_power_gated == gate)
+               return;
+
+       if (rdev->family == CHIP_KABINI)
+               return;
+
+       pi->acp_power_gated = gate;
+
+       if (gate) {
+               kv_update_acp_dpm(rdev, true);
+               if (pi->caps_acp_pg)
+                       kv_notify_message_to_smu(rdev, PPSMC_MSG_ACPPowerOFF);
+       } else {
+               if (pi->caps_acp_pg)
+                       kv_notify_message_to_smu(rdev, PPSMC_MSG_ACPPowerON);
+               kv_update_acp_dpm(rdev, false);
+       }
+}
+
+static void kv_set_valid_clock_range(struct radeon_device *rdev,
+                                    struct radeon_ps *new_rps)
+{
+       struct kv_ps *new_ps = kv_get_ps(new_rps);
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 i;
+       struct radeon_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+
+       if (table && table->count) {
+               for (i = 0; i < pi->graphics_dpm_level_count; i++) {
+                       if ((table->entries[i].clk >= new_ps->levels[0].sclk) ||
+                           (i == (pi->graphics_dpm_level_count - 1))) {
+                               pi->lowest_valid = i;
+                               break;
+                       }
+               }
+
+               for (i = pi->graphics_dpm_level_count - 1; i >= 0; i--) {
+                       if ((table->entries[i].clk <= new_ps->levels[new_ps->num_levels -1].sclk) ||
+                           (i == 0)) {
+                               pi->highest_valid = i;
+                               break;
+                       }
+               }
+
+               if (pi->lowest_valid > pi->highest_valid) {
+                       if ((new_ps->levels[0].sclk - table->entries[pi->highest_valid].clk) >
+                           (table->entries[pi->lowest_valid].clk - new_ps->levels[new_ps->num_levels - 1].sclk))
+                               pi->highest_valid = pi->lowest_valid;
+                       else
+                               pi->lowest_valid =  pi->highest_valid;
+               }
+       } else {
+               struct sumo_sclk_voltage_mapping_table *table =
+                       &pi->sys_info.sclk_voltage_mapping_table;
+
+               for (i = 0; i < (int)pi->graphics_dpm_level_count; i++) {
+                       if (table->entries[i].sclk_frequency >= new_ps->levels[0].sclk ||
+                           i == (int)(pi->graphics_dpm_level_count - 1)) {
+                               pi->lowest_valid = i;
+                               break;
+                       }
+               }
+
+               for (i = pi->graphics_dpm_level_count - 1; i >= 0; i--) {
+                       if (table->entries[i].sclk_frequency <=
+                           new_ps->levels[new_ps->num_levels - 1].sclk ||
+                           i == 0) {
+                               pi->highest_valid = i;
+                               break;
+                       }
+               }
+
+               if (pi->lowest_valid > pi->highest_valid) {
+                       if ((new_ps->levels[0].sclk -
+                            table->entries[pi->highest_valid].sclk_frequency) >
+                           (table->entries[pi->lowest_valid].sclk_frequency -
+                            new_ps->levels[new_ps->num_levels -1].sclk))
+                               pi->highest_valid = pi->lowest_valid;
+                       else
+                               pi->lowest_valid =  pi->highest_valid;
+               }
+       }
+}
+
+static int kv_update_dfs_bypass_settings(struct radeon_device *rdev,
+                                        struct radeon_ps *new_rps)
+{
+       struct kv_ps *new_ps = kv_get_ps(new_rps);
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret = 0;
+       u8 clk_bypass_cntl;
+
+       if (pi->caps_enable_dfs_bypass) {
+               clk_bypass_cntl = new_ps->need_dfs_bypass ?
+                       pi->graphics_level[pi->graphics_boot_level].ClkBypassCntl : 0;
+               ret = kv_copy_bytes_to_smc(rdev,
+                                          (pi->dpm_table_start +
+                                           offsetof(SMU7_Fusion_DpmTable, GraphicsLevel) +
+                                           (pi->graphics_boot_level * sizeof(SMU7_Fusion_GraphicsLevel)) +
+                                           offsetof(SMU7_Fusion_GraphicsLevel, ClkBypassCntl)),
+                                          &clk_bypass_cntl,
+                                          sizeof(u8), pi->sram_end);
+       }
+
+       return ret;
+}
+
+static int kv_enable_nb_dpm(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       int ret = 0;
+
+       if (pi->enable_nb_dpm && !pi->nb_dpm_enabled) {
+               ret = kv_notify_message_to_smu(rdev, PPSMC_MSG_NBDPM_Enable);
+               if (ret == 0)
+                       pi->nb_dpm_enabled = true;
+       }
+
+       return ret;
+}
+
+int kv_dpm_pre_set_power_state(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_ps requested_ps = *rdev->pm.dpm.requested_ps;
+       struct radeon_ps *new_ps = &requested_ps;
+
+       kv_update_requested_ps(rdev, new_ps);
+
+       kv_apply_state_adjust_rules(rdev,
+                                   &pi->requested_rps,
+                                   &pi->current_rps);
+
+       return 0;
+}
+
+int kv_dpm_set_power_state(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_ps *new_ps = &pi->requested_rps;
+       /*struct radeon_ps *old_ps = &pi->current_rps;*/
+       int ret;
+
+       if (rdev->family == CHIP_KABINI) {
+               if (pi->enable_dpm) {
+                       kv_set_valid_clock_range(rdev, new_ps);
+                       kv_update_dfs_bypass_settings(rdev, new_ps);
+                       ret = kv_calculate_ds_divider(rdev);
+                       if (ret) {
+                               DRM_ERROR("kv_calculate_ds_divider failed\n");
+                               return ret;
+                       }
+                       kv_calculate_nbps_level_settings(rdev);
+                       kv_calculate_dpm_settings(rdev);
+                       kv_force_lowest_valid(rdev);
+                       kv_enable_new_levels(rdev);
+                       kv_upload_dpm_settings(rdev);
+                       kv_program_nbps_index_settings(rdev, new_ps);
+                       kv_unforce_levels(rdev);
+                       kv_set_enabled_levels(rdev);
+                       kv_force_lowest_valid(rdev);
+                       kv_unforce_levels(rdev);
+#if 0
+                       ret = kv_update_vce_dpm(rdev, new_ps, old_ps);
+                       if (ret) {
+                               DRM_ERROR("kv_update_vce_dpm failed\n");
+                               return ret;
+                       }
+#endif
+                       kv_update_uvd_dpm(rdev, false);
+                       kv_update_sclk_t(rdev);
+               }
+       } else {
+               if (pi->enable_dpm) {
+                       kv_set_valid_clock_range(rdev, new_ps);
+                       kv_update_dfs_bypass_settings(rdev, new_ps);
+                       ret = kv_calculate_ds_divider(rdev);
+                       if (ret) {
+                               DRM_ERROR("kv_calculate_ds_divider failed\n");
+                               return ret;
+                       }
+                       kv_calculate_nbps_level_settings(rdev);
+                       kv_calculate_dpm_settings(rdev);
+                       kv_freeze_sclk_dpm(rdev, true);
+                       kv_upload_dpm_settings(rdev);
+                       kv_program_nbps_index_settings(rdev, new_ps);
+                       kv_freeze_sclk_dpm(rdev, false);
+                       kv_set_enabled_levels(rdev);
+#if 0
+                       ret = kv_update_vce_dpm(rdev, new_ps, old_ps);
+                       if (ret) {
+                               DRM_ERROR("kv_update_vce_dpm failed\n");
+                               return ret;
+                       }
+#endif
+                       kv_update_uvd_dpm(rdev, false);
+                       kv_update_sclk_t(rdev);
+                       kv_enable_nb_dpm(rdev);
+               }
+       }
+       return 0;
+}
+
+void kv_dpm_post_set_power_state(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_ps *new_ps = &pi->requested_rps;
+
+       kv_update_current_ps(rdev, new_ps);
+}
+
+void kv_dpm_setup_asic(struct radeon_device *rdev)
+{
+       sumo_take_smu_control(rdev, true);
+       kv_init_powergate_state(rdev);
+       kv_init_sclk_t(rdev);
+}
+
+void kv_dpm_reset_asic(struct radeon_device *rdev)
+{
+       kv_force_lowest_valid(rdev);
+       kv_init_graphics_levels(rdev);
+       kv_program_bootup_state(rdev);
+       kv_upload_dpm_settings(rdev);
+       kv_force_lowest_valid(rdev);
+       kv_unforce_levels(rdev);
+}
+
+//XXX use sumo_dpm_display_configuration_changed
+
+static void kv_construct_max_power_limits_table(struct radeon_device *rdev,
+                                               struct radeon_clock_and_voltage_limits *table)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       if (pi->sys_info.sclk_voltage_mapping_table.num_max_dpm_entries > 0) {
+               int idx = pi->sys_info.sclk_voltage_mapping_table.num_max_dpm_entries - 1;
+               table->sclk =
+                       pi->sys_info.sclk_voltage_mapping_table.entries[idx].sclk_frequency;
+               table->vddc =
+                       kv_convert_2bit_index_to_voltage(rdev,
+                                                        pi->sys_info.sclk_voltage_mapping_table.entries[idx].vid_2bit);
+       }
+
+       table->mclk = pi->sys_info.nbp_memory_clock[0];
+}
+
+static void kv_patch_voltage_values(struct radeon_device *rdev)
+{
+       int i;
+       struct radeon_uvd_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table;
+
+       if (table->count) {
+               for (i = 0; i < table->count; i++)
+                       table->entries[i].v =
+                               kv_convert_8bit_index_to_voltage(rdev,
+                                                                table->entries[i].v);
+       }
+
+}
+
+static void kv_construct_boot_state(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       pi->boot_pl.sclk = pi->sys_info.bootup_sclk;
+       pi->boot_pl.vddc_index = pi->sys_info.bootup_nb_voltage_index;
+       pi->boot_pl.ds_divider_index = 0;
+       pi->boot_pl.ss_divider_index = 0;
+       pi->boot_pl.allow_gnb_slow = 1;
+       pi->boot_pl.force_nbp_state = 0;
+       pi->boot_pl.display_wm = 0;
+       pi->boot_pl.vce_wm = 0;
+}
+
+static int kv_force_dpm_lowest(struct radeon_device *rdev)
+{
+       int ret;
+       u32 enable_mask, i;
+
+       ret = kv_dpm_get_enable_mask(rdev, &enable_mask);
+       if (ret)
+               return ret;
+
+       for (i = 0; i < SMU7_MAX_LEVELS_GRAPHICS; i++) {
+               if (enable_mask & (1 << i))
+                       break;
+       }
+
+       return kv_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_DPM_ForceState, i);
+}
+
+static u8 kv_get_sleep_divider_id_from_clock(struct radeon_device *rdev,
+                                            u32 sclk, u32 min_sclk_in_sr)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 i;
+       u32 temp;
+       u32 min = (min_sclk_in_sr > KV_MINIMUM_ENGINE_CLOCK) ?
+               min_sclk_in_sr : KV_MINIMUM_ENGINE_CLOCK;
+
+       if (sclk < min)
+               return 0;
+
+       if (!pi->caps_sclk_ds)
+               return 0;
+
+       for (i = KV_MAX_DEEPSLEEP_DIVIDER_ID; i <= 0; i--) {
+               temp = sclk / sumo_get_sleep_divider_from_id(i);
+               if ((temp >= min) || (i == 0))
+                       break;
+       }
+
+       return (u8)i;
+}
+
+static int kv_get_high_voltage_limit(struct radeon_device *rdev, int *limit)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+       int i;
+
+       if (table && table->count) {
+               for (i = table->count - 1; i >= 0; i--) {
+                       if (pi->high_voltage_t &&
+                           (kv_convert_8bit_index_to_voltage(rdev, table->entries[i].v) <=
+                            pi->high_voltage_t)) {
+                               *limit = i;
+                               return 0;
+                       }
+               }
+       } else {
+               struct sumo_sclk_voltage_mapping_table *table =
+                       &pi->sys_info.sclk_voltage_mapping_table;
+
+               for (i = table->num_max_dpm_entries - 1; i >= 0; i--) {
+                       if (pi->high_voltage_t &&
+                           (kv_convert_2bit_index_to_voltage(rdev, table->entries[i].vid_2bit) <=
+                            pi->high_voltage_t)) {
+                               *limit = i;
+                               return 0;
+                       }
+               }
+       }
+
+       *limit = 0;
+       return 0;
+}
+
+static void kv_apply_state_adjust_rules(struct radeon_device *rdev,
+                                       struct radeon_ps *new_rps,
+                                       struct radeon_ps *old_rps)
+{
+       struct kv_ps *ps = kv_get_ps(new_rps);
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 min_sclk = 10000; /* ??? */
+       u32 sclk, mclk = 0;
+       int i, limit;
+       bool force_high;
+       struct radeon_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+       u32 stable_p_state_sclk = 0;
+       struct radeon_clock_and_voltage_limits *max_limits =
+               &rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac;
+
+       mclk = max_limits->mclk;
+       sclk = min_sclk;
+
+       if (pi->caps_stable_p_state) {
+               stable_p_state_sclk = (max_limits->sclk * 75) / 100;
+
+               for (i = table->count - 1; i >= 0; i++) {
+                       if (stable_p_state_sclk >= table->entries[i].clk) {
+                               stable_p_state_sclk = table->entries[i].clk;
+                               break;
+                       }
+               }
+
+               if (i > 0)
+                       stable_p_state_sclk = table->entries[0].clk;
+
+               sclk = stable_p_state_sclk;
+       }
+
+       ps->need_dfs_bypass = true;
+
+       for (i = 0; i < ps->num_levels; i++) {
+               if (ps->levels[i].sclk < sclk)
+                       ps->levels[i].sclk = sclk;
+       }
+
+       if (table && table->count) {
+               for (i = 0; i < ps->num_levels; i++) {
+                       if (pi->high_voltage_t &&
+                           (pi->high_voltage_t <
+                            kv_convert_8bit_index_to_voltage(rdev, ps->levels[i].vddc_index))) {
+                               kv_get_high_voltage_limit(rdev, &limit);
+                               ps->levels[i].sclk = table->entries[limit].clk;
+                       }
+               }
+       } else {
+               struct sumo_sclk_voltage_mapping_table *table =
+                       &pi->sys_info.sclk_voltage_mapping_table;
+
+               for (i = 0; i < ps->num_levels; i++) {
+                       if (pi->high_voltage_t &&
+                           (pi->high_voltage_t <
+                            kv_convert_8bit_index_to_voltage(rdev, ps->levels[i].vddc_index))) {
+                               kv_get_high_voltage_limit(rdev, &limit);
+                               ps->levels[i].sclk = table->entries[limit].sclk_frequency;
+                       }
+               }
+       }
+
+       if (pi->caps_stable_p_state) {
+               for (i = 0; i < ps->num_levels; i++) {
+                       ps->levels[i].sclk = stable_p_state_sclk;
+               }
+       }
+
+       pi->video_start = new_rps->dclk || new_rps->vclk;
+
+       if ((new_rps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) ==
+           ATOM_PPLIB_CLASSIFICATION_UI_BATTERY)
+               pi->battery_state = true;
+       else
+               pi->battery_state = false;
+
+       if (rdev->family == CHIP_KABINI) {
+               ps->dpm0_pg_nb_ps_lo = 0x1;
+               ps->dpm0_pg_nb_ps_hi = 0x0;
+               ps->dpmx_nb_ps_lo = 0x1;
+               ps->dpmx_nb_ps_hi = 0x0;
+       } else {
+               ps->dpm0_pg_nb_ps_lo = 0x1;
+               ps->dpm0_pg_nb_ps_hi = 0x0;
+               ps->dpmx_nb_ps_lo = 0x2;
+               ps->dpmx_nb_ps_hi = 0x1;
+
+               if (pi->sys_info.nb_dpm_enable && pi->battery_state) {
+                       force_high = (mclk >= pi->sys_info.nbp_memory_clock[3]) ||
+                               pi->video_start || (rdev->pm.dpm.new_active_crtc_count >= 3) ||
+                               pi->disable_nb_ps3_in_battery;
+                       ps->dpm0_pg_nb_ps_lo = force_high ? 0x2 : 0x3;
+                       ps->dpm0_pg_nb_ps_hi = 0x2;
+                       ps->dpmx_nb_ps_lo = force_high ? 0x2 : 0x3;
+                       ps->dpmx_nb_ps_hi = 0x2;
+               }
+       }
+}
+
+static void kv_dpm_power_level_enabled_for_throttle(struct radeon_device *rdev,
+                                                   u32 index, bool enable)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       pi->graphics_level[index].EnabledForThrottle = enable ? 1 : 0;
+}
+
+static int kv_calculate_ds_divider(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 sclk_in_sr = 10000; /* ??? */
+       u32 i;
+
+       if (pi->lowest_valid > pi->highest_valid)
+               return -EINVAL;
+
+       for (i = pi->lowest_valid; i <= pi->highest_valid; i++) {
+               pi->graphics_level[i].DeepSleepDivId =
+                       kv_get_sleep_divider_id_from_clock(rdev,
+                                                          be32_to_cpu(pi->graphics_level[i].SclkFrequency),
+                                                          sclk_in_sr);
+       }
+       return 0;
+}
+
+static int kv_calculate_nbps_level_settings(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 i;
+       bool force_high;
+       struct radeon_clock_and_voltage_limits *max_limits =
+               &rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac;
+       u32 mclk = max_limits->mclk;
+
+       if (pi->lowest_valid > pi->highest_valid)
+               return -EINVAL;
+
+       if (rdev->family == CHIP_KABINI) {
+               for (i = pi->lowest_valid; i <= pi->highest_valid; i++) {
+                       pi->graphics_level[i].GnbSlow = 1;
+                       pi->graphics_level[i].ForceNbPs1 = 0;
+                       pi->graphics_level[i].UpH = 0;
+               }
+
+               if (!pi->sys_info.nb_dpm_enable)
+                       return 0;
+
+               force_high = ((mclk >= pi->sys_info.nbp_memory_clock[3]) ||
+                             (rdev->pm.dpm.new_active_crtc_count >= 3) || pi->video_start);
+
+               if (force_high) {
+                       for (i = pi->lowest_valid; i <= pi->highest_valid; i++)
+                               pi->graphics_level[i].GnbSlow = 0;
+               } else {
+                       if (pi->battery_state)
+                               pi->graphics_level[0].ForceNbPs1 = 1;
+
+                       pi->graphics_level[1].GnbSlow = 0;
+                       pi->graphics_level[2].GnbSlow = 0;
+                       pi->graphics_level[3].GnbSlow = 0;
+                       pi->graphics_level[4].GnbSlow = 0;
+               }
+       } else {
+               for (i = pi->lowest_valid; i <= pi->highest_valid; i++) {
+                       pi->graphics_level[i].GnbSlow = 1;
+                       pi->graphics_level[i].ForceNbPs1 = 0;
+                       pi->graphics_level[i].UpH = 0;
+               }
+
+               if (pi->sys_info.nb_dpm_enable && pi->battery_state) {
+                       pi->graphics_level[pi->lowest_valid].UpH = 0x28;
+                       pi->graphics_level[pi->lowest_valid].GnbSlow = 0;
+                       if (pi->lowest_valid != pi->highest_valid)
+                               pi->graphics_level[pi->lowest_valid].ForceNbPs1 = 1;
+               }
+       }
+       return 0;
+}
+
+static int kv_calculate_dpm_settings(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 i;
+
+       if (pi->lowest_valid > pi->highest_valid)
+               return -EINVAL;
+
+       for (i = pi->lowest_valid; i <= pi->highest_valid; i++)
+               pi->graphics_level[i].DisplayWatermark = (i == pi->highest_valid) ? 1 : 0;
+
+       return 0;
+}
+
+static void kv_init_graphics_levels(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 i;
+       struct radeon_clock_voltage_dependency_table *table =
+               &rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+
+       if (table && table->count) {
+               u32 vid_2bit;
+
+               pi->graphics_dpm_level_count = 0;
+               for (i = 0; i < table->count; i++) {
+                       if (pi->high_voltage_t &&
+                           (pi->high_voltage_t <
+                            kv_convert_8bit_index_to_voltage(rdev, table->entries[i].v)))
+                               break;
+
+                       kv_set_divider_value(rdev, i, table->entries[i].clk);
+                       vid_2bit = sumo_convert_vid7_to_vid2(rdev,
+                                                            &pi->sys_info.vid_mapping_table,
+                                                            table->entries[i].v);
+                       kv_set_vid(rdev, i, vid_2bit);
+                       kv_set_at(rdev, i, pi->at[i]);
+                       kv_dpm_power_level_enabled_for_throttle(rdev, i, true);
+                       pi->graphics_dpm_level_count++;
+               }
+       } else {
+               struct sumo_sclk_voltage_mapping_table *table =
+                       &pi->sys_info.sclk_voltage_mapping_table;
+
+               pi->graphics_dpm_level_count = 0;
+               for (i = 0; i < table->num_max_dpm_entries; i++) {
+                       if (pi->high_voltage_t &&
+                           pi->high_voltage_t <
+                           kv_convert_2bit_index_to_voltage(rdev, table->entries[i].vid_2bit))
+                               break;
+
+                       kv_set_divider_value(rdev, i, table->entries[i].sclk_frequency);
+                       kv_set_vid(rdev, i, table->entries[i].vid_2bit);
+                       kv_set_at(rdev, i, pi->at[i]);
+                       kv_dpm_power_level_enabled_for_throttle(rdev, i, true);
+                       pi->graphics_dpm_level_count++;
+               }
+       }
+
+       for (i = 0; i < SMU7_MAX_LEVELS_GRAPHICS; i++)
+               kv_dpm_power_level_enable(rdev, i, false);
+}
+
+static void kv_enable_new_levels(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 i;
+
+       for (i = 0; i < SMU7_MAX_LEVELS_GRAPHICS; i++) {
+               if (i >= pi->lowest_valid && i <= pi->highest_valid)
+                       kv_dpm_power_level_enable(rdev, i, true);
+       }
+}
+
+static int kv_set_enabled_levels(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 i, new_mask = 0;
+
+       for (i = pi->lowest_valid; i <= pi->highest_valid; i++)
+               new_mask |= (1 << i);
+
+       return kv_send_msg_to_smc_with_parameter(rdev,
+                                                PPSMC_MSG_SCLKDPM_SetEnabledMask,
+                                                new_mask);
+}
+
+static void kv_program_nbps_index_settings(struct radeon_device *rdev,
+                                          struct radeon_ps *new_rps)
+{
+       struct kv_ps *new_ps = kv_get_ps(new_rps);
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       u32 nbdpmconfig1;
+
+       if (rdev->family == CHIP_KABINI)
+               return;
+
+       if (pi->sys_info.nb_dpm_enable) {
+               nbdpmconfig1 = RREG32_SMC(NB_DPM_CONFIG_1);
+               nbdpmconfig1 &= ~(Dpm0PgNbPsLo_MASK | Dpm0PgNbPsHi_MASK |
+                                 DpmXNbPsLo_MASK | DpmXNbPsHi_MASK);
+               nbdpmconfig1 |= (Dpm0PgNbPsLo(new_ps->dpm0_pg_nb_ps_lo) |
+                                Dpm0PgNbPsHi(new_ps->dpm0_pg_nb_ps_hi) |
+                                DpmXNbPsLo(new_ps->dpmx_nb_ps_lo) |
+                                DpmXNbPsHi(new_ps->dpmx_nb_ps_hi));
+               WREG32_SMC(NB_DPM_CONFIG_1, nbdpmconfig1);
+       }
+}
+
+static int kv_set_thermal_temperature_range(struct radeon_device *rdev,
+                                           int min_temp, int max_temp)
+{
+       int low_temp = 0 * 1000;
+       int high_temp = 255 * 1000;
+       u32 tmp;
+
+       if (low_temp < min_temp)
+               low_temp = min_temp;
+       if (high_temp > max_temp)
+               high_temp = max_temp;
+       if (high_temp < low_temp) {
+               DRM_ERROR("invalid thermal range: %d - %d\n", low_temp, high_temp);
+               return -EINVAL;
+       }
+
+       tmp = RREG32_SMC(CG_THERMAL_INT_CTRL);
+       tmp &= ~(DIG_THERM_INTH_MASK | DIG_THERM_INTL_MASK);
+       tmp |= (DIG_THERM_INTH(49 + (high_temp / 1000)) |
+               DIG_THERM_INTL(49 + (low_temp / 1000)));
+       WREG32_SMC(CG_THERMAL_INT_CTRL, tmp);
+
+       rdev->pm.dpm.thermal.min_temp = low_temp;
+       rdev->pm.dpm.thermal.max_temp = high_temp;
+
+       return 0;
+}
+
+union igp_info {
+       struct _ATOM_INTEGRATED_SYSTEM_INFO info;
+       struct _ATOM_INTEGRATED_SYSTEM_INFO_V2 info_2;
+       struct _ATOM_INTEGRATED_SYSTEM_INFO_V5 info_5;
+       struct _ATOM_INTEGRATED_SYSTEM_INFO_V6 info_6;
+       struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_7 info_7;
+       struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_8 info_8;
+};
+
+static int kv_parse_sys_info_table(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct radeon_mode_info *mode_info = &rdev->mode_info;
+       int index = GetIndexIntoMasterTable(DATA, IntegratedSystemInfo);
+       union igp_info *igp_info;
+       u8 frev, crev;
+       u16 data_offset;
+       int i;
+
+       if (atom_parse_data_header(mode_info->atom_context, index, NULL,
+                                  &frev, &crev, &data_offset)) {
+               igp_info = (union igp_info *)(mode_info->atom_context->bios +
+                                             data_offset);
+
+               if (crev != 8) {
+                       DRM_ERROR("Unsupported IGP table: %d %d\n", frev, crev);
+                       return -EINVAL;
+               }
+               pi->sys_info.bootup_sclk = le32_to_cpu(igp_info->info_8.ulBootUpEngineClock);
+               pi->sys_info.bootup_uma_clk = le32_to_cpu(igp_info->info_8.ulBootUpUMAClock);
+               pi->sys_info.bootup_nb_voltage_index =
+                       le16_to_cpu(igp_info->info_8.usBootUpNBVoltage);
+               if (igp_info->info_8.ucHtcTmpLmt == 0)
+                       pi->sys_info.htc_tmp_lmt = 203;
+               else
+                       pi->sys_info.htc_tmp_lmt = igp_info->info_8.ucHtcTmpLmt;
+               if (igp_info->info_8.ucHtcHystLmt == 0)
+                       pi->sys_info.htc_hyst_lmt = 5;
+               else
+                       pi->sys_info.htc_hyst_lmt = igp_info->info_8.ucHtcHystLmt;
+               if (pi->sys_info.htc_tmp_lmt <= pi->sys_info.htc_hyst_lmt) {
+                       DRM_ERROR("The htcTmpLmt should be larger than htcHystLmt.\n");
+               }
+
+               if (le32_to_cpu(igp_info->info_8.ulSystemConfig) & (1 << 3))
+                       pi->sys_info.nb_dpm_enable = true;
+               else
+                       pi->sys_info.nb_dpm_enable = false;
+
+               for (i = 0; i < KV_NUM_NBPSTATES; i++) {
+                       pi->sys_info.nbp_memory_clock[i] =
+                               le32_to_cpu(igp_info->info_8.ulNbpStateMemclkFreq[i]);
+                       pi->sys_info.nbp_n_clock[i] =
+                               le32_to_cpu(igp_info->info_8.ulNbpStateNClkFreq[i]);
+               }
+               if (le32_to_cpu(igp_info->info_8.ulGPUCapInfo) &
+                   SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS)
+                       pi->caps_enable_dfs_bypass = true;
+
+               sumo_construct_sclk_voltage_mapping_table(rdev,
+                                                         &pi->sys_info.sclk_voltage_mapping_table,
+                                                         igp_info->info_8.sAvail_SCLK);
+
+               sumo_construct_vid_mapping_table(rdev,
+                                                &pi->sys_info.vid_mapping_table,
+                                                igp_info->info_8.sAvail_SCLK);
+
+               kv_construct_max_power_limits_table(rdev,
+                                                   &rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac);
+       }
+       return 0;
+}
+
+union power_info {
+       struct _ATOM_POWERPLAY_INFO info;
+       struct _ATOM_POWERPLAY_INFO_V2 info_2;
+       struct _ATOM_POWERPLAY_INFO_V3 info_3;
+       struct _ATOM_PPLIB_POWERPLAYTABLE pplib;
+       struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2;
+       struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3;
+};
+
+union pplib_clock_info {
+       struct _ATOM_PPLIB_R600_CLOCK_INFO r600;
+       struct _ATOM_PPLIB_RS780_CLOCK_INFO rs780;
+       struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen;
+       struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo;
+};
+
+union pplib_power_state {
+       struct _ATOM_PPLIB_STATE v1;
+       struct _ATOM_PPLIB_STATE_V2 v2;
+};
+
+static void kv_patch_boot_state(struct radeon_device *rdev,
+                               struct kv_ps *ps)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       ps->num_levels = 1;
+       ps->levels[0] = pi->boot_pl;
+}
+
+static void kv_parse_pplib_non_clock_info(struct radeon_device *rdev,
+                                         struct radeon_ps *rps,
+                                         struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info,
+                                         u8 table_rev)
+{
+       struct kv_ps *ps = kv_get_ps(rps);
+
+       rps->caps = le32_to_cpu(non_clock_info->ulCapsAndSettings);
+       rps->class = le16_to_cpu(non_clock_info->usClassification);
+       rps->class2 = le16_to_cpu(non_clock_info->usClassification2);
+
+       if (ATOM_PPLIB_NONCLOCKINFO_VER1 < table_rev) {
+               rps->vclk = le32_to_cpu(non_clock_info->ulVCLK);
+               rps->dclk = le32_to_cpu(non_clock_info->ulDCLK);
+       } else {
+               rps->vclk = 0;
+               rps->dclk = 0;
+       }
+
+       if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) {
+               rdev->pm.dpm.boot_ps = rps;
+               kv_patch_boot_state(rdev, ps);
+       }
+       if (rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE)
+               rdev->pm.dpm.uvd_ps = rps;
+}
+
+static void kv_parse_pplib_clock_info(struct radeon_device *rdev,
+                                     struct radeon_ps *rps, int index,
+                                       union pplib_clock_info *clock_info)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct kv_ps *ps = kv_get_ps(rps);
+       struct kv_pl *pl = &ps->levels[index];
+       u32 sclk;
+
+       sclk = le16_to_cpu(clock_info->sumo.usEngineClockLow);
+       sclk |= clock_info->sumo.ucEngineClockHigh << 16;
+       pl->sclk = sclk;
+       pl->vddc_index = clock_info->sumo.vddcIndex;
+
+       ps->num_levels = index + 1;
+
+       if (pi->caps_sclk_ds) {
+               pl->ds_divider_index = 5;
+               pl->ss_divider_index = 5;
+       }
+}
+
+static int kv_parse_power_table(struct radeon_device *rdev)
+{
+       struct radeon_mode_info *mode_info = &rdev->mode_info;
+       struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info;
+       union pplib_power_state *power_state;
+       int i, j, k, non_clock_array_index, clock_array_index;
+       union pplib_clock_info *clock_info;
+       struct _StateArray *state_array;
+       struct _ClockInfoArray *clock_info_array;
+       struct _NonClockInfoArray *non_clock_info_array;
+       union power_info *power_info;
+       int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo);
+        u16 data_offset;
+       u8 frev, crev;
+       u8 *power_state_offset;
+       struct kv_ps *ps;
+
+       if (!atom_parse_data_header(mode_info->atom_context, index, NULL,
+                                  &frev, &crev, &data_offset))
+               return -EINVAL;
+       power_info = (union power_info *)(mode_info->atom_context->bios + data_offset);
+
+       state_array = (struct _StateArray *)
+               (mode_info->atom_context->bios + data_offset +
+                le16_to_cpu(power_info->pplib.usStateArrayOffset));
+       clock_info_array = (struct _ClockInfoArray *)
+               (mode_info->atom_context->bios + data_offset +
+                le16_to_cpu(power_info->pplib.usClockInfoArrayOffset));
+       non_clock_info_array = (struct _NonClockInfoArray *)
+               (mode_info->atom_context->bios + data_offset +
+                le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset));
+
+       rdev->pm.dpm.ps = kzalloc(sizeof(struct radeon_ps) *
+                                 state_array->ucNumEntries, GFP_KERNEL);
+       if (!rdev->pm.dpm.ps)
+               return -ENOMEM;
+       power_state_offset = (u8 *)state_array->states;
+       rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps);
+       rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime);
+       rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime);
+       for (i = 0; i < state_array->ucNumEntries; i++) {
+               power_state = (union pplib_power_state *)power_state_offset;
+               non_clock_array_index = power_state->v2.nonClockInfoIndex;
+               non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *)
+                       &non_clock_info_array->nonClockInfo[non_clock_array_index];
+               if (!rdev->pm.power_state[i].clock_info)
+                       return -EINVAL;
+               ps = kzalloc(sizeof(struct kv_ps), GFP_KERNEL);
+               if (ps == NULL) {
+                       kfree(rdev->pm.dpm.ps);
+                       return -ENOMEM;
+               }
+               rdev->pm.dpm.ps[i].ps_priv = ps;
+               k = 0;
+               for (j = 0; j < power_state->v2.ucNumDPMLevels; j++) {
+                       clock_array_index = power_state->v2.clockInfoIndex[j];
+                       if (clock_array_index >= clock_info_array->ucNumEntries)
+                               continue;
+                       if (k >= SUMO_MAX_HARDWARE_POWERLEVELS)
+                               break;
+                       clock_info = (union pplib_clock_info *)
+                               &clock_info_array->clockInfo[clock_array_index * clock_info_array->ucEntrySize];
+                       kv_parse_pplib_clock_info(rdev,
+                                                 &rdev->pm.dpm.ps[i], k,
+                                                 clock_info);
+                       k++;
+               }
+               kv_parse_pplib_non_clock_info(rdev, &rdev->pm.dpm.ps[i],
+                                             non_clock_info,
+                                             non_clock_info_array->ucEntrySize);
+               power_state_offset += 2 + power_state->v2.ucNumDPMLevels;
+       }
+       rdev->pm.dpm.num_ps = state_array->ucNumEntries;
+       return 0;
+}
+
+int kv_dpm_init(struct radeon_device *rdev)
+{
+       struct kv_power_info *pi;
+       int ret, i;
+
+       pi = kzalloc(sizeof(struct kv_power_info), GFP_KERNEL);
+       if (pi == NULL)
+               return -ENOMEM;
+       rdev->pm.dpm.priv = pi;
+
+       ret = r600_parse_extended_power_table(rdev);
+       if (ret)
+               return ret;
+
+       for (i = 0; i < SUMO_MAX_HARDWARE_POWERLEVELS; i++)
+               pi->at[i] = TRINITY_AT_DFLT;
+
+        pi->sram_end = SMC_RAM_END;
+
+       if (rdev->family == CHIP_KABINI)
+               pi->high_voltage_t = 4001;
+
+       pi->enable_nb_dpm = true;
+
+       pi->caps_power_containment = true;
+       pi->caps_cac = true;
+       pi->enable_didt = false;
+       if (pi->enable_didt) {
+               pi->caps_sq_ramping = true;
+               pi->caps_db_ramping = true;
+               pi->caps_td_ramping = true;
+               pi->caps_tcp_ramping = true;
+       }
+
+       pi->caps_sclk_ds = true;
+       pi->enable_auto_thermal_throttling = true;
+       pi->disable_nb_ps3_in_battery = false;
+       pi->bapm_enable = true;
+       pi->voltage_drop_t = 0;
+       pi->caps_sclk_throttle_low_notification = false;
+       pi->caps_fps = false; /* true? */
+       pi->caps_uvd_pg = false; /* XXX */
+       pi->caps_uvd_dpm = true;
+       pi->caps_vce_pg = false;
+       pi->caps_samu_pg = false;
+       pi->caps_acp_pg = false;
+       pi->caps_stable_p_state = false;
+
+       ret = kv_parse_sys_info_table(rdev);
+       if (ret)
+               return ret;
+
+       kv_patch_voltage_values(rdev);
+       kv_construct_boot_state(rdev);
+
+       ret = kv_parse_power_table(rdev);
+       if (ret)
+               return ret;
+
+       pi->enable_dpm = true;
+
+       return 0;
+}
+
+void kv_dpm_print_power_state(struct radeon_device *rdev,
+                             struct radeon_ps *rps)
+{
+       int i;
+       struct kv_ps *ps = kv_get_ps(rps);
+
+       r600_dpm_print_class_info(rps->class, rps->class2);
+       r600_dpm_print_cap_info(rps->caps);
+       printk("\tuvd    vclk: %d dclk: %d\n", rps->vclk, rps->dclk);
+       for (i = 0; i < ps->num_levels; i++) {
+               struct kv_pl *pl = &ps->levels[i];
+               printk("\t\tpower level %d    sclk: %u vddc: %u\n",
+                      i, pl->sclk,
+                      kv_convert_8bit_index_to_voltage(rdev, pl->vddc_index));
+       }
+       r600_dpm_print_ps_status(rdev, rps);
+}
+
+void kv_dpm_fini(struct radeon_device *rdev)
+{
+       int i;
+
+       for (i = 0; i < rdev->pm.dpm.num_ps; i++) {
+               kfree(rdev->pm.dpm.ps[i].ps_priv);
+       }
+       kfree(rdev->pm.dpm.ps);
+       kfree(rdev->pm.dpm.priv);
+       r600_free_extended_power_table(rdev);
+}
+
+void kv_dpm_display_configuration_changed(struct radeon_device *rdev)
+{
+
+}
+
+u32 kv_dpm_get_sclk(struct radeon_device *rdev, bool low)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+       struct kv_ps *requested_state = kv_get_ps(&pi->requested_rps);
+
+       if (low)
+               return requested_state->levels[0].sclk;
+       else
+               return requested_state->levels[requested_state->num_levels - 1].sclk;
+}
+
+u32 kv_dpm_get_mclk(struct radeon_device *rdev, bool low)
+{
+       struct kv_power_info *pi = kv_get_pi(rdev);
+
+       return pi->sys_info.bootup_uma_clk;
+}
+
diff --git a/drivers/gpu/drm/radeon/kv_dpm.h b/drivers/gpu/drm/radeon/kv_dpm.h
new file mode 100644 (file)
index 0000000..32bb079
--- /dev/null
@@ -0,0 +1,199 @@
+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __KV_DPM_H__
+#define __KV_DPM_H__
+
+#define SMU__NUM_SCLK_DPM_STATE  8
+#define SMU__NUM_MCLK_DPM_LEVELS 4
+#define SMU__NUM_LCLK_DPM_LEVELS 8
+#define SMU__NUM_PCIE_DPM_LEVELS 0 /* ??? */
+#include "smu7_fusion.h"
+#include "trinity_dpm.h"
+#include "ppsmc.h"
+
+#define KV_NUM_NBPSTATES   4
+
+enum kv_pt_config_reg_type {
+       KV_CONFIGREG_MMR = 0,
+       KV_CONFIGREG_SMC_IND,
+       KV_CONFIGREG_DIDT_IND,
+       KV_CONFIGREG_CACHE,
+       KV_CONFIGREG_MAX
+};
+
+struct kv_pt_config_reg {
+       u32 offset;
+       u32 mask;
+       u32 shift;
+       u32 value;
+       enum kv_pt_config_reg_type type;
+};
+
+struct kv_lcac_config_values {
+       u32 block_id;
+       u32 signal_id;
+       u32 t;
+};
+
+struct kv_lcac_config_reg {
+       u32 cntl;
+       u32 block_mask;
+       u32 block_shift;
+       u32 signal_mask;
+       u32 signal_shift;
+       u32 t_mask;
+       u32 t_shift;
+       u32 enable_mask;
+       u32 enable_shift;
+};
+
+struct kv_pl {
+       u32 sclk;
+       u8 vddc_index;
+       u8 ds_divider_index;
+       u8 ss_divider_index;
+       u8 allow_gnb_slow;
+       u8 force_nbp_state;
+       u8 display_wm;
+       u8 vce_wm;
+};
+
+struct kv_ps {
+       struct kv_pl levels[SUMO_MAX_HARDWARE_POWERLEVELS];
+       u32 num_levels;
+       bool need_dfs_bypass;
+       u8 dpm0_pg_nb_ps_lo;
+       u8 dpm0_pg_nb_ps_hi;
+       u8 dpmx_nb_ps_lo;
+       u8 dpmx_nb_ps_hi;
+};
+
+struct kv_sys_info {
+       u32 bootup_uma_clk;
+       u32 bootup_sclk;
+       u32 dentist_vco_freq;
+       u32 nb_dpm_enable;
+       u32 nbp_memory_clock[KV_NUM_NBPSTATES];
+       u32 nbp_n_clock[KV_NUM_NBPSTATES];
+       u16 bootup_nb_voltage_index;
+       u8 htc_tmp_lmt;
+       u8 htc_hyst_lmt;
+       struct sumo_sclk_voltage_mapping_table sclk_voltage_mapping_table;
+       struct sumo_vid_mapping_table vid_mapping_table;
+       u32 uma_channel_number;
+};
+
+struct kv_power_info {
+       u32 at[SUMO_MAX_HARDWARE_POWERLEVELS];
+       u32 voltage_drop_t;
+       struct kv_sys_info sys_info;
+       struct kv_pl boot_pl;
+       bool enable_nb_ps_policy;
+       bool disable_nb_ps3_in_battery;
+       bool video_start;
+       bool battery_state;
+       u32 lowest_valid;
+       u32 highest_valid;
+       u16 high_voltage_t;
+       bool cac_enabled;
+       bool bapm_enable;
+       /* smc offsets */
+       u32 sram_end;
+       u32 dpm_table_start;
+       u32 soft_regs_start;
+       /* dpm SMU tables */
+       u8 graphics_dpm_level_count;
+       u8 uvd_level_count;
+       u8 vce_level_count;
+       u8 acp_level_count;
+       u8 samu_level_count;
+       u16 fps_high_t;
+       SMU7_Fusion_GraphicsLevel graphics_level[SMU__NUM_SCLK_DPM_STATE];
+       SMU7_Fusion_ACPILevel acpi_level;
+       SMU7_Fusion_UvdLevel uvd_level[SMU7_MAX_LEVELS_UVD];
+       SMU7_Fusion_ExtClkLevel vce_level[SMU7_MAX_LEVELS_VCE];
+       SMU7_Fusion_ExtClkLevel acp_level[SMU7_MAX_LEVELS_ACP];
+       SMU7_Fusion_ExtClkLevel samu_level[SMU7_MAX_LEVELS_SAMU];
+       u8 uvd_boot_level;
+       u8 vce_boot_level;
+       u8 acp_boot_level;
+       u8 samu_boot_level;
+       u8 uvd_interval;
+       u8 vce_interval;
+       u8 acp_interval;
+       u8 samu_interval;
+       u8 graphics_boot_level;
+       u8 graphics_interval;
+       u8 graphics_therm_throttle_enable;
+       u8 graphics_voltage_change_enable;
+       u8 graphics_clk_slow_enable;
+       u8 graphics_clk_slow_divider;
+       u8 fps_low_t;
+       u32 low_sclk_interrupt_t;
+       bool uvd_power_gated;
+       bool vce_power_gated;
+       bool acp_power_gated;
+       bool samu_power_gated;
+       bool nb_dpm_enabled;
+       /* flags */
+       bool enable_didt;
+       bool enable_dpm;
+       bool enable_auto_thermal_throttling;
+       bool enable_nb_dpm;
+       /* caps */
+       bool caps_cac;
+       bool caps_power_containment;
+       bool caps_sq_ramping;
+       bool caps_db_ramping;
+       bool caps_td_ramping;
+       bool caps_tcp_ramping;
+       bool caps_sclk_throttle_low_notification;
+       bool caps_fps;
+       bool caps_uvd_dpm;
+       bool caps_uvd_pg;
+       bool caps_vce_pg;
+       bool caps_samu_pg;
+       bool caps_acp_pg;
+       bool caps_stable_p_state;
+       bool caps_enable_dfs_bypass;
+       bool caps_sclk_ds;
+       struct radeon_ps current_rps;
+       struct kv_ps current_ps;
+       struct radeon_ps requested_rps;
+       struct kv_ps requested_ps;
+};
+
+
+/* kv_smc.c */
+int kv_notify_message_to_smu(struct radeon_device *rdev, u32 id);
+int kv_dpm_get_enable_mask(struct radeon_device *rdev, u32 *enable_mask);
+int kv_send_msg_to_smc_with_parameter(struct radeon_device *rdev,
+                                     PPSMC_Msg msg, u32 parameter);
+int kv_read_smc_sram_dword(struct radeon_device *rdev, u32 smc_address,
+                          u32 *value, u32 limit);
+int kv_smc_dpm_enable(struct radeon_device *rdev, bool enable);
+int kv_copy_bytes_to_smc(struct radeon_device *rdev,
+                        u32 smc_start_address,
+                        const u8 *src, u32 byte_count, u32 limit);
+
+#endif
diff --git a/drivers/gpu/drm/radeon/kv_smc.c b/drivers/gpu/drm/radeon/kv_smc.c
new file mode 100644 (file)
index 0000000..34a226d
--- /dev/null
@@ -0,0 +1,207 @@
+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Alex Deucher
+ */
+
+#include "drmP.h"
+#include "radeon.h"
+#include "cikd.h"
+#include "kv_dpm.h"
+
+int kv_notify_message_to_smu(struct radeon_device *rdev, u32 id)
+{
+       u32 i;
+       u32 tmp = 0;
+
+       WREG32(SMC_MESSAGE_0, id & SMC_MSG_MASK);
+
+       for (i = 0; i < rdev->usec_timeout; i++) {
+               if ((RREG32(SMC_RESP_0) & SMC_RESP_MASK) != 0)
+                       break;
+               udelay(1);
+       }
+       tmp = RREG32(SMC_RESP_0) & SMC_RESP_MASK;
+
+       if (tmp != 1) {
+               if (tmp == 0xFF)
+                       return -EINVAL;
+               else if (tmp == 0xFE)
+                       return -EINVAL;
+       }
+
+       return 0;
+}
+
+int kv_dpm_get_enable_mask(struct radeon_device *rdev, u32 *enable_mask)
+{
+       int ret;
+
+       ret = kv_notify_message_to_smu(rdev, PPSMC_MSG_SCLKDPM_GetEnabledMask);
+
+       if (ret == 0)
+               *enable_mask = RREG32_SMC(SMC_SYSCON_MSG_ARG_0);
+
+       return ret;
+}
+
+int kv_send_msg_to_smc_with_parameter(struct radeon_device *rdev,
+                                     PPSMC_Msg msg, u32 parameter)
+{
+
+       WREG32(SMC_MSG_ARG_0, parameter);
+
+       return kv_notify_message_to_smu(rdev, msg);
+}
+
+static int kv_set_smc_sram_address(struct radeon_device *rdev,
+                                  u32 smc_address, u32 limit)
+{
+       if (smc_address & 3)
+               return -EINVAL;
+       if ((smc_address + 3) > limit)
+               return -EINVAL;
+
+       WREG32(SMC_IND_INDEX_0, smc_address);
+       WREG32_P(SMC_IND_ACCESS_CNTL, 0, ~AUTO_INCREMENT_IND_0);
+
+       return 0;
+}
+
+int kv_read_smc_sram_dword(struct radeon_device *rdev, u32 smc_address,
+                          u32 *value, u32 limit)
+{
+       int ret;
+
+       ret = kv_set_smc_sram_address(rdev, smc_address, limit);
+       if (ret)
+               return ret;
+
+       *value = RREG32(SMC_IND_DATA_0);
+       return 0;
+}
+
+int kv_smc_dpm_enable(struct radeon_device *rdev, bool enable)
+{
+       if (enable)
+               return kv_notify_message_to_smu(rdev, PPSMC_MSG_DPM_Enable);
+       else
+               return kv_notify_message_to_smu(rdev, PPSMC_MSG_DPM_Disable);
+}
+
+int kv_copy_bytes_to_smc(struct radeon_device *rdev,
+                        u32 smc_start_address,
+                        const u8 *src, u32 byte_count, u32 limit)
+{
+       int ret;
+       u32 data, original_data, addr, extra_shift, t_byte, count, mask;
+
+       if ((smc_start_address + byte_count) > limit)
+               return -EINVAL;
+
+       addr = smc_start_address;
+       t_byte = addr & 3;
+
+       /* RMW for the initial bytes */
+       if  (t_byte != 0) {
+               addr -= t_byte;
+
+               ret = kv_set_smc_sram_address(rdev, addr, limit);
+               if (ret)
+                       return ret;
+
+               original_data = RREG32(SMC_IND_DATA_0);
+
+               data = 0;
+               mask = 0;
+               count = 4;
+               while (count > 0) {
+                       if (t_byte > 0) {
+                               mask = (mask << 8) | 0xff;
+                               t_byte--;
+                       } else if (byte_count > 0) {
+                               data = (data << 8) + *src++;
+                               byte_count--;
+                               mask <<= 8;
+                       } else {
+                               data <<= 8;
+                               mask = (mask << 8) | 0xff;
+                       }
+                       count--;
+               }
+
+               data |= original_data & mask;
+
+               ret = kv_set_smc_sram_address(rdev, addr, limit);
+               if (ret)
+                       return ret;
+
+               WREG32(SMC_IND_DATA_0, data);
+
+               addr += 4;
+       }
+
+       while (byte_count >= 4) {
+               /* SMC address space is BE */
+               data = (src[0] << 24) + (src[1] << 16) + (src[2] << 8) + src[3];
+
+               ret = kv_set_smc_sram_address(rdev, addr, limit);
+               if (ret)
+                       return ret;
+
+               WREG32(SMC_IND_DATA_0, data);
+
+               src += 4;
+               byte_count -= 4;
+               addr += 4;
+       }
+
+       /* RMW for the final bytes */
+       if (byte_count > 0) {
+               data = 0;
+
+               ret = kv_set_smc_sram_address(rdev, addr, limit);
+               if (ret)
+                       return ret;
+
+               original_data= RREG32(SMC_IND_DATA_0);
+
+               extra_shift = 8 * (4 - byte_count);
+
+               while (byte_count > 0) {
+                       /* SMC address space is BE */
+                       data = (data << 8) + *src++;
+                       byte_count--;
+               }
+
+               data <<= extra_shift;
+
+               data |= (original_data & ~((~0UL) << extra_shift));
+
+               ret = kv_set_smc_sram_address(rdev, addr, limit);
+               if (ret)
+                       return ret;
+
+               WREG32(SMC_IND_DATA_0, data);
+       }
+       return 0;
+}
+
index b5564a3645d2d4c2b90e27267253c344af9eb054..6db6e320bc7999d5224dcb1a1215ba6039ae630e 100644 (file)
@@ -99,11 +99,45 @@ typedef uint8_t PPSMC_Result;
 #define PPSMC_MSG_ThrottleOVRDSCLKDS        ((uint8_t)0x96)
 #define PPSMC_MSG_CancelThrottleOVRDSCLKDS  ((uint8_t)0x97)
 
+/* KV/KB */
+#define PPSMC_MSG_UVDDPM_SetEnabledMask       ((uint16_t) 0x12D)
+#define PPSMC_MSG_VCEDPM_SetEnabledMask       ((uint16_t) 0x12E)
+#define PPSMC_MSG_ACPDPM_SetEnabledMask       ((uint16_t) 0x12F)
+#define PPSMC_MSG_SAMUDPM_SetEnabledMask      ((uint16_t) 0x130)
+#define PPSMC_MSG_MCLKDPM_ForceState          ((uint16_t) 0x131)
+#define PPSMC_MSG_MCLKDPM_NoForcedLevel       ((uint16_t) 0x132)
+#define PPSMC_MSG_Voltage_Cntl_Disable        ((uint16_t) 0x135)
+#define PPSMC_MSG_PCIeDPM_Enable              ((uint16_t) 0x136)
+#define PPSMC_MSG_ACPPowerOFF                 ((uint16_t) 0x137)
+#define PPSMC_MSG_ACPPowerON                  ((uint16_t) 0x138)
+#define PPSMC_MSG_SAMPowerOFF                 ((uint16_t) 0x139)
+#define PPSMC_MSG_SAMPowerON                  ((uint16_t) 0x13a)
+#define PPSMC_MSG_PCIeDPM_Disable             ((uint16_t) 0x13d)
+#define PPSMC_MSG_NBDPM_Enable                ((uint16_t) 0x140)
+#define PPSMC_MSG_NBDPM_Disable               ((uint16_t) 0x141)
+#define PPSMC_MSG_SCLKDPM_SetEnabledMask      ((uint16_t) 0x145)
+#define PPSMC_MSG_DPM_Enable                  ((uint16_t) 0x14e)
+#define PPSMC_MSG_DPM_Disable                 ((uint16_t) 0x14f)
+#define PPSMC_MSG_UVDDPM_Enable               ((uint16_t) 0x154)
+#define PPSMC_MSG_UVDDPM_Disable              ((uint16_t) 0x155)
+#define PPSMC_MSG_SAMUDPM_Enable              ((uint16_t) 0x156)
+#define PPSMC_MSG_SAMUDPM_Disable             ((uint16_t) 0x157)
+#define PPSMC_MSG_ACPDPM_Enable               ((uint16_t) 0x158)
+#define PPSMC_MSG_ACPDPM_Disable              ((uint16_t) 0x159)
+#define PPSMC_MSG_VCEDPM_Enable               ((uint16_t) 0x15a)
+#define PPSMC_MSG_VCEDPM_Disable              ((uint16_t) 0x15b)
+#define PPSMC_MSG_SCLKDPM_GetEnabledMask      ((uint16_t) 0x162)
+#define PPSMC_MSG_SCLKDPM_FreezeLevel         ((uint16_t) 0x189)
+#define PPSMC_MSG_SCLKDPM_UnfreezeLevel       ((uint16_t) 0x18A)
+
 /* TN */
 #define PPSMC_MSG_DPM_Config                ((uint32_t) 0x102)
 #define PPSMC_MSG_DPM_ForceState            ((uint32_t) 0x104)
 #define PPSMC_MSG_PG_SIMD_Config            ((uint32_t) 0x108)
 #define PPSMC_MSG_DPM_N_LevelsDisabled      ((uint32_t) 0x112)
+#define PPSMC_MSG_Voltage_Cntl_Enable       ((uint32_t) 0x109)
+#define PPSMC_MSG_VCEPowerOFF               ((uint32_t) 0x10e)
+#define PPSMC_MSG_VCEPowerON                ((uint32_t) 0x10f)
 #define PPSMC_MSG_DCE_RemoveVoltageAdjustment   ((uint32_t) 0x11d)
 #define PPSMC_MSG_DCE_AllowVoltageAdjustment    ((uint32_t) 0x11e)
 #define PPSMC_MSG_UVD_DPM_Config            ((uint32_t) 0x124)
index 3a55540fe2801a2424854af0e7a89248b46486f1..9c83ecfd0eb77b19a4072e7192a7eb6ffaca12fc 100644 (file)
@@ -2610,6 +2610,20 @@ static struct radeon_asic kv_asic = {
                .set_uvd_clocks = &cik_set_uvd_clocks,
                .get_temperature = &kv_get_temp,
        },
+       .dpm = {
+               .init = &kv_dpm_init,
+               .setup_asic = &kv_dpm_setup_asic,
+               .enable = &kv_dpm_enable,
+               .disable = &kv_dpm_disable,
+               .pre_set_power_state = &kv_dpm_pre_set_power_state,
+               .set_power_state = &kv_dpm_set_power_state,
+               .post_set_power_state = &kv_dpm_post_set_power_state,
+               .display_configuration_changed = &kv_dpm_display_configuration_changed,
+               .fini = &kv_dpm_fini,
+               .get_sclk = &kv_dpm_get_sclk,
+               .get_mclk = &kv_dpm_get_mclk,
+               .print_power_state = &kv_dpm_print_power_state,
+       },
        .pflip = {
                .pre_page_flip = &evergreen_pre_page_flip,
                .page_flip = &evergreen_page_flip,
index d5c6c5b10edfe2b6d4566c5f143b0546494312ce..68a1a1fb371d2d7ecea268ea5468535f425d2654 100644 (file)
@@ -750,4 +750,18 @@ void cik_compute_ring_set_wptr(struct radeon_device *rdev,
 int ci_get_temp(struct radeon_device *rdev);
 int kv_get_temp(struct radeon_device *rdev);
 
+int kv_dpm_init(struct radeon_device *rdev);
+int kv_dpm_enable(struct radeon_device *rdev);
+void kv_dpm_disable(struct radeon_device *rdev);
+int kv_dpm_pre_set_power_state(struct radeon_device *rdev);
+int kv_dpm_set_power_state(struct radeon_device *rdev);
+void kv_dpm_post_set_power_state(struct radeon_device *rdev);
+void kv_dpm_setup_asic(struct radeon_device *rdev);
+void kv_dpm_display_configuration_changed(struct radeon_device *rdev);
+void kv_dpm_fini(struct radeon_device *rdev);
+u32 kv_dpm_get_sclk(struct radeon_device *rdev, bool low);
+u32 kv_dpm_get_mclk(struct radeon_device *rdev, bool low);
+void kv_dpm_print_power_state(struct radeon_device *rdev,
+                             struct radeon_ps *ps);
+
 #endif
index 1408014dce8fbe91193b30f2b024edce8b053f0f..37d3d343f687ddd188dfa5140e503dffd70ded77 100644 (file)
@@ -1202,6 +1202,8 @@ int radeon_pm_init(struct radeon_device *rdev)
        case CHIP_VERDE:
        case CHIP_OLAND:
        case CHIP_HAINAN:
+       case CHIP_KABINI:
+       case CHIP_KAVERI:
                /* DPM requires the RLC, RV770+ dGPU requires SMC */
                if (!rdev->rlc_fw)
                        rdev->pm.pm_method = PM_METHOD_PROFILE;
diff --git a/drivers/gpu/drm/radeon/smu7.h b/drivers/gpu/drm/radeon/smu7.h
new file mode 100644 (file)
index 0000000..75a380a
--- /dev/null
@@ -0,0 +1,170 @@
+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef SMU7_H
+#define SMU7_H
+
+#pragma pack(push, 1)
+
+#define SMU7_CONTEXT_ID_SMC        1
+#define SMU7_CONTEXT_ID_VBIOS      2
+
+
+#define SMU7_CONTEXT_ID_SMC        1
+#define SMU7_CONTEXT_ID_VBIOS      2
+
+#define SMU7_MAX_LEVELS_VDDC            8
+#define SMU7_MAX_LEVELS_VDDCI           4
+#define SMU7_MAX_LEVELS_MVDD            4
+#define SMU7_MAX_LEVELS_VDDNB           8
+
+#define SMU7_MAX_LEVELS_GRAPHICS        SMU__NUM_SCLK_DPM_STATE   // SCLK + SQ DPM + ULV
+#define SMU7_MAX_LEVELS_MEMORY          SMU__NUM_MCLK_DPM_LEVELS   // MCLK Levels DPM
+#define SMU7_MAX_LEVELS_GIO             SMU__NUM_LCLK_DPM_LEVELS  // LCLK Levels
+#define SMU7_MAX_LEVELS_LINK            SMU__NUM_PCIE_DPM_LEVELS  // PCIe speed and number of lanes.
+#define SMU7_MAX_LEVELS_UVD             8   // VCLK/DCLK levels for UVD.
+#define SMU7_MAX_LEVELS_VCE             8   // ECLK levels for VCE.
+#define SMU7_MAX_LEVELS_ACP             8   // ACLK levels for ACP.
+#define SMU7_MAX_LEVELS_SAMU            8   // SAMCLK levels for SAMU.
+#define SMU7_MAX_ENTRIES_SMIO           32  // Number of entries in SMIO table.
+
+#define DPM_NO_LIMIT 0
+#define DPM_NO_UP 1
+#define DPM_GO_DOWN 2
+#define DPM_GO_UP 3
+
+#define SMU7_FIRST_DPM_GRAPHICS_LEVEL    0
+#define SMU7_FIRST_DPM_MEMORY_LEVEL      0
+
+#define GPIO_CLAMP_MODE_VRHOT      1
+#define GPIO_CLAMP_MODE_THERM      2
+#define GPIO_CLAMP_MODE_DC         4
+
+#define SCRATCH_B_TARG_PCIE_INDEX_SHIFT 0
+#define SCRATCH_B_TARG_PCIE_INDEX_MASK  (0x7<<SCRATCH_B_TARG_PCIE_INDEX_SHIFT)
+#define SCRATCH_B_CURR_PCIE_INDEX_SHIFT 3
+#define SCRATCH_B_CURR_PCIE_INDEX_MASK  (0x7<<SCRATCH_B_CURR_PCIE_INDEX_SHIFT)
+#define SCRATCH_B_TARG_UVD_INDEX_SHIFT  6
+#define SCRATCH_B_TARG_UVD_INDEX_MASK   (0x7<<SCRATCH_B_TARG_UVD_INDEX_SHIFT)
+#define SCRATCH_B_CURR_UVD_INDEX_SHIFT  9
+#define SCRATCH_B_CURR_UVD_INDEX_MASK   (0x7<<SCRATCH_B_CURR_UVD_INDEX_SHIFT)
+#define SCRATCH_B_TARG_VCE_INDEX_SHIFT  12
+#define SCRATCH_B_TARG_VCE_INDEX_MASK   (0x7<<SCRATCH_B_TARG_VCE_INDEX_SHIFT)
+#define SCRATCH_B_CURR_VCE_INDEX_SHIFT  15
+#define SCRATCH_B_CURR_VCE_INDEX_MASK   (0x7<<SCRATCH_B_CURR_VCE_INDEX_SHIFT)
+#define SCRATCH_B_TARG_ACP_INDEX_SHIFT  18
+#define SCRATCH_B_TARG_ACP_INDEX_MASK   (0x7<<SCRATCH_B_TARG_ACP_INDEX_SHIFT)
+#define SCRATCH_B_CURR_ACP_INDEX_SHIFT  21
+#define SCRATCH_B_CURR_ACP_INDEX_MASK   (0x7<<SCRATCH_B_CURR_ACP_INDEX_SHIFT)
+#define SCRATCH_B_TARG_SAMU_INDEX_SHIFT 24
+#define SCRATCH_B_TARG_SAMU_INDEX_MASK  (0x7<<SCRATCH_B_TARG_SAMU_INDEX_SHIFT)
+#define SCRATCH_B_CURR_SAMU_INDEX_SHIFT 27
+#define SCRATCH_B_CURR_SAMU_INDEX_MASK  (0x7<<SCRATCH_B_CURR_SAMU_INDEX_SHIFT)
+
+
+struct SMU7_PIDController
+{
+    uint32_t Ki;
+    int32_t LFWindupUL;
+    int32_t LFWindupLL;
+    uint32_t StatePrecision;
+    uint32_t LfPrecision;
+    uint32_t LfOffset;
+    uint32_t MaxState;
+    uint32_t MaxLfFraction;
+    uint32_t StateShift;
+};
+
+typedef struct SMU7_PIDController SMU7_PIDController;
+
+// -------------------------------------------------------------------------------------------------------------------------
+#define SMU7_MAX_PCIE_LINK_SPEEDS 3 /* 0:Gen1 1:Gen2 2:Gen3 */
+
+#define SMU7_SCLK_DPM_CONFIG_MASK                        0x01
+#define SMU7_VOLTAGE_CONTROLLER_CONFIG_MASK              0x02
+#define SMU7_THERMAL_CONTROLLER_CONFIG_MASK              0x04
+#define SMU7_MCLK_DPM_CONFIG_MASK                        0x08
+#define SMU7_UVD_DPM_CONFIG_MASK                         0x10
+#define SMU7_VCE_DPM_CONFIG_MASK                         0x20
+#define SMU7_ACP_DPM_CONFIG_MASK                         0x40
+#define SMU7_SAMU_DPM_CONFIG_MASK                        0x80
+#define SMU7_PCIEGEN_DPM_CONFIG_MASK                    0x100
+
+#define SMU7_ACP_MCLK_HANDSHAKE_DISABLE                  0x00000001
+#define SMU7_ACP_SCLK_HANDSHAKE_DISABLE                  0x00000002
+#define SMU7_UVD_MCLK_HANDSHAKE_DISABLE                  0x00000100
+#define SMU7_UVD_SCLK_HANDSHAKE_DISABLE                  0x00000200
+#define SMU7_VCE_MCLK_HANDSHAKE_DISABLE                  0x00010000
+#define SMU7_VCE_SCLK_HANDSHAKE_DISABLE                  0x00020000
+
+struct SMU7_Firmware_Header
+{
+    uint32_t Digest[5];
+    uint32_t Version;
+    uint32_t HeaderSize;
+    uint32_t Flags;
+    uint32_t EntryPoint;
+    uint32_t CodeSize;
+    uint32_t ImageSize;
+
+    uint32_t Rtos;
+    uint32_t SoftRegisters;
+    uint32_t DpmTable;
+    uint32_t FanTable;
+    uint32_t CacConfigTable;
+    uint32_t CacStatusTable;
+
+    uint32_t mcRegisterTable;
+
+    uint32_t mcArbDramTimingTable;
+
+    uint32_t PmFuseTable;
+    uint32_t Globals;
+    uint32_t Reserved[42];
+    uint32_t Signature;
+};
+
+typedef struct SMU7_Firmware_Header SMU7_Firmware_Header;
+
+#define SMU7_FIRMWARE_HEADER_LOCATION 0x20000
+
+enum  DisplayConfig {
+    PowerDown = 1,
+    DP54x4,
+    DP54x2,
+    DP54x1,
+    DP27x4,
+    DP27x2,
+    DP27x1,
+    HDMI297,
+    HDMI162,
+    LVDS,
+    DP324x4,
+    DP324x2,
+    DP324x1
+};
+
+#pragma pack(pop)
+
+#endif
+
diff --git a/drivers/gpu/drm/radeon/smu7_fusion.h b/drivers/gpu/drm/radeon/smu7_fusion.h
new file mode 100644 (file)
index 0000000..78ada9f
--- /dev/null
@@ -0,0 +1,300 @@
+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef SMU7_FUSION_H
+#define SMU7_FUSION_H
+
+#include "smu7.h"
+
+#pragma pack(push, 1)
+
+#define SMU7_DTE_ITERATIONS 5
+#define SMU7_DTE_SOURCES 5
+#define SMU7_DTE_SINKS 3
+#define SMU7_NUM_CPU_TES 2
+#define SMU7_NUM_GPU_TES 1
+#define SMU7_NUM_NON_TES 2
+
+// All 'soft registers' should be uint32_t.
+struct SMU7_SoftRegisters
+{
+    uint32_t        RefClockFrequency;
+    uint32_t        PmTimerP;
+    uint32_t        FeatureEnables;
+    uint32_t        HandshakeDisables;
+
+    uint8_t         DisplayPhy1Config;
+    uint8_t         DisplayPhy2Config;
+    uint8_t         DisplayPhy3Config;
+    uint8_t         DisplayPhy4Config;
+
+    uint8_t         DisplayPhy5Config;
+    uint8_t         DisplayPhy6Config;
+    uint8_t         DisplayPhy7Config;
+    uint8_t         DisplayPhy8Config;
+
+    uint32_t        AverageGraphicsA;
+    uint32_t        AverageMemoryA;
+    uint32_t        AverageGioA;
+
+    uint8_t         SClkDpmEnabledLevels;
+    uint8_t         MClkDpmEnabledLevels;
+    uint8_t         LClkDpmEnabledLevels;
+    uint8_t         PCIeDpmEnabledLevels;
+
+    uint8_t         UVDDpmEnabledLevels;
+    uint8_t         SAMUDpmEnabledLevels;
+    uint8_t         ACPDpmEnabledLevels;
+    uint8_t         VCEDpmEnabledLevels;
+
+    uint32_t        DRAM_LOG_ADDR_H;
+    uint32_t        DRAM_LOG_ADDR_L;
+    uint32_t        DRAM_LOG_PHY_ADDR_H;
+    uint32_t        DRAM_LOG_PHY_ADDR_L;
+    uint32_t        DRAM_LOG_BUFF_SIZE;
+    uint32_t        UlvEnterC;
+    uint32_t        UlvTime;
+    uint32_t        Reserved[3];
+
+};
+
+typedef struct SMU7_SoftRegisters SMU7_SoftRegisters;
+
+struct SMU7_Fusion_GraphicsLevel
+{
+    uint32_t    MinVddNb;
+
+    uint32_t    SclkFrequency;
+
+    uint8_t     Vid;
+    uint8_t     VidOffset;
+    uint16_t    AT;
+
+    uint8_t     PowerThrottle;
+    uint8_t     GnbSlow;
+    uint8_t     ForceNbPs1;
+    uint8_t     SclkDid;
+
+    uint8_t     DisplayWatermark;
+    uint8_t     EnabledForActivity;
+    uint8_t     EnabledForThrottle;
+    uint8_t     UpH;
+
+    uint8_t     DownH;
+    uint8_t     VoltageDownH;
+    uint8_t     DeepSleepDivId;
+
+    uint8_t     ClkBypassCntl;
+
+    uint32_t    reserved;
+};
+
+typedef struct SMU7_Fusion_GraphicsLevel SMU7_Fusion_GraphicsLevel;
+
+struct SMU7_Fusion_GIOLevel
+{
+    uint8_t     EnabledForActivity;
+    uint8_t     LclkDid;
+    uint8_t     Vid;
+    uint8_t     VoltageDownH;
+
+    uint32_t    MinVddNb;
+
+    uint16_t    ResidencyCounter;
+    uint8_t     UpH;
+    uint8_t     DownH;
+
+    uint32_t    LclkFrequency;
+
+    uint8_t     ActivityLevel;
+    uint8_t     EnabledForThrottle;
+
+    uint8_t     ClkBypassCntl;
+
+    uint8_t     padding;
+};
+
+typedef struct SMU7_Fusion_GIOLevel SMU7_Fusion_GIOLevel;
+
+// UVD VCLK/DCLK state (level) definition.
+struct SMU7_Fusion_UvdLevel
+{
+    uint32_t VclkFrequency;
+    uint32_t DclkFrequency;
+    uint16_t MinVddNb;
+    uint8_t  VclkDivider;
+    uint8_t  DclkDivider;
+
+    uint8_t     VClkBypassCntl;
+    uint8_t     DClkBypassCntl;
+
+    uint8_t     padding[2];
+
+};
+
+typedef struct SMU7_Fusion_UvdLevel SMU7_Fusion_UvdLevel;
+
+// Clocks for other external blocks (VCE, ACP, SAMU).
+struct SMU7_Fusion_ExtClkLevel
+{
+    uint32_t Frequency;
+    uint16_t MinVoltage;
+    uint8_t  Divider;
+    uint8_t  ClkBypassCntl;
+
+    uint32_t Reserved;
+};
+typedef struct SMU7_Fusion_ExtClkLevel SMU7_Fusion_ExtClkLevel;
+
+struct SMU7_Fusion_ACPILevel
+{
+    uint32_t    Flags;
+    uint32_t    MinVddNb;
+    uint32_t    SclkFrequency;
+    uint8_t     SclkDid;
+    uint8_t     GnbSlow;
+    uint8_t     ForceNbPs1;
+    uint8_t     DisplayWatermark;
+    uint8_t     DeepSleepDivId;
+    uint8_t     padding[3];
+};
+
+typedef struct SMU7_Fusion_ACPILevel SMU7_Fusion_ACPILevel;
+
+struct SMU7_Fusion_NbDpm
+{
+    uint8_t DpmXNbPsHi;
+    uint8_t DpmXNbPsLo;
+    uint8_t Dpm0PgNbPsHi;
+    uint8_t Dpm0PgNbPsLo;
+    uint8_t EnablePsi1;
+    uint8_t SkipDPM0;
+    uint8_t SkipPG;
+    uint8_t Hysteresis;
+    uint8_t EnableDpmPstatePoll;
+    uint8_t padding[3];
+};
+
+typedef struct SMU7_Fusion_NbDpm SMU7_Fusion_NbDpm;
+
+struct SMU7_Fusion_StateInfo
+{
+    uint32_t SclkFrequency;
+    uint32_t LclkFrequency;
+    uint32_t VclkFrequency;
+    uint32_t DclkFrequency;
+    uint32_t SamclkFrequency;
+    uint32_t AclkFrequency;
+    uint32_t EclkFrequency;
+    uint8_t  DisplayWatermark;
+    uint8_t  McArbIndex;
+    int8_t   SclkIndex;
+    int8_t   MclkIndex;
+};
+
+typedef struct SMU7_Fusion_StateInfo SMU7_Fusion_StateInfo;
+
+struct SMU7_Fusion_DpmTable
+{
+    uint32_t                            SystemFlags;
+
+    SMU7_PIDController                  GraphicsPIDController;
+    SMU7_PIDController                  GioPIDController;
+
+    uint8_t                            GraphicsDpmLevelCount;
+    uint8_t                            GIOLevelCount;
+    uint8_t                            UvdLevelCount;
+    uint8_t                            VceLevelCount;
+
+    uint8_t                            AcpLevelCount;
+    uint8_t                            SamuLevelCount;
+    uint16_t                           FpsHighT;
+
+    SMU7_Fusion_GraphicsLevel         GraphicsLevel           [SMU__NUM_SCLK_DPM_STATE];
+    SMU7_Fusion_ACPILevel             ACPILevel;
+    SMU7_Fusion_UvdLevel              UvdLevel                [SMU7_MAX_LEVELS_UVD];
+    SMU7_Fusion_ExtClkLevel           VceLevel                [SMU7_MAX_LEVELS_VCE];
+    SMU7_Fusion_ExtClkLevel           AcpLevel                [SMU7_MAX_LEVELS_ACP];
+    SMU7_Fusion_ExtClkLevel           SamuLevel               [SMU7_MAX_LEVELS_SAMU];
+
+    uint8_t                           UvdBootLevel;
+    uint8_t                           VceBootLevel;
+    uint8_t                           AcpBootLevel;
+    uint8_t                           SamuBootLevel;
+    uint8_t                           UVDInterval;
+    uint8_t                           VCEInterval;
+    uint8_t                           ACPInterval;
+    uint8_t                           SAMUInterval;
+
+    uint8_t                           GraphicsBootLevel;
+    uint8_t                           GraphicsInterval;
+    uint8_t                           GraphicsThermThrottleEnable;
+    uint8_t                           GraphicsVoltageChangeEnable;
+
+    uint8_t                           GraphicsClkSlowEnable;
+    uint8_t                           GraphicsClkSlowDivider;
+    uint16_t                          FpsLowT;
+
+    uint32_t                          DisplayCac;
+    uint32_t                          LowSclkInterruptT;
+
+    uint32_t                          DRAM_LOG_ADDR_H;
+    uint32_t                          DRAM_LOG_ADDR_L;
+    uint32_t                          DRAM_LOG_PHY_ADDR_H;
+    uint32_t                          DRAM_LOG_PHY_ADDR_L;
+    uint32_t                          DRAM_LOG_BUFF_SIZE;
+
+};
+
+struct SMU7_Fusion_GIODpmTable
+{
+
+    SMU7_Fusion_GIOLevel              GIOLevel                [SMU7_MAX_LEVELS_GIO];
+
+    SMU7_PIDController                GioPIDController;
+
+    uint32_t                          GIOLevelCount;
+
+    uint8_t                           Enable;
+    uint8_t                           GIOVoltageChangeEnable;
+    uint8_t                           GIOBootLevel;
+    uint8_t                           padding;
+    uint8_t                           padding1[2];
+    uint8_t                           TargetState;
+    uint8_t                           CurrenttState;
+    uint8_t                           ThrottleOnHtc;
+    uint8_t                           ThermThrottleStatus;
+    uint8_t                           ThermThrottleTempSelect;
+    uint8_t                           ThermThrottleEnable;
+    uint16_t                          TemperatureLimitHigh;
+    uint16_t                          TemperatureLimitLow;
+
+};
+
+typedef struct SMU7_Fusion_DpmTable SMU7_Fusion_DpmTable;
+typedef struct SMU7_Fusion_GIODpmTable SMU7_Fusion_GIODpmTable;
+
+#pragma pack(pop)
+
+#endif
+