Merge branch 'drm-next' of git://people.freedesktop.org/~airlied/linux

Pull drm updates from Dave Airlie:
 "Highlights:

   - AMD KFD driver merge

     This is the AMD HSA interface for exposing a lowlevel interface for
     GPGPU use.  They have an open source userspace built on top of this
     interface, and the code looks as good as it was going to get out of
     tree.

   - Initial atomic modesetting work

     The need for an atomic modesetting interface to allow userspace to
     try and send a complete set of modesetting state to the driver has
     arisen, and been suffering from neglect this past year.  No more,
     the start of the common code and changes for msm driver to use it
     are in this tree.  Ongoing work to get the userspace ioctl finished
     and the code clean will probably wait until next kernel.

   - DisplayID 1.3 and tiled monitor exposed to userspace.

     Tiled monitor property is now exposed for userspace to make use of.

   - Rockchip drm driver merged.

   - imx gpu driver moved out of staging

  Other stuff:

   - core:
        panel - MIPI DSI + new panels.
        expose suggested x/y properties for virtual GPUs

   - i915:
        Initial Skylake (SKL) support
        gen3/4 reset work
        start of dri1/ums removal
        infoframe tracking
        fixes for lots of things.

   - nouveau:
        tegra k1 voltage support
        GM204 modesetting support
        GT21x memory reclocking work

   - radeon:
        CI dpm fixes
        GPUVM improvements
        Initial DPM fan control

   - rcar-du:
        HDMI support added
        removed some support for old boards
        slave encoder driver for Analog Devices adv7511

   - exynos:
        Exynos4415 SoC support

   - msm:
        a4xx gpu support
        atomic helper conversion

   - tegra:
        iommu support
        universal plane support
        ganged-mode DSI support

   - sti:
        HDMI i2c improvements

   - vmwgfx:
        some late fixes.

   - qxl:
        use suggested x/y properties"

* 'drm-next' of git://people.freedesktop.org/~airlied/linux: (969 commits)
  drm: sti: fix module compilation issue
  drm/i915: save/restore GMBUS freq across suspend/resume on gen4
  drm: sti: correctly cleanup CRTC and planes
  drm: sti: add HQVDP plane
  drm: sti: add cursor plane
  drm: sti: enable auxiliary CRTC
  drm: sti: fix delay in VTG programming
  drm: sti: prepare sti_tvout to support auxiliary crtc
  drm: sti: use drm_crtc_vblank_{on/off} instead of drm_vblank_{on/off}
  drm: sti: fix hdmi avi infoframe
  drm: sti: remove event lock while disabling vblank
  drm: sti: simplify gdp code
  drm: sti: clear all mixer control
  drm: sti: remove gpio for HDMI hot plug detection
  drm: sti: allow to change hdmi ddc i2c adapter
  drm/doc: Document drm_add_modes_noedid() usage
  drm/i915: Remove '& 0xffff' from the mask given to WA_REG()
  drm/i915: Invert the mask and val arguments in wa_add() and WA_REG()
  drm: Zero out DRM object memory upon cleanup
  drm/i915/bdw: Fix the write setting up the WIZ hashing mode
  ...

50 files changed, 5578 insertions, 1230 deletions

diff --git a/drivers/gpu/drm/msm/Kconfig b/drivers/gpu/drm/msm/Kconfig
index 9d907c526c94..5b2a1ff95d3d 100644
--- a/drivers/gpu/drm/msm/Kconfig
+++ b/drivers/gpu/drm/msm/Kconfig
@@ -3,6 +3,7 @@ config DRM_MSM
         tristate "MSM DRM"
         depends on DRM
         depends on ARCH_QCOM || (ARM && COMPILE_TEST)
+        select REGULATOR
         select DRM_KMS_HELPER
         select DRM_PANEL
         select SHMEM
diff --git a/drivers/gpu/drm/msm/Makefile b/drivers/gpu/drm/msm/Makefile
index 6283dcb96af5..143d988f8add 100644
--- a/drivers/gpu/drm/msm/Makefile
+++ b/drivers/gpu/drm/msm/Makefile
@@ -7,6 +7,7 @@ msm-y := \
         adreno/adreno_device.o \
         adreno/adreno_gpu.o \
         adreno/a3xx_gpu.o \
+        adreno/a4xx_gpu.o \
         hdmi/hdmi.o \
         hdmi/hdmi_audio.o \
         hdmi/hdmi_bridge.o \
@@ -24,12 +25,15 @@ msm-y := \
         mdp/mdp4/mdp4_irq.o \
         mdp/mdp4/mdp4_kms.o \
         mdp/mdp4/mdp4_plane.o \
+        mdp/mdp5/mdp5_cfg.o \
+        mdp/mdp5/mdp5_ctl.o \
         mdp/mdp5/mdp5_crtc.o \
         mdp/mdp5/mdp5_encoder.o \
         mdp/mdp5/mdp5_irq.o \
         mdp/mdp5/mdp5_kms.o \
         mdp/mdp5/mdp5_plane.o \
         mdp/mdp5/mdp5_smp.o \
+        msm_atomic.o \
         msm_drv.o \
         msm_fb.o \
         msm_gem.o \
diff --git a/drivers/gpu/drm/msm/adreno/a2xx.xml.h b/drivers/gpu/drm/msm/adreno/a2xx.xml.h
index a3104598c27f..22882cc0a573 100644
--- a/drivers/gpu/drm/msm/adreno/a2xx.xml.h
+++ b/drivers/gpu/drm/msm/adreno/a2xx.xml.h
@@ -11,10 +11,10 @@ The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/adreno.xml               (    364 bytes, from 2013-11-30 14:47:15)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml  (   1453 bytes, from 2013-03-31 16:51:27)
 - /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml          (  32901 bytes, from 2014-06-02 15:21:30)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml (   9859 bytes, from 2014-06-02 15:21:30)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml    (  14960 bytes, from 2014-07-27 17:22:13)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml          (  58020 bytes, from 2014-08-01 12:22:48)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml          (  41068 bytes, from 2014-08-01 12:22:48)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml (  10551 bytes, from 2014-11-13 22:44:30)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml    (  15053 bytes, from 2014-11-09 15:45:47)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml          (  63169 bytes, from 2014-11-13 22:44:18)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml          (  49097 bytes, from 2014-11-14 15:38:00)
 
 Copyright (C) 2013-2014 by the following authors:
 - Rob Clark <robdclark@gmail.com> (robclark)
@@ -926,11 +926,11 @@ static inline uint32_t A2XX_VGT_DRAW_INITIATOR_INDEX_SIZE(enum pc_di_index_size
 #define A2XX_VGT_DRAW_INITIATOR_NOT_EOP                         0x00001000
 #define A2XX_VGT_DRAW_INITIATOR_SMALL_INDEX                     0x00002000
 #define A2XX_VGT_DRAW_INITIATOR_PRE_DRAW_INITIATOR_ENABLE       0x00004000
-#define A2XX_VGT_DRAW_INITIATOR_NUM_INDICES__MASK               0xffff0000
-#define A2XX_VGT_DRAW_INITIATOR_NUM_INDICES__SHIFT              16
-static inline uint32_t A2XX_VGT_DRAW_INITIATOR_NUM_INDICES(uint32_t val)
+#define A2XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__MASK             0xff000000
+#define A2XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__SHIFT            24
+static inline uint32_t A2XX_VGT_DRAW_INITIATOR_NUM_INSTANCES(uint32_t val)
 {
-        return ((val) << A2XX_VGT_DRAW_INITIATOR_NUM_INDICES__SHIFT) & A2XX_VGT_DRAW_INITIATOR_NUM_INDICES__MASK;
+        return ((val) << A2XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__SHIFT) & A2XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__MASK;
 }
 
 #define REG_A2XX_VGT_IMMED_DATA                                 0x000021fd
@@ -1243,13 +1243,13 @@ static inline uint32_t A2XX_CLEAR_COLOR_ALPHA(uint32_t val)
 #define A2XX_PA_SU_POINT_SIZE_HEIGHT__SHIFT                     0
 static inline uint32_t A2XX_PA_SU_POINT_SIZE_HEIGHT(float val)
 {
-        return ((((uint32_t)(val * 8.0))) << A2XX_PA_SU_POINT_SIZE_HEIGHT__SHIFT) & A2XX_PA_SU_POINT_SIZE_HEIGHT__MASK;
+        return ((((uint32_t)(val * 16.0))) << A2XX_PA_SU_POINT_SIZE_HEIGHT__SHIFT) & A2XX_PA_SU_POINT_SIZE_HEIGHT__MASK;
 }
 #define A2XX_PA_SU_POINT_SIZE_WIDTH__MASK                       0xffff0000
 #define A2XX_PA_SU_POINT_SIZE_WIDTH__SHIFT                      16
 static inline uint32_t A2XX_PA_SU_POINT_SIZE_WIDTH(float val)
 {
-        return ((((uint32_t)(val * 8.0))) << A2XX_PA_SU_POINT_SIZE_WIDTH__SHIFT) & A2XX_PA_SU_POINT_SIZE_WIDTH__MASK;
+        return ((((uint32_t)(val * 16.0))) << A2XX_PA_SU_POINT_SIZE_WIDTH__SHIFT) & A2XX_PA_SU_POINT_SIZE_WIDTH__MASK;
 }
 
 #define REG_A2XX_PA_SU_POINT_MINMAX                             0x00002281
@@ -1257,13 +1257,13 @@ static inline uint32_t A2XX_PA_SU_POINT_SIZE_WIDTH(float val)
 #define A2XX_PA_SU_POINT_MINMAX_MIN__SHIFT                      0
 static inline uint32_t A2XX_PA_SU_POINT_MINMAX_MIN(float val)
 {
-        return ((((uint32_t)(val * 8.0))) << A2XX_PA_SU_POINT_MINMAX_MIN__SHIFT) & A2XX_PA_SU_POINT_MINMAX_MIN__MASK;
+        return ((((uint32_t)(val * 16.0))) << A2XX_PA_SU_POINT_MINMAX_MIN__SHIFT) & A2XX_PA_SU_POINT_MINMAX_MIN__MASK;
 }
 #define A2XX_PA_SU_POINT_MINMAX_MAX__MASK                       0xffff0000
 #define A2XX_PA_SU_POINT_MINMAX_MAX__SHIFT                      16
 static inline uint32_t A2XX_PA_SU_POINT_MINMAX_MAX(float val)
 {
-        return ((((uint32_t)(val * 8.0))) << A2XX_PA_SU_POINT_MINMAX_MAX__SHIFT) & A2XX_PA_SU_POINT_MINMAX_MAX__MASK;
+        return ((((uint32_t)(val * 16.0))) << A2XX_PA_SU_POINT_MINMAX_MAX__SHIFT) & A2XX_PA_SU_POINT_MINMAX_MAX__MASK;
 }
 
 #define REG_A2XX_PA_SU_LINE_CNTL                                0x00002282
@@ -1271,7 +1271,7 @@ static inline uint32_t A2XX_PA_SU_POINT_MINMAX_MAX(float val)
 #define A2XX_PA_SU_LINE_CNTL_WIDTH__SHIFT                       0
 static inline uint32_t A2XX_PA_SU_LINE_CNTL_WIDTH(float val)
 {
-        return ((((uint32_t)(val * 8.0))) << A2XX_PA_SU_LINE_CNTL_WIDTH__SHIFT) & A2XX_PA_SU_LINE_CNTL_WIDTH__MASK;
+        return ((((uint32_t)(val * 16.0))) << A2XX_PA_SU_LINE_CNTL_WIDTH__SHIFT) & A2XX_PA_SU_LINE_CNTL_WIDTH__MASK;
 }
 
 #define REG_A2XX_PA_SC_LINE_STIPPLE                             0x00002283
diff --git a/drivers/gpu/drm/msm/adreno/a3xx.xml.h b/drivers/gpu/drm/msm/adreno/a3xx.xml.h
index 82d015279b47..109e9a263daf 100644
--- a/drivers/gpu/drm/msm/adreno/a3xx.xml.h
+++ b/drivers/gpu/drm/msm/adreno/a3xx.xml.h
@@ -11,10 +11,10 @@ The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/adreno.xml               (    364 bytes, from 2013-11-30 14:47:15)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml  (   1453 bytes, from 2013-03-31 16:51:27)
 - /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml          (  32901 bytes, from 2014-06-02 15:21:30)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml (   9859 bytes, from 2014-06-02 15:21:30)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml    (  14960 bytes, from 2014-07-27 17:22:13)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml          (  58020 bytes, from 2014-08-01 12:22:48)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml          (  41068 bytes, from 2014-08-01 12:22:48)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml (  10551 bytes, from 2014-11-13 22:44:30)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml    (  15053 bytes, from 2014-11-09 15:45:47)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml          (  63169 bytes, from 2014-11-13 22:44:18)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml          (  49097 bytes, from 2014-11-14 15:38:00)
 
 Copyright (C) 2013-2014 by the following authors:
 - Rob Clark <robdclark@gmail.com> (robclark)
@@ -86,6 +86,14 @@ enum a3xx_vtx_fmt {
         VFMT_NORM_USHORT_16_16 = 29,
         VFMT_NORM_USHORT_16_16_16 = 30,
         VFMT_NORM_USHORT_16_16_16_16 = 31,
+        VFMT_UINT_32 = 32,
+        VFMT_UINT_32_32 = 33,
+        VFMT_UINT_32_32_32 = 34,
+        VFMT_UINT_32_32_32_32 = 35,
+        VFMT_INT_32 = 36,
+        VFMT_INT_32_32 = 37,
+        VFMT_INT_32_32_32 = 38,
+        VFMT_INT_32_32_32_32 = 39,
         VFMT_UBYTE_8 = 40,
         VFMT_UBYTE_8_8 = 41,
         VFMT_UBYTE_8_8_8 = 42,
@@ -112,7 +120,9 @@ enum a3xx_tex_fmt {
         TFMT_NORM_USHORT_565 = 4,
         TFMT_NORM_USHORT_5551 = 6,
         TFMT_NORM_USHORT_4444 = 7,
+        TFMT_NORM_USHORT_Z16 = 9,
         TFMT_NORM_UINT_X8Z24 = 10,
+        TFMT_FLOAT_Z32 = 11,
         TFMT_NORM_UINT_NV12_UV_TILED = 17,
         TFMT_NORM_UINT_NV12_Y_TILED = 19,
         TFMT_NORM_UINT_NV12_UV = 21,
@@ -121,18 +131,38 @@ enum a3xx_tex_fmt {
         TFMT_NORM_UINT_I420_U = 26,
         TFMT_NORM_UINT_I420_V = 27,
         TFMT_NORM_UINT_2_10_10_10 = 41,
+        TFMT_FLOAT_9_9_9_E5 = 42,
+        TFMT_FLOAT_10_11_11 = 43,
         TFMT_NORM_UINT_A8 = 44,
         TFMT_NORM_UINT_L8_A8 = 47,
         TFMT_NORM_UINT_8 = 48,
         TFMT_NORM_UINT_8_8 = 49,
         TFMT_NORM_UINT_8_8_8 = 50,
         TFMT_NORM_UINT_8_8_8_8 = 51,
+        TFMT_NORM_SINT_8_8 = 53,
+        TFMT_NORM_SINT_8_8_8_8 = 55,
+        TFMT_UINT_8_8 = 57,
+        TFMT_UINT_8_8_8_8 = 59,
+        TFMT_SINT_8_8 = 61,
+        TFMT_SINT_8_8_8_8 = 63,
         TFMT_FLOAT_16 = 64,
         TFMT_FLOAT_16_16 = 65,
         TFMT_FLOAT_16_16_16_16 = 67,
+        TFMT_UINT_16 = 68,
+        TFMT_UINT_16_16 = 69,
+        TFMT_UINT_16_16_16_16 = 71,
+        TFMT_SINT_16 = 72,
+        TFMT_SINT_16_16 = 73,
+        TFMT_SINT_16_16_16_16 = 75,
         TFMT_FLOAT_32 = 84,
         TFMT_FLOAT_32_32 = 85,
         TFMT_FLOAT_32_32_32_32 = 87,
+        TFMT_UINT_32 = 88,
+        TFMT_UINT_32_32 = 89,
+        TFMT_UINT_32_32_32_32 = 91,
+        TFMT_SINT_32 = 92,
+        TFMT_SINT_32_32 = 93,
+        TFMT_SINT_32_32_32_32 = 95,
 };
 
 enum a3xx_tex_fetchsize {
@@ -145,19 +175,34 @@ enum a3xx_tex_fetchsize {
 };
 
 enum a3xx_color_fmt {
+        RB_R5G6B5_UNORM = 0,
+        RB_R5G5B5A1_UNORM = 1,
+        RB_R4G4B4A4_UNORM = 3,
         RB_R8G8B8_UNORM = 4,
         RB_R8G8B8A8_UNORM = 8,
-        RB_Z16_UNORM = 12,
+        RB_R8G8B8A8_UINT = 10,
+        RB_R8G8B8A8_SINT = 11,
+        RB_R8G8_UNORM = 12,
+        RB_R8_UINT = 14,
+        RB_R8_SINT = 15,
+        RB_R10G10B10A2_UNORM = 16,
         RB_A8_UNORM = 20,
+        RB_R8_UNORM = 21,
         RB_R16G16B16A16_FLOAT = 27,
+        RB_R11G11B10_FLOAT = 28,
+        RB_R16_SINT = 40,
+        RB_R16G16_SINT = 41,
+        RB_R16G16B16A16_SINT = 43,
+        RB_R16_UINT = 44,
+        RB_R16G16_UINT = 45,
+        RB_R16G16B16A16_UINT = 47,
         RB_R32G32B32A32_FLOAT = 51,
-};
-
-enum a3xx_color_swap {
-        WZYX = 0,
-        WXYZ = 1,
-        ZYXW = 2,
-        XYZW = 3,
+        RB_R32_SINT = 52,
+        RB_R32G32_SINT = 53,
+        RB_R32G32B32A32_SINT = 55,
+        RB_R32_UINT = 56,
+        RB_R32G32_UINT = 57,
+        RB_R32G32B32A32_UINT = 59,
 };
 
 enum a3xx_sp_perfcounter_select {
@@ -194,6 +239,11 @@ enum a3xx_rb_blend_opcode {
         BLEND_MAX_DST_SRC = 4,
 };
 
+enum a3xx_intp_mode {
+        SMOOTH = 0,
+        FLAT = 1,
+};
+
 enum a3xx_tex_filter {
         A3XX_TEX_NEAREST = 0,
         A3XX_TEX_LINEAR = 1,
@@ -536,6 +586,10 @@ enum a3xx_tex_type {
 
 #define REG_A3XX_CP_MEQ_DATA                                    0x000001db
 
+#define REG_A3XX_CP_WFI_PEND_CTR                                0x000001f5
+
+#define REG_A3XX_RBBM_PM_OVERRIDE2                              0x0000039d
+
 #define REG_A3XX_CP_PERFCOUNTER_SELECT                          0x00000445
 
 #define REG_A3XX_CP_HW_FAULT                                    0x0000045c
@@ -550,6 +604,12 @@ static inline uint32_t REG_A3XX_CP_PROTECT_REG(uint32_t i0) { return 0x00000460
 
 #define REG_A3XX_CP_AHB_FAULT                                   0x0000054d
 
+#define REG_A3XX_SQ_GPR_MANAGEMENT                              0x00000d00
+
+#define REG_A3XX_SQ_INST_STORE_MANAGMENT                        0x00000d02
+
+#define REG_A3XX_TP0_CHICKEN                                    0x00000e1e
+
 #define REG_A3XX_SP_GLOBAL_MEM_SIZE                             0x00000e22
 
 #define REG_A3XX_SP_GLOBAL_MEM_ADDR                             0x00000e23
@@ -632,13 +692,13 @@ static inline uint32_t A3XX_GRAS_CL_VPORT_ZSCALE(float val)
 #define A3XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT                    0
 static inline uint32_t A3XX_GRAS_SU_POINT_MINMAX_MIN(float val)
 {
-        return ((((uint32_t)(val * 8.0))) << A3XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT) & A3XX_GRAS_SU_POINT_MINMAX_MIN__MASK;
+        return ((((uint32_t)(val * 16.0))) << A3XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT) & A3XX_GRAS_SU_POINT_MINMAX_MIN__MASK;
 }
 #define A3XX_GRAS_SU_POINT_MINMAX_MAX__MASK                     0xffff0000
 #define A3XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT                    16
 static inline uint32_t A3XX_GRAS_SU_POINT_MINMAX_MAX(float val)
 {
-        return ((((uint32_t)(val * 8.0))) << A3XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT) & A3XX_GRAS_SU_POINT_MINMAX_MAX__MASK;
+        return ((((uint32_t)(val * 16.0))) << A3XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT) & A3XX_GRAS_SU_POINT_MINMAX_MAX__MASK;
 }
 
 #define REG_A3XX_GRAS_SU_POINT_SIZE                             0x00002069
@@ -646,7 +706,7 @@ static inline uint32_t A3XX_GRAS_SU_POINT_MINMAX_MAX(float val)
 #define A3XX_GRAS_SU_POINT_SIZE__SHIFT                          0
 static inline uint32_t A3XX_GRAS_SU_POINT_SIZE(float val)
 {
-        return ((((uint32_t)(val * 8.0))) << A3XX_GRAS_SU_POINT_SIZE__SHIFT) & A3XX_GRAS_SU_POINT_SIZE__MASK;
+        return ((((int32_t)(val * 16.0))) << A3XX_GRAS_SU_POINT_SIZE__SHIFT) & A3XX_GRAS_SU_POINT_SIZE__MASK;
 }
 
 #define REG_A3XX_GRAS_SU_POLY_OFFSET_SCALE                      0x0000206c
@@ -654,7 +714,7 @@ static inline uint32_t A3XX_GRAS_SU_POINT_SIZE(float val)
 #define A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL__SHIFT               0
 static inline uint32_t A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL(float val)
 {
-        return ((((uint32_t)(val * 28.0))) << A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL__SHIFT) & A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL__MASK;
+        return ((((int32_t)(val * 16384.0))) << A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL__SHIFT) & A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL__MASK;
 }
 
 #define REG_A3XX_GRAS_SU_POLY_OFFSET_OFFSET                     0x0000206d
@@ -662,7 +722,7 @@ static inline uint32_t A3XX_GRAS_SU_POLY_OFFSET_SCALE_VAL(float val)
 #define A3XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT                  0
 static inline uint32_t A3XX_GRAS_SU_POLY_OFFSET_OFFSET(float val)
 {
-        return ((((uint32_t)(val * 28.0))) << A3XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT) & A3XX_GRAS_SU_POLY_OFFSET_OFFSET__MASK;
+        return ((((int32_t)(val * 16384.0))) << A3XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT) & A3XX_GRAS_SU_POLY_OFFSET_OFFSET__MASK;
 }
 
 #define REG_A3XX_GRAS_SU_MODE_CONTROL                           0x00002070
@@ -673,7 +733,7 @@ static inline uint32_t A3XX_GRAS_SU_POLY_OFFSET_OFFSET(float val)
 #define A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__SHIFT          3
 static inline uint32_t A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH(float val)
 {
-        return ((((uint32_t)(val * 4.0))) << A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__SHIFT) & A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__MASK;
+        return ((((int32_t)(val * 4.0))) << A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__SHIFT) & A3XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__MASK;
 }
 #define A3XX_GRAS_SU_MODE_CONTROL_POLY_OFFSET                   0x00000800
 
@@ -863,6 +923,7 @@ static inline uint32_t A3XX_RB_MRT_BUF_INFO_COLOR_SWAP(enum a3xx_color_swap val)
 {
         return ((val) << A3XX_RB_MRT_BUF_INFO_COLOR_SWAP__SHIFT) & A3XX_RB_MRT_BUF_INFO_COLOR_SWAP__MASK;
 }
+#define A3XX_RB_MRT_BUF_INFO_COLOR_SRGB                         0x00004000
 #define A3XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__MASK              0xfffe0000
 #define A3XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__SHIFT             17
 static inline uint32_t A3XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH(uint32_t val)
@@ -1001,6 +1062,7 @@ static inline uint32_t A3XX_RB_COPY_CONTROL_FASTCLEAR(uint32_t val)
 {
         return ((val) << A3XX_RB_COPY_CONTROL_FASTCLEAR__SHIFT) & A3XX_RB_COPY_CONTROL_FASTCLEAR__MASK;
 }
+#define A3XX_RB_COPY_CONTROL_UNK12                              0x00001000
 #define A3XX_RB_COPY_CONTROL_GMEM_BASE__MASK                    0xffffc000
 #define A3XX_RB_COPY_CONTROL_GMEM_BASE__SHIFT                   14
 static inline uint32_t A3XX_RB_COPY_CONTROL_GMEM_BASE(uint32_t val)
@@ -1079,7 +1141,7 @@ static inline uint32_t A3XX_RB_DEPTH_CONTROL_ZFUNC(enum adreno_compare_func val)
 #define REG_A3XX_RB_DEPTH_CLEAR                                 0x00002101
 
 #define REG_A3XX_RB_DEPTH_INFO                                  0x00002102
-#define A3XX_RB_DEPTH_INFO_DEPTH_FORMAT__MASK                   0x00000001
+#define A3XX_RB_DEPTH_INFO_DEPTH_FORMAT__MASK                   0x00000003
 #define A3XX_RB_DEPTH_INFO_DEPTH_FORMAT__SHIFT                  0
 static inline uint32_t A3XX_RB_DEPTH_INFO_DEPTH_FORMAT(enum adreno_rb_depth_format val)
 {
@@ -1265,6 +1327,7 @@ static inline uint32_t A3XX_PC_PRIM_VTX_CNTL_POLYMODE_BACK_PTYPE(enum adreno_pa_
 {
         return ((val) << A3XX_PC_PRIM_VTX_CNTL_POLYMODE_BACK_PTYPE__SHIFT) & A3XX_PC_PRIM_VTX_CNTL_POLYMODE_BACK_PTYPE__MASK;
 }
+#define A3XX_PC_PRIM_VTX_CNTL_PRIMITIVE_RESTART                 0x00100000
 #define A3XX_PC_PRIM_VTX_CNTL_PROVOKING_VTX_LAST                0x02000000
 #define A3XX_PC_PRIM_VTX_CNTL_PSIZE                             0x04000000
 
@@ -1281,7 +1344,12 @@ static inline uint32_t A3XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE(enum a3xx_threadsize
 #define A3XX_HLSQ_CONTROL_0_REG_SPSHADERRESTART                 0x00000200
 #define A3XX_HLSQ_CONTROL_0_REG_RESERVED2                       0x00000400
 #define A3XX_HLSQ_CONTROL_0_REG_CHUNKDISABLE                    0x04000000
-#define A3XX_HLSQ_CONTROL_0_REG_CONSTSWITCHMODE                 0x08000000
+#define A3XX_HLSQ_CONTROL_0_REG_CONSTMODE__MASK                 0x08000000
+#define A3XX_HLSQ_CONTROL_0_REG_CONSTMODE__SHIFT                27
+static inline uint32_t A3XX_HLSQ_CONTROL_0_REG_CONSTMODE(uint32_t val)
+{
+        return ((val) << A3XX_HLSQ_CONTROL_0_REG_CONSTMODE__SHIFT) & A3XX_HLSQ_CONTROL_0_REG_CONSTMODE__MASK;
+}
 #define A3XX_HLSQ_CONTROL_0_REG_LAZYUPDATEDISABLE               0x10000000
 #define A3XX_HLSQ_CONTROL_0_REG_SPCONSTFULLUPDATE               0x20000000
 #define A3XX_HLSQ_CONTROL_0_REG_TPFULLUPDATE                    0x40000000
@@ -1484,6 +1552,8 @@ static inline uint32_t A3XX_VFD_CONTROL_1_REGID4INST(uint32_t val)
 
 #define REG_A3XX_VFD_INDEX_OFFSET                               0x00002245
 
+#define REG_A3XX_VFD_INDEX_OFFSET                               0x00002245
+
 static inline uint32_t REG_A3XX_VFD_FETCH(uint32_t i0) { return 0x00002246 + 0x2*i0; }
 
 static inline uint32_t REG_A3XX_VFD_FETCH_INSTR_0(uint32_t i0) { return 0x00002246 + 0x2*i0; }
@@ -1537,6 +1607,7 @@ static inline uint32_t A3XX_VFD_DECODE_INSTR_REGID(uint32_t val)
 {
         return ((val) << A3XX_VFD_DECODE_INSTR_REGID__SHIFT) & A3XX_VFD_DECODE_INSTR_REGID__MASK;
 }
+#define A3XX_VFD_DECODE_INSTR_INT                               0x00100000
 #define A3XX_VFD_DECODE_INSTR_SWAP__MASK                        0x00c00000
 #define A3XX_VFD_DECODE_INSTR_SWAP__SHIFT                       22
 static inline uint32_t A3XX_VFD_DECODE_INSTR_SWAP(enum a3xx_color_swap val)
@@ -1604,6 +1675,102 @@ static inline uint32_t A3XX_VPC_PACK_NUMNONPOSVSVAR(uint32_t val)
 static inline uint32_t REG_A3XX_VPC_VARYING_INTERP(uint32_t i0) { return 0x00002282 + 0x1*i0; }
 
 static inline uint32_t REG_A3XX_VPC_VARYING_INTERP_MODE(uint32_t i0) { return 0x00002282 + 0x1*i0; }
+#define A3XX_VPC_VARYING_INTERP_MODE_C0__MASK                   0x00000003
+#define A3XX_VPC_VARYING_INTERP_MODE_C0__SHIFT                  0
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C0(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C0__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C0__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_C1__MASK                   0x0000000c
+#define A3XX_VPC_VARYING_INTERP_MODE_C1__SHIFT                  2
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C1(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C1__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C1__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_C2__MASK                   0x00000030
+#define A3XX_VPC_VARYING_INTERP_MODE_C2__SHIFT                  4
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C2(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C2__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C2__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_C3__MASK                   0x000000c0
+#define A3XX_VPC_VARYING_INTERP_MODE_C3__SHIFT                  6
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C3(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C3__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C3__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_C4__MASK                   0x00000300
+#define A3XX_VPC_VARYING_INTERP_MODE_C4__SHIFT                  8
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C4(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C4__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C4__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_C5__MASK                   0x00000c00
+#define A3XX_VPC_VARYING_INTERP_MODE_C5__SHIFT                  10
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C5(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C5__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C5__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_C6__MASK                   0x00003000
+#define A3XX_VPC_VARYING_INTERP_MODE_C6__SHIFT                  12
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C6(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C6__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C6__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_C7__MASK                   0x0000c000
+#define A3XX_VPC_VARYING_INTERP_MODE_C7__SHIFT                  14
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C7(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C7__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C7__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_C8__MASK                   0x00030000
+#define A3XX_VPC_VARYING_INTERP_MODE_C8__SHIFT                  16
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C8(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C8__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C8__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_C9__MASK                   0x000c0000
+#define A3XX_VPC_VARYING_INTERP_MODE_C9__SHIFT                  18
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_C9(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_C9__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_C9__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_CA__MASK                   0x00300000
+#define A3XX_VPC_VARYING_INTERP_MODE_CA__SHIFT                  20
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CA(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CA__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CA__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_CB__MASK                   0x00c00000
+#define A3XX_VPC_VARYING_INTERP_MODE_CB__SHIFT                  22
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CB(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CB__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CB__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_CC__MASK                   0x03000000
+#define A3XX_VPC_VARYING_INTERP_MODE_CC__SHIFT                  24
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CC(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CC__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CC__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_CD__MASK                   0x0c000000
+#define A3XX_VPC_VARYING_INTERP_MODE_CD__SHIFT                  26
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CD(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CD__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CD__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_CE__MASK                   0x30000000
+#define A3XX_VPC_VARYING_INTERP_MODE_CE__SHIFT                  28
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CE(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CE__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CE__MASK;
+}
+#define A3XX_VPC_VARYING_INTERP_MODE_CF__MASK                   0xc0000000
+#define A3XX_VPC_VARYING_INTERP_MODE_CF__SHIFT                  30
+static inline uint32_t A3XX_VPC_VARYING_INTERP_MODE_CF(enum a3xx_intp_mode val)
+{
+        return ((val) << A3XX_VPC_VARYING_INTERP_MODE_CF__SHIFT) & A3XX_VPC_VARYING_INTERP_MODE_CF__MASK;
+}
 
 static inline uint32_t REG_A3XX_VPC_VARYING_PS_REPL(uint32_t i0) { return 0x00002286 + 0x1*i0; }
 
@@ -1928,6 +2095,8 @@ static inline uint32_t A3XX_SP_FS_MRT_REG_REGID(uint32_t val)
         return ((val) << A3XX_SP_FS_MRT_REG_REGID__SHIFT) & A3XX_SP_FS_MRT_REG_REGID__MASK;
 }
 #define A3XX_SP_FS_MRT_REG_HALF_PRECISION                       0x00000100
+#define A3XX_SP_FS_MRT_REG_SINT                                 0x00000400
+#define A3XX_SP_FS_MRT_REG_UINT                                 0x00000800
 
 static inline uint32_t REG_A3XX_SP_FS_IMAGE_OUTPUT(uint32_t i0) { return 0x000022f4 + 0x1*i0; }
 
@@ -1947,6 +2116,8 @@ static inline uint32_t A3XX_SP_FS_LENGTH_REG_SHADERLENGTH(uint32_t val)
         return ((val) << A3XX_SP_FS_LENGTH_REG_SHADERLENGTH__SHIFT) & A3XX_SP_FS_LENGTH_REG_SHADERLENGTH__MASK;
 }
 
+#define REG_A3XX_PA_SC_AA_CONFIG                                0x00002301
+
 #define REG_A3XX_TPL1_TP_VS_TEX_OFFSET                          0x00002340
 #define A3XX_TPL1_TP_VS_TEX_OFFSET_SAMPLEROFFSET__MASK          0x000000ff
 #define A3XX_TPL1_TP_VS_TEX_OFFSET_SAMPLEROFFSET__SHIFT         0
@@ -2297,11 +2468,11 @@ static inline uint32_t A3XX_VGT_DRAW_INITIATOR_INDEX_SIZE(enum pc_di_index_size
 #define A3XX_VGT_DRAW_INITIATOR_NOT_EOP                         0x00001000
 #define A3XX_VGT_DRAW_INITIATOR_SMALL_INDEX                     0x00002000
 #define A3XX_VGT_DRAW_INITIATOR_PRE_DRAW_INITIATOR_ENABLE       0x00004000
-#define A3XX_VGT_DRAW_INITIATOR_NUM_INDICES__MASK               0xffff0000
-#define A3XX_VGT_DRAW_INITIATOR_NUM_INDICES__SHIFT              16
-static inline uint32_t A3XX_VGT_DRAW_INITIATOR_NUM_INDICES(uint32_t val)
+#define A3XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__MASK             0xff000000
+#define A3XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__SHIFT            24
+static inline uint32_t A3XX_VGT_DRAW_INITIATOR_NUM_INSTANCES(uint32_t val)
 {
-        return ((val) << A3XX_VGT_DRAW_INITIATOR_NUM_INDICES__SHIFT) & A3XX_VGT_DRAW_INITIATOR_NUM_INDICES__MASK;
+        return ((val) << A3XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__SHIFT) & A3XX_VGT_DRAW_INITIATOR_NUM_INSTANCES__MASK;
 }
 
 #define REG_A3XX_VGT_IMMED_DATA                                 0x000021fd
@@ -2347,17 +2518,23 @@ static inline uint32_t A3XX_TEX_SAMP_0_COMPARE_FUNC(enum adreno_compare_func val
 #define A3XX_TEX_SAMP_0_UNNORM_COORDS                           0x80000000
 
 #define REG_A3XX_TEX_SAMP_1                                     0x00000001
+#define A3XX_TEX_SAMP_1_LOD_BIAS__MASK                          0x000007ff
+#define A3XX_TEX_SAMP_1_LOD_BIAS__SHIFT                         0
+static inline uint32_t A3XX_TEX_SAMP_1_LOD_BIAS(float val)
+{
+        return ((((int32_t)(val * 64.0))) << A3XX_TEX_SAMP_1_LOD_BIAS__SHIFT) & A3XX_TEX_SAMP_1_LOD_BIAS__MASK;
+}
 #define A3XX_TEX_SAMP_1_MAX_LOD__MASK                           0x003ff000
 #define A3XX_TEX_SAMP_1_MAX_LOD__SHIFT                          12
 static inline uint32_t A3XX_TEX_SAMP_1_MAX_LOD(float val)
 {
-        return ((((uint32_t)(val * 12.0))) << A3XX_TEX_SAMP_1_MAX_LOD__SHIFT) & A3XX_TEX_SAMP_1_MAX_LOD__MASK;
+        return ((((uint32_t)(val * 64.0))) << A3XX_TEX_SAMP_1_MAX_LOD__SHIFT) & A3XX_TEX_SAMP_1_MAX_LOD__MASK;
 }
 #define A3XX_TEX_SAMP_1_MIN_LOD__MASK                           0xffc00000
 #define A3XX_TEX_SAMP_1_MIN_LOD__SHIFT                          22
 static inline uint32_t A3XX_TEX_SAMP_1_MIN_LOD(float val)
 {
-        return ((((uint32_t)(val * 12.0))) << A3XX_TEX_SAMP_1_MIN_LOD__SHIFT) & A3XX_TEX_SAMP_1_MIN_LOD__MASK;
+        return ((((uint32_t)(val * 64.0))) << A3XX_TEX_SAMP_1_MIN_LOD__SHIFT) & A3XX_TEX_SAMP_1_MIN_LOD__MASK;
 }
 
 #define REG_A3XX_TEX_CONST_0                                    0x00000000
@@ -2448,6 +2625,24 @@ static inline uint32_t A3XX_TEX_CONST_2_SWAP(enum a3xx_color_swap val)
 }
 
 #define REG_A3XX_TEX_CONST_3                                    0x00000003
+#define A3XX_TEX_CONST_3_LAYERSZ1__MASK                         0x0000000f
+#define A3XX_TEX_CONST_3_LAYERSZ1__SHIFT                        0
+static inline uint32_t A3XX_TEX_CONST_3_LAYERSZ1(uint32_t val)
+{
+        return ((val >> 12) << A3XX_TEX_CONST_3_LAYERSZ1__SHIFT) & A3XX_TEX_CONST_3_LAYERSZ1__MASK;
+}
+#define A3XX_TEX_CONST_3_DEPTH__MASK                            0x0ffe0000
+#define A3XX_TEX_CONST_3_DEPTH__SHIFT                           17
+static inline uint32_t A3XX_TEX_CONST_3_DEPTH(uint32_t val)
+{
+        return ((val) << A3XX_TEX_CONST_3_DEPTH__SHIFT) & A3XX_TEX_CONST_3_DEPTH__MASK;
+}
+#define A3XX_TEX_CONST_3_LAYERSZ2__MASK                         0xf0000000
+#define A3XX_TEX_CONST_3_LAYERSZ2__SHIFT                        28
+static inline uint32_t A3XX_TEX_CONST_3_LAYERSZ2(uint32_t val)
+{
+        return ((val >> 12) << A3XX_TEX_CONST_3_LAYERSZ2__SHIFT) & A3XX_TEX_CONST_3_LAYERSZ2__MASK;
+}
 
 
 #endif /* A3XX_XML */
diff --git a/drivers/gpu/drm/msm/adreno/a3xx_gpu.c b/drivers/gpu/drm/msm/adreno/a3xx_gpu.c
index 218c5b060398..b66c53bdc039 100644
--- a/drivers/gpu/drm/msm/adreno/a3xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a3xx_gpu.c
@@ -2,6 +2,8 @@
  * Copyright (C) 2013 Red Hat
  * Author: Rob Clark <robdclark@gmail.com>
  *
+ * Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published by
  * the Free Software Foundation.
@@ -406,6 +408,94 @@ static void a3xx_dump(struct msm_gpu *gpu)
                         gpu_read(gpu, REG_A3XX_RBBM_STATUS));
         adreno_dump(gpu);
 }
+/* Register offset defines for A3XX */
+static const unsigned int a3xx_register_offsets[REG_ADRENO_REGISTER_MAX] = {
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_DEBUG, REG_AXXX_CP_DEBUG),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_WADDR, REG_AXXX_CP_ME_RAM_WADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_DATA, REG_AXXX_CP_ME_RAM_DATA),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_PFP_UCODE_DATA,
+                        REG_A3XX_CP_PFP_UCODE_DATA),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_PFP_UCODE_ADDR,
+                        REG_A3XX_CP_PFP_UCODE_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_WFI_PEND_CTR, REG_A3XX_CP_WFI_PEND_CTR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_BASE, REG_AXXX_CP_RB_BASE),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_RPTR_ADDR, REG_AXXX_CP_RB_RPTR_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_RPTR, REG_AXXX_CP_RB_RPTR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_WPTR, REG_AXXX_CP_RB_WPTR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_PROTECT_CTRL, REG_A3XX_CP_PROTECT_CTRL),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_CNTL, REG_AXXX_CP_ME_CNTL),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_CNTL, REG_AXXX_CP_RB_CNTL),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_IB1_BASE, REG_AXXX_CP_IB1_BASE),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_IB1_BUFSZ, REG_AXXX_CP_IB1_BUFSZ),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_IB2_BASE, REG_AXXX_CP_IB2_BASE),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_IB2_BUFSZ, REG_AXXX_CP_IB2_BUFSZ),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_TIMESTAMP, REG_AXXX_CP_SCRATCH_REG0),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_RADDR, REG_AXXX_CP_ME_RAM_RADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_SCRATCH_ADDR, REG_AXXX_SCRATCH_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_SCRATCH_UMSK, REG_AXXX_SCRATCH_UMSK),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ROQ_ADDR, REG_A3XX_CP_ROQ_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ROQ_DATA, REG_A3XX_CP_ROQ_DATA),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_ADDR, REG_A3XX_CP_MERCIU_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_DATA, REG_A3XX_CP_MERCIU_DATA),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_DATA2, REG_A3XX_CP_MERCIU_DATA2),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_MEQ_ADDR, REG_A3XX_CP_MEQ_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_MEQ_DATA, REG_A3XX_CP_MEQ_DATA),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_HW_FAULT, REG_A3XX_CP_HW_FAULT),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_PROTECT_STATUS,
+                        REG_A3XX_CP_PROTECT_STATUS),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_STATUS, REG_A3XX_RBBM_STATUS),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_CTL,
+                        REG_A3XX_RBBM_PERFCTR_CTL),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_CMD0,
+                        REG_A3XX_RBBM_PERFCTR_LOAD_CMD0),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_CMD1,
+                        REG_A3XX_RBBM_PERFCTR_LOAD_CMD1),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_PWR_1_LO,
+                        REG_A3XX_RBBM_PERFCTR_PWR_1_LO),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_0_MASK, REG_A3XX_RBBM_INT_0_MASK),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_0_STATUS,
+                        REG_A3XX_RBBM_INT_0_STATUS),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_ERROR_STATUS,
+                        REG_A3XX_RBBM_AHB_ERROR_STATUS),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_CMD, REG_A3XX_RBBM_AHB_CMD),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_CLEAR_CMD,
+                        REG_A3XX_RBBM_INT_CLEAR_CMD),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_CLOCK_CTL, REG_A3XX_RBBM_CLOCK_CTL),
+        REG_ADRENO_DEFINE(REG_ADRENO_VPC_DEBUG_RAM_SEL,
+                        REG_A3XX_VPC_VPC_DEBUG_RAM_SEL),
+        REG_ADRENO_DEFINE(REG_ADRENO_VPC_DEBUG_RAM_READ,
+                        REG_A3XX_VPC_VPC_DEBUG_RAM_READ),
+        REG_ADRENO_DEFINE(REG_ADRENO_VSC_SIZE_ADDRESS,
+                        REG_A3XX_VSC_SIZE_ADDRESS),
+        REG_ADRENO_DEFINE(REG_ADRENO_VFD_CONTROL_0, REG_A3XX_VFD_CONTROL_0),
+        REG_ADRENO_DEFINE(REG_ADRENO_VFD_INDEX_MAX, REG_A3XX_VFD_INDEX_MAX),
+        REG_ADRENO_DEFINE(REG_ADRENO_SP_VS_PVT_MEM_ADDR_REG,
+                        REG_A3XX_SP_VS_PVT_MEM_ADDR_REG),
+        REG_ADRENO_DEFINE(REG_ADRENO_SP_FS_PVT_MEM_ADDR_REG,
+                        REG_A3XX_SP_FS_PVT_MEM_ADDR_REG),
+        REG_ADRENO_DEFINE(REG_ADRENO_SP_VS_OBJ_START_REG,
+                        REG_A3XX_SP_VS_OBJ_START_REG),
+        REG_ADRENO_DEFINE(REG_ADRENO_SP_FS_OBJ_START_REG,
+                        REG_A3XX_SP_FS_OBJ_START_REG),
+        REG_ADRENO_DEFINE(REG_ADRENO_PA_SC_AA_CONFIG, REG_A3XX_PA_SC_AA_CONFIG),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PM_OVERRIDE2,
+                        REG_A3XX_RBBM_PM_OVERRIDE2),
+        REG_ADRENO_DEFINE(REG_ADRENO_SCRATCH_REG2, REG_AXXX_CP_SCRATCH_REG2),
+        REG_ADRENO_DEFINE(REG_ADRENO_SQ_GPR_MANAGEMENT,
+                        REG_A3XX_SQ_GPR_MANAGEMENT),
+        REG_ADRENO_DEFINE(REG_ADRENO_SQ_INST_STORE_MANAGMENT,
+                        REG_A3XX_SQ_INST_STORE_MANAGMENT),
+        REG_ADRENO_DEFINE(REG_ADRENO_TP0_CHICKEN, REG_A3XX_TP0_CHICKEN),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_RBBM_CTL, REG_A3XX_RBBM_RBBM_CTL),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_SW_RESET_CMD,
+                        REG_A3XX_RBBM_SW_RESET_CMD),
+        REG_ADRENO_DEFINE(REG_ADRENO_UCHE_INVALIDATE0,
+                        REG_A3XX_UCHE_CACHE_INVALIDATE0_REG),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_LO,
+                        REG_A3XX_RBBM_PERFCTR_LOAD_VALUE_LO),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_HI,
+                        REG_A3XX_RBBM_PERFCTR_LOAD_VALUE_HI),
+};
 
 static const struct adreno_gpu_funcs funcs = {
         .base = {
@@ -463,6 +553,7 @@ struct msm_gpu *a3xx_gpu_init(struct drm_device *dev)
         gpu->num_perfcntrs = ARRAY_SIZE(perfcntrs);
 
         adreno_gpu->registers = a3xx_registers;
+        adreno_gpu->reg_offsets = a3xx_register_offsets;
 
         ret = adreno_gpu_init(dev, pdev, adreno_gpu, &funcs);
         if (ret)
diff --git a/drivers/gpu/drm/msm/adreno/a4xx.xml.h b/drivers/gpu/drm/msm/adreno/a4xx.xml.h
new file mode 100644
index 000000000000..5a24c416d2dd
--- /dev/null
+++ b/drivers/gpu/drm/msm/adreno/a4xx.xml.h
@@ -0,0 +1,2144 @@
+#ifndef A4XX_XML
+#define A4XX_XML
+
+/* Autogenerated file, DO NOT EDIT manually!
+
+This file was generated by the rules-ng-ng headergen tool in this git repository:
+http://github.com/freedreno/envytools/
+git clone https://github.com/freedreno/envytools.git
+
+The rules-ng-ng source files this header was generated from are:
+- /home/robclark/src/freedreno/envytools/rnndb/adreno.xml               (    364 bytes, from 2013-11-30 14:47:15)
+- /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml  (   1453 bytes, from 2013-03-31 16:51:27)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml          (  32901 bytes, from 2014-06-02 15:21:30)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml (  10551 bytes, from 2014-11-13 22:44:30)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml    (  15053 bytes, from 2014-11-09 15:45:47)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml          (  63169 bytes, from 2014-11-13 22:44:18)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml          (  49097 bytes, from 2014-11-14 15:38:00)
+
+Copyright (C) 2013-2014 by the following authors:
+- Rob Clark <robdclark@gmail.com> (robclark)
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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.
+*/
+
+
+enum a4xx_color_fmt {
+        RB4_A8_UNORM = 1,
+        RB4_R5G6R5_UNORM = 14,
+        RB4_Z16_UNORM = 15,
+        RB4_R8G8B8_UNORM = 25,
+        RB4_R8G8B8A8_UNORM = 26,
+};
+
+enum a4xx_tile_mode {
+        TILE4_LINEAR = 0,
+        TILE4_3 = 3,
+};
+
+enum a4xx_rb_blend_opcode {
+        BLEND_DST_PLUS_SRC = 0,
+        BLEND_SRC_MINUS_DST = 1,
+        BLEND_DST_MINUS_SRC = 2,
+        BLEND_MIN_DST_SRC = 3,
+        BLEND_MAX_DST_SRC = 4,
+};
+
+enum a4xx_vtx_fmt {
+        VFMT4_FLOAT_32 = 1,
+        VFMT4_FLOAT_32_32 = 2,
+        VFMT4_FLOAT_32_32_32 = 3,
+        VFMT4_FLOAT_32_32_32_32 = 4,
+        VFMT4_FLOAT_16 = 5,
+        VFMT4_FLOAT_16_16 = 6,
+        VFMT4_FLOAT_16_16_16 = 7,
+        VFMT4_FLOAT_16_16_16_16 = 8,
+        VFMT4_FIXED_32 = 9,
+        VFMT4_FIXED_32_32 = 10,
+        VFMT4_FIXED_32_32_32 = 11,
+        VFMT4_FIXED_32_32_32_32 = 12,
+        VFMT4_SHORT_16 = 16,
+        VFMT4_SHORT_16_16 = 17,
+        VFMT4_SHORT_16_16_16 = 18,
+        VFMT4_SHORT_16_16_16_16 = 19,
+        VFMT4_USHORT_16 = 20,
+        VFMT4_USHORT_16_16 = 21,
+        VFMT4_USHORT_16_16_16 = 22,
+        VFMT4_USHORT_16_16_16_16 = 23,
+        VFMT4_NORM_SHORT_16 = 24,
+        VFMT4_NORM_SHORT_16_16 = 25,
+        VFMT4_NORM_SHORT_16_16_16 = 26,
+        VFMT4_NORM_SHORT_16_16_16_16 = 27,
+        VFMT4_NORM_USHORT_16 = 28,
+        VFMT4_NORM_USHORT_16_16 = 29,
+        VFMT4_NORM_USHORT_16_16_16 = 30,
+        VFMT4_NORM_USHORT_16_16_16_16 = 31,
+        VFMT4_UBYTE_8 = 40,
+        VFMT4_UBYTE_8_8 = 41,
+        VFMT4_UBYTE_8_8_8 = 42,
+        VFMT4_UBYTE_8_8_8_8 = 43,
+        VFMT4_NORM_UBYTE_8 = 44,
+        VFMT4_NORM_UBYTE_8_8 = 45,
+        VFMT4_NORM_UBYTE_8_8_8 = 46,
+        VFMT4_NORM_UBYTE_8_8_8_8 = 47,
+        VFMT4_BYTE_8 = 48,
+        VFMT4_BYTE_8_8 = 49,
+        VFMT4_BYTE_8_8_8 = 50,
+        VFMT4_BYTE_8_8_8_8 = 51,
+        VFMT4_NORM_BYTE_8 = 52,
+        VFMT4_NORM_BYTE_8_8 = 53,
+        VFMT4_NORM_BYTE_8_8_8 = 54,
+        VFMT4_NORM_BYTE_8_8_8_8 = 55,
+        VFMT4_UINT_10_10_10_2 = 60,
+        VFMT4_NORM_UINT_10_10_10_2 = 61,
+        VFMT4_INT_10_10_10_2 = 62,
+        VFMT4_NORM_INT_10_10_10_2 = 63,
+};
+
+enum a4xx_tex_fmt {
+        TFMT4_NORM_USHORT_565 = 11,
+        TFMT4_NORM_USHORT_5551 = 10,
+        TFMT4_NORM_USHORT_4444 = 8,
+        TFMT4_NORM_UINT_X8Z24 = 71,
+        TFMT4_NORM_UINT_2_10_10_10 = 33,
+        TFMT4_NORM_UINT_A8 = 3,
+        TFMT4_NORM_UINT_L8_A8 = 13,
+        TFMT4_NORM_UINT_8 = 4,
+        TFMT4_NORM_UINT_8_8_8_8 = 28,
+        TFMT4_FLOAT_16 = 20,
+        TFMT4_FLOAT_16_16 = 40,
+        TFMT4_FLOAT_16_16_16_16 = 53,
+        TFMT4_FLOAT_32 = 43,
+        TFMT4_FLOAT_32_32 = 56,
+        TFMT4_FLOAT_32_32_32_32 = 63,
+};
+
+enum a4xx_depth_format {
+        DEPTH4_NONE = 0,
+        DEPTH4_16 = 1,
+        DEPTH4_24_8 = 2,
+};
+
+enum a4xx_tex_filter {
+        A4XX_TEX_NEAREST = 0,
+        A4XX_TEX_LINEAR = 1,
+};
+
+enum a4xx_tex_clamp {
+        A4XX_TEX_REPEAT = 0,
+        A4XX_TEX_CLAMP_TO_EDGE = 1,
+        A4XX_TEX_MIRROR_REPEAT = 2,
+        A4XX_TEX_CLAMP_NONE = 3,
+};
+
+enum a4xx_tex_swiz {
+        A4XX_TEX_X = 0,
+        A4XX_TEX_Y = 1,
+        A4XX_TEX_Z = 2,
+        A4XX_TEX_W = 3,
+        A4XX_TEX_ZERO = 4,
+        A4XX_TEX_ONE = 5,
+};
+
+enum a4xx_tex_type {
+        A4XX_TEX_1D = 0,
+        A4XX_TEX_2D = 1,
+        A4XX_TEX_CUBE = 2,
+        A4XX_TEX_3D = 3,
+};
+
+#define A4XX_CGC_HLSQ_EARLY_CYC__MASK                           0x00700000
+#define A4XX_CGC_HLSQ_EARLY_CYC__SHIFT                          20
+static inline uint32_t A4XX_CGC_HLSQ_EARLY_CYC(uint32_t val)
+{
+        return ((val) << A4XX_CGC_HLSQ_EARLY_CYC__SHIFT) & A4XX_CGC_HLSQ_EARLY_CYC__MASK;
+}
+#define A4XX_INT0_RBBM_GPU_IDLE                                 0x00000001
+#define A4XX_INT0_RBBM_AHB_ERROR                                0x00000002
+#define A4XX_INT0_RBBM_REG_TIMEOUT                              0x00000004
+#define A4XX_INT0_RBBM_ME_MS_TIMEOUT                            0x00000008
+#define A4XX_INT0_RBBM_PFP_MS_TIMEOUT                           0x00000010
+#define A4XX_INT0_RBBM_ATB_BUS_OVERFLOW                         0x00000020
+#define A4XX_INT0_VFD_ERROR                                     0x00000040
+#define A4XX_INT0_CP_SW_INT                                     0x00000080
+#define A4XX_INT0_CP_T0_PACKET_IN_IB                            0x00000100
+#define A4XX_INT0_CP_OPCODE_ERROR                               0x00000200
+#define A4XX_INT0_CP_RESERVED_BIT_ERROR                         0x00000400
+#define A4XX_INT0_CP_HW_FAULT                                   0x00000800
+#define A4XX_INT0_CP_DMA                                        0x00001000
+#define A4XX_INT0_CP_IB2_INT                                    0x00002000
+#define A4XX_INT0_CP_IB1_INT                                    0x00004000
+#define A4XX_INT0_CP_RB_INT                                     0x00008000
+#define A4XX_INT0_CP_REG_PROTECT_FAULT                          0x00010000
+#define A4XX_INT0_CP_RB_DONE_TS                                 0x00020000
+#define A4XX_INT0_CP_VS_DONE_TS                                 0x00040000
+#define A4XX_INT0_CP_PS_DONE_TS                                 0x00080000
+#define A4XX_INT0_CACHE_FLUSH_TS                                0x00100000
+#define A4XX_INT0_CP_AHB_ERROR_HALT                             0x00200000
+#define A4XX_INT0_MISC_HANG_DETECT                              0x01000000
+#define A4XX_INT0_UCHE_OOB_ACCESS                               0x02000000
+#define REG_A4XX_RB_GMEM_BASE_ADDR                              0x00000cc0
+
+#define REG_A4XX_RB_PERFCTR_RB_SEL_0                            0x00000cc7
+
+#define REG_A4XX_RB_PERFCTR_RB_SEL_1                            0x00000cc8
+
+#define REG_A4XX_RB_PERFCTR_RB_SEL_2                            0x00000cc9
+
+#define REG_A4XX_RB_PERFCTR_RB_SEL_3                            0x00000cca
+
+#define REG_A4XX_RB_PERFCTR_RB_SEL_4                            0x00000ccb
+
+#define REG_A4XX_RB_PERFCTR_RB_SEL_5                            0x00000ccc
+
+#define REG_A4XX_RB_PERFCTR_RB_SEL_6                            0x00000ccd
+
+#define REG_A4XX_RB_PERFCTR_RB_SEL_7                            0x00000cce
+
+#define REG_A4XX_RB_PERFCTR_CCU_SEL_3                           0x00000cd2
+
+#define REG_A4XX_RB_FRAME_BUFFER_DIMENSION                      0x00000ce0
+#define A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH__MASK              0x00003fff
+#define A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH__SHIFT             0
+static inline uint32_t A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH(uint32_t val)
+{
+        return ((val) << A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH__SHIFT) & A4XX_RB_FRAME_BUFFER_DIMENSION_WIDTH__MASK;
+}
+#define A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT__MASK             0x3fff0000
+#define A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT__SHIFT            16
+static inline uint32_t A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT(uint32_t val)
+{
+        return ((val) << A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT__SHIFT) & A4XX_RB_FRAME_BUFFER_DIMENSION_HEIGHT__MASK;
+}
+
+#define REG_A4XX_RB_CLEAR_COLOR_DW0                             0x000020cc
+
+#define REG_A4XX_RB_CLEAR_COLOR_DW1                             0x000020cd
+
+#define REG_A4XX_RB_CLEAR_COLOR_DW2                             0x000020ce
+
+#define REG_A4XX_RB_CLEAR_COLOR_DW3                             0x000020cf
+
+#define REG_A4XX_RB_MODE_CONTROL                                0x000020a0
+#define A4XX_RB_MODE_CONTROL_WIDTH__MASK                        0x0000003f
+#define A4XX_RB_MODE_CONTROL_WIDTH__SHIFT                       0
+static inline uint32_t A4XX_RB_MODE_CONTROL_WIDTH(uint32_t val)
+{
+        return ((val >> 5) << A4XX_RB_MODE_CONTROL_WIDTH__SHIFT) & A4XX_RB_MODE_CONTROL_WIDTH__MASK;
+}
+#define A4XX_RB_MODE_CONTROL_HEIGHT__MASK                       0x00003f00
+#define A4XX_RB_MODE_CONTROL_HEIGHT__SHIFT                      8
+static inline uint32_t A4XX_RB_MODE_CONTROL_HEIGHT(uint32_t val)
+{
+        return ((val >> 5) << A4XX_RB_MODE_CONTROL_HEIGHT__SHIFT) & A4XX_RB_MODE_CONTROL_HEIGHT__MASK;
+}
+
+#define REG_A4XX_RB_RENDER_CONTROL                              0x000020a1
+#define A4XX_RB_RENDER_CONTROL_BINNING_PASS                     0x00000001
+#define A4XX_RB_RENDER_CONTROL_DISABLE_COLOR_PIPE               0x00000020
+
+#define REG_A4XX_RB_MSAA_CONTROL                                0x000020a2
+#define A4XX_RB_MSAA_CONTROL_DISABLE                            0x00001000
+#define A4XX_RB_MSAA_CONTROL_SAMPLES__MASK                      0x0000e000
+#define A4XX_RB_MSAA_CONTROL_SAMPLES__SHIFT                     13
+static inline uint32_t A4XX_RB_MSAA_CONTROL_SAMPLES(uint32_t val)
+{
+        return ((val) << A4XX_RB_MSAA_CONTROL_SAMPLES__SHIFT) & A4XX_RB_MSAA_CONTROL_SAMPLES__MASK;
+}
+
+#define REG_A4XX_RB_MSAA_CONTROL2                               0x000020a3
+#define A4XX_RB_MSAA_CONTROL2_MSAA_SAMPLES__MASK                0x00000380
+#define A4XX_RB_MSAA_CONTROL2_MSAA_SAMPLES__SHIFT               7
+static inline uint32_t A4XX_RB_MSAA_CONTROL2_MSAA_SAMPLES(uint32_t val)
+{
+        return ((val) << A4XX_RB_MSAA_CONTROL2_MSAA_SAMPLES__SHIFT) & A4XX_RB_MSAA_CONTROL2_MSAA_SAMPLES__MASK;
+}
+#define A4XX_RB_MSAA_CONTROL2_VARYING                           0x00001000
+
+static inline uint32_t REG_A4XX_RB_MRT(uint32_t i0) { return 0x000020a4 + 0x5*i0; }
+
+static inline uint32_t REG_A4XX_RB_MRT_CONTROL(uint32_t i0) { return 0x000020a4 + 0x5*i0; }
+#define A4XX_RB_MRT_CONTROL_READ_DEST_ENABLE                    0x00000008
+#define A4XX_RB_MRT_CONTROL_BLEND                               0x00000010
+#define A4XX_RB_MRT_CONTROL_BLEND2                              0x00000020
+#define A4XX_RB_MRT_CONTROL_FASTCLEAR                           0x00000400
+#define A4XX_RB_MRT_CONTROL_B11                                 0x00000800
+#define A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE__MASK              0x0f000000
+#define A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE__SHIFT             24
+static inline uint32_t A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE(uint32_t val)
+{
+        return ((val) << A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE__SHIFT) & A4XX_RB_MRT_CONTROL_COMPONENT_ENABLE__MASK;
+}
+
+static inline uint32_t REG_A4XX_RB_MRT_BUF_INFO(uint32_t i0) { return 0x000020a5 + 0x5*i0; }
+#define A4XX_RB_MRT_BUF_INFO_COLOR_FORMAT__MASK                 0x0000003f
+#define A4XX_RB_MRT_BUF_INFO_COLOR_FORMAT__SHIFT                0
+static inline uint32_t A4XX_RB_MRT_BUF_INFO_COLOR_FORMAT(enum a4xx_color_fmt val)
+{
+        return ((val) << A4XX_RB_MRT_BUF_INFO_COLOR_FORMAT__SHIFT) & A4XX_RB_MRT_BUF_INFO_COLOR_FORMAT__MASK;
+}
+#define A4XX_RB_MRT_BUF_INFO_DITHER_MODE__MASK                  0x00000600
+#define A4XX_RB_MRT_BUF_INFO_DITHER_MODE__SHIFT                 9
+static inline uint32_t A4XX_RB_MRT_BUF_INFO_DITHER_MODE(enum adreno_rb_dither_mode val)
+{
+        return ((val) << A4XX_RB_MRT_BUF_INFO_DITHER_MODE__SHIFT) & A4XX_RB_MRT_BUF_INFO_DITHER_MODE__MASK;
+}
+#define A4XX_RB_MRT_BUF_INFO_COLOR_SWAP__MASK                   0x00001800
+#define A4XX_RB_MRT_BUF_INFO_COLOR_SWAP__SHIFT                  11
+static inline uint32_t A4XX_RB_MRT_BUF_INFO_COLOR_SWAP(enum a3xx_color_swap val)
+{
+        return ((val) << A4XX_RB_MRT_BUF_INFO_COLOR_SWAP__SHIFT) & A4XX_RB_MRT_BUF_INFO_COLOR_SWAP__MASK;
+}
+#define A4XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__MASK              0x007fc000
+#define A4XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__SHIFT             14
+static inline uint32_t A4XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH(uint32_t val)
+{
+        return ((val >> 4) << A4XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__SHIFT) & A4XX_RB_MRT_BUF_INFO_COLOR_BUF_PITCH__MASK;
+}
+
+static inline uint32_t REG_A4XX_RB_MRT_BASE(uint32_t i0) { return 0x000020a6 + 0x5*i0; }
+
+static inline uint32_t REG_A4XX_RB_MRT_CONTROL3(uint32_t i0) { return 0x000020a7 + 0x5*i0; }
+#define A4XX_RB_MRT_CONTROL3_STRIDE__MASK                       0x0001fff8
+#define A4XX_RB_MRT_CONTROL3_STRIDE__SHIFT                      3
+static inline uint32_t A4XX_RB_MRT_CONTROL3_STRIDE(uint32_t val)
+{
+        return ((val) << A4XX_RB_MRT_CONTROL3_STRIDE__SHIFT) & A4XX_RB_MRT_CONTROL3_STRIDE__MASK;
+}
+
+static inline uint32_t REG_A4XX_RB_MRT_BLEND_CONTROL(uint32_t i0) { return 0x000020a8 + 0x5*i0; }
+#define A4XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__MASK          0x0000001f
+#define A4XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__SHIFT         0
+static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR(enum adreno_rb_blend_factor val)
+{
+        return ((val) << A4XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_RGB_SRC_FACTOR__MASK;
+}
+#define A4XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__MASK        0x000000e0
+#define A4XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__SHIFT       5
+static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE(enum a4xx_rb_blend_opcode val)
+{
+        return ((val) << A4XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_RGB_BLEND_OPCODE__MASK;
+}
+#define A4XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__MASK         0x00001f00
+#define A4XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__SHIFT        8
+static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR(enum adreno_rb_blend_factor val)
+{
+        return ((val) << A4XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_RGB_DEST_FACTOR__MASK;
+}
+#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__MASK        0x001f0000
+#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__SHIFT       16
+static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR(enum adreno_rb_blend_factor val)
+{
+        return ((val) << A4XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_ALPHA_SRC_FACTOR__MASK;
+}
+#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__MASK      0x00e00000
+#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__SHIFT     21
+static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE(enum a4xx_rb_blend_opcode val)
+{
+        return ((val) << A4XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_ALPHA_BLEND_OPCODE__MASK;
+}
+#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__MASK       0x1f000000
+#define A4XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__SHIFT      24
+static inline uint32_t A4XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR(enum adreno_rb_blend_factor val)
+{
+        return ((val) << A4XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__SHIFT) & A4XX_RB_MRT_BLEND_CONTROL_ALPHA_DEST_FACTOR__MASK;
+}
+
+#define REG_A4XX_RB_ALPHA_CONTROL                               0x000020f8
+#define A4XX_RB_ALPHA_CONTROL_ALPHA_TEST                        0x00000100
+#define A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__MASK             0x00000e00
+#define A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__SHIFT            9
+static inline uint32_t A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC(enum adreno_compare_func val)
+{
+        return ((val) << A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__SHIFT) & A4XX_RB_ALPHA_CONTROL_ALPHA_TEST_FUNC__MASK;
+}
+
+#define REG_A4XX_RB_FS_OUTPUT                                   0x000020f9
+#define A4XX_RB_FS_OUTPUT_ENABLE_COLOR_PIPE                     0x00000001
+#define A4XX_RB_FS_OUTPUT_FAST_CLEAR                            0x00000100
+#define A4XX_RB_FS_OUTPUT_SAMPLE_MASK__MASK                     0xffff0000
+#define A4XX_RB_FS_OUTPUT_SAMPLE_MASK__SHIFT                    16
+static inline uint32_t A4XX_RB_FS_OUTPUT_SAMPLE_MASK(uint32_t val)
+{
+        return ((val) << A4XX_RB_FS_OUTPUT_SAMPLE_MASK__SHIFT) & A4XX_RB_FS_OUTPUT_SAMPLE_MASK__MASK;
+}
+
+#define REG_A4XX_RB_RENDER_CONTROL3                             0x000020fb
+#define A4XX_RB_RENDER_CONTROL3_COMPONENT_ENABLE__MASK          0x0000001f
+#define A4XX_RB_RENDER_CONTROL3_COMPONENT_ENABLE__SHIFT         0
+static inline uint32_t A4XX_RB_RENDER_CONTROL3_COMPONENT_ENABLE(uint32_t val)
+{
+        return ((val) << A4XX_RB_RENDER_CONTROL3_COMPONENT_ENABLE__SHIFT) & A4XX_RB_RENDER_CONTROL3_COMPONENT_ENABLE__MASK;
+}
+
+#define REG_A4XX_RB_COPY_CONTROL                                0x000020fc
+#define A4XX_RB_COPY_CONTROL_MSAA_RESOLVE__MASK                 0x00000003
+#define A4XX_RB_COPY_CONTROL_MSAA_RESOLVE__SHIFT                0
+static inline uint32_t A4XX_RB_COPY_CONTROL_MSAA_RESOLVE(enum a3xx_msaa_samples val)
+{
+        return ((val) << A4XX_RB_COPY_CONTROL_MSAA_RESOLVE__SHIFT) & A4XX_RB_COPY_CONTROL_MSAA_RESOLVE__MASK;
+}
+#define A4XX_RB_COPY_CONTROL_MODE__MASK                         0x00000070
+#define A4XX_RB_COPY_CONTROL_MODE__SHIFT                        4
+static inline uint32_t A4XX_RB_COPY_CONTROL_MODE(enum adreno_rb_copy_control_mode val)
+{
+        return ((val) << A4XX_RB_COPY_CONTROL_MODE__SHIFT) & A4XX_RB_COPY_CONTROL_MODE__MASK;
+}
+#define A4XX_RB_COPY_CONTROL_FASTCLEAR__MASK                    0x00000f00
+#define A4XX_RB_COPY_CONTROL_FASTCLEAR__SHIFT                   8
+static inline uint32_t A4XX_RB_COPY_CONTROL_FASTCLEAR(uint32_t val)
+{
+        return ((val) << A4XX_RB_COPY_CONTROL_FASTCLEAR__SHIFT) & A4XX_RB_COPY_CONTROL_FASTCLEAR__MASK;
+}
+#define A4XX_RB_COPY_CONTROL_GMEM_BASE__MASK                    0xffffc000
+#define A4XX_RB_COPY_CONTROL_GMEM_BASE__SHIFT                   14
+static inline uint32_t A4XX_RB_COPY_CONTROL_GMEM_BASE(uint32_t val)
+{
+        return ((val >> 14) << A4XX_RB_COPY_CONTROL_GMEM_BASE__SHIFT) & A4XX_RB_COPY_CONTROL_GMEM_BASE__MASK;
+}
+
+#define REG_A4XX_RB_COPY_DEST_BASE                              0x000020fd
+#define A4XX_RB_COPY_DEST_BASE_BASE__MASK                       0xfffffff0
+#define A4XX_RB_COPY_DEST_BASE_BASE__SHIFT                      4
+static inline uint32_t A4XX_RB_COPY_DEST_BASE_BASE(uint32_t val)
+{
+        return ((val >> 4) << A4XX_RB_COPY_DEST_BASE_BASE__SHIFT) & A4XX_RB_COPY_DEST_BASE_BASE__MASK;
+}
+
+#define REG_A4XX_RB_COPY_DEST_PITCH                             0x000020fe
+#define A4XX_RB_COPY_DEST_PITCH_PITCH__MASK                     0xffffffff
+#define A4XX_RB_COPY_DEST_PITCH_PITCH__SHIFT                    0
+static inline uint32_t A4XX_RB_COPY_DEST_PITCH_PITCH(uint32_t val)
+{
+        return ((val >> 5) << A4XX_RB_COPY_DEST_PITCH_PITCH__SHIFT) & A4XX_RB_COPY_DEST_PITCH_PITCH__MASK;
+}
+
+#define REG_A4XX_RB_COPY_DEST_INFO                              0x000020ff
+#define A4XX_RB_COPY_DEST_INFO_FORMAT__MASK                     0x000000fc
+#define A4XX_RB_COPY_DEST_INFO_FORMAT__SHIFT                    2
+static inline uint32_t A4XX_RB_COPY_DEST_INFO_FORMAT(enum a4xx_color_fmt val)
+{
+        return ((val) << A4XX_RB_COPY_DEST_INFO_FORMAT__SHIFT) & A4XX_RB_COPY_DEST_INFO_FORMAT__MASK;
+}
+#define A4XX_RB_COPY_DEST_INFO_SWAP__MASK                       0x00000300
+#define A4XX_RB_COPY_DEST_INFO_SWAP__SHIFT                      8
+static inline uint32_t A4XX_RB_COPY_DEST_INFO_SWAP(enum a3xx_color_swap val)
+{
+        return ((val) << A4XX_RB_COPY_DEST_INFO_SWAP__SHIFT) & A4XX_RB_COPY_DEST_INFO_SWAP__MASK;
+}
+#define A4XX_RB_COPY_DEST_INFO_DITHER_MODE__MASK                0x00000c00
+#define A4XX_RB_COPY_DEST_INFO_DITHER_MODE__SHIFT               10
+static inline uint32_t A4XX_RB_COPY_DEST_INFO_DITHER_MODE(enum adreno_rb_dither_mode val)
+{
+        return ((val) << A4XX_RB_COPY_DEST_INFO_DITHER_MODE__SHIFT) & A4XX_RB_COPY_DEST_INFO_DITHER_MODE__MASK;
+}
+#define A4XX_RB_COPY_DEST_INFO_COMPONENT_ENABLE__MASK           0x0003c000
+#define A4XX_RB_COPY_DEST_INFO_COMPONENT_ENABLE__SHIFT          14
+static inline uint32_t A4XX_RB_COPY_DEST_INFO_COMPONENT_ENABLE(uint32_t val)
+{
+        return ((val) << A4XX_RB_COPY_DEST_INFO_COMPONENT_ENABLE__SHIFT) & A4XX_RB_COPY_DEST_INFO_COMPONENT_ENABLE__MASK;
+}
+#define A4XX_RB_COPY_DEST_INFO_ENDIAN__MASK                     0x001c0000
+#define A4XX_RB_COPY_DEST_INFO_ENDIAN__SHIFT                    18
+static inline uint32_t A4XX_RB_COPY_DEST_INFO_ENDIAN(enum adreno_rb_surface_endian val)
+{
+        return ((val) << A4XX_RB_COPY_DEST_INFO_ENDIAN__SHIFT) & A4XX_RB_COPY_DEST_INFO_ENDIAN__MASK;
+}
+#define A4XX_RB_COPY_DEST_INFO_TILE__MASK                       0x03000000
+#define A4XX_RB_COPY_DEST_INFO_TILE__SHIFT                      24
+static inline uint32_t A4XX_RB_COPY_DEST_INFO_TILE(enum a4xx_tile_mode val)
+{
+        return ((val) << A4XX_RB_COPY_DEST_INFO_TILE__SHIFT) & A4XX_RB_COPY_DEST_INFO_TILE__MASK;
+}
+
+#define REG_A4XX_RB_FS_OUTPUT_REG                               0x00002100
+#define A4XX_RB_FS_OUTPUT_REG_COLOR_PIPE_ENABLE                 0x00000001
+#define A4XX_RB_FS_OUTPUT_REG_FRAG_WRITES_Z                     0x00000020
+
+#define REG_A4XX_RB_DEPTH_CONTROL                               0x00002101
+#define A4XX_RB_DEPTH_CONTROL_FRAG_WRITES_Z                     0x00000001
+#define A4XX_RB_DEPTH_CONTROL_Z_ENABLE                          0x00000002
+#define A4XX_RB_DEPTH_CONTROL_Z_WRITE_ENABLE                    0x00000004
+#define A4XX_RB_DEPTH_CONTROL_ZFUNC__MASK                       0x00000070
+#define A4XX_RB_DEPTH_CONTROL_ZFUNC__SHIFT                      4
+static inline uint32_t A4XX_RB_DEPTH_CONTROL_ZFUNC(enum adreno_compare_func val)
+{
+        return ((val) << A4XX_RB_DEPTH_CONTROL_ZFUNC__SHIFT) & A4XX_RB_DEPTH_CONTROL_ZFUNC__MASK;
+}
+#define A4XX_RB_DEPTH_CONTROL_BF_ENABLE                         0x00000080
+#define A4XX_RB_DEPTH_CONTROL_EARLY_Z_DISABLE                   0x00010000
+#define A4XX_RB_DEPTH_CONTROL_Z_TEST_ENABLE                     0x80000000
+
+#define REG_A4XX_RB_DEPTH_CLEAR                                 0x00002102
+
+#define REG_A4XX_RB_DEPTH_INFO                                  0x00002103
+#define A4XX_RB_DEPTH_INFO_DEPTH_FORMAT__MASK                   0x00000003
+#define A4XX_RB_DEPTH_INFO_DEPTH_FORMAT__SHIFT                  0
+static inline uint32_t A4XX_RB_DEPTH_INFO_DEPTH_FORMAT(enum a4xx_depth_format val)
+{
+        return ((val) << A4XX_RB_DEPTH_INFO_DEPTH_FORMAT__SHIFT) & A4XX_RB_DEPTH_INFO_DEPTH_FORMAT__MASK;
+}
+#define A4XX_RB_DEPTH_INFO_DEPTH_BASE__MASK                     0xfffff000
+#define A4XX_RB_DEPTH_INFO_DEPTH_BASE__SHIFT                    12
+static inline uint32_t A4XX_RB_DEPTH_INFO_DEPTH_BASE(uint32_t val)
+{
+        return ((val >> 12) << A4XX_RB_DEPTH_INFO_DEPTH_BASE__SHIFT) & A4XX_RB_DEPTH_INFO_DEPTH_BASE__MASK;
+}
+
+#define REG_A4XX_RB_DEPTH_PITCH                                 0x00002104
+#define A4XX_RB_DEPTH_PITCH__MASK                               0xffffffff
+#define A4XX_RB_DEPTH_PITCH__SHIFT                              0
+static inline uint32_t A4XX_RB_DEPTH_PITCH(uint32_t val)
+{
+        return ((val >> 4) << A4XX_RB_DEPTH_PITCH__SHIFT) & A4XX_RB_DEPTH_PITCH__MASK;
+}
+
+#define REG_A4XX_RB_DEPTH_PITCH2                                0x00002105
+#define A4XX_RB_DEPTH_PITCH2__MASK                              0xffffffff
+#define A4XX_RB_DEPTH_PITCH2__SHIFT                             0
+static inline uint32_t A4XX_RB_DEPTH_PITCH2(uint32_t val)
+{
+        return ((val >> 4) << A4XX_RB_DEPTH_PITCH2__SHIFT) & A4XX_RB_DEPTH_PITCH2__MASK;
+}
+
+#define REG_A4XX_RB_STENCIL_CONTROL                             0x00002106
+#define A4XX_RB_STENCIL_CONTROL_STENCIL_ENABLE                  0x00000001
+#define A4XX_RB_STENCIL_CONTROL_STENCIL_ENABLE_BF               0x00000002
+#define A4XX_RB_STENCIL_CONTROL_STENCIL_READ                    0x00000004
+#define A4XX_RB_STENCIL_CONTROL_FUNC__MASK                      0x00000700
+#define A4XX_RB_STENCIL_CONTROL_FUNC__SHIFT                     8
+static inline uint32_t A4XX_RB_STENCIL_CONTROL_FUNC(enum adreno_compare_func val)
+{
+        return ((val) << A4XX_RB_STENCIL_CONTROL_FUNC__SHIFT) & A4XX_RB_STENCIL_CONTROL_FUNC__MASK;
+}
+#define A4XX_RB_STENCIL_CONTROL_FAIL__MASK                      0x00003800
+#define A4XX_RB_STENCIL_CONTROL_FAIL__SHIFT                     11
+static inline uint32_t A4XX_RB_STENCIL_CONTROL_FAIL(enum adreno_stencil_op val)
+{
+        return ((val) << A4XX_RB_STENCIL_CONTROL_FAIL__SHIFT) & A4XX_RB_STENCIL_CONTROL_FAIL__MASK;
+}
+#define A4XX_RB_STENCIL_CONTROL_ZPASS__MASK                     0x0001c000
+#define A4XX_RB_STENCIL_CONTROL_ZPASS__SHIFT                    14
+static inline uint32_t A4XX_RB_STENCIL_CONTROL_ZPASS(enum adreno_stencil_op val)
+{
+        return ((val) << A4XX_RB_STENCIL_CONTROL_ZPASS__SHIFT) & A4XX_RB_STENCIL_CONTROL_ZPASS__MASK;
+}
+#define A4XX_RB_STENCIL_CONTROL_ZFAIL__MASK                     0x000e0000
+#define A4XX_RB_STENCIL_CONTROL_ZFAIL__SHIFT                    17
+static inline uint32_t A4XX_RB_STENCIL_CONTROL_ZFAIL(enum adreno_stencil_op val)
+{
+        return ((val) << A4XX_RB_STENCIL_CONTROL_ZFAIL__SHIFT) & A4XX_RB_STENCIL_CONTROL_ZFAIL__MASK;
+}
+#define A4XX_RB_STENCIL_CONTROL_FUNC_BF__MASK                   0x00700000
+#define A4XX_RB_STENCIL_CONTROL_FUNC_BF__SHIFT                  20
+static inline uint32_t A4XX_RB_STENCIL_CONTROL_FUNC_BF(enum adreno_compare_func val)
+{
+        return ((val) << A4XX_RB_STENCIL_CONTROL_FUNC_BF__SHIFT) & A4XX_RB_STENCIL_CONTROL_FUNC_BF__MASK;
+}
+#define A4XX_RB_STENCIL_CONTROL_FAIL_BF__MASK                   0x03800000
+#define A4XX_RB_STENCIL_CONTROL_FAIL_BF__SHIFT                  23
+static inline uint32_t A4XX_RB_STENCIL_CONTROL_FAIL_BF(enum adreno_stencil_op val)
+{
+        return ((val) << A4XX_RB_STENCIL_CONTROL_FAIL_BF__SHIFT) & A4XX_RB_STENCIL_CONTROL_FAIL_BF__MASK;
+}
+#define A4XX_RB_STENCIL_CONTROL_ZPASS_BF__MASK                  0x1c000000
+#define A4XX_RB_STENCIL_CONTROL_ZPASS_BF__SHIFT                 26
+static inline uint32_t A4XX_RB_STENCIL_CONTROL_ZPASS_BF(enum adreno_stencil_op val)
+{
+        return ((val) << A4XX_RB_STENCIL_CONTROL_ZPASS_BF__SHIFT) & A4XX_RB_STENCIL_CONTROL_ZPASS_BF__MASK;
+}
+#define A4XX_RB_STENCIL_CONTROL_ZFAIL_BF__MASK                  0xe0000000
+#define A4XX_RB_STENCIL_CONTROL_ZFAIL_BF__SHIFT                 29
+static inline uint32_t A4XX_RB_STENCIL_CONTROL_ZFAIL_BF(enum adreno_stencil_op val)
+{
+        return ((val) << A4XX_RB_STENCIL_CONTROL_ZFAIL_BF__SHIFT) & A4XX_RB_STENCIL_CONTROL_ZFAIL_BF__MASK;
+}
+
+#define REG_A4XX_RB_STENCIL_CONTROL2                            0x00002107
+#define A4XX_RB_STENCIL_CONTROL2_STENCIL_BUFFER                 0x00000001
+
+#define REG_A4XX_RB_STENCILREFMASK                              0x0000210b
+#define A4XX_RB_STENCILREFMASK_STENCILREF__MASK                 0x000000ff
+#define A4XX_RB_STENCILREFMASK_STENCILREF__SHIFT                0
+static inline uint32_t A4XX_RB_STENCILREFMASK_STENCILREF(uint32_t val)
+{
+        return ((val) << A4XX_RB_STENCILREFMASK_STENCILREF__SHIFT) & A4XX_RB_STENCILREFMASK_STENCILREF__MASK;
+}
+#define A4XX_RB_STENCILREFMASK_STENCILMASK__MASK                0x0000ff00
+#define A4XX_RB_STENCILREFMASK_STENCILMASK__SHIFT               8
+static inline uint32_t A4XX_RB_STENCILREFMASK_STENCILMASK(uint32_t val)
+{
+        return ((val) << A4XX_RB_STENCILREFMASK_STENCILMASK__SHIFT) & A4XX_RB_STENCILREFMASK_STENCILMASK__MASK;
+}
+#define A4XX_RB_STENCILREFMASK_STENCILWRITEMASK__MASK           0x00ff0000
+#define A4XX_RB_STENCILREFMASK_STENCILWRITEMASK__SHIFT          16
+static inline uint32_t A4XX_RB_STENCILREFMASK_STENCILWRITEMASK(uint32_t val)
+{
+        return ((val) << A4XX_RB_STENCILREFMASK_STENCILWRITEMASK__SHIFT) & A4XX_RB_STENCILREFMASK_STENCILWRITEMASK__MASK;
+}
+
+#define REG_A4XX_RB_STENCILREFMASK_BF                           0x0000210c
+#define A4XX_RB_STENCILREFMASK_BF_STENCILREF__MASK              0x000000ff
+#define A4XX_RB_STENCILREFMASK_BF_STENCILREF__SHIFT             0
+static inline uint32_t A4XX_RB_STENCILREFMASK_BF_STENCILREF(uint32_t val)
+{
+        return ((val) << A4XX_RB_STENCILREFMASK_BF_STENCILREF__SHIFT) & A4XX_RB_STENCILREFMASK_BF_STENCILREF__MASK;
+}
+#define A4XX_RB_STENCILREFMASK_BF_STENCILMASK__MASK             0x0000ff00
+#define A4XX_RB_STENCILREFMASK_BF_STENCILMASK__SHIFT            8
+static inline uint32_t A4XX_RB_STENCILREFMASK_BF_STENCILMASK(uint32_t val)
+{
+        return ((val) << A4XX_RB_STENCILREFMASK_BF_STENCILMASK__SHIFT) & A4XX_RB_STENCILREFMASK_BF_STENCILMASK__MASK;
+}
+#define A4XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__MASK        0x00ff0000
+#define A4XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__SHIFT       16
+static inline uint32_t A4XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK(uint32_t val)
+{
+        return ((val) << A4XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__SHIFT) & A4XX_RB_STENCILREFMASK_BF_STENCILWRITEMASK__MASK;
+}
+
+#define REG_A4XX_RB_BIN_OFFSET                                  0x0000210d
+#define A4XX_RB_BIN_OFFSET_WINDOW_OFFSET_DISABLE                0x80000000
+#define A4XX_RB_BIN_OFFSET_X__MASK                              0x00007fff
+#define A4XX_RB_BIN_OFFSET_X__SHIFT                             0
+static inline uint32_t A4XX_RB_BIN_OFFSET_X(uint32_t val)
+{
+        return ((val) << A4XX_RB_BIN_OFFSET_X__SHIFT) & A4XX_RB_BIN_OFFSET_X__MASK;
+}
+#define A4XX_RB_BIN_OFFSET_Y__MASK                              0x7fff0000
+#define A4XX_RB_BIN_OFFSET_Y__SHIFT                             16
+static inline uint32_t A4XX_RB_BIN_OFFSET_Y(uint32_t val)
+{
+        return ((val) << A4XX_RB_BIN_OFFSET_Y__SHIFT) & A4XX_RB_BIN_OFFSET_Y__MASK;
+}
+
+#define REG_A4XX_RB_VPORT_Z_CLAMP_MAX_15                        0x0000213f
+
+#define REG_A4XX_RBBM_HW_VERSION                                0x00000000
+
+#define REG_A4XX_RBBM_HW_CONFIGURATION                          0x00000002
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_TP(uint32_t i0) { return 0x00000004 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_TP_REG(uint32_t i0) { return 0x00000004 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_TP(uint32_t i0) { return 0x00000008 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_TP_REG(uint32_t i0) { return 0x00000008 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_TP(uint32_t i0) { return 0x0000000c + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_TP_REG(uint32_t i0) { return 0x0000000c + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_TP(uint32_t i0) { return 0x00000010 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_TP_REG(uint32_t i0) { return 0x00000010 + 0x1*i0; }
+
+#define REG_A4XX_RBBM_CLOCK_CTL_UCHE                            0x00000014
+
+#define REG_A4XX_RBBM_CLOCK_CTL2_UCHE                           0x00000015
+
+#define REG_A4XX_RBBM_CLOCK_CTL3_UCHE                           0x00000016
+
+#define REG_A4XX_RBBM_CLOCK_CTL4_UCHE                           0x00000017
+
+#define REG_A4XX_RBBM_CLOCK_HYST_UCHE                           0x00000018
+
+#define REG_A4XX_RBBM_CLOCK_DELAY_UCHE                          0x00000019
+
+#define REG_A4XX_RBBM_CLOCK_MODE_GPC                            0x0000001a
+
+#define REG_A4XX_RBBM_CLOCK_DELAY_GPC                           0x0000001b
+
+#define REG_A4XX_RBBM_CLOCK_HYST_GPC                            0x0000001c
+
+#define REG_A4XX_RBBM_CLOCK_CTL_TSE_RAS_RBBM                    0x0000001d
+
+#define REG_A4XX_RBBM_CLOCK_HYST_TSE_RAS_RBBM                   0x0000001e
+
+#define REG_A4XX_RBBM_CLOCK_DELAY_TSE_RAS_RBBM                  0x0000001f
+
+#define REG_A4XX_RBBM_CLOCK_CTL                                 0x00000020
+
+#define REG_A4XX_RBBM_SP_HYST_CNT                               0x00000021
+
+#define REG_A4XX_RBBM_SW_RESET_CMD                              0x00000022
+
+#define REG_A4XX_RBBM_AHB_CTL0                                  0x00000023
+
+#define REG_A4XX_RBBM_AHB_CTL1                                  0x00000024
+
+#define REG_A4XX_RBBM_AHB_CMD                                   0x00000025
+
+#define REG_A4XX_RBBM_RB_SUB_BLOCK_SEL_CTL                      0x00000026
+
+#define REG_A4XX_RBBM_RAM_ACC_63_32                             0x00000028
+
+#define REG_A4XX_RBBM_WAIT_IDLE_CLOCKS_CTL                      0x0000002b
+
+#define REG_A4XX_RBBM_INTERFACE_HANG_INT_CTL                    0x0000002f
+
+#define REG_A4XX_RBBM_INTERFACE_HANG_MASK_CTL4                  0x00000034
+
+#define REG_A4XX_RBBM_INT_CLEAR_CMD                             0x00000036
+
+#define REG_A4XX_RBBM_INT_0_MASK                                0x00000037
+
+#define REG_A4XX_RBBM_RBBM_CTL                                  0x0000003e
+
+#define REG_A4XX_RBBM_AHB_DEBUG_CTL                             0x0000003f
+
+#define REG_A4XX_RBBM_VBIF_DEBUG_CTL                            0x00000041
+
+#define REG_A4XX_RBBM_CLOCK_CTL2                                0x00000042
+
+#define REG_A4XX_RBBM_BLOCK_SW_RESET_CMD                        0x00000045
+
+#define REG_A4XX_RBBM_RESET_CYCLES                              0x00000047
+
+#define REG_A4XX_RBBM_EXT_TRACE_BUS_CTL                         0x00000049
+
+#define REG_A4XX_RBBM_CFG_DEBBUS_SEL_A                          0x0000004a
+
+#define REG_A4XX_RBBM_CFG_DEBBUS_SEL_B                          0x0000004b
+
+#define REG_A4XX_RBBM_CFG_DEBBUS_SEL_C                          0x0000004c
+
+#define REG_A4XX_RBBM_CFG_DEBBUS_SEL_D                          0x0000004d
+
+#define REG_A4XX_RBBM_PERFCTR_CP_0_LO                           0x0000009c
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_SP(uint32_t i0) { return 0x00000068 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_SP_REG(uint32_t i0) { return 0x00000068 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_SP(uint32_t i0) { return 0x0000006c + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_SP_REG(uint32_t i0) { return 0x0000006c + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_SP(uint32_t i0) { return 0x00000070 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_SP_REG(uint32_t i0) { return 0x00000070 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_SP(uint32_t i0) { return 0x00000074 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_SP_REG(uint32_t i0) { return 0x00000074 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_RB(uint32_t i0) { return 0x00000078 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_RB_REG(uint32_t i0) { return 0x00000078 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_RB(uint32_t i0) { return 0x0000007c + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL2_RB_REG(uint32_t i0) { return 0x0000007c + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_MARB_CCU(uint32_t i0) { return 0x00000082 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_CTL_MARB_CCU_REG(uint32_t i0) { return 0x00000082 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_RB_MARB_CCU(uint32_t i0) { return 0x00000086 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_HYST_RB_MARB_CCU_REG(uint32_t i0) { return 0x00000086 + 0x1*i0; }
+
+#define REG_A4XX_RBBM_CLOCK_HYST_COM_DCOM                       0x00000080
+
+#define REG_A4XX_RBBM_CLOCK_CTL_COM_DCOM                        0x00000081
+
+#define REG_A4XX_RBBM_CLOCK_CTL_HLSQ                            0x0000008a
+
+#define REG_A4XX_RBBM_CLOCK_HYST_HLSQ                           0x0000008b
+
+#define REG_A4XX_RBBM_CLOCK_DELAY_HLSQ                          0x0000008c
+
+#define REG_A4XX_RBBM_CLOCK_DELAY_COM_DCOM                      0x0000008d
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_RB_MARB_CCU_L1(uint32_t i0) { return 0x0000008e + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_RBBM_CLOCK_DELAY_RB_MARB_CCU_L1_REG(uint32_t i0) { return 0x0000008e + 0x1*i0; }
+
+#define REG_A4XX_RBBM_PERFCTR_PWR_1_LO                          0x00000168
+
+#define REG_A4XX_RBBM_PERFCTR_CTL                               0x00000170
+
+#define REG_A4XX_RBBM_PERFCTR_LOAD_CMD0                         0x00000171
+
+#define REG_A4XX_RBBM_PERFCTR_LOAD_CMD1                         0x00000172
+
+#define REG_A4XX_RBBM_PERFCTR_LOAD_CMD2                         0x00000173
+
+#define REG_A4XX_RBBM_PERFCTR_LOAD_VALUE_LO                     0x00000174
+
+#define REG_A4XX_RBBM_PERFCTR_LOAD_VALUE_HI                     0x00000175
+
+#define REG_A4XX_RBBM_GPU_BUSY_MASKED                           0x0000017a
+
+#define REG_A4XX_RBBM_INT_0_STATUS                              0x0000017d
+
+#define REG_A4XX_RBBM_CLOCK_STATUS                              0x00000182
+
+#define REG_A4XX_RBBM_AHB_STATUS                                0x00000189
+
+#define REG_A4XX_RBBM_AHB_ME_SPLIT_STATUS                       0x0000018c
+
+#define REG_A4XX_RBBM_AHB_PFP_SPLIT_STATUS                      0x0000018d
+
+#define REG_A4XX_RBBM_AHB_ERROR_STATUS                          0x0000018f
+
+#define REG_A4XX_RBBM_STATUS                                    0x00000191
+#define A4XX_RBBM_STATUS_HI_BUSY                                0x00000001
+#define A4XX_RBBM_STATUS_CP_ME_BUSY                             0x00000002
+#define A4XX_RBBM_STATUS_CP_PFP_BUSY                            0x00000004
+#define A4XX_RBBM_STATUS_CP_NRT_BUSY                            0x00004000
+#define A4XX_RBBM_STATUS_VBIF_BUSY                              0x00008000
+#define A4XX_RBBM_STATUS_TSE_BUSY                               0x00010000
+#define A4XX_RBBM_STATUS_RAS_BUSY                               0x00020000
+#define A4XX_RBBM_STATUS_RB_BUSY                                0x00040000
+#define A4XX_RBBM_STATUS_PC_DCALL_BUSY                          0x00080000
+#define A4XX_RBBM_STATUS_PC_VSD_BUSY                            0x00100000
+#define A4XX_RBBM_STATUS_VFD_BUSY                               0x00200000
+#define A4XX_RBBM_STATUS_VPC_BUSY                               0x00400000
+#define A4XX_RBBM_STATUS_UCHE_BUSY                              0x00800000
+#define A4XX_RBBM_STATUS_SP_BUSY                                0x01000000
+#define A4XX_RBBM_STATUS_TPL1_BUSY                              0x02000000
+#define A4XX_RBBM_STATUS_MARB_BUSY                              0x04000000
+#define A4XX_RBBM_STATUS_VSC_BUSY                               0x08000000
+#define A4XX_RBBM_STATUS_ARB_BUSY                               0x10000000
+#define A4XX_RBBM_STATUS_HLSQ_BUSY                              0x20000000
+#define A4XX_RBBM_STATUS_GPU_BUSY_NOHC                          0x40000000
+#define A4XX_RBBM_STATUS_GPU_BUSY                               0x80000000
+
+#define REG_A4XX_RBBM_INTERFACE_RRDY_STATUS5                    0x0000019f
+
+#define REG_A4XX_CP_SCRATCH_UMASK                               0x00000228
+
+#define REG_A4XX_CP_SCRATCH_ADDR                                0x00000229
+
+#define REG_A4XX_CP_RB_BASE                                     0x00000200
+
+#define REG_A4XX_CP_RB_CNTL                                     0x00000201
+
+#define REG_A4XX_CP_RB_WPTR                                     0x00000205
+
+#define REG_A4XX_CP_RB_RPTR_ADDR                                0x00000203
+
+#define REG_A4XX_CP_RB_RPTR                                     0x00000204
+
+#define REG_A4XX_CP_IB1_BASE                                    0x00000206
+
+#define REG_A4XX_CP_IB1_BUFSZ                                   0x00000207
+
+#define REG_A4XX_CP_IB2_BASE                                    0x00000208
+
+#define REG_A4XX_CP_IB2_BUFSZ                                   0x00000209
+
+#define REG_A4XX_CP_ME_RB_DONE_DATA                             0x00000217
+
+#define REG_A4XX_CP_QUEUE_THRESH2                               0x00000219
+
+#define REG_A4XX_CP_MERCIU_SIZE                                 0x0000021b
+
+#define REG_A4XX_CP_ROQ_ADDR                                    0x0000021c
+
+#define REG_A4XX_CP_ROQ_DATA                                    0x0000021d
+
+#define REG_A4XX_CP_MEQ_ADDR                                    0x0000021e
+
+#define REG_A4XX_CP_MEQ_DATA                                    0x0000021f
+
+#define REG_A4XX_CP_MERCIU_ADDR                                 0x00000220
+
+#define REG_A4XX_CP_MERCIU_DATA                                 0x00000221
+
+#define REG_A4XX_CP_MERCIU_DATA2                                0x00000222
+
+#define REG_A4XX_CP_PFP_UCODE_ADDR                              0x00000223
+
+#define REG_A4XX_CP_PFP_UCODE_DATA                              0x00000224
+
+#define REG_A4XX_CP_ME_RAM_WADDR                                0x00000225
+
+#define REG_A4XX_CP_ME_RAM_RADDR                                0x00000226
+
+#define REG_A4XX_CP_ME_RAM_DATA                                 0x00000227
+
+#define REG_A4XX_CP_PREEMPT                                     0x0000022a
+
+#define REG_A4XX_CP_CNTL                                        0x0000022c
+
+#define REG_A4XX_CP_ME_CNTL                                     0x0000022d
+
+#define REG_A4XX_CP_DEBUG                                       0x0000022e
+
+#define REG_A4XX_CP_DEBUG_ECO_CONTROL                           0x00000231
+
+#define REG_A4XX_CP_DRAW_STATE_ADDR                             0x00000232
+
+#define REG_A4XX_CP_PROTECT_REG_0                               0x00000240
+
+static inline uint32_t REG_A4XX_CP_PROTECT(uint32_t i0) { return 0x00000240 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_CP_PROTECT_REG(uint32_t i0) { return 0x00000240 + 0x1*i0; }
+
+#define REG_A4XX_CP_PROTECT_CTRL                                0x00000250
+
+#define REG_A4XX_CP_ST_BASE                                     0x000004c0
+
+#define REG_A4XX_CP_STQ_AVAIL                                   0x000004ce
+
+#define REG_A4XX_CP_MERCIU_STAT                                 0x000004d0
+
+#define REG_A4XX_CP_WFI_PEND_CTR                                0x000004d2
+
+#define REG_A4XX_CP_HW_FAULT                                    0x000004d8
+
+#define REG_A4XX_CP_PROTECT_STATUS                              0x000004da
+
+#define REG_A4XX_CP_EVENTS_IN_FLIGHT                            0x000004dd
+
+#define REG_A4XX_CP_PERFCTR_CP_SEL_0                            0x00000500
+
+#define REG_A4XX_CP_PERFCOMBINER_SELECT                         0x0000050b
+
+static inline uint32_t REG_A4XX_CP_SCRATCH(uint32_t i0) { return 0x00000578 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_CP_SCRATCH_REG(uint32_t i0) { return 0x00000578 + 0x1*i0; }
+
+#define REG_A4XX_SP_VS_STATUS                                   0x00000ec0
+
+#define REG_A4XX_SP_PERFCTR_SP_SEL_11                           0x00000ecf
+
+#define REG_A4XX_SP_SP_CTRL_REG                                 0x000022c0
+#define A4XX_SP_SP_CTRL_REG_BINNING_PASS                        0x00080000
+
+#define REG_A4XX_SP_INSTR_CACHE_CTRL                            0x000022c1
+
+#define REG_A4XX_SP_VS_CTRL_REG0                                0x000022c4
+#define A4XX_SP_VS_CTRL_REG0_THREADMODE__MASK                   0x00000001
+#define A4XX_SP_VS_CTRL_REG0_THREADMODE__SHIFT                  0
+static inline uint32_t A4XX_SP_VS_CTRL_REG0_THREADMODE(enum a3xx_threadmode val)
+{
+        return ((val) << A4XX_SP_VS_CTRL_REG0_THREADMODE__SHIFT) & A4XX_SP_VS_CTRL_REG0_THREADMODE__MASK;
+}
+#define A4XX_SP_VS_CTRL_REG0_VARYING                            0x00000002
+#define A4XX_SP_VS_CTRL_REG0_CACHEINVALID                       0x00000004
+#define A4XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__MASK             0x000003f0
+#define A4XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT            4
+static inline uint32_t A4XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT) & A4XX_SP_VS_CTRL_REG0_HALFREGFOOTPRINT__MASK;
+}
+#define A4XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__MASK             0x0003fc00
+#define A4XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT            10
+static inline uint32_t A4XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT) & A4XX_SP_VS_CTRL_REG0_FULLREGFOOTPRINT__MASK;
+}
+#define A4XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__MASK              0x000c0000
+#define A4XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__SHIFT             18
+static inline uint32_t A4XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__SHIFT) & A4XX_SP_VS_CTRL_REG0_INOUTREGOVERLAP__MASK;
+}
+#define A4XX_SP_VS_CTRL_REG0_THREADSIZE__MASK                   0x00100000
+#define A4XX_SP_VS_CTRL_REG0_THREADSIZE__SHIFT                  20
+static inline uint32_t A4XX_SP_VS_CTRL_REG0_THREADSIZE(enum a3xx_threadsize val)
+{
+        return ((val) << A4XX_SP_VS_CTRL_REG0_THREADSIZE__SHIFT) & A4XX_SP_VS_CTRL_REG0_THREADSIZE__MASK;
+}
+#define A4XX_SP_VS_CTRL_REG0_SUPERTHREADMODE                    0x00200000
+#define A4XX_SP_VS_CTRL_REG0_PIXLODENABLE                       0x00400000
+
+#define REG_A4XX_SP_VS_CTRL_REG1                                0x000022c5
+#define A4XX_SP_VS_CTRL_REG1_CONSTLENGTH__MASK                  0x000000ff
+#define A4XX_SP_VS_CTRL_REG1_CONSTLENGTH__SHIFT                 0
+static inline uint32_t A4XX_SP_VS_CTRL_REG1_CONSTLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_CTRL_REG1_CONSTLENGTH__SHIFT) & A4XX_SP_VS_CTRL_REG1_CONSTLENGTH__MASK;
+}
+#define A4XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING__MASK           0x7f000000
+#define A4XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING__SHIFT          24
+static inline uint32_t A4XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING__SHIFT) & A4XX_SP_VS_CTRL_REG1_INITIALOUTSTANDING__MASK;
+}
+
+#define REG_A4XX_SP_VS_PARAM_REG                                0x000022c6
+#define A4XX_SP_VS_PARAM_REG_POSREGID__MASK                     0x000000ff
+#define A4XX_SP_VS_PARAM_REG_POSREGID__SHIFT                    0
+static inline uint32_t A4XX_SP_VS_PARAM_REG_POSREGID(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_PARAM_REG_POSREGID__SHIFT) & A4XX_SP_VS_PARAM_REG_POSREGID__MASK;
+}
+#define A4XX_SP_VS_PARAM_REG_PSIZEREGID__MASK                   0x0000ff00
+#define A4XX_SP_VS_PARAM_REG_PSIZEREGID__SHIFT                  8
+static inline uint32_t A4XX_SP_VS_PARAM_REG_PSIZEREGID(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_PARAM_REG_PSIZEREGID__SHIFT) & A4XX_SP_VS_PARAM_REG_PSIZEREGID__MASK;
+}
+#define A4XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__MASK                0xfff00000
+#define A4XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__SHIFT               20
+static inline uint32_t A4XX_SP_VS_PARAM_REG_TOTALVSOUTVAR(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__SHIFT) & A4XX_SP_VS_PARAM_REG_TOTALVSOUTVAR__MASK;
+}
+
+static inline uint32_t REG_A4XX_SP_VS_OUT(uint32_t i0) { return 0x000022c7 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_SP_VS_OUT_REG(uint32_t i0) { return 0x000022c7 + 0x1*i0; }
+#define A4XX_SP_VS_OUT_REG_A_REGID__MASK                        0x000001ff
+#define A4XX_SP_VS_OUT_REG_A_REGID__SHIFT                       0
+static inline uint32_t A4XX_SP_VS_OUT_REG_A_REGID(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_OUT_REG_A_REGID__SHIFT) & A4XX_SP_VS_OUT_REG_A_REGID__MASK;
+}
+#define A4XX_SP_VS_OUT_REG_A_COMPMASK__MASK                     0x00001e00
+#define A4XX_SP_VS_OUT_REG_A_COMPMASK__SHIFT                    9
+static inline uint32_t A4XX_SP_VS_OUT_REG_A_COMPMASK(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_OUT_REG_A_COMPMASK__SHIFT) & A4XX_SP_VS_OUT_REG_A_COMPMASK__MASK;
+}
+#define A4XX_SP_VS_OUT_REG_B_REGID__MASK                        0x01ff0000
+#define A4XX_SP_VS_OUT_REG_B_REGID__SHIFT                       16
+static inline uint32_t A4XX_SP_VS_OUT_REG_B_REGID(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_OUT_REG_B_REGID__SHIFT) & A4XX_SP_VS_OUT_REG_B_REGID__MASK;
+}
+#define A4XX_SP_VS_OUT_REG_B_COMPMASK__MASK                     0x1e000000
+#define A4XX_SP_VS_OUT_REG_B_COMPMASK__SHIFT                    25
+static inline uint32_t A4XX_SP_VS_OUT_REG_B_COMPMASK(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_OUT_REG_B_COMPMASK__SHIFT) & A4XX_SP_VS_OUT_REG_B_COMPMASK__MASK;
+}
+
+static inline uint32_t REG_A4XX_SP_VS_VPC_DST(uint32_t i0) { return 0x000022d8 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_SP_VS_VPC_DST_REG(uint32_t i0) { return 0x000022d8 + 0x1*i0; }
+#define A4XX_SP_VS_VPC_DST_REG_OUTLOC0__MASK                    0x000000ff
+#define A4XX_SP_VS_VPC_DST_REG_OUTLOC0__SHIFT                   0
+static inline uint32_t A4XX_SP_VS_VPC_DST_REG_OUTLOC0(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_VPC_DST_REG_OUTLOC0__SHIFT) & A4XX_SP_VS_VPC_DST_REG_OUTLOC0__MASK;
+}
+#define A4XX_SP_VS_VPC_DST_REG_OUTLOC1__MASK                    0x0000ff00
+#define A4XX_SP_VS_VPC_DST_REG_OUTLOC1__SHIFT                   8
+static inline uint32_t A4XX_SP_VS_VPC_DST_REG_OUTLOC1(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_VPC_DST_REG_OUTLOC1__SHIFT) & A4XX_SP_VS_VPC_DST_REG_OUTLOC1__MASK;
+}
+#define A4XX_SP_VS_VPC_DST_REG_OUTLOC2__MASK                    0x00ff0000
+#define A4XX_SP_VS_VPC_DST_REG_OUTLOC2__SHIFT                   16
+static inline uint32_t A4XX_SP_VS_VPC_DST_REG_OUTLOC2(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_VPC_DST_REG_OUTLOC2__SHIFT) & A4XX_SP_VS_VPC_DST_REG_OUTLOC2__MASK;
+}
+#define A4XX_SP_VS_VPC_DST_REG_OUTLOC3__MASK                    0xff000000
+#define A4XX_SP_VS_VPC_DST_REG_OUTLOC3__SHIFT                   24
+static inline uint32_t A4XX_SP_VS_VPC_DST_REG_OUTLOC3(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_VPC_DST_REG_OUTLOC3__SHIFT) & A4XX_SP_VS_VPC_DST_REG_OUTLOC3__MASK;
+}
+
+#define REG_A4XX_SP_VS_OBJ_OFFSET_REG                           0x000022e0
+#define A4XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK       0x01ff0000
+#define A4XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT      16
+static inline uint32_t A4XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_SP_VS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK;
+}
+#define A4XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK         0xfe000000
+#define A4XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT        25
+static inline uint32_t A4XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT) & A4XX_SP_VS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK;
+}
+
+#define REG_A4XX_SP_VS_OBJ_START                                0x000022e1
+
+#define REG_A4XX_SP_VS_PVT_MEM_PARAM                            0x000022e2
+
+#define REG_A4XX_SP_VS_PVT_MEM_ADDR                             0x000022e3
+
+#define REG_A4XX_SP_VS_LENGTH_REG                               0x000022e5
+
+#define REG_A4XX_SP_FS_CTRL_REG0                                0x000022e8
+#define A4XX_SP_FS_CTRL_REG0_THREADMODE__MASK                   0x00000001
+#define A4XX_SP_FS_CTRL_REG0_THREADMODE__SHIFT                  0
+static inline uint32_t A4XX_SP_FS_CTRL_REG0_THREADMODE(enum a3xx_threadmode val)
+{
+        return ((val) << A4XX_SP_FS_CTRL_REG0_THREADMODE__SHIFT) & A4XX_SP_FS_CTRL_REG0_THREADMODE__MASK;
+}
+#define A4XX_SP_FS_CTRL_REG0_VARYING                            0x00000002
+#define A4XX_SP_FS_CTRL_REG0_CACHEINVALID                       0x00000004
+#define A4XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__MASK             0x000003f0
+#define A4XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT            4
+static inline uint32_t A4XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT(uint32_t val)
+{
+        return ((val) << A4XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__SHIFT) & A4XX_SP_FS_CTRL_REG0_HALFREGFOOTPRINT__MASK;
+}
+#define A4XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__MASK             0x0003fc00
+#define A4XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT            10
+static inline uint32_t A4XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT(uint32_t val)
+{
+        return ((val) << A4XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__SHIFT) & A4XX_SP_FS_CTRL_REG0_FULLREGFOOTPRINT__MASK;
+}
+#define A4XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__MASK              0x000c0000
+#define A4XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__SHIFT             18
+static inline uint32_t A4XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP(uint32_t val)
+{
+        return ((val) << A4XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__SHIFT) & A4XX_SP_FS_CTRL_REG0_INOUTREGOVERLAP__MASK;
+}
+#define A4XX_SP_FS_CTRL_REG0_THREADSIZE__MASK                   0x00100000
+#define A4XX_SP_FS_CTRL_REG0_THREADSIZE__SHIFT                  20
+static inline uint32_t A4XX_SP_FS_CTRL_REG0_THREADSIZE(enum a3xx_threadsize val)
+{
+        return ((val) << A4XX_SP_FS_CTRL_REG0_THREADSIZE__SHIFT) & A4XX_SP_FS_CTRL_REG0_THREADSIZE__MASK;
+}
+#define A4XX_SP_FS_CTRL_REG0_SUPERTHREADMODE                    0x00200000
+#define A4XX_SP_FS_CTRL_REG0_PIXLODENABLE                       0x00400000
+
+#define REG_A4XX_SP_FS_CTRL_REG1                                0x000022e9
+#define A4XX_SP_FS_CTRL_REG1_CONSTLENGTH__MASK                  0x000000ff
+#define A4XX_SP_FS_CTRL_REG1_CONSTLENGTH__SHIFT                 0
+static inline uint32_t A4XX_SP_FS_CTRL_REG1_CONSTLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_SP_FS_CTRL_REG1_CONSTLENGTH__SHIFT) & A4XX_SP_FS_CTRL_REG1_CONSTLENGTH__MASK;
+}
+#define A4XX_SP_FS_CTRL_REG1_VARYING                            0x00100000
+
+#define REG_A4XX_SP_FS_OBJ_OFFSET_REG                           0x000022ea
+#define A4XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK       0x01ff0000
+#define A4XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT      16
+static inline uint32_t A4XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_SP_FS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK;
+}
+#define A4XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK         0xfe000000
+#define A4XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT        25
+static inline uint32_t A4XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT) & A4XX_SP_FS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK;
+}
+
+#define REG_A4XX_SP_FS_OBJ_START                                0x000022eb
+
+#define REG_A4XX_SP_FS_PVT_MEM_PARAM                            0x000022ec
+
+#define REG_A4XX_SP_FS_PVT_MEM_ADDR                             0x000022ed
+
+#define REG_A4XX_SP_FS_LENGTH_REG                               0x000022ef
+
+#define REG_A4XX_SP_FS_OUTPUT_REG                               0x000022f0
+#define A4XX_SP_FS_OUTPUT_REG_DEPTH_ENABLE                      0x00000080
+#define A4XX_SP_FS_OUTPUT_REG_DEPTH_REGID__MASK                 0x0000ff00
+#define A4XX_SP_FS_OUTPUT_REG_DEPTH_REGID__SHIFT                8
+static inline uint32_t A4XX_SP_FS_OUTPUT_REG_DEPTH_REGID(uint32_t val)
+{
+        return ((val) << A4XX_SP_FS_OUTPUT_REG_DEPTH_REGID__SHIFT) & A4XX_SP_FS_OUTPUT_REG_DEPTH_REGID__MASK;
+}
+
+static inline uint32_t REG_A4XX_SP_FS_MRT(uint32_t i0) { return 0x000022f1 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_SP_FS_MRT_REG(uint32_t i0) { return 0x000022f1 + 0x1*i0; }
+#define A4XX_SP_FS_MRT_REG_REGID__MASK                          0x000000ff
+#define A4XX_SP_FS_MRT_REG_REGID__SHIFT                         0
+static inline uint32_t A4XX_SP_FS_MRT_REG_REGID(uint32_t val)
+{
+        return ((val) << A4XX_SP_FS_MRT_REG_REGID__SHIFT) & A4XX_SP_FS_MRT_REG_REGID__MASK;
+}
+#define A4XX_SP_FS_MRT_REG_HALF_PRECISION                       0x00000100
+#define A4XX_SP_FS_MRT_REG_MRTFORMAT__MASK                      0x0003f000
+#define A4XX_SP_FS_MRT_REG_MRTFORMAT__SHIFT                     12
+static inline uint32_t A4XX_SP_FS_MRT_REG_MRTFORMAT(enum a4xx_color_fmt val)
+{
+        return ((val) << A4XX_SP_FS_MRT_REG_MRTFORMAT__SHIFT) & A4XX_SP_FS_MRT_REG_MRTFORMAT__MASK;
+}
+
+#define REG_A4XX_SP_HS_OBJ_OFFSET_REG                           0x0000230d
+#define A4XX_SP_HS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK       0x01ff0000
+#define A4XX_SP_HS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT      16
+static inline uint32_t A4XX_SP_HS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_SP_HS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_SP_HS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK;
+}
+#define A4XX_SP_HS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK         0xfe000000
+#define A4XX_SP_HS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT        25
+static inline uint32_t A4XX_SP_HS_OBJ_OFFSET_REG_SHADEROBJOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_SP_HS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT) & A4XX_SP_HS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK;
+}
+
+#define REG_A4XX_SP_DS_OBJ_OFFSET_REG                           0x00002334
+#define A4XX_SP_DS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK       0x01ff0000
+#define A4XX_SP_DS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT      16
+static inline uint32_t A4XX_SP_DS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_SP_DS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_SP_DS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK;
+}
+#define A4XX_SP_DS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK         0xfe000000
+#define A4XX_SP_DS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT        25
+static inline uint32_t A4XX_SP_DS_OBJ_OFFSET_REG_SHADEROBJOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_SP_DS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT) & A4XX_SP_DS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK;
+}
+
+#define REG_A4XX_SP_GS_OBJ_OFFSET_REG                           0x0000235b
+#define A4XX_SP_GS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK       0x01ff0000
+#define A4XX_SP_GS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT      16
+static inline uint32_t A4XX_SP_GS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_SP_GS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_SP_GS_OBJ_OFFSET_REG_CONSTOBJECTOFFSET__MASK;
+}
+#define A4XX_SP_GS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK         0xfe000000
+#define A4XX_SP_GS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT        25
+static inline uint32_t A4XX_SP_GS_OBJ_OFFSET_REG_SHADEROBJOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_SP_GS_OBJ_OFFSET_REG_SHADEROBJOFFSET__SHIFT) & A4XX_SP_GS_OBJ_OFFSET_REG_SHADEROBJOFFSET__MASK;
+}
+
+#define REG_A4XX_SP_GS_LENGTH_REG                               0x00002360
+
+#define REG_A4XX_VPC_DEBUG_RAM_SEL                              0x00000e60
+
+#define REG_A4XX_VPC_DEBUG_RAM_READ                             0x00000e61
+
+#define REG_A4XX_VPC_DEBUG_ECO_CONTROL                          0x00000e64
+
+#define REG_A4XX_VPC_PERFCTR_VPC_SEL_3                          0x00000e68
+
+#define REG_A4XX_VPC_ATTR                                       0x00002140
+#define A4XX_VPC_ATTR_TOTALATTR__MASK                           0x000001ff
+#define A4XX_VPC_ATTR_TOTALATTR__SHIFT                          0
+static inline uint32_t A4XX_VPC_ATTR_TOTALATTR(uint32_t val)
+{
+        return ((val) << A4XX_VPC_ATTR_TOTALATTR__SHIFT) & A4XX_VPC_ATTR_TOTALATTR__MASK;
+}
+#define A4XX_VPC_ATTR_PSIZE                                     0x00000200
+#define A4XX_VPC_ATTR_THRDASSIGN__MASK                          0x00003000
+#define A4XX_VPC_ATTR_THRDASSIGN__SHIFT                         12
+static inline uint32_t A4XX_VPC_ATTR_THRDASSIGN(uint32_t val)
+{
+        return ((val) << A4XX_VPC_ATTR_THRDASSIGN__SHIFT) & A4XX_VPC_ATTR_THRDASSIGN__MASK;
+}
+#define A4XX_VPC_ATTR_ENABLE                                    0x02000000
+
+#define REG_A4XX_VPC_PACK                                       0x00002141
+#define A4XX_VPC_PACK_NUMBYPASSVAR__MASK                        0x000000ff
+#define A4XX_VPC_PACK_NUMBYPASSVAR__SHIFT                       0
+static inline uint32_t A4XX_VPC_PACK_NUMBYPASSVAR(uint32_t val)
+{
+        return ((val) << A4XX_VPC_PACK_NUMBYPASSVAR__SHIFT) & A4XX_VPC_PACK_NUMBYPASSVAR__MASK;
+}
+#define A4XX_VPC_PACK_NUMFPNONPOSVAR__MASK                      0x0000ff00
+#define A4XX_VPC_PACK_NUMFPNONPOSVAR__SHIFT                     8
+static inline uint32_t A4XX_VPC_PACK_NUMFPNONPOSVAR(uint32_t val)
+{
+        return ((val) << A4XX_VPC_PACK_NUMFPNONPOSVAR__SHIFT) & A4XX_VPC_PACK_NUMFPNONPOSVAR__MASK;
+}
+#define A4XX_VPC_PACK_NUMNONPOSVSVAR__MASK                      0x00ff0000
+#define A4XX_VPC_PACK_NUMNONPOSVSVAR__SHIFT                     16
+static inline uint32_t A4XX_VPC_PACK_NUMNONPOSVSVAR(uint32_t val)
+{
+        return ((val) << A4XX_VPC_PACK_NUMNONPOSVSVAR__SHIFT) & A4XX_VPC_PACK_NUMNONPOSVSVAR__MASK;
+}
+
+static inline uint32_t REG_A4XX_VPC_VARYING_INTERP(uint32_t i0) { return 0x00002142 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_VPC_VARYING_INTERP_MODE(uint32_t i0) { return 0x00002142 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_VPC_VARYING_PS_REPL(uint32_t i0) { return 0x0000214a + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_VPC_VARYING_PS_REPL_MODE(uint32_t i0) { return 0x0000214a + 0x1*i0; }
+
+#define REG_A4XX_VPC_SO_FLUSH_WADDR_3                           0x0000216e
+
+#define REG_A4XX_VSC_BIN_SIZE                                   0x00000c00
+#define A4XX_VSC_BIN_SIZE_WIDTH__MASK                           0x0000001f
+#define A4XX_VSC_BIN_SIZE_WIDTH__SHIFT                          0
+static inline uint32_t A4XX_VSC_BIN_SIZE_WIDTH(uint32_t val)
+{
+        return ((val >> 5) << A4XX_VSC_BIN_SIZE_WIDTH__SHIFT) & A4XX_VSC_BIN_SIZE_WIDTH__MASK;
+}
+#define A4XX_VSC_BIN_SIZE_HEIGHT__MASK                          0x000003e0
+#define A4XX_VSC_BIN_SIZE_HEIGHT__SHIFT                         5
+static inline uint32_t A4XX_VSC_BIN_SIZE_HEIGHT(uint32_t val)
+{
+        return ((val >> 5) << A4XX_VSC_BIN_SIZE_HEIGHT__SHIFT) & A4XX_VSC_BIN_SIZE_HEIGHT__MASK;
+}
+
+#define REG_A4XX_VSC_SIZE_ADDRESS                               0x00000c01
+
+#define REG_A4XX_VSC_SIZE_ADDRESS2                              0x00000c02
+
+#define REG_A4XX_VSC_DEBUG_ECO_CONTROL                          0x00000c03
+
+static inline uint32_t REG_A4XX_VSC_PIPE_CONFIG(uint32_t i0) { return 0x00000c08 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_VSC_PIPE_CONFIG_REG(uint32_t i0) { return 0x00000c08 + 0x1*i0; }
+#define A4XX_VSC_PIPE_CONFIG_REG_X__MASK                        0x000003ff
+#define A4XX_VSC_PIPE_CONFIG_REG_X__SHIFT                       0
+static inline uint32_t A4XX_VSC_PIPE_CONFIG_REG_X(uint32_t val)
+{
+        return ((val) << A4XX_VSC_PIPE_CONFIG_REG_X__SHIFT) & A4XX_VSC_PIPE_CONFIG_REG_X__MASK;
+}
+#define A4XX_VSC_PIPE_CONFIG_REG_Y__MASK                        0x000ffc00
+#define A4XX_VSC_PIPE_CONFIG_REG_Y__SHIFT                       10
+static inline uint32_t A4XX_VSC_PIPE_CONFIG_REG_Y(uint32_t val)
+{
+        return ((val) << A4XX_VSC_PIPE_CONFIG_REG_Y__SHIFT) & A4XX_VSC_PIPE_CONFIG_REG_Y__MASK;
+}
+#define A4XX_VSC_PIPE_CONFIG_REG_W__MASK                        0x00f00000
+#define A4XX_VSC_PIPE_CONFIG_REG_W__SHIFT                       20
+static inline uint32_t A4XX_VSC_PIPE_CONFIG_REG_W(uint32_t val)
+{
+        return ((val) << A4XX_VSC_PIPE_CONFIG_REG_W__SHIFT) & A4XX_VSC_PIPE_CONFIG_REG_W__MASK;
+}
+#define A4XX_VSC_PIPE_CONFIG_REG_H__MASK                        0x0f000000
+#define A4XX_VSC_PIPE_CONFIG_REG_H__SHIFT                       24
+static inline uint32_t A4XX_VSC_PIPE_CONFIG_REG_H(uint32_t val)
+{
+        return ((val) << A4XX_VSC_PIPE_CONFIG_REG_H__SHIFT) & A4XX_VSC_PIPE_CONFIG_REG_H__MASK;
+}
+
+static inline uint32_t REG_A4XX_VSC_PIPE_DATA_ADDRESS(uint32_t i0) { return 0x00000c10 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_VSC_PIPE_DATA_ADDRESS_REG(uint32_t i0) { return 0x00000c10 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_VSC_PIPE_DATA_LENGTH(uint32_t i0) { return 0x00000c18 + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_VSC_PIPE_DATA_LENGTH_REG(uint32_t i0) { return 0x00000c18 + 0x1*i0; }
+
+#define REG_A4XX_VSC_PIPE_PARTIAL_POSN_1                        0x00000c41
+
+#define REG_A4XX_VSC_PERFCTR_VSC_SEL_0                          0x00000c50
+
+#define REG_A4XX_VSC_PERFCTR_VSC_SEL_1                          0x00000c51
+
+#define REG_A4XX_VFD_DEBUG_CONTROL                              0x00000e40
+
+#define REG_A4XX_VFD_PERFCTR_VFD_SEL_7                          0x00000e4a
+
+#define REG_A4XX_VFD_CONTROL_0                                  0x00002200
+#define A4XX_VFD_CONTROL_0_TOTALATTRTOVS__MASK                  0x000000ff
+#define A4XX_VFD_CONTROL_0_TOTALATTRTOVS__SHIFT                 0
+static inline uint32_t A4XX_VFD_CONTROL_0_TOTALATTRTOVS(uint32_t val)
+{
+        return ((val) << A4XX_VFD_CONTROL_0_TOTALATTRTOVS__SHIFT) & A4XX_VFD_CONTROL_0_TOTALATTRTOVS__MASK;
+}
+#define A4XX_VFD_CONTROL_0_BYPASSATTROVS__MASK                  0x0001fe00
+#define A4XX_VFD_CONTROL_0_BYPASSATTROVS__SHIFT                 9
+static inline uint32_t A4XX_VFD_CONTROL_0_BYPASSATTROVS(uint32_t val)
+{
+        return ((val) << A4XX_VFD_CONTROL_0_BYPASSATTROVS__SHIFT) & A4XX_VFD_CONTROL_0_BYPASSATTROVS__MASK;
+}
+#define A4XX_VFD_CONTROL_0_STRMDECINSTRCNT__MASK                0x03f00000
+#define A4XX_VFD_CONTROL_0_STRMDECINSTRCNT__SHIFT               20
+static inline uint32_t A4XX_VFD_CONTROL_0_STRMDECINSTRCNT(uint32_t val)
+{
+        return ((val) << A4XX_VFD_CONTROL_0_STRMDECINSTRCNT__SHIFT) & A4XX_VFD_CONTROL_0_STRMDECINSTRCNT__MASK;
+}
+#define A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT__MASK              0xfc000000
+#define A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT__SHIFT             26
+static inline uint32_t A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT(uint32_t val)
+{
+        return ((val) << A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT__SHIFT) & A4XX_VFD_CONTROL_0_STRMFETCHINSTRCNT__MASK;
+}
+
+#define REG_A4XX_VFD_CONTROL_1                                  0x00002201
+#define A4XX_VFD_CONTROL_1_MAXSTORAGE__MASK                     0x0000ffff
+#define A4XX_VFD_CONTROL_1_MAXSTORAGE__SHIFT                    0
+static inline uint32_t A4XX_VFD_CONTROL_1_MAXSTORAGE(uint32_t val)
+{
+        return ((val) << A4XX_VFD_CONTROL_1_MAXSTORAGE__SHIFT) & A4XX_VFD_CONTROL_1_MAXSTORAGE__MASK;
+}
+#define A4XX_VFD_CONTROL_1_REGID4VTX__MASK                      0x00ff0000
+#define A4XX_VFD_CONTROL_1_REGID4VTX__SHIFT                     16
+static inline uint32_t A4XX_VFD_CONTROL_1_REGID4VTX(uint32_t val)
+{
+        return ((val) << A4XX_VFD_CONTROL_1_REGID4VTX__SHIFT) & A4XX_VFD_CONTROL_1_REGID4VTX__MASK;
+}
+#define A4XX_VFD_CONTROL_1_REGID4INST__MASK                     0xff000000
+#define A4XX_VFD_CONTROL_1_REGID4INST__SHIFT                    24
+static inline uint32_t A4XX_VFD_CONTROL_1_REGID4INST(uint32_t val)
+{
+        return ((val) << A4XX_VFD_CONTROL_1_REGID4INST__SHIFT) & A4XX_VFD_CONTROL_1_REGID4INST__MASK;
+}
+
+#define REG_A4XX_VFD_CONTROL_2                                  0x00002202
+
+#define REG_A4XX_VFD_CONTROL_3                                  0x00002203
+
+#define REG_A4XX_VFD_CONTROL_4                                  0x00002204
+
+#define REG_A4XX_VFD_INDEX_OFFSET                               0x00002208
+
+static inline uint32_t REG_A4XX_VFD_FETCH(uint32_t i0) { return 0x0000220a + 0x4*i0; }
+
+static inline uint32_t REG_A4XX_VFD_FETCH_INSTR_0(uint32_t i0) { return 0x0000220a + 0x4*i0; }
+#define A4XX_VFD_FETCH_INSTR_0_FETCHSIZE__MASK                  0x0000007f
+#define A4XX_VFD_FETCH_INSTR_0_FETCHSIZE__SHIFT                 0
+static inline uint32_t A4XX_VFD_FETCH_INSTR_0_FETCHSIZE(uint32_t val)
+{
+        return ((val) << A4XX_VFD_FETCH_INSTR_0_FETCHSIZE__SHIFT) & A4XX_VFD_FETCH_INSTR_0_FETCHSIZE__MASK;
+}
+#define A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE__MASK                  0x0001ff80
+#define A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE__SHIFT                 7
+static inline uint32_t A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE(uint32_t val)
+{
+        return ((val) << A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE__SHIFT) & A4XX_VFD_FETCH_INSTR_0_BUFSTRIDE__MASK;
+}
+#define A4XX_VFD_FETCH_INSTR_0_SWITCHNEXT                       0x00080000
+#define A4XX_VFD_FETCH_INSTR_0_STEPRATE__MASK                   0xff000000
+#define A4XX_VFD_FETCH_INSTR_0_STEPRATE__SHIFT                  24
+static inline uint32_t A4XX_VFD_FETCH_INSTR_0_STEPRATE(uint32_t val)
+{
+        return ((val) << A4XX_VFD_FETCH_INSTR_0_STEPRATE__SHIFT) & A4XX_VFD_FETCH_INSTR_0_STEPRATE__MASK;
+}
+
+static inline uint32_t REG_A4XX_VFD_FETCH_INSTR_1(uint32_t i0) { return 0x0000220b + 0x4*i0; }
+
+static inline uint32_t REG_A4XX_VFD_FETCH_INSTR_2(uint32_t i0) { return 0x0000220c + 0x4*i0; }
+#define A4XX_VFD_FETCH_INSTR_2_SIZE__MASK                       0xfffffff0
+#define A4XX_VFD_FETCH_INSTR_2_SIZE__SHIFT                      4
+static inline uint32_t A4XX_VFD_FETCH_INSTR_2_SIZE(uint32_t val)
+{
+        return ((val >> 4) << A4XX_VFD_FETCH_INSTR_2_SIZE__SHIFT) & A4XX_VFD_FETCH_INSTR_2_SIZE__MASK;
+}
+
+static inline uint32_t REG_A4XX_VFD_FETCH_INSTR_3(uint32_t i0) { return 0x0000220d + 0x4*i0; }
+
+static inline uint32_t REG_A4XX_VFD_DECODE(uint32_t i0) { return 0x0000228a + 0x1*i0; }
+
+static inline uint32_t REG_A4XX_VFD_DECODE_INSTR(uint32_t i0) { return 0x0000228a + 0x1*i0; }
+#define A4XX_VFD_DECODE_INSTR_WRITEMASK__MASK                   0x0000000f
+#define A4XX_VFD_DECODE_INSTR_WRITEMASK__SHIFT                  0
+static inline uint32_t A4XX_VFD_DECODE_INSTR_WRITEMASK(uint32_t val)
+{
+        return ((val) << A4XX_VFD_DECODE_INSTR_WRITEMASK__SHIFT) & A4XX_VFD_DECODE_INSTR_WRITEMASK__MASK;
+}
+#define A4XX_VFD_DECODE_INSTR_CONSTFILL                         0x00000010
+#define A4XX_VFD_DECODE_INSTR_FORMAT__MASK                      0x00000fc0
+#define A4XX_VFD_DECODE_INSTR_FORMAT__SHIFT                     6
+static inline uint32_t A4XX_VFD_DECODE_INSTR_FORMAT(enum a4xx_vtx_fmt val)
+{
+        return ((val) << A4XX_VFD_DECODE_INSTR_FORMAT__SHIFT) & A4XX_VFD_DECODE_INSTR_FORMAT__MASK;
+}
+#define A4XX_VFD_DECODE_INSTR_REGID__MASK                       0x000ff000
+#define A4XX_VFD_DECODE_INSTR_REGID__SHIFT                      12
+static inline uint32_t A4XX_VFD_DECODE_INSTR_REGID(uint32_t val)
+{
+        return ((val) << A4XX_VFD_DECODE_INSTR_REGID__SHIFT) & A4XX_VFD_DECODE_INSTR_REGID__MASK;
+}
+#define A4XX_VFD_DECODE_INSTR_SWAP__MASK                        0x00c00000
+#define A4XX_VFD_DECODE_INSTR_SWAP__SHIFT                       22
+static inline uint32_t A4XX_VFD_DECODE_INSTR_SWAP(enum a3xx_color_swap val)
+{
+        return ((val) << A4XX_VFD_DECODE_INSTR_SWAP__SHIFT) & A4XX_VFD_DECODE_INSTR_SWAP__MASK;
+}
+#define A4XX_VFD_DECODE_INSTR_SHIFTCNT__MASK                    0x1f000000
+#define A4XX_VFD_DECODE_INSTR_SHIFTCNT__SHIFT                   24
+static inline uint32_t A4XX_VFD_DECODE_INSTR_SHIFTCNT(uint32_t val)
+{
+        return ((val) << A4XX_VFD_DECODE_INSTR_SHIFTCNT__SHIFT) & A4XX_VFD_DECODE_INSTR_SHIFTCNT__MASK;
+}
+#define A4XX_VFD_DECODE_INSTR_LASTCOMPVALID                     0x20000000
+#define A4XX_VFD_DECODE_INSTR_SWITCHNEXT                        0x40000000
+
+#define REG_A4XX_TPL1_DEBUG_ECO_CONTROL                         0x00000f00
+
+#define REG_A4XX_TPL1_PERFCTR_TP_SEL_7                          0x00000f0b
+
+#define REG_A4XX_TPL1_TP_TEX_OFFSET                             0x00002380
+
+#define REG_A4XX_TPL1_TP_CS_TEXMEMOBJ_BASE_ADDR                 0x000023a6
+
+#define REG_A4XX_GRAS_TSE_STATUS                                0x00000c80
+
+#define REG_A4XX_GRAS_DEBUG_ECO_CONTROL                         0x00000c81
+
+#define REG_A4XX_GRAS_PERFCTR_TSE_SEL_0                         0x00000c88
+
+#define REG_A4XX_GRAS_PERFCTR_TSE_SEL_3                         0x00000c8b
+
+#define REG_A4XX_GRAS_CL_CLIP_CNTL                              0x00002000
+
+#define REG_A4XX_GRAS_CLEAR_CNTL                                0x00002003
+#define A4XX_GRAS_CLEAR_CNTL_NOT_FASTCLEAR                      0x00000001
+
+#define REG_A4XX_GRAS_CL_GB_CLIP_ADJ                            0x00002004
+#define A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ__MASK                     0x000003ff
+#define A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ__SHIFT                    0
+static inline uint32_t A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ__SHIFT) & A4XX_GRAS_CL_GB_CLIP_ADJ_HORZ__MASK;
+}
+#define A4XX_GRAS_CL_GB_CLIP_ADJ_VERT__MASK                     0x000ffc00
+#define A4XX_GRAS_CL_GB_CLIP_ADJ_VERT__SHIFT                    10
+static inline uint32_t A4XX_GRAS_CL_GB_CLIP_ADJ_VERT(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_CL_GB_CLIP_ADJ_VERT__SHIFT) & A4XX_GRAS_CL_GB_CLIP_ADJ_VERT__MASK;
+}
+
+#define REG_A4XX_GRAS_CL_VPORT_XOFFSET_0                        0x00002008
+#define A4XX_GRAS_CL_VPORT_XOFFSET_0__MASK                      0xffffffff
+#define A4XX_GRAS_CL_VPORT_XOFFSET_0__SHIFT                     0
+static inline uint32_t A4XX_GRAS_CL_VPORT_XOFFSET_0(float val)
+{
+        return ((fui(val)) << A4XX_GRAS_CL_VPORT_XOFFSET_0__SHIFT) & A4XX_GRAS_CL_VPORT_XOFFSET_0__MASK;
+}
+
+#define REG_A4XX_GRAS_CL_VPORT_XSCALE_0                         0x00002009
+#define A4XX_GRAS_CL_VPORT_XSCALE_0__MASK                       0xffffffff
+#define A4XX_GRAS_CL_VPORT_XSCALE_0__SHIFT                      0
+static inline uint32_t A4XX_GRAS_CL_VPORT_XSCALE_0(float val)
+{
+        return ((fui(val)) << A4XX_GRAS_CL_VPORT_XSCALE_0__SHIFT) & A4XX_GRAS_CL_VPORT_XSCALE_0__MASK;
+}
+
+#define REG_A4XX_GRAS_CL_VPORT_YOFFSET_0                        0x0000200a
+#define A4XX_GRAS_CL_VPORT_YOFFSET_0__MASK                      0xffffffff
+#define A4XX_GRAS_CL_VPORT_YOFFSET_0__SHIFT                     0
+static inline uint32_t A4XX_GRAS_CL_VPORT_YOFFSET_0(float val)
+{
+        return ((fui(val)) << A4XX_GRAS_CL_VPORT_YOFFSET_0__SHIFT) & A4XX_GRAS_CL_VPORT_YOFFSET_0__MASK;
+}
+
+#define REG_A4XX_GRAS_CL_VPORT_YSCALE_0                         0x0000200b
+#define A4XX_GRAS_CL_VPORT_YSCALE_0__MASK                       0xffffffff
+#define A4XX_GRAS_CL_VPORT_YSCALE_0__SHIFT                      0
+static inline uint32_t A4XX_GRAS_CL_VPORT_YSCALE_0(float val)
+{
+        return ((fui(val)) << A4XX_GRAS_CL_VPORT_YSCALE_0__SHIFT) & A4XX_GRAS_CL_VPORT_YSCALE_0__MASK;
+}
+
+#define REG_A4XX_GRAS_CL_VPORT_ZOFFSET_0                        0x0000200c
+#define A4XX_GRAS_CL_VPORT_ZOFFSET_0__MASK                      0xffffffff
+#define A4XX_GRAS_CL_VPORT_ZOFFSET_0__SHIFT                     0
+static inline uint32_t A4XX_GRAS_CL_VPORT_ZOFFSET_0(float val)
+{
+        return ((fui(val)) << A4XX_GRAS_CL_VPORT_ZOFFSET_0__SHIFT) & A4XX_GRAS_CL_VPORT_ZOFFSET_0__MASK;
+}
+
+#define REG_A4XX_GRAS_CL_VPORT_ZSCALE_0                         0x0000200d
+#define A4XX_GRAS_CL_VPORT_ZSCALE_0__MASK                       0xffffffff
+#define A4XX_GRAS_CL_VPORT_ZSCALE_0__SHIFT                      0
+static inline uint32_t A4XX_GRAS_CL_VPORT_ZSCALE_0(float val)
+{
+        return ((fui(val)) << A4XX_GRAS_CL_VPORT_ZSCALE_0__SHIFT) & A4XX_GRAS_CL_VPORT_ZSCALE_0__MASK;
+}
+
+#define REG_A4XX_GRAS_SU_POINT_MINMAX                           0x00002070
+#define A4XX_GRAS_SU_POINT_MINMAX_MIN__MASK                     0x0000ffff
+#define A4XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT                    0
+static inline uint32_t A4XX_GRAS_SU_POINT_MINMAX_MIN(float val)
+{
+        return ((((uint32_t)(val * 16.0))) << A4XX_GRAS_SU_POINT_MINMAX_MIN__SHIFT) & A4XX_GRAS_SU_POINT_MINMAX_MIN__MASK;
+}
+#define A4XX_GRAS_SU_POINT_MINMAX_MAX__MASK                     0xffff0000
+#define A4XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT                    16
+static inline uint32_t A4XX_GRAS_SU_POINT_MINMAX_MAX(float val)
+{
+        return ((((uint32_t)(val * 16.0))) << A4XX_GRAS_SU_POINT_MINMAX_MAX__SHIFT) & A4XX_GRAS_SU_POINT_MINMAX_MAX__MASK;
+}
+
+#define REG_A4XX_GRAS_SU_POINT_SIZE                             0x00002071
+#define A4XX_GRAS_SU_POINT_SIZE__MASK                           0xffffffff
+#define A4XX_GRAS_SU_POINT_SIZE__SHIFT                          0
+static inline uint32_t A4XX_GRAS_SU_POINT_SIZE(float val)
+{
+        return ((((int32_t)(val * 16.0))) << A4XX_GRAS_SU_POINT_SIZE__SHIFT) & A4XX_GRAS_SU_POINT_SIZE__MASK;
+}
+
+#define REG_A4XX_GRAS_ALPHA_CONTROL                             0x00002073
+#define A4XX_GRAS_ALPHA_CONTROL_ALPHA_TEST_ENABLE               0x00000004
+
+#define REG_A4XX_GRAS_SU_POLY_OFFSET_SCALE                      0x00002074
+#define A4XX_GRAS_SU_POLY_OFFSET_SCALE__MASK                    0xffffffff
+#define A4XX_GRAS_SU_POLY_OFFSET_SCALE__SHIFT                   0
+static inline uint32_t A4XX_GRAS_SU_POLY_OFFSET_SCALE(float val)
+{
+        return ((fui(val)) << A4XX_GRAS_SU_POLY_OFFSET_SCALE__SHIFT) & A4XX_GRAS_SU_POLY_OFFSET_SCALE__MASK;
+}
+
+#define REG_A4XX_GRAS_SU_POLY_OFFSET_OFFSET                     0x00002075
+#define A4XX_GRAS_SU_POLY_OFFSET_OFFSET__MASK                   0xffffffff
+#define A4XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT                  0
+static inline uint32_t A4XX_GRAS_SU_POLY_OFFSET_OFFSET(float val)
+{
+        return ((fui(val)) << A4XX_GRAS_SU_POLY_OFFSET_OFFSET__SHIFT) & A4XX_GRAS_SU_POLY_OFFSET_OFFSET__MASK;
+}
+
+#define REG_A4XX_GRAS_SC_EXTENT_WINDOW_TL                       0x0000209f
+
+#define REG_A4XX_GRAS_SC_SCREEN_SCISSOR_TL                      0x0000207c
+#define A4XX_GRAS_SC_SCREEN_SCISSOR_TL_WINDOW_OFFSET_DISABLE    0x80000000
+#define A4XX_GRAS_SC_SCREEN_SCISSOR_TL_X__MASK                  0x00007fff
+#define A4XX_GRAS_SC_SCREEN_SCISSOR_TL_X__SHIFT                 0
+static inline uint32_t A4XX_GRAS_SC_SCREEN_SCISSOR_TL_X(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_SC_SCREEN_SCISSOR_TL_X__SHIFT) & A4XX_GRAS_SC_SCREEN_SCISSOR_TL_X__MASK;
+}
+#define A4XX_GRAS_SC_SCREEN_SCISSOR_TL_Y__MASK                  0x7fff0000
+#define A4XX_GRAS_SC_SCREEN_SCISSOR_TL_Y__SHIFT                 16
+static inline uint32_t A4XX_GRAS_SC_SCREEN_SCISSOR_TL_Y(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_SC_SCREEN_SCISSOR_TL_Y__SHIFT) & A4XX_GRAS_SC_SCREEN_SCISSOR_TL_Y__MASK;
+}
+
+#define REG_A4XX_GRAS_SC_SCREEN_SCISSOR_BR                      0x0000207d
+#define A4XX_GRAS_SC_SCREEN_SCISSOR_BR_WINDOW_OFFSET_DISABLE    0x80000000
+#define A4XX_GRAS_SC_SCREEN_SCISSOR_BR_X__MASK                  0x00007fff
+#define A4XX_GRAS_SC_SCREEN_SCISSOR_BR_X__SHIFT                 0
+static inline uint32_t A4XX_GRAS_SC_SCREEN_SCISSOR_BR_X(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_SC_SCREEN_SCISSOR_BR_X__SHIFT) & A4XX_GRAS_SC_SCREEN_SCISSOR_BR_X__MASK;
+}
+#define A4XX_GRAS_SC_SCREEN_SCISSOR_BR_Y__MASK                  0x7fff0000
+#define A4XX_GRAS_SC_SCREEN_SCISSOR_BR_Y__SHIFT                 16
+static inline uint32_t A4XX_GRAS_SC_SCREEN_SCISSOR_BR_Y(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_SC_SCREEN_SCISSOR_BR_Y__SHIFT) & A4XX_GRAS_SC_SCREEN_SCISSOR_BR_Y__MASK;
+}
+
+#define REG_A4XX_GRAS_SC_WINDOW_SCISSOR_BR                      0x0000209c
+#define A4XX_GRAS_SC_WINDOW_SCISSOR_BR_WINDOW_OFFSET_DISABLE    0x80000000
+#define A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X__MASK                  0x00007fff
+#define A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X__SHIFT                 0
+static inline uint32_t A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X__SHIFT) & A4XX_GRAS_SC_WINDOW_SCISSOR_BR_X__MASK;
+}
+#define A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__MASK                  0x7fff0000
+#define A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__SHIFT                 16
+static inline uint32_t A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__SHIFT) & A4XX_GRAS_SC_WINDOW_SCISSOR_BR_Y__MASK;
+}
+
+#define REG_A4XX_GRAS_SC_WINDOW_SCISSOR_TL                      0x0000209d
+#define A4XX_GRAS_SC_WINDOW_SCISSOR_TL_WINDOW_OFFSET_DISABLE    0x80000000
+#define A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X__MASK                  0x00007fff
+#define A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X__SHIFT                 0
+static inline uint32_t A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X__SHIFT) & A4XX_GRAS_SC_WINDOW_SCISSOR_TL_X__MASK;
+}
+#define A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__MASK                  0x7fff0000
+#define A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__SHIFT                 16
+static inline uint32_t A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__SHIFT) & A4XX_GRAS_SC_WINDOW_SCISSOR_TL_Y__MASK;
+}
+
+#define REG_A4XX_GRAS_DEPTH_CONTROL                             0x00002077
+#define A4XX_GRAS_DEPTH_CONTROL_FORMAT__MASK                    0x00000003
+#define A4XX_GRAS_DEPTH_CONTROL_FORMAT__SHIFT                   0
+static inline uint32_t A4XX_GRAS_DEPTH_CONTROL_FORMAT(enum a4xx_depth_format val)
+{
+        return ((val) << A4XX_GRAS_DEPTH_CONTROL_FORMAT__SHIFT) & A4XX_GRAS_DEPTH_CONTROL_FORMAT__MASK;
+}
+
+#define REG_A4XX_GRAS_SU_MODE_CONTROL                           0x00002078
+#define A4XX_GRAS_SU_MODE_CONTROL_CULL_FRONT                    0x00000001
+#define A4XX_GRAS_SU_MODE_CONTROL_CULL_BACK                     0x00000002
+#define A4XX_GRAS_SU_MODE_CONTROL_FRONT_CW                      0x00000004
+#define A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__MASK           0x000007f8
+#define A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__SHIFT          3
+static inline uint32_t A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH(float val)
+{
+        return ((((int32_t)(val * 4.0))) << A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__SHIFT) & A4XX_GRAS_SU_MODE_CONTROL_LINEHALFWIDTH__MASK;
+}
+#define A4XX_GRAS_SU_MODE_CONTROL_POLY_OFFSET                   0x00000800
+#define A4XX_GRAS_SU_MODE_CONTROL_RENDERING_PASS                0x00100000
+
+#define REG_A4XX_GRAS_SC_CONTROL                                0x0000207b
+#define A4XX_GRAS_SC_CONTROL_RENDER_MODE__MASK                  0x0000000c
+#define A4XX_GRAS_SC_CONTROL_RENDER_MODE__SHIFT                 2
+static inline uint32_t A4XX_GRAS_SC_CONTROL_RENDER_MODE(enum a3xx_render_mode val)
+{
+        return ((val) << A4XX_GRAS_SC_CONTROL_RENDER_MODE__SHIFT) & A4XX_GRAS_SC_CONTROL_RENDER_MODE__MASK;
+}
+#define A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES__MASK                 0x00000380
+#define A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES__SHIFT                7
+static inline uint32_t A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES__SHIFT) & A4XX_GRAS_SC_CONTROL_MSAA_SAMPLES__MASK;
+}
+#define A4XX_GRAS_SC_CONTROL_MSAA_DISABLE                       0x00000800
+#define A4XX_GRAS_SC_CONTROL_RASTER_MODE__MASK                  0x0000f000
+#define A4XX_GRAS_SC_CONTROL_RASTER_MODE__SHIFT                 12
+static inline uint32_t A4XX_GRAS_SC_CONTROL_RASTER_MODE(uint32_t val)
+{
+        return ((val) << A4XX_GRAS_SC_CONTROL_RASTER_MODE__SHIFT) & A4XX_GRAS_SC_CONTROL_RASTER_MODE__MASK;
+}
+
+#define REG_A4XX_UCHE_CACHE_MODE_CONTROL                        0x00000e80
+
+#define REG_A4XX_UCHE_TRAP_BASE_LO                              0x00000e83
+
+#define REG_A4XX_UCHE_TRAP_BASE_HI                              0x00000e84
+
+#define REG_A4XX_UCHE_CACHE_STATUS                              0x00000e88
+
+#define REG_A4XX_UCHE_INVALIDATE0                               0x00000e8a
+
+#define REG_A4XX_UCHE_INVALIDATE1                               0x00000e8b
+
+#define REG_A4XX_UCHE_CACHE_WAYS_VFD                            0x00000e8c
+
+#define REG_A4XX_UCHE_PERFCTR_UCHE_SEL_7                        0x00000e95
+
+#define REG_A4XX_HLSQ_TIMEOUT_THRESHOLD                         0x00000e00
+
+#define REG_A4XX_HLSQ_DEBUG_ECO_CONTROL                         0x00000e04
+
+#define REG_A4XX_HLSQ_PERF_PIPE_MASK                            0x00000e0e
+
+#define REG_A4XX_HLSQ_CONTROL_0_REG                             0x000023c0
+#define A4XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__MASK              0x00000010
+#define A4XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__SHIFT             4
+static inline uint32_t A4XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE(enum a3xx_threadsize val)
+{
+        return ((val) << A4XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__SHIFT) & A4XX_HLSQ_CONTROL_0_REG_FSTHREADSIZE__MASK;
+}
+#define A4XX_HLSQ_CONTROL_0_REG_FSSUPERTHREADENABLE             0x00000040
+#define A4XX_HLSQ_CONTROL_0_REG_SPSHADERRESTART                 0x00000200
+#define A4XX_HLSQ_CONTROL_0_REG_RESERVED2                       0x00000400
+#define A4XX_HLSQ_CONTROL_0_REG_CHUNKDISABLE                    0x04000000
+#define A4XX_HLSQ_CONTROL_0_REG_CONSTMODE__MASK                 0x08000000
+#define A4XX_HLSQ_CONTROL_0_REG_CONSTMODE__SHIFT                27
+static inline uint32_t A4XX_HLSQ_CONTROL_0_REG_CONSTMODE(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_CONTROL_0_REG_CONSTMODE__SHIFT) & A4XX_HLSQ_CONTROL_0_REG_CONSTMODE__MASK;
+}
+#define A4XX_HLSQ_CONTROL_0_REG_LAZYUPDATEDISABLE               0x10000000
+#define A4XX_HLSQ_CONTROL_0_REG_SPCONSTFULLUPDATE               0x20000000
+#define A4XX_HLSQ_CONTROL_0_REG_TPFULLUPDATE                    0x40000000
+#define A4XX_HLSQ_CONTROL_0_REG_SINGLECONTEXT                   0x80000000
+
+#define REG_A4XX_HLSQ_CONTROL_1_REG                             0x000023c1
+#define A4XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__MASK              0x00000040
+#define A4XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__SHIFT             6
+static inline uint32_t A4XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE(enum a3xx_threadsize val)
+{
+        return ((val) << A4XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__SHIFT) & A4XX_HLSQ_CONTROL_1_REG_VSTHREADSIZE__MASK;
+}
+#define A4XX_HLSQ_CONTROL_1_REG_VSSUPERTHREADENABLE             0x00000100
+#define A4XX_HLSQ_CONTROL_1_REG_RESERVED1                       0x00000200
+#define A4XX_HLSQ_CONTROL_1_REG_ZWCOORD                         0x02000000
+
+#define REG_A4XX_HLSQ_CONTROL_2_REG                             0x000023c2
+#define A4XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD__MASK        0xfc000000
+#define A4XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD__SHIFT       26
+static inline uint32_t A4XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD__SHIFT) & A4XX_HLSQ_CONTROL_2_REG_PRIMALLOCTHRESHOLD__MASK;
+}
+
+#define REG_A4XX_HLSQ_CONTROL_3_REG                             0x000023c3
+#define A4XX_HLSQ_CONTROL_3_REG_REGID__MASK                     0x000000ff
+#define A4XX_HLSQ_CONTROL_3_REG_REGID__SHIFT                    0
+static inline uint32_t A4XX_HLSQ_CONTROL_3_REG_REGID(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_CONTROL_3_REG_REGID__SHIFT) & A4XX_HLSQ_CONTROL_3_REG_REGID__MASK;
+}
+
+#define REG_A4XX_HLSQ_VS_CONTROL_REG                            0x000023c5
+#define A4XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__MASK              0x000000ff
+#define A4XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__SHIFT             0
+static inline uint32_t A4XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__SHIFT) & A4XX_HLSQ_VS_CONTROL_REG_CONSTLENGTH__MASK;
+}
+#define A4XX_HLSQ_VS_CONTROL_REG_CONSTOBJECTOFFSET__MASK        0x0000ff00
+#define A4XX_HLSQ_VS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT       8
+static inline uint32_t A4XX_HLSQ_VS_CONTROL_REG_CONSTOBJECTOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_VS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_HLSQ_VS_CONTROL_REG_CONSTOBJECTOFFSET__MASK;
+}
+#define A4XX_HLSQ_VS_CONTROL_REG_SHADEROBJOFFSET__MASK          0x00fe0000
+#define A4XX_HLSQ_VS_CONTROL_REG_SHADEROBJOFFSET__SHIFT         17
+static inline uint32_t A4XX_HLSQ_VS_CONTROL_REG_SHADEROBJOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_VS_CONTROL_REG_SHADEROBJOFFSET__SHIFT) & A4XX_HLSQ_VS_CONTROL_REG_SHADEROBJOFFSET__MASK;
+}
+#define A4XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH__MASK              0xff000000
+#define A4XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH__SHIFT             24
+static inline uint32_t A4XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH__SHIFT) & A4XX_HLSQ_VS_CONTROL_REG_INSTRLENGTH__MASK;
+}
+
+#define REG_A4XX_HLSQ_FS_CONTROL_REG                            0x000023c6
+#define A4XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__MASK              0x000000ff
+#define A4XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__SHIFT             0
+static inline uint32_t A4XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__SHIFT) & A4XX_HLSQ_FS_CONTROL_REG_CONSTLENGTH__MASK;
+}
+#define A4XX_HLSQ_FS_CONTROL_REG_CONSTOBJECTOFFSET__MASK        0x0000ff00
+#define A4XX_HLSQ_FS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT       8
+static inline uint32_t A4XX_HLSQ_FS_CONTROL_REG_CONSTOBJECTOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_FS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_HLSQ_FS_CONTROL_REG_CONSTOBJECTOFFSET__MASK;
+}
+#define A4XX_HLSQ_FS_CONTROL_REG_SHADEROBJOFFSET__MASK          0x00fe0000
+#define A4XX_HLSQ_FS_CONTROL_REG_SHADEROBJOFFSET__SHIFT         17
+static inline uint32_t A4XX_HLSQ_FS_CONTROL_REG_SHADEROBJOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_FS_CONTROL_REG_SHADEROBJOFFSET__SHIFT) & A4XX_HLSQ_FS_CONTROL_REG_SHADEROBJOFFSET__MASK;
+}
+#define A4XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH__MASK              0xff000000
+#define A4XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH__SHIFT             24
+static inline uint32_t A4XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH__SHIFT) & A4XX_HLSQ_FS_CONTROL_REG_INSTRLENGTH__MASK;
+}
+
+#define REG_A4XX_HLSQ_HS_CONTROL_REG                            0x000023c7
+#define A4XX_HLSQ_HS_CONTROL_REG_CONSTLENGTH__MASK              0x000000ff
+#define A4XX_HLSQ_HS_CONTROL_REG_CONSTLENGTH__SHIFT             0
+static inline uint32_t A4XX_HLSQ_HS_CONTROL_REG_CONSTLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_HS_CONTROL_REG_CONSTLENGTH__SHIFT) & A4XX_HLSQ_HS_CONTROL_REG_CONSTLENGTH__MASK;
+}
+#define A4XX_HLSQ_HS_CONTROL_REG_CONSTOBJECTOFFSET__MASK        0x0000ff00
+#define A4XX_HLSQ_HS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT       8
+static inline uint32_t A4XX_HLSQ_HS_CONTROL_REG_CONSTOBJECTOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_HS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_HLSQ_HS_CONTROL_REG_CONSTOBJECTOFFSET__MASK;
+}
+#define A4XX_HLSQ_HS_CONTROL_REG_SHADEROBJOFFSET__MASK          0x00fe0000
+#define A4XX_HLSQ_HS_CONTROL_REG_SHADEROBJOFFSET__SHIFT         17
+static inline uint32_t A4XX_HLSQ_HS_CONTROL_REG_SHADEROBJOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_HS_CONTROL_REG_SHADEROBJOFFSET__SHIFT) & A4XX_HLSQ_HS_CONTROL_REG_SHADEROBJOFFSET__MASK;
+}
+#define A4XX_HLSQ_HS_CONTROL_REG_INSTRLENGTH__MASK              0xff000000
+#define A4XX_HLSQ_HS_CONTROL_REG_INSTRLENGTH__SHIFT             24
+static inline uint32_t A4XX_HLSQ_HS_CONTROL_REG_INSTRLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_HS_CONTROL_REG_INSTRLENGTH__SHIFT) & A4XX_HLSQ_HS_CONTROL_REG_INSTRLENGTH__MASK;
+}
+
+#define REG_A4XX_HLSQ_DS_CONTROL_REG                            0x000023c8
+#define A4XX_HLSQ_DS_CONTROL_REG_CONSTLENGTH__MASK              0x000000ff
+#define A4XX_HLSQ_DS_CONTROL_REG_CONSTLENGTH__SHIFT             0
+static inline uint32_t A4XX_HLSQ_DS_CONTROL_REG_CONSTLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_DS_CONTROL_REG_CONSTLENGTH__SHIFT) & A4XX_HLSQ_DS_CONTROL_REG_CONSTLENGTH__MASK;
+}
+#define A4XX_HLSQ_DS_CONTROL_REG_CONSTOBJECTOFFSET__MASK        0x0000ff00
+#define A4XX_HLSQ_DS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT       8
+static inline uint32_t A4XX_HLSQ_DS_CONTROL_REG_CONSTOBJECTOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_DS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_HLSQ_DS_CONTROL_REG_CONSTOBJECTOFFSET__MASK;
+}
+#define A4XX_HLSQ_DS_CONTROL_REG_SHADEROBJOFFSET__MASK          0x00fe0000
+#define A4XX_HLSQ_DS_CONTROL_REG_SHADEROBJOFFSET__SHIFT         17
+static inline uint32_t A4XX_HLSQ_DS_CONTROL_REG_SHADEROBJOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_DS_CONTROL_REG_SHADEROBJOFFSET__SHIFT) & A4XX_HLSQ_DS_CONTROL_REG_SHADEROBJOFFSET__MASK;
+}
+#define A4XX_HLSQ_DS_CONTROL_REG_INSTRLENGTH__MASK              0xff000000
+#define A4XX_HLSQ_DS_CONTROL_REG_INSTRLENGTH__SHIFT             24
+static inline uint32_t A4XX_HLSQ_DS_CONTROL_REG_INSTRLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_DS_CONTROL_REG_INSTRLENGTH__SHIFT) & A4XX_HLSQ_DS_CONTROL_REG_INSTRLENGTH__MASK;
+}
+
+#define REG_A4XX_HLSQ_GS_CONTROL_REG                            0x000023c9
+#define A4XX_HLSQ_GS_CONTROL_REG_CONSTLENGTH__MASK              0x000000ff
+#define A4XX_HLSQ_GS_CONTROL_REG_CONSTLENGTH__SHIFT             0
+static inline uint32_t A4XX_HLSQ_GS_CONTROL_REG_CONSTLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_GS_CONTROL_REG_CONSTLENGTH__SHIFT) & A4XX_HLSQ_GS_CONTROL_REG_CONSTLENGTH__MASK;
+}
+#define A4XX_HLSQ_GS_CONTROL_REG_CONSTOBJECTOFFSET__MASK        0x0000ff00
+#define A4XX_HLSQ_GS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT       8
+static inline uint32_t A4XX_HLSQ_GS_CONTROL_REG_CONSTOBJECTOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_GS_CONTROL_REG_CONSTOBJECTOFFSET__SHIFT) & A4XX_HLSQ_GS_CONTROL_REG_CONSTOBJECTOFFSET__MASK;
+}
+#define A4XX_HLSQ_GS_CONTROL_REG_SHADEROBJOFFSET__MASK          0x00fe0000
+#define A4XX_HLSQ_GS_CONTROL_REG_SHADEROBJOFFSET__SHIFT         17
+static inline uint32_t A4XX_HLSQ_GS_CONTROL_REG_SHADEROBJOFFSET(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_GS_CONTROL_REG_SHADEROBJOFFSET__SHIFT) & A4XX_HLSQ_GS_CONTROL_REG_SHADEROBJOFFSET__MASK;
+}
+#define A4XX_HLSQ_GS_CONTROL_REG_INSTRLENGTH__MASK              0xff000000
+#define A4XX_HLSQ_GS_CONTROL_REG_INSTRLENGTH__SHIFT             24
+static inline uint32_t A4XX_HLSQ_GS_CONTROL_REG_INSTRLENGTH(uint32_t val)
+{
+        return ((val) << A4XX_HLSQ_GS_CONTROL_REG_INSTRLENGTH__SHIFT) & A4XX_HLSQ_GS_CONTROL_REG_INSTRLENGTH__MASK;
+}
+
+#define REG_A4XX_HLSQ_UPDATE_CONTROL                            0x000023db
+
+#define REG_A4XX_PC_BINNING_COMMAND                             0x00000d00
+#define A4XX_PC_BINNING_COMMAND_BINNING_ENABLE                  0x00000001
+
+#define REG_A4XX_PC_DRAWCALL_SETUP_OVERRIDE                     0x00000d0c
+
+#define REG_A4XX_PC_PERFCTR_PC_SEL_0                            0x00000d10
+
+#define REG_A4XX_PC_PERFCTR_PC_SEL_7                            0x00000d17
+
+#define REG_A4XX_PC_BIN_BASE                                    0x000021c0
+
+#define REG_A4XX_PC_PRIM_VTX_CNTL                               0x000021c4
+#define A4XX_PC_PRIM_VTX_CNTL_VAROUT                            0x00000001
+#define A4XX_PC_PRIM_VTX_CNTL_PROVOKING_VTX_LAST                0x02000000
+#define A4XX_PC_PRIM_VTX_CNTL_PSIZE                             0x04000000
+
+#define REG_A4XX_UNKNOWN_21C5                                   0x000021c5
+
+#define REG_A4XX_PC_RESTART_INDEX                               0x000021c6
+
+#define REG_A4XX_PC_GS_PARAM                                    0x000021e5
+
+#define REG_A4XX_PC_HS_PARAM                                    0x000021e7
+
+#define REG_A4XX_VBIF_VERSION                                   0x00003000
+
+#define REG_A4XX_VBIF_CLKON                                     0x00003001
+#define A4XX_VBIF_CLKON_FORCE_ON_TESTBUS                        0x00000001
+
+#define REG_A4XX_VBIF_ABIT_SORT                                 0x0000301c
+
+#define REG_A4XX_VBIF_ABIT_SORT_CONF                            0x0000301d
+
+#define REG_A4XX_VBIF_GATE_OFF_WRREQ_EN                         0x0000302a
+
+#define REG_A4XX_VBIF_IN_RD_LIM_CONF0                           0x0000302c
+
+#define REG_A4XX_VBIF_IN_RD_LIM_CONF1                           0x0000302d
+
+#define REG_A4XX_VBIF_IN_WR_LIM_CONF0                           0x00003030
+
+#define REG_A4XX_VBIF_IN_WR_LIM_CONF1                           0x00003031
+
+#define REG_A4XX_VBIF_ROUND_ROBIN_QOS_ARB                       0x00003049
+
+#define REG_A4XX_UNKNOWN_0CC5                                   0x00000cc5
+
+#define REG_A4XX_UNKNOWN_0CC6                                   0x00000cc6
+
+#define REG_A4XX_UNKNOWN_0D01                                   0x00000d01
+
+#define REG_A4XX_UNKNOWN_0E05                                   0x00000e05
+
+#define REG_A4XX_UNKNOWN_0E42                                   0x00000e42
+
+#define REG_A4XX_UNKNOWN_0EC2                                   0x00000ec2
+
+#define REG_A4XX_UNKNOWN_0EC3                                   0x00000ec3
+
+#define REG_A4XX_UNKNOWN_0F03                                   0x00000f03
+
+#define REG_A4XX_UNKNOWN_2001                                   0x00002001
+
+#define REG_A4XX_UNKNOWN_209B                                   0x0000209b
+
+#define REG_A4XX_UNKNOWN_20EF                                   0x000020ef
+
+#define REG_A4XX_UNKNOWN_20F0                                   0x000020f0
+
+#define REG_A4XX_UNKNOWN_20F1                                   0x000020f1
+
+#define REG_A4XX_UNKNOWN_20F2                                   0x000020f2
+
+#define REG_A4XX_UNKNOWN_20F3                                   0x000020f3
+
+#define REG_A4XX_UNKNOWN_20F4                                   0x000020f4
+
+#define REG_A4XX_UNKNOWN_20F5                                   0x000020f5
+
+#define REG_A4XX_UNKNOWN_20F6                                   0x000020f6
+
+#define REG_A4XX_UNKNOWN_20F7                                   0x000020f7
+
+#define REG_A4XX_UNKNOWN_2152                                   0x00002152
+
+#define REG_A4XX_UNKNOWN_2153                                   0x00002153
+
+#define REG_A4XX_UNKNOWN_2154                                   0x00002154
+
+#define REG_A4XX_UNKNOWN_2155                                   0x00002155
+
+#define REG_A4XX_UNKNOWN_2156                                   0x00002156
+
+#define REG_A4XX_UNKNOWN_2157                                   0x00002157
+
+#define REG_A4XX_UNKNOWN_21C3                                   0x000021c3
+
+#define REG_A4XX_UNKNOWN_21E6                                   0x000021e6
+
+#define REG_A4XX_UNKNOWN_2209                                   0x00002209
+
+#define REG_A4XX_UNKNOWN_22D7                                   0x000022d7
+
+#define REG_A4XX_UNKNOWN_2381                                   0x00002381
+
+#define REG_A4XX_UNKNOWN_23A0                                   0x000023a0
+
+#define REG_A4XX_TEX_SAMP_0                                     0x00000000
+#define A4XX_TEX_SAMP_0_XY_MAG__MASK                            0x00000006
+#define A4XX_TEX_SAMP_0_XY_MAG__SHIFT                           1
+static inline uint32_t A4XX_TEX_SAMP_0_XY_MAG(enum a4xx_tex_filter val)
+{
+        return ((val) << A4XX_TEX_SAMP_0_XY_MAG__SHIFT) & A4XX_TEX_SAMP_0_XY_MAG__MASK;
+}
+#define A4XX_TEX_SAMP_0_XY_MIN__MASK                            0x00000018
+#define A4XX_TEX_SAMP_0_XY_MIN__SHIFT                           3
+static inline uint32_t A4XX_TEX_SAMP_0_XY_MIN(enum a4xx_tex_filter val)
+{
+        return ((val) << A4XX_TEX_SAMP_0_XY_MIN__SHIFT) & A4XX_TEX_SAMP_0_XY_MIN__MASK;
+}
+#define A4XX_TEX_SAMP_0_WRAP_S__MASK                            0x000000e0
+#define A4XX_TEX_SAMP_0_WRAP_S__SHIFT                           5
+static inline uint32_t A4XX_TEX_SAMP_0_WRAP_S(enum a4xx_tex_clamp val)
+{
+        return ((val) << A4XX_TEX_SAMP_0_WRAP_S__SHIFT) & A4XX_TEX_SAMP_0_WRAP_S__MASK;
+}
+#define A4XX_TEX_SAMP_0_WRAP_T__MASK                            0x00000700
+#define A4XX_TEX_SAMP_0_WRAP_T__SHIFT                           8
+static inline uint32_t A4XX_TEX_SAMP_0_WRAP_T(enum a4xx_tex_clamp val)
+{
+        return ((val) << A4XX_TEX_SAMP_0_WRAP_T__SHIFT) & A4XX_TEX_SAMP_0_WRAP_T__MASK;
+}
+#define A4XX_TEX_SAMP_0_WRAP_R__MASK                            0x00003800
+#define A4XX_TEX_SAMP_0_WRAP_R__SHIFT                           11
+static inline uint32_t A4XX_TEX_SAMP_0_WRAP_R(enum a4xx_tex_clamp val)
+{
+        return ((val) << A4XX_TEX_SAMP_0_WRAP_R__SHIFT) & A4XX_TEX_SAMP_0_WRAP_R__MASK;
+}
+
+#define REG_A4XX_TEX_SAMP_1                                     0x00000001
+#define A4XX_TEX_SAMP_1_COMPARE_FUNC__MASK                      0x0000000e
+#define A4XX_TEX_SAMP_1_COMPARE_FUNC__SHIFT                     1
+static inline uint32_t A4XX_TEX_SAMP_1_COMPARE_FUNC(enum adreno_compare_func val)
+{
+        return ((val) << A4XX_TEX_SAMP_1_COMPARE_FUNC__SHIFT) & A4XX_TEX_SAMP_1_COMPARE_FUNC__MASK;
+}
+#define A4XX_TEX_SAMP_1_MAX_LOD__MASK                           0x000fff00
+#define A4XX_TEX_SAMP_1_MAX_LOD__SHIFT                          8
+static inline uint32_t A4XX_TEX_SAMP_1_MAX_LOD(float val)
+{
+        return ((((uint32_t)(val * 64.0))) << A4XX_TEX_SAMP_1_MAX_LOD__SHIFT) & A4XX_TEX_SAMP_1_MAX_LOD__MASK;
+}
+#define A4XX_TEX_SAMP_1_MIN_LOD__MASK                           0xfff00000
+#define A4XX_TEX_SAMP_1_MIN_LOD__SHIFT                          20
+static inline uint32_t A4XX_TEX_SAMP_1_MIN_LOD(float val)
+{
+        return ((((uint32_t)(val * 64.0))) << A4XX_TEX_SAMP_1_MIN_LOD__SHIFT) & A4XX_TEX_SAMP_1_MIN_LOD__MASK;
+}
+
+#define REG_A4XX_TEX_CONST_0                                    0x00000000
+#define A4XX_TEX_CONST_0_TILED                                  0x00000001
+#define A4XX_TEX_CONST_0_SWIZ_X__MASK                           0x00000070
+#define A4XX_TEX_CONST_0_SWIZ_X__SHIFT                          4
+static inline uint32_t A4XX_TEX_CONST_0_SWIZ_X(enum a4xx_tex_swiz val)
+{
+        return ((val) << A4XX_TEX_CONST_0_SWIZ_X__SHIFT) & A4XX_TEX_CONST_0_SWIZ_X__MASK;
+}
+#define A4XX_TEX_CONST_0_SWIZ_Y__MASK                           0x00000380
+#define A4XX_TEX_CONST_0_SWIZ_Y__SHIFT                          7
+static inline uint32_t A4XX_TEX_CONST_0_SWIZ_Y(enum a4xx_tex_swiz val)
+{
+        return ((val) << A4XX_TEX_CONST_0_SWIZ_Y__SHIFT) & A4XX_TEX_CONST_0_SWIZ_Y__MASK;
+}
+#define A4XX_TEX_CONST_0_SWIZ_Z__MASK                           0x00001c00
+#define A4XX_TEX_CONST_0_SWIZ_Z__SHIFT                          10
+static inline uint32_t A4XX_TEX_CONST_0_SWIZ_Z(enum a4xx_tex_swiz val)
+{
+        return ((val) << A4XX_TEX_CONST_0_SWIZ_Z__SHIFT) & A4XX_TEX_CONST_0_SWIZ_Z__MASK;
+}
+#define A4XX_TEX_CONST_0_SWIZ_W__MASK                           0x0000e000
+#define A4XX_TEX_CONST_0_SWIZ_W__SHIFT                          13
+static inline uint32_t A4XX_TEX_CONST_0_SWIZ_W(enum a4xx_tex_swiz val)
+{
+        return ((val) << A4XX_TEX_CONST_0_SWIZ_W__SHIFT) & A4XX_TEX_CONST_0_SWIZ_W__MASK;
+}
+#define A4XX_TEX_CONST_0_FMT__MASK                              0x1fc00000
+#define A4XX_TEX_CONST_0_FMT__SHIFT                             22
+static inline uint32_t A4XX_TEX_CONST_0_FMT(enum a4xx_tex_fmt val)
+{
+        return ((val) << A4XX_TEX_CONST_0_FMT__SHIFT) & A4XX_TEX_CONST_0_FMT__MASK;
+}
+#define A4XX_TEX_CONST_0_TYPE__MASK                             0x60000000
+#define A4XX_TEX_CONST_0_TYPE__SHIFT                            29
+static inline uint32_t A4XX_TEX_CONST_0_TYPE(enum a4xx_tex_type val)
+{
+        return ((val) << A4XX_TEX_CONST_0_TYPE__SHIFT) & A4XX_TEX_CONST_0_TYPE__MASK;
+}
+
+#define REG_A4XX_TEX_CONST_1                                    0x00000001
+#define A4XX_TEX_CONST_1_HEIGHT__MASK                           0x00007fff
+#define A4XX_TEX_CONST_1_HEIGHT__SHIFT                          0
+static inline uint32_t A4XX_TEX_CONST_1_HEIGHT(uint32_t val)
+{
+        return ((val) << A4XX_TEX_CONST_1_HEIGHT__SHIFT) & A4XX_TEX_CONST_1_HEIGHT__MASK;
+}
+#define A4XX_TEX_CONST_1_WIDTH__MASK                            0x1fff8000
+#define A4XX_TEX_CONST_1_WIDTH__SHIFT                           15
+static inline uint32_t A4XX_TEX_CONST_1_WIDTH(uint32_t val)
+{
+        return ((val) << A4XX_TEX_CONST_1_WIDTH__SHIFT) & A4XX_TEX_CONST_1_WIDTH__MASK;
+}
+
+#define REG_A4XX_TEX_CONST_2                                    0x00000002
+#define A4XX_TEX_CONST_2_PITCH__MASK                            0x3ffffe00
+#define A4XX_TEX_CONST_2_PITCH__SHIFT                           9
+static inline uint32_t A4XX_TEX_CONST_2_PITCH(uint32_t val)
+{
+        return ((val) << A4XX_TEX_CONST_2_PITCH__SHIFT) & A4XX_TEX_CONST_2_PITCH__MASK;
+}
+#define A4XX_TEX_CONST_2_SWAP__MASK                             0xc0000000
+#define A4XX_TEX_CONST_2_SWAP__SHIFT                            30
+static inline uint32_t A4XX_TEX_CONST_2_SWAP(enum a3xx_color_swap val)
+{
+        return ((val) << A4XX_TEX_CONST_2_SWAP__SHIFT) & A4XX_TEX_CONST_2_SWAP__MASK;
+}
+
+#define REG_A4XX_TEX_CONST_3                                    0x00000003
+#define A4XX_TEX_CONST_3_LAYERSZ__MASK                          0x0000000f
+#define A4XX_TEX_CONST_3_LAYERSZ__SHIFT                         0
+static inline uint32_t A4XX_TEX_CONST_3_LAYERSZ(uint32_t val)
+{
+        return ((val >> 12) << A4XX_TEX_CONST_3_LAYERSZ__SHIFT) & A4XX_TEX_CONST_3_LAYERSZ__MASK;
+}
+
+#define REG_A4XX_TEX_CONST_4                                    0x00000004
+#define A4XX_TEX_CONST_4_BASE__MASK                             0xffffffff
+#define A4XX_TEX_CONST_4_BASE__SHIFT                            0
+static inline uint32_t A4XX_TEX_CONST_4_BASE(uint32_t val)
+{
+        return ((val) << A4XX_TEX_CONST_4_BASE__SHIFT) & A4XX_TEX_CONST_4_BASE__MASK;
+}
+
+#define REG_A4XX_TEX_CONST_5                                    0x00000005
+
+#define REG_A4XX_TEX_CONST_6                                    0x00000006
+
+#define REG_A4XX_TEX_CONST_7                                    0x00000007
+
+
+#endif /* A4XX_XML */
diff --git a/drivers/gpu/drm/msm/adreno/a4xx_gpu.c b/drivers/gpu/drm/msm/adreno/a4xx_gpu.c
new file mode 100644
index 000000000000..91221836c5ad
--- /dev/null
+++ b/drivers/gpu/drm/msm/adreno/a4xx_gpu.c
@@ -0,0 +1,604 @@
+/* Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ */
+#include "a4xx_gpu.h"
+#ifdef CONFIG_MSM_OCMEM
+#  include <soc/qcom/ocmem.h>
+#endif
+
+#define A4XX_INT0_MASK \
+        (A4XX_INT0_RBBM_AHB_ERROR |        \
+         A4XX_INT0_RBBM_ATB_BUS_OVERFLOW | \
+         A4XX_INT0_CP_T0_PACKET_IN_IB |    \
+         A4XX_INT0_CP_OPCODE_ERROR |       \
+         A4XX_INT0_CP_RESERVED_BIT_ERROR | \
+         A4XX_INT0_CP_HW_FAULT |           \
+         A4XX_INT0_CP_IB1_INT |            \
+         A4XX_INT0_CP_IB2_INT |            \
+         A4XX_INT0_CP_RB_INT |             \
+         A4XX_INT0_CP_REG_PROTECT_FAULT |  \
+         A4XX_INT0_CP_AHB_ERROR_HALT |     \
+         A4XX_INT0_UCHE_OOB_ACCESS)
+
+extern bool hang_debug;
+static void a4xx_dump(struct msm_gpu *gpu);
+
+/*
+ * a4xx_enable_hwcg() - Program the clock control registers
+ * @device: The adreno device pointer
+ */
+static void a4xx_enable_hwcg(struct msm_gpu *gpu)
+{
+        struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
+        unsigned int i;
+        for (i = 0; i < 4; i++)
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_TP(i), 0x02222202);
+        for (i = 0; i < 4; i++)
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_TP(i), 0x00002222);
+        for (i = 0; i < 4; i++)
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_TP(i), 0x0E739CE7);
+        for (i = 0; i < 4; i++)
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_TP(i), 0x00111111);
+        for (i = 0; i < 4; i++)
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_SP(i), 0x22222222);
+        for (i = 0; i < 4; i++)
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_SP(i), 0x00222222);
+        for (i = 0; i < 4; i++)
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_SP(i), 0x00000104);
+        for (i = 0; i < 4; i++)
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_SP(i), 0x00000081);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_UCHE, 0x22222222);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_UCHE, 0x02222222);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL3_UCHE, 0x00000000);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL4_UCHE, 0x00000000);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_UCHE, 0x00004444);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_UCHE, 0x00001112);
+        for (i = 0; i < 4; i++)
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_RB(i), 0x22222222);
+
+        /* Disable L1 clocking in A420 due to CCU issues with it */
+        for (i = 0; i < 4; i++) {
+                if (adreno_is_a420(adreno_gpu)) {
+                        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_RB(i),
+                                        0x00002020);
+                } else {
+                        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_RB(i),
+                                        0x00022020);
+                }
+        }
+
+        for (i = 0; i < 4; i++) {
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_MARB_CCU(i),
+                                0x00000922);
+        }
+
+        for (i = 0; i < 4; i++) {
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_RB_MARB_CCU(i),
+                                0x00000000);
+        }
+
+        for (i = 0; i < 4; i++) {
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_RB_MARB_CCU_L1(i),
+                                0x00000001);
+        }
+
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_MODE_GPC, 0x02222222);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_GPC, 0x04100104);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_GPC, 0x00022222);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_COM_DCOM, 0x00000022);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_COM_DCOM, 0x0000010F);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_COM_DCOM, 0x00000022);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_TSE_RAS_RBBM, 0x00222222);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_TSE_RAS_RBBM, 0x00004104);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_TSE_RAS_RBBM, 0x00000222);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_HLSQ , 0x00000000);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_HLSQ, 0x00000000);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_HLSQ, 0x00020000);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL, 0xAAAAAAAA);
+        gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2, 0);
+}
+
+static void a4xx_me_init(struct msm_gpu *gpu)
+{
+        struct msm_ringbuffer *ring = gpu->rb;
+
+        OUT_PKT3(ring, CP_ME_INIT, 17);
+        OUT_RING(ring, 0x000003f7);
+        OUT_RING(ring, 0x00000000);
+        OUT_RING(ring, 0x00000000);
+        OUT_RING(ring, 0x00000000);
+        OUT_RING(ring, 0x00000080);
+        OUT_RING(ring, 0x00000100);
+        OUT_RING(ring, 0x00000180);
+        OUT_RING(ring, 0x00006600);
+        OUT_RING(ring, 0x00000150);
+        OUT_RING(ring, 0x0000014e);
+        OUT_RING(ring, 0x00000154);
+        OUT_RING(ring, 0x00000001);
+        OUT_RING(ring, 0x00000000);
+        OUT_RING(ring, 0x00000000);
+        OUT_RING(ring, 0x00000000);
+        OUT_RING(ring, 0x00000000);
+        OUT_RING(ring, 0x00000000);
+
+        gpu->funcs->flush(gpu);
+        gpu->funcs->idle(gpu);
+}
+
+static int a4xx_hw_init(struct msm_gpu *gpu)
+{
+        struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
+        struct a4xx_gpu *a4xx_gpu = to_a4xx_gpu(adreno_gpu);
+        uint32_t *ptr, len;
+        int i, ret;
+
+        if (adreno_is_a4xx(adreno_gpu)) {
+                gpu_write(gpu, REG_A4XX_VBIF_ABIT_SORT, 0x0001001F);
+                gpu_write(gpu, REG_A4XX_VBIF_ABIT_SORT_CONF, 0x000000A4);
+                gpu_write(gpu, REG_A4XX_VBIF_GATE_OFF_WRREQ_EN, 0x00000001);
+                gpu_write(gpu, REG_A4XX_VBIF_IN_RD_LIM_CONF0, 0x18181818);
+                gpu_write(gpu, REG_A4XX_VBIF_IN_RD_LIM_CONF1, 0x00000018);
+                gpu_write(gpu, REG_A4XX_VBIF_IN_WR_LIM_CONF0, 0x18181818);
+                gpu_write(gpu, REG_A4XX_VBIF_IN_WR_LIM_CONF1, 0x00000018);
+                gpu_write(gpu, REG_A4XX_VBIF_ROUND_ROBIN_QOS_ARB, 0x00000003);
+        } else {
+                BUG();
+        }
+
+        /* Make all blocks contribute to the GPU BUSY perf counter */
+        gpu_write(gpu, REG_A4XX_RBBM_GPU_BUSY_MASKED, 0xffffffff);
+
+        /* Tune the hystersis counters for SP and CP idle detection */
+        gpu_write(gpu, REG_A4XX_RBBM_SP_HYST_CNT, 0x10);
+        gpu_write(gpu, REG_A4XX_RBBM_WAIT_IDLE_CLOCKS_CTL, 0x10);
+
+         /* Enable the RBBM error reporting bits */
+        gpu_write(gpu, REG_A4XX_RBBM_AHB_CTL0, 0x00000001);
+
+        /* Enable AHB error reporting*/
+        gpu_write(gpu, REG_A4XX_RBBM_AHB_CTL1, 0xa6ffffff);
+
+        /* Enable power counters*/
+        gpu_write(gpu, REG_A4XX_RBBM_RBBM_CTL, 0x00000030);
+
+        /*
+         * Turn on hang detection - this spews a lot of useful information
+         * into the RBBM registers on a hang:
+         */
+        gpu_write(gpu, REG_A4XX_RBBM_INTERFACE_HANG_INT_CTL,
+                        (1 << 30) | 0xFFFF);
+
+        gpu_write(gpu, REG_A4XX_RB_GMEM_BASE_ADDR,
+                        (unsigned int)(a4xx_gpu->ocmem_base >> 14));
+
+        /* Turn on performance counters: */
+        gpu_write(gpu, REG_A4XX_RBBM_PERFCTR_CTL, 0x01);
+
+        /* Disable L2 bypass to avoid UCHE out of bounds errors */
+        gpu_write(gpu, REG_A4XX_UCHE_TRAP_BASE_LO, 0xffff0000);
+        gpu_write(gpu, REG_A4XX_UCHE_TRAP_BASE_HI, 0xffff0000);
+
+        gpu_write(gpu, REG_A4XX_CP_DEBUG, (1 << 25) |
+                        (adreno_is_a420(adreno_gpu) ? (1 << 29) : 0));
+
+        a4xx_enable_hwcg(gpu);
+
+        /*
+         * For A420 set RBBM_CLOCK_DELAY_HLSQ.CGC_HLSQ_TP_EARLY_CYC >= 2
+         * due to timing issue with HLSQ_TP_CLK_EN
+         */
+        if (adreno_is_a420(adreno_gpu)) {
+                unsigned int val;
+                val = gpu_read(gpu, REG_A4XX_RBBM_CLOCK_DELAY_HLSQ);
+                val &= ~A4XX_CGC_HLSQ_EARLY_CYC__MASK;
+                val |= 2 << A4XX_CGC_HLSQ_EARLY_CYC__SHIFT;
+                gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_HLSQ, val);
+        }
+
+        ret = adreno_hw_init(gpu);
+        if (ret)
+                return ret;
+
+        /* setup access protection: */
+        gpu_write(gpu, REG_A4XX_CP_PROTECT_CTRL, 0x00000007);
+
+        /* RBBM registers */
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(0), 0x62000010);
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(1), 0x63000020);
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(2), 0x64000040);
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(3), 0x65000080);
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(4), 0x66000100);
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(5), 0x64000200);
+
+        /* CP registers */
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(6), 0x67000800);
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(7), 0x64001600);
+
+
+        /* RB registers */
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(8), 0x60003300);
+
+        /* HLSQ registers */
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(9), 0x60003800);
+
+        /* VPC registers */
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(10), 0x61003980);
+
+        /* SMMU registers */
+        gpu_write(gpu, REG_A4XX_CP_PROTECT(11), 0x6e010000);
+
+        gpu_write(gpu, REG_A4XX_RBBM_INT_0_MASK, A4XX_INT0_MASK);
+
+        ret = adreno_hw_init(gpu);
+        if (ret)
+                return ret;
+
+        /* Load PM4: */
+        ptr = (uint32_t *)(adreno_gpu->pm4->data);
+        len = adreno_gpu->pm4->size / 4;
+        DBG("loading PM4 ucode version: %u", ptr[0]);
+        gpu_write(gpu, REG_A4XX_CP_ME_RAM_WADDR, 0);
+        for (i = 1; i < len; i++)
+                gpu_write(gpu, REG_A4XX_CP_ME_RAM_DATA, ptr[i]);
+
+        /* Load PFP: */
+        ptr = (uint32_t *)(adreno_gpu->pfp->data);
+        len = adreno_gpu->pfp->size / 4;
+        DBG("loading PFP ucode version: %u", ptr[0]);
+
+        gpu_write(gpu, REG_A4XX_CP_PFP_UCODE_ADDR, 0);
+        for (i = 1; i < len; i++)
+                gpu_write(gpu, REG_A4XX_CP_PFP_UCODE_DATA, ptr[i]);
+
+        /* clear ME_HALT to start micro engine */
+        gpu_write(gpu, REG_A4XX_CP_ME_CNTL, 0);
+
+        a4xx_me_init(gpu);
+        return 0;
+}
+
+static void a4xx_recover(struct msm_gpu *gpu)
+{
+        /* dump registers before resetting gpu, if enabled: */
+        if (hang_debug)
+                a4xx_dump(gpu);
+
+        gpu_write(gpu, REG_A4XX_RBBM_SW_RESET_CMD, 1);
+        gpu_read(gpu, REG_A4XX_RBBM_SW_RESET_CMD);
+        gpu_write(gpu, REG_A4XX_RBBM_SW_RESET_CMD, 0);
+        adreno_recover(gpu);
+}
+
+static void a4xx_destroy(struct msm_gpu *gpu)
+{
+        struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
+        struct a4xx_gpu *a4xx_gpu = to_a4xx_gpu(adreno_gpu);
+
+        DBG("%s", gpu->name);
+
+        adreno_gpu_cleanup(adreno_gpu);
+
+#ifdef CONFIG_MSM_OCMEM
+        if (a4xx_gpu->ocmem_base)
+                ocmem_free(OCMEM_GRAPHICS, a4xx_gpu->ocmem_hdl);
+#endif
+
+        kfree(a4xx_gpu);
+}
+
+static void a4xx_idle(struct msm_gpu *gpu)
+{
+        /* wait for ringbuffer to drain: */
+        adreno_idle(gpu);
+
+        /* then wait for GPU to finish: */
+        if (spin_until(!(gpu_read(gpu, REG_A4XX_RBBM_STATUS) &
+                                        A4XX_RBBM_STATUS_GPU_BUSY)))
+                DRM_ERROR("%s: timeout waiting for GPU to idle!\n", gpu->name);
+
+        /* TODO maybe we need to reset GPU here to recover from hang? */
+}
+
+static irqreturn_t a4xx_irq(struct msm_gpu *gpu)
+{
+        uint32_t status;
+
+        status = gpu_read(gpu, REG_A4XX_RBBM_INT_0_STATUS);
+        DBG("%s: Int status %08x", gpu->name, status);
+
+        gpu_write(gpu, REG_A4XX_RBBM_INT_CLEAR_CMD, status);
+
+        msm_gpu_retire(gpu);
+
+        return IRQ_HANDLED;
+}
+
+static const unsigned int a4xx_registers[] = {
+        /* RBBM */
+        0x0000, 0x0002, 0x0004, 0x0021, 0x0023, 0x0024, 0x0026, 0x0026,
+        0x0028, 0x002B, 0x002E, 0x0034, 0x0037, 0x0044, 0x0047, 0x0066,
+        0x0068, 0x0095, 0x009C, 0x0170, 0x0174, 0x01AF,
+        /* CP */
+        0x0200, 0x0233, 0x0240, 0x0250, 0x04C0, 0x04DD, 0x0500, 0x050B,
+        0x0578, 0x058F,
+        /* VSC */
+        0x0C00, 0x0C03, 0x0C08, 0x0C41, 0x0C50, 0x0C51,
+        /* GRAS */
+        0x0C80, 0x0C81, 0x0C88, 0x0C8F,
+        /* RB */
+        0x0CC0, 0x0CC0, 0x0CC4, 0x0CD2,
+        /* PC */
+        0x0D00, 0x0D0C, 0x0D10, 0x0D17, 0x0D20, 0x0D23,
+        /* VFD */
+        0x0E40, 0x0E4A,
+        /* VPC */
+        0x0E60, 0x0E61, 0x0E63, 0x0E68,
+        /* UCHE */
+        0x0E80, 0x0E84, 0x0E88, 0x0E95,
+        /* VMIDMT */
+        0x1000, 0x1000, 0x1002, 0x1002, 0x1004, 0x1004, 0x1008, 0x100A,
+        0x100C, 0x100D, 0x100F, 0x1010, 0x1012, 0x1016, 0x1024, 0x1024,
+        0x1027, 0x1027, 0x1100, 0x1100, 0x1102, 0x1102, 0x1104, 0x1104,
+        0x1110, 0x1110, 0x1112, 0x1116, 0x1124, 0x1124, 0x1300, 0x1300,
+        0x1380, 0x1380,
+        /* GRAS CTX 0 */
+        0x2000, 0x2004, 0x2008, 0x2067, 0x2070, 0x2078, 0x207B, 0x216E,
+        /* PC CTX 0 */
+        0x21C0, 0x21C6, 0x21D0, 0x21D0, 0x21D9, 0x21D9, 0x21E5, 0x21E7,
+        /* VFD CTX 0 */
+        0x2200, 0x2204, 0x2208, 0x22A9,
+        /* GRAS CTX 1 */
+        0x2400, 0x2404, 0x2408, 0x2467, 0x2470, 0x2478, 0x247B, 0x256E,
+        /* PC CTX 1 */
+        0x25C0, 0x25C6, 0x25D0, 0x25D0, 0x25D9, 0x25D9, 0x25E5, 0x25E7,
+        /* VFD CTX 1 */
+        0x2600, 0x2604, 0x2608, 0x26A9,
+        /* XPU */
+        0x2C00, 0x2C01, 0x2C10, 0x2C10, 0x2C12, 0x2C16, 0x2C1D, 0x2C20,
+        0x2C28, 0x2C28, 0x2C30, 0x2C30, 0x2C32, 0x2C36, 0x2C40, 0x2C40,
+        0x2C50, 0x2C50, 0x2C52, 0x2C56, 0x2C80, 0x2C80, 0x2C94, 0x2C95,
+        /* VBIF */
+        0x3000, 0x3007, 0x300C, 0x3014, 0x3018, 0x301D, 0x3020, 0x3022,
+        0x3024, 0x3026, 0x3028, 0x302A, 0x302C, 0x302D, 0x3030, 0x3031,
+        0x3034, 0x3036, 0x3038, 0x3038, 0x303C, 0x303D, 0x3040, 0x3040,
+        0x3049, 0x3049, 0x3058, 0x3058, 0x305B, 0x3061, 0x3064, 0x3068,
+        0x306C, 0x306D, 0x3080, 0x3088, 0x308B, 0x308C, 0x3090, 0x3094,
+        0x3098, 0x3098, 0x309C, 0x309C, 0x30C0, 0x30C0, 0x30C8, 0x30C8,
+        0x30D0, 0x30D0, 0x30D8, 0x30D8, 0x30E0, 0x30E0, 0x3100, 0x3100,
+        0x3108, 0x3108, 0x3110, 0x3110, 0x3118, 0x3118, 0x3120, 0x3120,
+        0x3124, 0x3125, 0x3129, 0x3129, 0x3131, 0x3131, 0x330C, 0x330C,
+        0x3310, 0x3310, 0x3400, 0x3401, 0x3410, 0x3410, 0x3412, 0x3416,
+        0x341D, 0x3420, 0x3428, 0x3428, 0x3430, 0x3430, 0x3432, 0x3436,
+        0x3440, 0x3440, 0x3450, 0x3450, 0x3452, 0x3456, 0x3480, 0x3480,
+        0x3494, 0x3495, 0x4000, 0x4000, 0x4002, 0x4002, 0x4004, 0x4004,
+        0x4008, 0x400A, 0x400C, 0x400D, 0x400F, 0x4012, 0x4014, 0x4016,
+        0x401D, 0x401D, 0x4020, 0x4027, 0x4060, 0x4062, 0x4200, 0x4200,
+        0x4300, 0x4300, 0x4400, 0x4400, 0x4500, 0x4500, 0x4800, 0x4802,
+        0x480F, 0x480F, 0x4811, 0x4811, 0x4813, 0x4813, 0x4815, 0x4816,
+        0x482B, 0x482B, 0x4857, 0x4857, 0x4883, 0x4883, 0x48AF, 0x48AF,
+        0x48C5, 0x48C5, 0x48E5, 0x48E5, 0x4905, 0x4905, 0x4925, 0x4925,
+        0x4945, 0x4945, 0x4950, 0x4950, 0x495B, 0x495B, 0x4980, 0x498E,
+        0x4B00, 0x4B00, 0x4C00, 0x4C00, 0x4D00, 0x4D00, 0x4E00, 0x4E00,
+        0x4E80, 0x4E80, 0x4F00, 0x4F00, 0x4F08, 0x4F08, 0x4F10, 0x4F10,
+        0x4F18, 0x4F18, 0x4F20, 0x4F20, 0x4F30, 0x4F30, 0x4F60, 0x4F60,
+        0x4F80, 0x4F81, 0x4F88, 0x4F89, 0x4FEE, 0x4FEE, 0x4FF3, 0x4FF3,
+        0x6000, 0x6001, 0x6008, 0x600F, 0x6014, 0x6016, 0x6018, 0x601B,
+        0x61FD, 0x61FD, 0x623C, 0x623C, 0x6380, 0x6380, 0x63A0, 0x63A0,
+        0x63C0, 0x63C1, 0x63C8, 0x63C9, 0x63D0, 0x63D4, 0x63D6, 0x63D6,
+        0x63EE, 0x63EE, 0x6400, 0x6401, 0x6408, 0x640F, 0x6414, 0x6416,
+        0x6418, 0x641B, 0x65FD, 0x65FD, 0x663C, 0x663C, 0x6780, 0x6780,
+        0x67A0, 0x67A0, 0x67C0, 0x67C1, 0x67C8, 0x67C9, 0x67D0, 0x67D4,
+        0x67D6, 0x67D6, 0x67EE, 0x67EE, 0x6800, 0x6801, 0x6808, 0x680F,
+        0x6814, 0x6816, 0x6818, 0x681B, 0x69FD, 0x69FD, 0x6A3C, 0x6A3C,
+        0x6B80, 0x6B80, 0x6BA0, 0x6BA0, 0x6BC0, 0x6BC1, 0x6BC8, 0x6BC9,
+        0x6BD0, 0x6BD4, 0x6BD6, 0x6BD6, 0x6BEE, 0x6BEE,
+        ~0 /* sentinel */
+};
+
+#ifdef CONFIG_DEBUG_FS
+static void a4xx_show(struct msm_gpu *gpu, struct seq_file *m)
+{
+        gpu->funcs->pm_resume(gpu);
+
+        seq_printf(m, "status:   %08x\n",
+                        gpu_read(gpu, REG_A4XX_RBBM_STATUS));
+        gpu->funcs->pm_suspend(gpu);
+
+        adreno_show(gpu, m);
+
+}
+#endif
+
+/* Register offset defines for A4XX, in order of enum adreno_regs */
+static const unsigned int a4xx_register_offsets[REG_ADRENO_REGISTER_MAX] = {
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_DEBUG, REG_A4XX_CP_DEBUG),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_WADDR, REG_A4XX_CP_ME_RAM_WADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_DATA, REG_A4XX_CP_ME_RAM_DATA),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_PFP_UCODE_DATA,
+                        REG_A4XX_CP_PFP_UCODE_DATA),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_PFP_UCODE_ADDR,
+                        REG_A4XX_CP_PFP_UCODE_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_WFI_PEND_CTR, REG_A4XX_CP_WFI_PEND_CTR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_BASE, REG_A4XX_CP_RB_BASE),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_RPTR_ADDR, REG_A4XX_CP_RB_RPTR_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_RPTR, REG_A4XX_CP_RB_RPTR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_WPTR, REG_A4XX_CP_RB_WPTR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_PROTECT_CTRL, REG_A4XX_CP_PROTECT_CTRL),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_CNTL, REG_A4XX_CP_ME_CNTL),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_RB_CNTL, REG_A4XX_CP_RB_CNTL),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_IB1_BASE, REG_A4XX_CP_IB1_BASE),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_IB1_BUFSZ, REG_A4XX_CP_IB1_BUFSZ),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_IB2_BASE, REG_A4XX_CP_IB2_BASE),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_IB2_BUFSZ, REG_A4XX_CP_IB2_BUFSZ),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_TIMESTAMP, REG_AXXX_CP_SCRATCH_REG0),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ME_RAM_RADDR, REG_A4XX_CP_ME_RAM_RADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ROQ_ADDR, REG_A4XX_CP_ROQ_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_ROQ_DATA, REG_A4XX_CP_ROQ_DATA),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_ADDR, REG_A4XX_CP_MERCIU_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_DATA, REG_A4XX_CP_MERCIU_DATA),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_MERCIU_DATA2, REG_A4XX_CP_MERCIU_DATA2),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_MEQ_ADDR, REG_A4XX_CP_MEQ_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_MEQ_DATA, REG_A4XX_CP_MEQ_DATA),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_HW_FAULT, REG_A4XX_CP_HW_FAULT),
+        REG_ADRENO_DEFINE(REG_ADRENO_CP_PROTECT_STATUS,
+                        REG_A4XX_CP_PROTECT_STATUS),
+        REG_ADRENO_DEFINE(REG_ADRENO_SCRATCH_ADDR, REG_A4XX_CP_SCRATCH_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_SCRATCH_UMSK, REG_A4XX_CP_SCRATCH_UMASK),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_STATUS, REG_A4XX_RBBM_STATUS),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_CTL,
+                        REG_A4XX_RBBM_PERFCTR_CTL),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_CMD0,
+                        REG_A4XX_RBBM_PERFCTR_LOAD_CMD0),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_CMD1,
+                        REG_A4XX_RBBM_PERFCTR_LOAD_CMD1),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_CMD2,
+                        REG_A4XX_RBBM_PERFCTR_LOAD_CMD2),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_PWR_1_LO,
+                        REG_A4XX_RBBM_PERFCTR_PWR_1_LO),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_0_MASK, REG_A4XX_RBBM_INT_0_MASK),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_0_STATUS,
+                        REG_A4XX_RBBM_INT_0_STATUS),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_ERROR_STATUS,
+                        REG_A4XX_RBBM_AHB_ERROR_STATUS),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_CMD, REG_A4XX_RBBM_AHB_CMD),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_CLOCK_CTL, REG_A4XX_RBBM_CLOCK_CTL),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_ME_SPLIT_STATUS,
+                        REG_A4XX_RBBM_AHB_ME_SPLIT_STATUS),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_AHB_PFP_SPLIT_STATUS,
+                        REG_A4XX_RBBM_AHB_PFP_SPLIT_STATUS),
+        REG_ADRENO_DEFINE(REG_ADRENO_VPC_DEBUG_RAM_SEL,
+                        REG_A4XX_VPC_DEBUG_RAM_SEL),
+        REG_ADRENO_DEFINE(REG_ADRENO_VPC_DEBUG_RAM_READ,
+                        REG_A4XX_VPC_DEBUG_RAM_READ),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_INT_CLEAR_CMD,
+                        REG_A4XX_RBBM_INT_CLEAR_CMD),
+        REG_ADRENO_DEFINE(REG_ADRENO_VSC_SIZE_ADDRESS,
+                        REG_A4XX_VSC_SIZE_ADDRESS),
+        REG_ADRENO_DEFINE(REG_ADRENO_VFD_CONTROL_0, REG_A4XX_VFD_CONTROL_0),
+        REG_ADRENO_DEFINE(REG_ADRENO_SP_VS_PVT_MEM_ADDR_REG,
+                        REG_A4XX_SP_VS_PVT_MEM_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_SP_FS_PVT_MEM_ADDR_REG,
+                        REG_A4XX_SP_FS_PVT_MEM_ADDR),
+        REG_ADRENO_DEFINE(REG_ADRENO_SP_VS_OBJ_START_REG,
+                        REG_A4XX_SP_VS_OBJ_START),
+        REG_ADRENO_DEFINE(REG_ADRENO_SP_FS_OBJ_START_REG,
+                        REG_A4XX_SP_FS_OBJ_START),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_RBBM_CTL, REG_A4XX_RBBM_RBBM_CTL),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_SW_RESET_CMD,
+                        REG_A4XX_RBBM_SW_RESET_CMD),
+        REG_ADRENO_DEFINE(REG_ADRENO_UCHE_INVALIDATE0,
+                        REG_A4XX_UCHE_INVALIDATE0),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_LO,
+                        REG_A4XX_RBBM_PERFCTR_LOAD_VALUE_LO),
+        REG_ADRENO_DEFINE(REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_HI,
+                        REG_A4XX_RBBM_PERFCTR_LOAD_VALUE_HI),
+};
+
+static void a4xx_dump(struct msm_gpu *gpu)
+{
+        adreno_dump(gpu);
+        printk("status:   %08x\n",
+                        gpu_read(gpu, REG_A4XX_RBBM_STATUS));
+        adreno_dump(gpu);
+}
+
+static const struct adreno_gpu_funcs funcs = {
+        .base = {
+                .get_param = adreno_get_param,
+                .hw_init = a4xx_hw_init,
+                .pm_suspend = msm_gpu_pm_suspend,
+                .pm_resume = msm_gpu_pm_resume,
+                .recover = a4xx_recover,
+                .last_fence = adreno_last_fence,
+                .submit = adreno_submit,
+                .flush = adreno_flush,
+                .idle = a4xx_idle,
+                .irq = a4xx_irq,
+                .destroy = a4xx_destroy,
+#ifdef CONFIG_DEBUG_FS
+                .show = a4xx_show,
+#endif
+        },
+};
+
+struct msm_gpu *a4xx_gpu_init(struct drm_device *dev)
+{
+        struct a4xx_gpu *a4xx_gpu = NULL;
+        struct adreno_gpu *adreno_gpu;
+        struct msm_gpu *gpu;
+        struct msm_drm_private *priv = dev->dev_private;
+        struct platform_device *pdev = priv->gpu_pdev;
+        int ret;
+
+        if (!pdev) {
+                dev_err(dev->dev, "no a4xx device\n");
+                ret = -ENXIO;
+                goto fail;
+        }
+
+        a4xx_gpu = kzalloc(sizeof(*a4xx_gpu), GFP_KERNEL);
+        if (!a4xx_gpu) {
+                ret = -ENOMEM;
+                goto fail;
+        }
+
+        adreno_gpu = &a4xx_gpu->base;
+        gpu = &adreno_gpu->base;
+
+        a4xx_gpu->pdev = pdev;
+
+        gpu->perfcntrs = NULL;
+        gpu->num_perfcntrs = 0;
+
+        adreno_gpu->registers = a4xx_registers;
+        adreno_gpu->reg_offsets = a4xx_register_offsets;
+
+        ret = adreno_gpu_init(dev, pdev, adreno_gpu, &funcs);
+        if (ret)
+                goto fail;
+
+        /* if needed, allocate gmem: */
+        if (adreno_is_a4xx(adreno_gpu)) {
+#ifdef CONFIG_MSM_OCMEM
+                /* TODO this is different/missing upstream: */
+                struct ocmem_buf *ocmem_hdl =
+                                ocmem_allocate(OCMEM_GRAPHICS, adreno_gpu->gmem);
+
+                a4xx_gpu->ocmem_hdl = ocmem_hdl;
+                a4xx_gpu->ocmem_base = ocmem_hdl->addr;
+                adreno_gpu->gmem = ocmem_hdl->len;
+                DBG("using %dK of OCMEM at 0x%08x", adreno_gpu->gmem / 1024,
+                                a4xx_gpu->ocmem_base);
+#endif
+        }
+
+        if (!gpu->mmu) {
+                /* TODO we think it is possible to configure the GPU to
+                 * restrict access to VRAM carveout.  But the required
+                 * registers are unknown.  For now just bail out and
+                 * limp along with just modesetting.  If it turns out
+                 * to not be possible to restrict access, then we must
+                 * implement a cmdstream validator.
+                 */
+                dev_err(dev->dev, "No memory protection without IOMMU\n");
+                ret = -ENXIO;
+                goto fail;
+        }
+
+        return gpu;
+
+fail:
+        if (a4xx_gpu)
+                a4xx_destroy(&a4xx_gpu->base.base);
+
+        return ERR_PTR(ret);
+}
diff --git a/drivers/gpu/drm/msm/adreno/a4xx_gpu.h b/drivers/gpu/drm/msm/adreno/a4xx_gpu.h
new file mode 100644
index 000000000000..01247204ac92
--- /dev/null
+++ b/drivers/gpu/drm/msm/adreno/a4xx_gpu.h
@@ -0,0 +1,34 @@
+/* Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ */
+#ifndef __A4XX_GPU_H__
+#define __A4XX_GPU_H__
+
+#include "adreno_gpu.h"
+
+/* arrg, somehow fb.h is getting pulled in: */
+#undef ROP_COPY
+#undef ROP_XOR
+
+#include "a4xx.xml.h"
+
+struct a4xx_gpu {
+        struct adreno_gpu base;
+        struct platform_device *pdev;
+
+        /* if OCMEM is used for GMEM: */
+        uint32_t ocmem_base;
+        void *ocmem_hdl;
+};
+#define to_a4xx_gpu(x) container_of(x, struct a4xx_gpu, base)
+
+#endif /* __A4XX_GPU_H__ */
diff --git a/drivers/gpu/drm/msm/adreno/adreno_common.xml.h b/drivers/gpu/drm/msm/adreno/adreno_common.xml.h
index cc341bc62b51..a4b33af9338d 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_common.xml.h
+++ b/drivers/gpu/drm/msm/adreno/adreno_common.xml.h
@@ -11,10 +11,10 @@ The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/adreno.xml               (    364 bytes, from 2013-11-30 14:47:15)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml  (   1453 bytes, from 2013-03-31 16:51:27)
 - /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml          (  32901 bytes, from 2014-06-02 15:21:30)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml (   9859 bytes, from 2014-06-02 15:21:30)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml    (  14960 bytes, from 2014-07-27 17:22:13)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml          (  58020 bytes, from 2014-08-01 12:22:48)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml          (  41068 bytes, from 2014-08-01 12:22:48)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml (  10551 bytes, from 2014-11-13 22:44:30)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml    (  15053 bytes, from 2014-11-09 15:45:47)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml          (  63169 bytes, from 2014-11-13 22:44:18)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml          (  49097 bytes, from 2014-11-14 15:38:00)
 
 Copyright (C) 2013-2014 by the following authors:
 - Rob Clark <robdclark@gmail.com> (robclark)
@@ -105,6 +105,7 @@ enum adreno_rb_dither_mode {
 enum adreno_rb_depth_format {
         DEPTHX_16 = 0,
         DEPTHX_24_8 = 1,
+        DEPTHX_32 = 2,
 };
 
 enum adreno_rb_copy_control_mode {
@@ -132,6 +133,7 @@ enum a3xx_threadmode {
 };
 
 enum a3xx_instrbuffermode {
+        CACHE = 0,
         BUFFER = 1,
 };
 
@@ -140,6 +142,13 @@ enum a3xx_threadsize {
         FOUR_QUADS = 1,
 };
 
+enum a3xx_color_swap {
+        WZYX = 0,
+        WXYZ = 1,
+        ZYXW = 2,
+        XYZW = 3,
+};
+
 #define REG_AXXX_CP_RB_BASE                                     0x000001c0
 
 #define REG_AXXX_CP_RB_CNTL                                     0x000001c1
diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c
index 7ab85af3a7db..be83dee83d08 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_device.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_device.c
@@ -2,6 +2,8 @@
  * Copyright (C) 2013-2014 Red Hat
  * Author: Rob Clark <robdclark@gmail.com>
  *
+ * Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published by
  * the Free Software Foundation.
@@ -28,6 +30,7 @@ MODULE_PARM_DESC(hang_debug, "Dump registers when hang is detected (can be slow!
 module_param_named(hang_debug, hang_debug, bool, 0600);
 
 struct msm_gpu *a3xx_gpu_init(struct drm_device *dev);
+struct msm_gpu *a4xx_gpu_init(struct drm_device *dev);
 
 static const struct adreno_info gpulist[] = {
         {
@@ -54,6 +57,14 @@ static const struct adreno_info gpulist[] = {
                 .pfpfw = "a330_pfp.fw",
                 .gmem  = SZ_1M,
                 .init  = a3xx_gpu_init,
+        }, {
+                .rev   = ADRENO_REV(4, 2, 0, ANY_ID),
+                .revn  = 420,
+                .name  = "A420",
+                .pm4fw = "a420_pm4.fw",
+                .pfpfw = "a420_pfp.fw",
+                .gmem  = (SZ_1M + SZ_512K),
+                .init  = a4xx_gpu_init,
         },
 };
 
@@ -61,6 +72,8 @@ MODULE_FIRMWARE("a300_pm4.fw");
 MODULE_FIRMWARE("a300_pfp.fw");
 MODULE_FIRMWARE("a330_pm4.fw");
 MODULE_FIRMWARE("a330_pfp.fw");
+MODULE_FIRMWARE("a420_pm4.fw");
+MODULE_FIRMWARE("a420_pfp.fw");
 
 static inline bool _rev_match(uint8_t entry, uint8_t id)
 {
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
index 6afa29167fee..aa873048308b 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
@@ -2,6 +2,8 @@
  * Copyright (C) 2013 Red Hat
  * Author: Rob Clark <robdclark@gmail.com>
  *
+ * Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published by
  * the Free Software Foundation.
@@ -63,19 +65,21 @@ int adreno_hw_init(struct msm_gpu *gpu)
         }
 
         /* Setup REG_CP_RB_CNTL: */
-        gpu_write(gpu, REG_AXXX_CP_RB_CNTL,
+        adreno_gpu_write(adreno_gpu, REG_ADRENO_CP_RB_CNTL,
                         /* size is log2(quad-words): */
                         AXXX_CP_RB_CNTL_BUFSZ(ilog2(gpu->rb->size / 8)) |
                         AXXX_CP_RB_CNTL_BLKSZ(ilog2(RB_BLKSIZE / 8)));
 
         /* Setup ringbuffer address: */
-        gpu_write(gpu, REG_AXXX_CP_RB_BASE, gpu->rb_iova);
-        gpu_write(gpu, REG_AXXX_CP_RB_RPTR_ADDR, rbmemptr(adreno_gpu, rptr));
+        adreno_gpu_write(adreno_gpu, REG_ADRENO_CP_RB_BASE, gpu->rb_iova);
+        adreno_gpu_write(adreno_gpu, REG_ADRENO_CP_RB_RPTR_ADDR,
+                        rbmemptr(adreno_gpu, rptr));
 
         /* Setup scratch/timestamp: */
-        gpu_write(gpu, REG_AXXX_SCRATCH_ADDR, rbmemptr(adreno_gpu, fence));
+        adreno_gpu_write(adreno_gpu, REG_ADRENO_SCRATCH_ADDR,
+                        rbmemptr(adreno_gpu, fence));
 
-        gpu_write(gpu, REG_AXXX_SCRATCH_UMSK, 0x1);
+        adreno_gpu_write(adreno_gpu, REG_ADRENO_SCRATCH_UMSK, 0x1);
 
         return 0;
 }
@@ -151,7 +155,7 @@ int adreno_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
         OUT_PKT0(ring, REG_AXXX_CP_SCRATCH_REG2, 1);
         OUT_RING(ring, submit->fence);
 
-        if (adreno_is_a3xx(adreno_gpu)) {
+        if (adreno_is_a3xx(adreno_gpu) || adreno_is_a4xx(adreno_gpu)) {
                 /* Flush HLSQ lazy updates to make sure there is nothing
                  * pending for indirect loads after the timestamp has
                  * passed:
@@ -188,12 +192,13 @@ int adreno_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 
 void adreno_flush(struct msm_gpu *gpu)
 {
+        struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
         uint32_t wptr = get_wptr(gpu->rb);
 
         /* ensure writes to ringbuffer have hit system memory: */
         mb();
 
-        gpu_write(gpu, REG_AXXX_CP_RB_WPTR, wptr);
+        adreno_gpu_write(adreno_gpu, REG_ADRENO_CP_RB_WPTR, wptr);
 }
 
 void adreno_idle(struct msm_gpu *gpu)
@@ -319,6 +324,12 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
         DBG("fast_rate=%u, slow_rate=%u, bus_freq=%u",
                         gpu->fast_rate, gpu->slow_rate, gpu->bus_freq);
 
+        ret = msm_gpu_init(drm, pdev, &adreno_gpu->base, &funcs->base,
+                        adreno_gpu->info->name, "kgsl_3d0_reg_memory", "kgsl_3d0_irq",
+                        RB_SIZE);
+        if (ret)
+                return ret;
+
         ret = request_firmware(&adreno_gpu->pm4, adreno_gpu->info->pm4fw, drm->dev);
         if (ret) {
                 dev_err(drm->dev, "failed to load %s PM4 firmware: %d\n",
@@ -333,12 +344,6 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
                 return ret;
         }
 
-        ret = msm_gpu_init(drm, pdev, &adreno_gpu->base, &funcs->base,
-                        adreno_gpu->info->name, "kgsl_3d0_reg_memory", "kgsl_3d0_irq",
-                        RB_SIZE);
-        if (ret)
-                return ret;
-
         mmu = gpu->mmu;
         if (mmu) {
                 ret = mmu->funcs->attach(mmu, iommu_ports,
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.h b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
index 52f051579753..a0cc30977e67 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.h
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
@@ -2,6 +2,8 @@
  * Copyright (C) 2013 Red Hat
  * Author: Rob Clark <robdclark@gmail.com>
  *
+ * Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 as published by
  * the Free Software Foundation.
@@ -25,6 +27,81 @@
 #include "adreno_common.xml.h"
 #include "adreno_pm4.xml.h"
 
+#define REG_ADRENO_DEFINE(_offset, _reg) [_offset] = (_reg) + 1
+/**
+ * adreno_regs: List of registers that are used in across all
+ * 3D devices. Each device type has different offset value for the same
+ * register, so an array of register offsets are declared for every device
+ * and are indexed by the enumeration values defined in this enum
+ */
+enum adreno_regs {
+        REG_ADRENO_CP_DEBUG,
+        REG_ADRENO_CP_ME_RAM_WADDR,
+        REG_ADRENO_CP_ME_RAM_DATA,
+        REG_ADRENO_CP_PFP_UCODE_DATA,
+        REG_ADRENO_CP_PFP_UCODE_ADDR,
+        REG_ADRENO_CP_WFI_PEND_CTR,
+        REG_ADRENO_CP_RB_BASE,
+        REG_ADRENO_CP_RB_RPTR_ADDR,
+        REG_ADRENO_CP_RB_RPTR,
+        REG_ADRENO_CP_RB_WPTR,
+        REG_ADRENO_CP_PROTECT_CTRL,
+        REG_ADRENO_CP_ME_CNTL,
+        REG_ADRENO_CP_RB_CNTL,
+        REG_ADRENO_CP_IB1_BASE,
+        REG_ADRENO_CP_IB1_BUFSZ,
+        REG_ADRENO_CP_IB2_BASE,
+        REG_ADRENO_CP_IB2_BUFSZ,
+        REG_ADRENO_CP_TIMESTAMP,
+        REG_ADRENO_CP_ME_RAM_RADDR,
+        REG_ADRENO_CP_ROQ_ADDR,
+        REG_ADRENO_CP_ROQ_DATA,
+        REG_ADRENO_CP_MERCIU_ADDR,
+        REG_ADRENO_CP_MERCIU_DATA,
+        REG_ADRENO_CP_MERCIU_DATA2,
+        REG_ADRENO_CP_MEQ_ADDR,
+        REG_ADRENO_CP_MEQ_DATA,
+        REG_ADRENO_CP_HW_FAULT,
+        REG_ADRENO_CP_PROTECT_STATUS,
+        REG_ADRENO_SCRATCH_ADDR,
+        REG_ADRENO_SCRATCH_UMSK,
+        REG_ADRENO_SCRATCH_REG2,
+        REG_ADRENO_RBBM_STATUS,
+        REG_ADRENO_RBBM_PERFCTR_CTL,
+        REG_ADRENO_RBBM_PERFCTR_LOAD_CMD0,
+        REG_ADRENO_RBBM_PERFCTR_LOAD_CMD1,
+        REG_ADRENO_RBBM_PERFCTR_LOAD_CMD2,
+        REG_ADRENO_RBBM_PERFCTR_PWR_1_LO,
+        REG_ADRENO_RBBM_INT_0_MASK,
+        REG_ADRENO_RBBM_INT_0_STATUS,
+        REG_ADRENO_RBBM_AHB_ERROR_STATUS,
+        REG_ADRENO_RBBM_PM_OVERRIDE2,
+        REG_ADRENO_RBBM_AHB_CMD,
+        REG_ADRENO_RBBM_INT_CLEAR_CMD,
+        REG_ADRENO_RBBM_SW_RESET_CMD,
+        REG_ADRENO_RBBM_CLOCK_CTL,
+        REG_ADRENO_RBBM_AHB_ME_SPLIT_STATUS,
+        REG_ADRENO_RBBM_AHB_PFP_SPLIT_STATUS,
+        REG_ADRENO_VPC_DEBUG_RAM_SEL,
+        REG_ADRENO_VPC_DEBUG_RAM_READ,
+        REG_ADRENO_VSC_SIZE_ADDRESS,
+        REG_ADRENO_VFD_CONTROL_0,
+        REG_ADRENO_VFD_INDEX_MAX,
+        REG_ADRENO_SP_VS_PVT_MEM_ADDR_REG,
+        REG_ADRENO_SP_FS_PVT_MEM_ADDR_REG,
+        REG_ADRENO_SP_VS_OBJ_START_REG,
+        REG_ADRENO_SP_FS_OBJ_START_REG,
+        REG_ADRENO_PA_SC_AA_CONFIG,
+        REG_ADRENO_SQ_GPR_MANAGEMENT,
+        REG_ADRENO_SQ_INST_STORE_MANAGMENT,
+        REG_ADRENO_TP0_CHICKEN,
+        REG_ADRENO_RBBM_RBBM_CTL,
+        REG_ADRENO_UCHE_INVALIDATE0,
+        REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_LO,
+        REG_ADRENO_RBBM_PERFCTR_LOAD_VALUE_HI,
+        REG_ADRENO_REGISTER_MAX,
+};
+
 struct adreno_rev {
         uint8_t  core;
         uint8_t  major;
@@ -76,6 +153,13 @@ struct adreno_gpu {
         struct adreno_rbmemptrs *memptrs;
         struct drm_gem_object *memptrs_bo;
         uint32_t memptrs_iova;
+
+        /*
+         * Register offsets are different between some GPUs.
+         * GPU specific offsets will be exported by GPU specific
+         * code (a3xx_gpu.c) and stored in this common location.
+         */
+        const unsigned int *reg_offsets;
 };
 #define to_adreno_gpu(x) container_of(x, struct adreno_gpu, base)
 
@@ -128,6 +212,16 @@ static inline bool adreno_is_a330v2(struct adreno_gpu *gpu)
         return adreno_is_a330(gpu) && (gpu->rev.patchid > 0);
 }
 
+static inline bool adreno_is_a4xx(struct adreno_gpu *gpu)
+{
+        return (gpu->revn >= 400) && (gpu->revn < 500);
+}
+
+static inline int adreno_is_a420(struct adreno_gpu *gpu)
+{
+        return gpu->revn == 420;
+}
+
 int adreno_get_param(struct msm_gpu *gpu, uint32_t param, uint64_t *value);
 int adreno_hw_init(struct msm_gpu *gpu);
 uint32_t adreno_last_fence(struct msm_gpu *gpu);
@@ -171,5 +265,37 @@ OUT_PKT3(struct msm_ringbuffer *ring, uint8_t opcode, uint16_t cnt)
         OUT_RING(ring, CP_TYPE3_PKT | ((cnt-1) << 16) | ((opcode & 0xFF) << 8));
 }
 
+/*
+ * adreno_checkreg_off() - Checks the validity of a register enum
+ * @gpu:                Pointer to struct adreno_gpu
+ * @offset_name:        The register enum that is checked
+ */
+static inline bool adreno_reg_check(struct adreno_gpu *gpu,
+                enum adreno_regs offset_name)
+{
+        if (offset_name >= REG_ADRENO_REGISTER_MAX ||
+                        !gpu->reg_offsets[offset_name]) {
+                BUG();
+        }
+        return true;
+}
+
+static inline u32 adreno_gpu_read(struct adreno_gpu *gpu,
+                enum adreno_regs offset_name)
+{
+        u32 reg = gpu->reg_offsets[offset_name];
+        u32 val = 0;
+        if(adreno_reg_check(gpu,offset_name))
+                val = gpu_read(&gpu->base, reg - 1);
+        return val;
+}
+
+static inline void adreno_gpu_write(struct adreno_gpu *gpu,
+                enum adreno_regs offset_name, u32 data)
+{
+        u32 reg = gpu->reg_offsets[offset_name];
+        if(adreno_reg_check(gpu, offset_name))
+                gpu_write(&gpu->base, reg - 1, data);
+}
 
 #endif /* __ADRENO_GPU_H__ */
diff --git a/drivers/gpu/drm/msm/adreno/adreno_pm4.xml.h b/drivers/gpu/drm/msm/adreno/adreno_pm4.xml.h
index 6ef43f66c30a..6a75cee94d81 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_pm4.xml.h
+++ b/drivers/gpu/drm/msm/adreno/adreno_pm4.xml.h
@@ -11,10 +11,10 @@ The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/adreno.xml               (    364 bytes, from 2013-11-30 14:47:15)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml  (   1453 bytes, from 2013-03-31 16:51:27)
 - /home/robclark/src/freedreno/envytools/rnndb/adreno/a2xx.xml          (  32901 bytes, from 2014-06-02 15:21:30)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml (   9859 bytes, from 2014-06-02 15:21:30)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml    (  14960 bytes, from 2014-07-27 17:22:13)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml          (  58020 bytes, from 2014-08-01 12:22:48)
-- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml          (  41068 bytes, from 2014-08-01 12:22:48)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_common.xml (  10551 bytes, from 2014-11-13 22:44:30)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/adreno_pm4.xml    (  15053 bytes, from 2014-11-09 15:45:47)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a3xx.xml          (  63169 bytes, from 2014-11-13 22:44:18)
+- /home/robclark/src/freedreno/envytools/rnndb/adreno/a4xx.xml          (  49097 bytes, from 2014-11-14 15:38:00)
 
 Copyright (C) 2013-2014 by the following authors:
 - Rob Clark <robdclark@gmail.com> (robclark)
@@ -157,6 +157,7 @@ enum adreno_pm4_type3_packets {
         CP_IM_STORE = 44,
         CP_SET_DRAW_INIT_FLAGS = 75,
         CP_SET_PROTECTED_MODE = 95,
+        CP_BOOTSTRAP_UCODE = 111,
         CP_LOAD_STATE = 48,
         CP_COND_INDIRECT_BUFFER_PFE = 58,
         CP_COND_INDIRECT_BUFFER_PFD = 50,
@@ -278,11 +279,11 @@ static inline uint32_t CP_DRAW_INDX_1_INDEX_SIZE(enum pc_di_index_size val)
 #define CP_DRAW_INDX_1_NOT_EOP                                  0x00001000
 #define CP_DRAW_INDX_1_SMALL_INDEX                              0x00002000
 #define CP_DRAW_INDX_1_PRE_DRAW_INITIATOR_ENABLE                0x00004000
-#define CP_DRAW_INDX_1_NUM_INDICES__MASK                        0xffff0000
-#define CP_DRAW_INDX_1_NUM_INDICES__SHIFT                       16
-static inline uint32_t CP_DRAW_INDX_1_NUM_INDICES(uint32_t val)
+#define CP_DRAW_INDX_1_NUM_INSTANCES__MASK                      0xff000000
+#define CP_DRAW_INDX_1_NUM_INSTANCES__SHIFT                     24
+static inline uint32_t CP_DRAW_INDX_1_NUM_INSTANCES(uint32_t val)
 {
-        return ((val) << CP_DRAW_INDX_1_NUM_INDICES__SHIFT) & CP_DRAW_INDX_1_NUM_INDICES__MASK;
+        return ((val) << CP_DRAW_INDX_1_NUM_INSTANCES__SHIFT) & CP_DRAW_INDX_1_NUM_INSTANCES__MASK;
 }
 
 #define REG_CP_DRAW_INDX_2                                      0x00000002
@@ -293,20 +294,20 @@ static inline uint32_t CP_DRAW_INDX_2_NUM_INDICES(uint32_t val)
         return ((val) << CP_DRAW_INDX_2_NUM_INDICES__SHIFT) & CP_DRAW_INDX_2_NUM_INDICES__MASK;
 }
 
-#define REG_CP_DRAW_INDX_2                                      0x00000002
-#define CP_DRAW_INDX_2_INDX_BASE__MASK                          0xffffffff
-#define CP_DRAW_INDX_2_INDX_BASE__SHIFT                         0
-static inline uint32_t CP_DRAW_INDX_2_INDX_BASE(uint32_t val)
+#define REG_CP_DRAW_INDX_3                                      0x00000003
+#define CP_DRAW_INDX_3_INDX_BASE__MASK                          0xffffffff
+#define CP_DRAW_INDX_3_INDX_BASE__SHIFT                         0
+static inline uint32_t CP_DRAW_INDX_3_INDX_BASE(uint32_t val)
 {
-        return ((val) << CP_DRAW_INDX_2_INDX_BASE__SHIFT) & CP_DRAW_INDX_2_INDX_BASE__MASK;
+        return ((val) << CP_DRAW_INDX_3_INDX_BASE__SHIFT) & CP_DRAW_INDX_3_INDX_BASE__MASK;
 }
 
-#define REG_CP_DRAW_INDX_2                                      0x00000002
-#define CP_DRAW_INDX_2_INDX_SIZE__MASK                          0xffffffff
-#define CP_DRAW_INDX_2_INDX_SIZE__SHIFT                         0
-static inline uint32_t CP_DRAW_INDX_2_INDX_SIZE(uint32_t val)
+#define REG_CP_DRAW_INDX_4                                      0x00000004
+#define CP_DRAW_INDX_4_INDX_SIZE__MASK                          0xffffffff
+#define CP_DRAW_INDX_4_INDX_SIZE__SHIFT                         0
+static inline uint32_t CP_DRAW_INDX_4_INDX_SIZE(uint32_t val)
 {
-        return ((val) << CP_DRAW_INDX_2_INDX_SIZE__SHIFT) & CP_DRAW_INDX_2_INDX_SIZE__MASK;
+        return ((val) << CP_DRAW_INDX_4_INDX_SIZE__SHIFT) & CP_DRAW_INDX_4_INDX_SIZE__MASK;
 }
 
 #define REG_CP_DRAW_INDX_2_0                                    0x00000000
@@ -345,11 +346,11 @@ static inline uint32_t CP_DRAW_INDX_2_1_INDEX_SIZE(enum pc_di_index_size val)
 #define CP_DRAW_INDX_2_1_NOT_EOP                                0x00001000
 #define CP_DRAW_INDX_2_1_SMALL_INDEX                            0x00002000
 #define CP_DRAW_INDX_2_1_PRE_DRAW_INITIATOR_ENABLE              0x00004000
-#define CP_DRAW_INDX_2_1_NUM_INDICES__MASK                      0xffff0000
-#define CP_DRAW_INDX_2_1_NUM_INDICES__SHIFT                     16
-static inline uint32_t CP_DRAW_INDX_2_1_NUM_INDICES(uint32_t val)
+#define CP_DRAW_INDX_2_1_NUM_INSTANCES__MASK                    0xff000000
+#define CP_DRAW_INDX_2_1_NUM_INSTANCES__SHIFT                   24
+static inline uint32_t CP_DRAW_INDX_2_1_NUM_INSTANCES(uint32_t val)
 {
-        return ((val) << CP_DRAW_INDX_2_1_NUM_INDICES__SHIFT) & CP_DRAW_INDX_2_1_NUM_INDICES__MASK;
+        return ((val) << CP_DRAW_INDX_2_1_NUM_INSTANCES__SHIFT) & CP_DRAW_INDX_2_1_NUM_INSTANCES__MASK;
 }
 
 #define REG_CP_DRAW_INDX_2_2                                    0x00000002
@@ -388,11 +389,11 @@ static inline uint32_t CP_DRAW_INDX_OFFSET_0_INDEX_SIZE(enum pc_di_index_size va
 #define CP_DRAW_INDX_OFFSET_0_NOT_EOP                           0x00001000
 #define CP_DRAW_INDX_OFFSET_0_SMALL_INDEX                       0x00002000
 #define CP_DRAW_INDX_OFFSET_0_PRE_DRAW_INITIATOR_ENABLE         0x00004000
-#define CP_DRAW_INDX_OFFSET_0_NUM_INDICES__MASK                 0xffff0000
-#define CP_DRAW_INDX_OFFSET_0_NUM_INDICES__SHIFT                16
-static inline uint32_t CP_DRAW_INDX_OFFSET_0_NUM_INDICES(uint32_t val)
+#define CP_DRAW_INDX_OFFSET_0_NUM_INSTANCES__MASK               0xffff0000
+#define CP_DRAW_INDX_OFFSET_0_NUM_INSTANCES__SHIFT              16
+static inline uint32_t CP_DRAW_INDX_OFFSET_0_NUM_INSTANCES(uint32_t val)
 {
-        return ((val) << CP_DRAW_INDX_OFFSET_0_NUM_INDICES__SHIFT) & CP_DRAW_INDX_OFFSET_0_NUM_INDICES__MASK;
+        return ((val) << CP_DRAW_INDX_OFFSET_0_NUM_INSTANCES__SHIFT) & CP_DRAW_INDX_OFFSET_0_NUM_INSTANCES__MASK;
 }
 
 #define REG_CP_DRAW_INDX_OFFSET_1                               0x00000001
@@ -405,20 +406,22 @@ static inline uint32_t CP_DRAW_INDX_OFFSET_2_NUM_INDICES(uint32_t val)
         return ((val) << CP_DRAW_INDX_OFFSET_2_NUM_INDICES__SHIFT) & CP_DRAW_INDX_OFFSET_2_NUM_INDICES__MASK;
 }
 
-#define REG_CP_DRAW_INDX_OFFSET_2                               0x00000002
-#define CP_DRAW_INDX_OFFSET_2_INDX_BASE__MASK                   0xffffffff
-#define CP_DRAW_INDX_OFFSET_2_INDX_BASE__SHIFT                  0
-static inline uint32_t CP_DRAW_INDX_OFFSET_2_INDX_BASE(uint32_t val)
+#define REG_CP_DRAW_INDX_OFFSET_3                               0x00000003
+
+#define REG_CP_DRAW_INDX_OFFSET_4                               0x00000004
+#define CP_DRAW_INDX_OFFSET_4_INDX_BASE__MASK                   0xffffffff
+#define CP_DRAW_INDX_OFFSET_4_INDX_BASE__SHIFT                  0
+static inline uint32_t CP_DRAW_INDX_OFFSET_4_INDX_BASE(uint32_t val)
 {
-        return ((val) << CP_DRAW_INDX_OFFSET_2_INDX_BASE__SHIFT) & CP_DRAW_INDX_OFFSET_2_INDX_BASE__MASK;
+        return ((val) << CP_DRAW_INDX_OFFSET_4_INDX_BASE__SHIFT) & CP_DRAW_INDX_OFFSET_4_INDX_BASE__MASK;
 }
 
-#define REG_CP_DRAW_INDX_OFFSET_2                               0x00000002
-#define CP_DRAW_INDX_OFFSET_2_INDX_SIZE__MASK                   0xffffffff
-#define CP_DRAW_INDX_OFFSET_2_INDX_SIZE__SHIFT                  0
-static inline uint32_t CP_DRAW_INDX_OFFSET_2_INDX_SIZE(uint32_t val)
+#define REG_CP_DRAW_INDX_OFFSET_5                               0x00000005
+#define CP_DRAW_INDX_OFFSET_5_INDX_SIZE__MASK                   0xffffffff
+#define CP_DRAW_INDX_OFFSET_5_INDX_SIZE__SHIFT                  0
+static inline uint32_t CP_DRAW_INDX_OFFSET_5_INDX_SIZE(uint32_t val)
 {
-        return ((val) << CP_DRAW_INDX_OFFSET_2_INDX_SIZE__SHIFT) & CP_DRAW_INDX_OFFSET_2_INDX_SIZE__MASK;
+        return ((val) << CP_DRAW_INDX_OFFSET_5_INDX_SIZE__SHIFT) & CP_DRAW_INDX_OFFSET_5_INDX_SIZE__MASK;
 }
 
 #define REG_CP_SET_DRAW_STATE_0                                 0x00000000
diff --git a/drivers/gpu/drm/msm/dsi/dsi.xml.h b/drivers/gpu/drm/msm/dsi/dsi.xml.h
index e965898dfda6..448438b759b4 100644
--- a/drivers/gpu/drm/msm/dsi/dsi.xml.h
+++ b/drivers/gpu/drm/msm/dsi/dsi.xml.h
@@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git
 The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/msm.xml                 (    647 bytes, from 2013-11-30 14:45:35)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml (   1453 bytes, from 2013-03-31 16:51:27)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20457 bytes, from 2014-08-01 12:22:48)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1615 bytes, from 2014-07-17 15:34:33)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  22517 bytes, from 2014-07-17 15:34:33)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20136 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1940 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  23963 bytes, from 2014-10-31 16:51:46)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml             (  11712 bytes, from 2013-08-17 17:13:43)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml            (    344 bytes, from 2013-08-11 19:26:32)
-- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-08-01 12:23:53)
+- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-10-31 16:48:57)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml         (    600 bytes, from 2013-07-05 19:21:12)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml           (  23613 bytes, from 2014-07-17 15:33:30)
 
diff --git a/drivers/gpu/drm/msm/dsi/mmss_cc.xml.h b/drivers/gpu/drm/msm/dsi/mmss_cc.xml.h
index f2bdda957205..c102a7f074ac 100644
--- a/drivers/gpu/drm/msm/dsi/mmss_cc.xml.h
+++ b/drivers/gpu/drm/msm/dsi/mmss_cc.xml.h
@@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git
 The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/msm.xml                 (    647 bytes, from 2013-11-30 14:45:35)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml (   1453 bytes, from 2013-03-31 16:51:27)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20457 bytes, from 2014-08-01 12:22:48)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1615 bytes, from 2014-07-17 15:34:33)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  22517 bytes, from 2014-07-17 15:34:33)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20136 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1940 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  23963 bytes, from 2014-10-31 16:51:46)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml             (  11712 bytes, from 2013-08-17 17:13:43)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml            (    344 bytes, from 2013-08-11 19:26:32)
-- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-08-01 12:23:53)
+- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-10-31 16:48:57)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml         (    600 bytes, from 2013-07-05 19:21:12)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml           (  23613 bytes, from 2014-07-17 15:33:30)
 
diff --git a/drivers/gpu/drm/msm/dsi/sfpb.xml.h b/drivers/gpu/drm/msm/dsi/sfpb.xml.h
index e5b071ffd865..a900134bdf33 100644
--- a/drivers/gpu/drm/msm/dsi/sfpb.xml.h
+++ b/drivers/gpu/drm/msm/dsi/sfpb.xml.h
@@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git
 The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/msm.xml                 (    647 bytes, from 2013-11-30 14:45:35)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml (   1453 bytes, from 2013-03-31 16:51:27)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20457 bytes, from 2014-08-01 12:22:48)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1615 bytes, from 2014-07-17 15:34:33)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  22517 bytes, from 2014-07-17 15:34:33)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20136 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1940 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  23963 bytes, from 2014-10-31 16:51:46)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml             (  11712 bytes, from 2013-08-17 17:13:43)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml            (    344 bytes, from 2013-08-11 19:26:32)
-- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-08-01 12:23:53)
+- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-10-31 16:48:57)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml         (    600 bytes, from 2013-07-05 19:21:12)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml           (  23613 bytes, from 2014-07-17 15:33:30)
 
diff --git a/drivers/gpu/drm/msm/hdmi/hdmi.c b/drivers/gpu/drm/msm/hdmi/hdmi.c
index 9d00dcba6959..062c68725376 100644
--- a/drivers/gpu/drm/msm/hdmi/hdmi.c
+++ b/drivers/gpu/drm/msm/hdmi/hdmi.c
@@ -15,6 +15,7 @@
  * this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
+#include <linux/of_irq.h>
 #include "hdmi.h"
 
 void hdmi_set_mode(struct hdmi *hdmi, bool power_on)
@@ -39,7 +40,7 @@ void hdmi_set_mode(struct hdmi *hdmi, bool power_on)
                         power_on ? "Enable" : "Disable", ctrl);
 }
 
-irqreturn_t hdmi_irq(int irq, void *dev_id)
+static irqreturn_t hdmi_irq(int irq, void *dev_id)
 {
         struct hdmi *hdmi = dev_id;
 
@@ -54,9 +55,8 @@ irqreturn_t hdmi_irq(int irq, void *dev_id)
         return IRQ_HANDLED;
 }
 
-void hdmi_destroy(struct kref *kref)
+static void hdmi_destroy(struct hdmi *hdmi)
 {
-        struct hdmi *hdmi = container_of(kref, struct hdmi, refcount);
         struct hdmi_phy *phy = hdmi->phy;
 
         if (phy)
@@ -68,37 +68,24 @@ void hdmi_destroy(struct kref *kref)
         platform_set_drvdata(hdmi->pdev, NULL);
 }
 
-/* initialize connector */
-struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder)
+/* construct hdmi at bind/probe time, grab all the resources.  If
+ * we are to EPROBE_DEFER we want to do it here, rather than later
+ * at modeset_init() time
+ */
+static struct hdmi *hdmi_init(struct platform_device *pdev)
 {
+        struct hdmi_platform_config *config = pdev->dev.platform_data;
         struct hdmi *hdmi = NULL;
-        struct msm_drm_private *priv = dev->dev_private;
-        struct platform_device *pdev = priv->hdmi_pdev;
-        struct hdmi_platform_config *config;
         int i, ret;
 
-        if (!pdev) {
-                dev_err(dev->dev, "no hdmi device\n");
-                ret = -ENXIO;
-                goto fail;
-        }
-
-        config = pdev->dev.platform_data;
-
-        hdmi = kzalloc(sizeof(*hdmi), GFP_KERNEL);
+        hdmi = devm_kzalloc(&pdev->dev, sizeof(*hdmi), GFP_KERNEL);
         if (!hdmi) {
                 ret = -ENOMEM;
                 goto fail;
         }
 
-        kref_init(&hdmi->refcount);
-
-        hdmi->dev = dev;
         hdmi->pdev = pdev;
         hdmi->config = config;
-        hdmi->encoder = encoder;
-
-        hdmi_audio_infoframe_init(&hdmi->audio.infoframe);
 
         /* not sure about which phy maps to which msm.. probably I miss some */
         if (config->phy_init)
@@ -108,7 +95,7 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder)
 
         if (IS_ERR(hdmi->phy)) {
                 ret = PTR_ERR(hdmi->phy);
-                dev_err(dev->dev, "failed to load phy: %d\n", ret);
+                dev_err(&pdev->dev, "failed to load phy: %d\n", ret);
                 hdmi->phy = NULL;
                 goto fail;
         }
@@ -127,7 +114,7 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder)
                                 config->hpd_reg_names[i]);
                 if (IS_ERR(reg)) {
                         ret = PTR_ERR(reg);
-                        dev_err(dev->dev, "failed to get hpd regulator: %s (%d)\n",
+                        dev_err(&pdev->dev, "failed to get hpd regulator: %s (%d)\n",
                                         config->hpd_reg_names[i], ret);
                         goto fail;
                 }
@@ -143,7 +130,7 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder)
                                 config->pwr_reg_names[i]);
                 if (IS_ERR(reg)) {
                         ret = PTR_ERR(reg);
-                        dev_err(dev->dev, "failed to get pwr regulator: %s (%d)\n",
+                        dev_err(&pdev->dev, "failed to get pwr regulator: %s (%d)\n",
                                         config->pwr_reg_names[i], ret);
                         goto fail;
                 }
@@ -158,7 +145,7 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder)
                 clk = devm_clk_get(&pdev->dev, config->hpd_clk_names[i]);
                 if (IS_ERR(clk)) {
                         ret = PTR_ERR(clk);
-                        dev_err(dev->dev, "failed to get hpd clk: %s (%d)\n",
+                        dev_err(&pdev->dev, "failed to get hpd clk: %s (%d)\n",
                                         config->hpd_clk_names[i], ret);
                         goto fail;
                 }
@@ -173,7 +160,7 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder)
                 clk = devm_clk_get(&pdev->dev, config->pwr_clk_names[i]);
                 if (IS_ERR(clk)) {
                         ret = PTR_ERR(clk);
-                        dev_err(dev->dev, "failed to get pwr clk: %s (%d)\n",
+                        dev_err(&pdev->dev, "failed to get pwr clk: %s (%d)\n",
                                         config->pwr_clk_names[i], ret);
                         goto fail;
                 }
@@ -184,11 +171,40 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder)
         hdmi->i2c = hdmi_i2c_init(hdmi);
         if (IS_ERR(hdmi->i2c)) {
                 ret = PTR_ERR(hdmi->i2c);
-                dev_err(dev->dev, "failed to get i2c: %d\n", ret);
+                dev_err(&pdev->dev, "failed to get i2c: %d\n", ret);
                 hdmi->i2c = NULL;
                 goto fail;
         }
 
+        return hdmi;
+
+fail:
+        if (hdmi)
+                hdmi_destroy(hdmi);
+
+        return ERR_PTR(ret);
+}
+
+/* Second part of initialization, the drm/kms level modeset_init,
+ * constructs/initializes mode objects, etc, is called from master
+ * driver (not hdmi sub-device's probe/bind!)
+ *
+ * Any resource (regulator/clk/etc) which could be missing at boot
+ * should be handled in hdmi_init() so that failure happens from
+ * hdmi sub-device's probe.
+ */
+int hdmi_modeset_init(struct hdmi *hdmi,
+                struct drm_device *dev, struct drm_encoder *encoder)
+{
+        struct msm_drm_private *priv = dev->dev_private;
+        struct platform_device *pdev = hdmi->pdev;
+        int ret;
+
+        hdmi->dev = dev;
+        hdmi->encoder = encoder;
+
+        hdmi_audio_infoframe_init(&hdmi->audio.infoframe);
+
         hdmi->bridge = hdmi_bridge_init(hdmi);
         if (IS_ERR(hdmi->bridge)) {
                 ret = PTR_ERR(hdmi->bridge);
@@ -205,22 +221,20 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder)
                 goto fail;
         }
 
-        if (!config->shared_irq) {
-                hdmi->irq = platform_get_irq(pdev, 0);
-                if (hdmi->irq < 0) {
-                        ret = hdmi->irq;
-                        dev_err(dev->dev, "failed to get irq: %d\n", ret);
-                        goto fail;
-                }
+        hdmi->irq = irq_of_parse_and_map(pdev->dev.of_node, 0);
+        if (hdmi->irq < 0) {
+                ret = hdmi->irq;
+                dev_err(dev->dev, "failed to get irq: %d\n", ret);
+                goto fail;
+        }
 
-                ret = devm_request_threaded_irq(&pdev->dev, hdmi->irq,
-                                NULL, hdmi_irq, IRQF_TRIGGER_HIGH | IRQF_ONESHOT,
-                                "hdmi_isr", hdmi);
-                if (ret < 0) {
-                        dev_err(dev->dev, "failed to request IRQ%u: %d\n",
-                                        hdmi->irq, ret);
-                        goto fail;
-                }
+        ret = devm_request_irq(&pdev->dev, hdmi->irq,
+                        hdmi_irq, IRQF_TRIGGER_HIGH | IRQF_ONESHOT,
+                        "hdmi_isr", hdmi);
+        if (ret < 0) {
+                dev_err(dev->dev, "failed to request IRQ%u: %d\n",
+                                hdmi->irq, ret);
+                goto fail;
         }
 
         encoder->bridge = hdmi->bridge;
@@ -230,19 +244,20 @@ struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder)
 
         platform_set_drvdata(pdev, hdmi);
 
-        return hdmi;
+        return 0;
 
 fail:
-        if (hdmi) {
-                /* bridge/connector are normally destroyed by drm: */
-                if (hdmi->bridge)
-                        hdmi->bridge->funcs->destroy(hdmi->bridge);
-                if (hdmi->connector)
-                        hdmi->connector->funcs->destroy(hdmi->connector);
-                hdmi_destroy(&hdmi->refcount);
+        /* bridge/connector are normally destroyed by drm: */
+        if (hdmi->bridge) {
+                hdmi->bridge->funcs->destroy(hdmi->bridge);
+                hdmi->bridge = NULL;
+        }
+        if (hdmi->connector) {
+                hdmi->connector->funcs->destroy(hdmi->connector);
+                hdmi->connector = NULL;
         }
 
-        return ERR_PTR(ret);
+        return ret;
 }
 
 /*
@@ -251,13 +266,6 @@ fail:
 
 #include <linux/of_gpio.h>
 
-static void set_hdmi_pdev(struct drm_device *dev,
-                struct platform_device *pdev)
-{
-        struct msm_drm_private *priv = dev->dev_private;
-        priv->hdmi_pdev = pdev;
-}
-
 #ifdef CONFIG_OF
 static int get_gpio(struct device *dev, struct device_node *of_node, const char *name)
 {
@@ -278,7 +286,10 @@ static int get_gpio(struct device *dev, struct device_node *of_node, const char
 
 static int hdmi_bind(struct device *dev, struct device *master, void *data)
 {
+        struct drm_device *drm = dev_get_drvdata(master);
+        struct msm_drm_private *priv = drm->dev_private;
         static struct hdmi_platform_config config = {};
+        struct hdmi *hdmi;
 #ifdef CONFIG_OF
         struct device_node *of_node = dev->of_node;
 
@@ -298,7 +309,6 @@ static int hdmi_bind(struct device *dev, struct device *master, void *data)
                 config.hpd_clk_cnt   = ARRAY_SIZE(hpd_clk_names);
                 config.pwr_clk_names = pwr_clk_names;
                 config.pwr_clk_cnt   = ARRAY_SIZE(pwr_clk_names);
-                config.shared_irq    = true;
         } else if (of_device_is_compatible(of_node, "qcom,hdmi-tx-8960")) {
                 static const char *hpd_clk_names[] = {"core_clk", "master_iface_clk", "slave_iface_clk"};
                 static const char *hpd_reg_names[] = {"core-vdda", "hdmi-mux"};
@@ -369,14 +379,22 @@ static int hdmi_bind(struct device *dev, struct device *master, void *data)
         }
 #endif
         dev->platform_data = &config;
-        set_hdmi_pdev(dev_get_drvdata(master), to_platform_device(dev));
+        hdmi = hdmi_init(to_platform_device(dev));
+        if (IS_ERR(hdmi))
+                return PTR_ERR(hdmi);
+        priv->hdmi = hdmi;
         return 0;
 }
 
 static void hdmi_unbind(struct device *dev, struct device *master,
                 void *data)
 {
-        set_hdmi_pdev(dev_get_drvdata(master), NULL);
+        struct drm_device *drm = dev_get_drvdata(master);
+        struct msm_drm_private *priv = drm->dev_private;
+        if (priv->hdmi) {
+                hdmi_destroy(priv->hdmi);
+                priv->hdmi = NULL;
+        }
 }
 
 static const struct component_ops hdmi_ops = {
diff --git a/drivers/gpu/drm/msm/hdmi/hdmi.h b/drivers/gpu/drm/msm/hdmi/hdmi.h
index b981995410b5..43e654f751b7 100644
--- a/drivers/gpu/drm/msm/hdmi/hdmi.h
+++ b/drivers/gpu/drm/msm/hdmi/hdmi.h
@@ -38,8 +38,6 @@ struct hdmi_audio {
 };
 
 struct hdmi {
-        struct kref refcount;
-
         struct drm_device *dev;
         struct platform_device *pdev;
 
@@ -97,13 +95,9 @@ struct hdmi_platform_config {
         /* gpio's: */
         int ddc_clk_gpio, ddc_data_gpio, hpd_gpio, mux_en_gpio, mux_sel_gpio;
         int mux_lpm_gpio;
-
-        /* older devices had their own irq, mdp5+ it is shared w/ mdp: */
-        bool shared_irq;
 };
 
 void hdmi_set_mode(struct hdmi *hdmi, bool power_on);
-void hdmi_destroy(struct kref *kref);
 
 static inline void hdmi_write(struct hdmi *hdmi, u32 reg, u32 data)
 {
@@ -115,17 +109,6 @@ static inline u32 hdmi_read(struct hdmi *hdmi, u32 reg)
         return msm_readl(hdmi->mmio + reg);
 }
 
-static inline struct hdmi * hdmi_reference(struct hdmi *hdmi)
-{
-        kref_get(&hdmi->refcount);
-        return hdmi;
-}
-
-static inline void hdmi_unreference(struct hdmi *hdmi)
-{
-        kref_put(&hdmi->refcount, hdmi_destroy);
-}
-
 /*
  * The phy appears to be different, for example between 8960 and 8x60,
  * so split the phy related functions out and load the correct one at
diff --git a/drivers/gpu/drm/msm/hdmi/hdmi.xml.h b/drivers/gpu/drm/msm/hdmi/hdmi.xml.h
index 76fd0cfc6558..5b0844befbab 100644
--- a/drivers/gpu/drm/msm/hdmi/hdmi.xml.h
+++ b/drivers/gpu/drm/msm/hdmi/hdmi.xml.h
@@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git
 The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/msm.xml                 (    647 bytes, from 2013-11-30 14:45:35)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml (   1453 bytes, from 2013-03-31 16:51:27)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20457 bytes, from 2014-08-01 12:22:48)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1615 bytes, from 2014-07-17 15:34:33)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  22517 bytes, from 2014-07-17 15:34:33)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20136 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1940 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  23963 bytes, from 2014-10-31 16:51:46)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml             (  11712 bytes, from 2013-08-17 17:13:43)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml            (    344 bytes, from 2013-08-11 19:26:32)
-- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-08-01 12:23:53)
+- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-10-31 16:48:57)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml         (    600 bytes, from 2013-07-05 19:21:12)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml           (  23613 bytes, from 2014-07-17 15:33:30)
 
diff --git a/drivers/gpu/drm/msm/hdmi/hdmi_bridge.c b/drivers/gpu/drm/msm/hdmi/hdmi_bridge.c
index f6cf745c249e..6902ad6da710 100644
--- a/drivers/gpu/drm/msm/hdmi/hdmi_bridge.c
+++ b/drivers/gpu/drm/msm/hdmi/hdmi_bridge.c
@@ -26,7 +26,6 @@ struct hdmi_bridge {
 static void hdmi_bridge_destroy(struct drm_bridge *bridge)
 {
         struct hdmi_bridge *hdmi_bridge = to_hdmi_bridge(bridge);
-        hdmi_unreference(hdmi_bridge->hdmi);
         drm_bridge_cleanup(bridge);
         kfree(hdmi_bridge);
 }
@@ -218,7 +217,7 @@ struct drm_bridge *hdmi_bridge_init(struct hdmi *hdmi)
                 goto fail;
         }
 
-        hdmi_bridge->hdmi = hdmi_reference(hdmi);
+        hdmi_bridge->hdmi = hdmi;
 
         bridge = &hdmi_bridge->base;
 
diff --git a/drivers/gpu/drm/msm/hdmi/hdmi_connector.c b/drivers/gpu/drm/msm/hdmi/hdmi_connector.c
index 4aca2a3c667c..fbebb0405d76 100644
--- a/drivers/gpu/drm/msm/hdmi/hdmi_connector.c
+++ b/drivers/gpu/drm/msm/hdmi/hdmi_connector.c
@@ -330,8 +330,6 @@ static void hdmi_connector_destroy(struct drm_connector *connector)
         drm_connector_unregister(connector);
         drm_connector_cleanup(connector);
 
-        hdmi_unreference(hdmi_connector->hdmi);
-
         kfree(hdmi_connector);
 }
 
@@ -401,6 +399,9 @@ static const struct drm_connector_funcs hdmi_connector_funcs = {
         .detect = hdmi_connector_detect,
         .fill_modes = drm_helper_probe_single_connector_modes,
         .destroy = hdmi_connector_destroy,
+        .reset = drm_atomic_helper_connector_reset,
+        .atomic_duplicate_state = drm_atomic_helper_connector_duplicate_state,
+        .atomic_destroy_state = drm_atomic_helper_connector_destroy_state,
 };
 
 static const struct drm_connector_helper_funcs hdmi_connector_helper_funcs = {
@@ -422,7 +423,7 @@ struct drm_connector *hdmi_connector_init(struct hdmi *hdmi)
                 goto fail;
         }
 
-        hdmi_connector->hdmi = hdmi_reference(hdmi);
+        hdmi_connector->hdmi = hdmi;
         INIT_WORK(&hdmi_connector->hpd_work, hotplug_work);
 
         connector = &hdmi_connector->base;
diff --git a/drivers/gpu/drm/msm/hdmi/hdmi_phy_8960.c b/drivers/gpu/drm/msm/hdmi/hdmi_phy_8960.c
index f408b69486a8..eeed006eed13 100644
--- a/drivers/gpu/drm/msm/hdmi/hdmi_phy_8960.c
+++ b/drivers/gpu/drm/msm/hdmi/hdmi_phy_8960.c
@@ -510,7 +510,7 @@ struct hdmi_phy *hdmi_phy_8960_init(struct hdmi *hdmi)
 
 #ifdef CONFIG_COMMON_CLK
         phy_8960->pll_hw.init = &pll_init;
-        phy_8960->pll = devm_clk_register(hdmi->dev->dev, &phy_8960->pll_hw);
+        phy_8960->pll = devm_clk_register(&hdmi->pdev->dev, &phy_8960->pll_hw);
         if (IS_ERR(phy_8960->pll)) {
                 ret = PTR_ERR(phy_8960->pll);
                 phy_8960->pll = NULL;
diff --git a/drivers/gpu/drm/msm/hdmi/qfprom.xml.h b/drivers/gpu/drm/msm/hdmi/qfprom.xml.h
index d53c29327df9..29bd796797de 100644
--- a/drivers/gpu/drm/msm/hdmi/qfprom.xml.h
+++ b/drivers/gpu/drm/msm/hdmi/qfprom.xml.h
@@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git
 The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/msm.xml                 (    647 bytes, from 2013-11-30 14:45:35)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml (   1453 bytes, from 2013-03-31 16:51:27)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20457 bytes, from 2014-08-01 12:22:48)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1615 bytes, from 2014-07-17 15:34:33)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  22517 bytes, from 2014-07-17 15:34:33)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20136 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1940 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  23963 bytes, from 2014-10-31 16:51:46)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml             (  11712 bytes, from 2013-08-17 17:13:43)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml            (    344 bytes, from 2013-08-11 19:26:32)
-- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-08-01 12:23:53)
+- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-10-31 16:48:57)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml         (    600 bytes, from 2013-07-05 19:21:12)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml           (  23613 bytes, from 2014-07-17 15:33:30)
 
diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4.xml.h b/drivers/gpu/drm/msm/mdp/mdp4/mdp4.xml.h
index 03c0bd9cd5b9..a4a7f8c7122a 100644
--- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4.xml.h
+++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4.xml.h
@@ -10,12 +10,12 @@ git clone https://github.com/freedreno/envytools.git
 The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/msm.xml                 (    647 bytes, from 2013-11-30 14:45:35)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml (   1453 bytes, from 2013-03-31 16:51:27)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20457 bytes, from 2014-08-01 12:22:48)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1615 bytes, from 2014-07-17 15:34:33)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  22517 bytes, from 2014-07-17 15:34:33)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20136 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1940 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  23963 bytes, from 2014-10-31 16:51:46)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml             (  11712 bytes, from 2013-08-17 17:13:43)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml            (    344 bytes, from 2013-08-11 19:26:32)
-- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-08-01 12:23:53)
+- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-10-31 16:48:57)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml         (    600 bytes, from 2013-07-05 19:21:12)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml           (  23613 bytes, from 2014-07-17 15:33:30)
 
diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c
index 7d00f7fb5773..a7672e100d8b 100644
--- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c
+++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c
@@ -25,8 +25,6 @@
 struct mdp4_crtc {
         struct drm_crtc base;
         char name[8];
-        struct drm_plane *plane;
-        struct drm_plane *planes[8];
         int id;
         int ovlp;
         enum mdp4_dma dma;
@@ -52,25 +50,11 @@ struct mdp4_crtc {
 
         /* if there is a pending flip, these will be non-null: */
         struct drm_pending_vblank_event *event;
-        struct msm_fence_cb pageflip_cb;
 
 #define PENDING_CURSOR 0x1
 #define PENDING_FLIP   0x2
         atomic_t pending;
 
-        /* the fb that we logically (from PoV of KMS API) hold a ref
-         * to.  Which we may not yet be scanning out (we may still
-         * be scanning out previous in case of page_flip while waiting
-         * for gpu rendering to complete:
-         */
-        struct drm_framebuffer *fb;
-
-        /* the fb that we currently hold a scanout ref to: */
-        struct drm_framebuffer *scanout_fb;
-
-        /* for unref'ing framebuffers after scanout completes: */
-        struct drm_flip_work unref_fb_work;
-
         /* for unref'ing cursor bo's after scanout completes: */
         struct drm_flip_work unref_cursor_work;
 
@@ -97,15 +81,14 @@ static void crtc_flush(struct drm_crtc *crtc)
 {
         struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
         struct mdp4_kms *mdp4_kms = get_kms(crtc);
-        uint32_t i, flush = 0;
+        struct drm_plane *plane;
+        uint32_t flush = 0;
 
-        for (i = 0; i < ARRAY_SIZE(mdp4_crtc->planes); i++) {
-                struct drm_plane *plane = mdp4_crtc->planes[i];
-                if (plane) {
-                        enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane);
-                        flush |= pipe2flush(pipe_id);
-                }
+        drm_atomic_crtc_for_each_plane(plane, crtc) {
+                enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane);
+                flush |= pipe2flush(pipe_id);
         }
+
         flush |= ovlp2flush(mdp4_crtc->ovlp);
 
         DBG("%s: flush=%08x", mdp4_crtc->name, flush);
@@ -113,47 +96,6 @@ static void crtc_flush(struct drm_crtc *crtc)
         mdp4_write(mdp4_kms, REG_MDP4_OVERLAY_FLUSH, flush);
 }
 
-static void update_fb(struct drm_crtc *crtc, struct drm_framebuffer *new_fb)
-{
-        struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
-        struct drm_framebuffer *old_fb = mdp4_crtc->fb;
-
-        /* grab reference to incoming scanout fb: */
-        drm_framebuffer_reference(new_fb);
-        mdp4_crtc->base.primary->fb = new_fb;
-        mdp4_crtc->fb = new_fb;
-
-        if (old_fb)
-                drm_flip_work_queue(&mdp4_crtc->unref_fb_work, old_fb);
-}
-
-/* unlike update_fb(), take a ref to the new scanout fb *before* updating
- * plane, then call this.  Needed to ensure we don't unref the buffer that
- * is actually still being scanned out.
- *
- * Note that this whole thing goes away with atomic.. since we can defer
- * calling into driver until rendering is done.
- */
-static void update_scanout(struct drm_crtc *crtc, struct drm_framebuffer *fb)
-{
-        struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
-
-        /* flush updates, to make sure hw is updated to new scanout fb,
-         * so that we can safely queue unref to current fb (ie. next
-         * vblank we know hw is done w/ previous scanout_fb).
-         */
-        crtc_flush(crtc);
-
-        if (mdp4_crtc->scanout_fb)
-                drm_flip_work_queue(&mdp4_crtc->unref_fb_work,
-                                mdp4_crtc->scanout_fb);
-
-        mdp4_crtc->scanout_fb = fb;
-
-        /* enable vblank to complete flip: */
-        request_pending(crtc, PENDING_FLIP);
-}
-
 /* if file!=NULL, this is preclose potential cancel-flip path */
 static void complete_flip(struct drm_crtc *crtc, struct drm_file *file)
 {
@@ -171,38 +113,13 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file)
                  */
                 if (!file || (event->base.file_priv == file)) {
                         mdp4_crtc->event = NULL;
+                        DBG("%s: send event: %p", mdp4_crtc->name, event);
                         drm_send_vblank_event(dev, mdp4_crtc->id, event);
                 }
         }
         spin_unlock_irqrestore(&dev->event_lock, flags);
 }
 
-static void pageflip_cb(struct msm_fence_cb *cb)
-{
-        struct mdp4_crtc *mdp4_crtc =
-                container_of(cb, struct mdp4_crtc, pageflip_cb);
-        struct drm_crtc *crtc = &mdp4_crtc->base;
-        struct drm_framebuffer *fb = crtc->primary->fb;
-
-        if (!fb)
-                return;
-
-        drm_framebuffer_reference(fb);
-        mdp4_plane_set_scanout(mdp4_crtc->plane, fb);
-        update_scanout(crtc, fb);
-}
-
-static void unref_fb_worker(struct drm_flip_work *work, void *val)
-{
-        struct mdp4_crtc *mdp4_crtc =
-                container_of(work, struct mdp4_crtc, unref_fb_work);
-        struct drm_device *dev = mdp4_crtc->base.dev;
-
-        mutex_lock(&dev->mode_config.mutex);
-        drm_framebuffer_unreference(val);
-        mutex_unlock(&dev->mode_config.mutex);
-}
-
 static void unref_cursor_worker(struct drm_flip_work *work, void *val)
 {
         struct mdp4_crtc *mdp4_crtc =
@@ -218,7 +135,6 @@ static void mdp4_crtc_destroy(struct drm_crtc *crtc)
         struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
 
         drm_crtc_cleanup(crtc);
-        drm_flip_work_cleanup(&mdp4_crtc->unref_fb_work);
         drm_flip_work_cleanup(&mdp4_crtc->unref_cursor_work);
 
         kfree(mdp4_crtc);
@@ -251,57 +167,70 @@ static bool mdp4_crtc_mode_fixup(struct drm_crtc *crtc,
         return true;
 }
 
-static void blend_setup(struct drm_crtc *crtc)
+/* statically (for now) map planes to mixer stage (z-order): */
+static const int idxs[] = {
+                [VG1]  = 1,
+                [VG2]  = 2,
+                [RGB1] = 0,
+                [RGB2] = 0,
+                [RGB3] = 0,
+                [VG3]  = 3,
+                [VG4]  = 4,
+
+};
+
+/* setup mixer config, for which we need to consider all crtc's and
+ * the planes attached to them
+ *
+ * TODO may possibly need some extra locking here
+ */
+static void setup_mixer(struct mdp4_kms *mdp4_kms)
 {
-        struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
-        struct mdp4_kms *mdp4_kms = get_kms(crtc);
-        int i, ovlp = mdp4_crtc->ovlp;
+        struct drm_mode_config *config = &mdp4_kms->dev->mode_config;
+        struct drm_crtc *crtc;
         uint32_t mixer_cfg = 0;
         static const enum mdp_mixer_stage_id stages[] = {
                         STAGE_BASE, STAGE0, STAGE1, STAGE2, STAGE3,
         };
-        /* statically (for now) map planes to mixer stage (z-order): */
-        static const int idxs[] = {
-                        [VG1]  = 1,
-                        [VG2]  = 2,
-                        [RGB1] = 0,
-                        [RGB2] = 0,
-                        [RGB3] = 0,
-                        [VG3]  = 3,
-                        [VG4]  = 4,
 
-        };
-        bool alpha[4]= { false, false, false, false };
+        list_for_each_entry(crtc, &config->crtc_list, head) {
+                struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
+                struct drm_plane *plane;
 
-        /* Don't rely on value read back from hw, but instead use our
-         * own shadowed value.  Possibly disable/reenable looses the
-         * previous value and goes back to power-on default?
-         */
-        mixer_cfg = mdp4_kms->mixer_cfg;
+                drm_atomic_crtc_for_each_plane(plane, crtc) {
+                        enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane);
+                        int idx = idxs[pipe_id];
+                        mixer_cfg = mixercfg(mixer_cfg, mdp4_crtc->mixer,
+                                        pipe_id, stages[idx]);
+                }
+        }
+
+        mdp4_write(mdp4_kms, REG_MDP4_LAYERMIXER_IN_CFG, mixer_cfg);
+}
+
+static void blend_setup(struct drm_crtc *crtc)
+{
+        struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
+        struct mdp4_kms *mdp4_kms = get_kms(crtc);
+        struct drm_plane *plane;
+        int i, ovlp = mdp4_crtc->ovlp;
+        bool alpha[4]= { false, false, false, false };
 
         mdp4_write(mdp4_kms, REG_MDP4_OVLP_TRANSP_LOW0(ovlp), 0);
         mdp4_write(mdp4_kms, REG_MDP4_OVLP_TRANSP_LOW1(ovlp), 0);
         mdp4_write(mdp4_kms, REG_MDP4_OVLP_TRANSP_HIGH0(ovlp), 0);
         mdp4_write(mdp4_kms, REG_MDP4_OVLP_TRANSP_HIGH1(ovlp), 0);
 
-        for (i = 0; i < ARRAY_SIZE(mdp4_crtc->planes); i++) {
-                struct drm_plane *plane = mdp4_crtc->planes[i];
-                if (plane) {
-                        enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane);
-                        int idx = idxs[pipe_id];
-                        if (idx > 0) {
-                                const struct mdp_format *format =
+        drm_atomic_crtc_for_each_plane(plane, crtc) {
+                enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane);
+                int idx = idxs[pipe_id];
+                if (idx > 0) {
+                        const struct mdp_format *format =
                                         to_mdp_format(msm_framebuffer_format(plane->fb));
-                                alpha[idx-1] = format->alpha_enable;
-                        }
-                        mixer_cfg = mixercfg(mixer_cfg, mdp4_crtc->mixer,
-                                        pipe_id, stages[idx]);
+                        alpha[idx-1] = format->alpha_enable;
                 }
         }
 
-        /* this shouldn't happen.. and seems to cause underflow: */
-        WARN_ON(!mixer_cfg);
-
         for (i = 0; i < 4; i++) {
                 uint32_t op;
 
@@ -324,22 +253,21 @@ static void blend_setup(struct drm_crtc *crtc)
                 mdp4_write(mdp4_kms, REG_MDP4_OVLP_STAGE_TRANSP_HIGH1(ovlp, i), 0);
         }
 
-        mdp4_kms->mixer_cfg = mixer_cfg;
-        mdp4_write(mdp4_kms, REG_MDP4_LAYERMIXER_IN_CFG, mixer_cfg);
+        setup_mixer(mdp4_kms);
 }
 
-static int mdp4_crtc_mode_set(struct drm_crtc *crtc,
-                struct drm_display_mode *mode,
-                struct drm_display_mode *adjusted_mode,
-                int x, int y,
-                struct drm_framebuffer *old_fb)
+static void mdp4_crtc_mode_set_nofb(struct drm_crtc *crtc)
 {
         struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
         struct mdp4_kms *mdp4_kms = get_kms(crtc);
         enum mdp4_dma dma = mdp4_crtc->dma;
-        int ret, ovlp = mdp4_crtc->ovlp;
+        int ovlp = mdp4_crtc->ovlp;
+        struct drm_display_mode *mode;
+
+        if (WARN_ON(!crtc->state))
+                return;
 
-        mode = adjusted_mode;
+        mode = &crtc->state->adjusted_mode;
 
         DBG("%s: set mode: %d:\"%s\" %d %d %d %d %d %d %d %d %d %d 0x%x 0x%x",
                         mdp4_crtc->name, mode->base.id, mode->name,
@@ -350,28 +278,13 @@ static int mdp4_crtc_mode_set(struct drm_crtc *crtc,
                         mode->vsync_end, mode->vtotal,
                         mode->type, mode->flags);
 
-        /* grab extra ref for update_scanout() */
-        drm_framebuffer_reference(crtc->primary->fb);
-
-        ret = mdp4_plane_mode_set(mdp4_crtc->plane, crtc, crtc->primary->fb,
-                        0, 0, mode->hdisplay, mode->vdisplay,
-                        x << 16, y << 16,
-                        mode->hdisplay << 16, mode->vdisplay << 16);
-        if (ret) {
-                drm_framebuffer_unreference(crtc->primary->fb);
-                dev_err(crtc->dev->dev, "%s: failed to set mode on plane: %d\n",
-                                mdp4_crtc->name, ret);
-                return ret;
-        }
-
         mdp4_write(mdp4_kms, REG_MDP4_DMA_SRC_SIZE(dma),
                         MDP4_DMA_SRC_SIZE_WIDTH(mode->hdisplay) |
                         MDP4_DMA_SRC_SIZE_HEIGHT(mode->vdisplay));
 
         /* take data from pipe: */
         mdp4_write(mdp4_kms, REG_MDP4_DMA_SRC_BASE(dma), 0);
-        mdp4_write(mdp4_kms, REG_MDP4_DMA_SRC_STRIDE(dma),
-                        crtc->primary->fb->pitches[0]);
+        mdp4_write(mdp4_kms, REG_MDP4_DMA_SRC_STRIDE(dma), 0);
         mdp4_write(mdp4_kms, REG_MDP4_DMA_DST_SIZE(dma),
                         MDP4_DMA_DST_SIZE_WIDTH(0) |
                         MDP4_DMA_DST_SIZE_HEIGHT(0));
@@ -380,8 +293,7 @@ static int mdp4_crtc_mode_set(struct drm_crtc *crtc,
         mdp4_write(mdp4_kms, REG_MDP4_OVLP_SIZE(ovlp),
                         MDP4_OVLP_SIZE_WIDTH(mode->hdisplay) |
                         MDP4_OVLP_SIZE_HEIGHT(mode->vdisplay));
-        mdp4_write(mdp4_kms, REG_MDP4_OVLP_STRIDE(ovlp),
-                        crtc->primary->fb->pitches[0]);
+        mdp4_write(mdp4_kms, REG_MDP4_OVLP_STRIDE(ovlp), 0);
 
         mdp4_write(mdp4_kms, REG_MDP4_OVLP_CFG(ovlp), 1);
 
@@ -390,11 +302,6 @@ static int mdp4_crtc_mode_set(struct drm_crtc *crtc,
                 mdp4_write(mdp4_kms, REG_MDP4_DMA_E_QUANT(1), 0x00ff0000);
                 mdp4_write(mdp4_kms, REG_MDP4_DMA_E_QUANT(2), 0x00ff0000);
         }
-
-        update_fb(crtc, crtc->primary->fb);
-        update_scanout(crtc, crtc->primary->fb);
-
-        return 0;
 }
 
 static void mdp4_crtc_prepare(struct drm_crtc *crtc)
@@ -416,60 +323,51 @@ static void mdp4_crtc_commit(struct drm_crtc *crtc)
         drm_crtc_vblank_put(crtc);
 }
 
-static int mdp4_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y,
-                struct drm_framebuffer *old_fb)
+static void mdp4_crtc_load_lut(struct drm_crtc *crtc)
+{
+}
+
+static int mdp4_crtc_atomic_check(struct drm_crtc *crtc,
+                struct drm_crtc_state *state)
 {
         struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
-        struct drm_plane *plane = mdp4_crtc->plane;
-        struct drm_display_mode *mode = &crtc->mode;
-        int ret;
+        struct drm_device *dev = crtc->dev;
 
-        /* grab extra ref for update_scanout() */
-        drm_framebuffer_reference(crtc->primary->fb);
+        DBG("%s: check", mdp4_crtc->name);
 
-        ret = mdp4_plane_mode_set(plane, crtc, crtc->primary->fb,
-                        0, 0, mode->hdisplay, mode->vdisplay,
-                        x << 16, y << 16,
-                        mode->hdisplay << 16, mode->vdisplay << 16);
-        if (ret) {
-                drm_framebuffer_unreference(crtc->primary->fb);
-                return ret;
+        if (mdp4_crtc->event) {
+                dev_err(dev->dev, "already pending flip!\n");
+                return -EBUSY;
         }
 
-        update_fb(crtc, crtc->primary->fb);
-        update_scanout(crtc, crtc->primary->fb);
+        // TODO anything else to check?
 
         return 0;
 }
 
-static void mdp4_crtc_load_lut(struct drm_crtc *crtc)
+static void mdp4_crtc_atomic_begin(struct drm_crtc *crtc)
 {
+        struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
+        DBG("%s: begin", mdp4_crtc->name);
 }
 
-static int mdp4_crtc_page_flip(struct drm_crtc *crtc,
-                struct drm_framebuffer *new_fb,
-                struct drm_pending_vblank_event *event,
-                uint32_t page_flip_flags)
+static void mdp4_crtc_atomic_flush(struct drm_crtc *crtc)
 {
         struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
         struct drm_device *dev = crtc->dev;
-        struct drm_gem_object *obj;
         unsigned long flags;
 
-        if (mdp4_crtc->event) {
-                dev_err(dev->dev, "already pending flip!\n");
-                return -EBUSY;
-        }
+        DBG("%s: flush", mdp4_crtc->name);
 
-        obj = msm_framebuffer_bo(new_fb, 0);
+        WARN_ON(mdp4_crtc->event);
 
         spin_lock_irqsave(&dev->event_lock, flags);
-        mdp4_crtc->event = event;
+        mdp4_crtc->event = crtc->state->event;
         spin_unlock_irqrestore(&dev->event_lock, flags);
 
-        update_fb(crtc, new_fb);
-
-        return msm_gem_queue_inactive_cb(obj, &mdp4_crtc->pageflip_cb);
+        blend_setup(crtc);
+        crtc_flush(crtc);
+        request_pending(crtc, PENDING_FLIP);
 }
 
 static int mdp4_crtc_set_property(struct drm_crtc *crtc,
@@ -607,22 +505,29 @@ static int mdp4_crtc_cursor_move(struct drm_crtc *crtc, int x, int y)
 }
 
 static const struct drm_crtc_funcs mdp4_crtc_funcs = {
-        .set_config = drm_crtc_helper_set_config,
+        .set_config = drm_atomic_helper_set_config,
         .destroy = mdp4_crtc_destroy,
-        .page_flip = mdp4_crtc_page_flip,
+        .page_flip = drm_atomic_helper_page_flip,
         .set_property = mdp4_crtc_set_property,
         .cursor_set = mdp4_crtc_cursor_set,
         .cursor_move = mdp4_crtc_cursor_move,
+        .reset = drm_atomic_helper_crtc_reset,
+        .atomic_duplicate_state = drm_atomic_helper_crtc_duplicate_state,
+        .atomic_destroy_state = drm_atomic_helper_crtc_destroy_state,
 };
 
 static const struct drm_crtc_helper_funcs mdp4_crtc_helper_funcs = {
         .dpms = mdp4_crtc_dpms,
         .mode_fixup = mdp4_crtc_mode_fixup,
-        .mode_set = mdp4_crtc_mode_set,
+        .mode_set_nofb = mdp4_crtc_mode_set_nofb,
+        .mode_set = drm_helper_crtc_mode_set,
+        .mode_set_base = drm_helper_crtc_mode_set_base,
         .prepare = mdp4_crtc_prepare,
         .commit = mdp4_crtc_commit,
-        .mode_set_base = mdp4_crtc_mode_set_base,
         .load_lut = mdp4_crtc_load_lut,
+        .atomic_check = mdp4_crtc_atomic_check,
+        .atomic_begin = mdp4_crtc_atomic_begin,
+        .atomic_flush = mdp4_crtc_atomic_flush,
 };
 
 static void mdp4_crtc_vblank_irq(struct mdp_irq *irq, uint32_t irqstatus)
@@ -638,7 +543,6 @@ static void mdp4_crtc_vblank_irq(struct mdp_irq *irq, uint32_t irqstatus)
 
         if (pending & PENDING_FLIP) {
                 complete_flip(crtc, NULL);
-                drm_flip_work_commit(&mdp4_crtc->unref_fb_work, priv->wq);
         }
 
         if (pending & PENDING_CURSOR) {
@@ -663,7 +567,8 @@ uint32_t mdp4_crtc_vblank(struct drm_crtc *crtc)
 
 void mdp4_crtc_cancel_pending_flip(struct drm_crtc *crtc, struct drm_file *file)
 {
-        DBG("cancel: %p", file);
+        struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
+        DBG("%s: cancel: %p", mdp4_crtc->name, file);
         complete_flip(crtc, file);
 }
 
@@ -717,35 +622,6 @@ void mdp4_crtc_set_intf(struct drm_crtc *crtc, enum mdp4_intf intf, int mixer)
         mdp4_write(mdp4_kms, REG_MDP4_DISP_INTF_SEL, intf_sel);
 }
 
-static void set_attach(struct drm_crtc *crtc, enum mdp4_pipe pipe_id,
-                struct drm_plane *plane)
-{
-        struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
-
-        BUG_ON(pipe_id >= ARRAY_SIZE(mdp4_crtc->planes));
-
-        if (mdp4_crtc->planes[pipe_id] == plane)
-                return;
-
-        mdp4_crtc->planes[pipe_id] = plane;
-        blend_setup(crtc);
-        if (mdp4_crtc->enabled && (plane != mdp4_crtc->plane))
-                crtc_flush(crtc);
-}
-
-void mdp4_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane)
-{
-        set_attach(crtc, mdp4_plane_pipe(plane), plane);
-}
-
-void mdp4_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane)
-{
-        /* don't actually detatch our primary plane: */
-        if (to_mdp4_crtc(crtc)->plane == plane)
-                return;
-        set_attach(crtc, mdp4_plane_pipe(plane), NULL);
-}
-
 static const char *dma_names[] = {
                 "DMA_P", "DMA_S", "DMA_E",
 };
@@ -757,17 +633,13 @@ struct drm_crtc *mdp4_crtc_init(struct drm_device *dev,
 {
         struct drm_crtc *crtc = NULL;
         struct mdp4_crtc *mdp4_crtc;
-        int ret;
 
         mdp4_crtc = kzalloc(sizeof(*mdp4_crtc), GFP_KERNEL);
-        if (!mdp4_crtc) {
-                ret = -ENOMEM;
-                goto fail;
-        }
+        if (!mdp4_crtc)
+                return ERR_PTR(-ENOMEM);
 
         crtc = &mdp4_crtc->base;
 
-        mdp4_crtc->plane = plane;
         mdp4_crtc->id = id;
 
         mdp4_crtc->ovlp = ovlp_id;
@@ -784,26 +656,14 @@ struct drm_crtc *mdp4_crtc_init(struct drm_device *dev,
 
         spin_lock_init(&mdp4_crtc->cursor.lock);
 
-        ret = drm_flip_work_init(&mdp4_crtc->unref_fb_work, 16,
-                        "unref fb", unref_fb_worker);
-        if (ret)
-                goto fail;
-
-        ret = drm_flip_work_init(&mdp4_crtc->unref_cursor_work, 64,
+        drm_flip_work_init(&mdp4_crtc->unref_cursor_work,
                         "unref cursor", unref_cursor_worker);
 
-        INIT_FENCE_CB(&mdp4_crtc->pageflip_cb, pageflip_cb);
-
         drm_crtc_init_with_planes(dev, crtc, plane, NULL, &mdp4_crtc_funcs);
         drm_crtc_helper_add(crtc, &mdp4_crtc_helper_funcs);
+        plane->crtc = crtc;
 
-        mdp4_plane_install_properties(mdp4_crtc->plane, &crtc->base);
+        mdp4_plane_install_properties(plane, &crtc->base);
 
         return crtc;
-
-fail:
-        if (crtc)
-                mdp4_crtc_destroy(crtc);
-
-        return ERR_PTR(ret);
 }
diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c
index 79d804e61cc4..a62109e4ae0d 100644
--- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c
+++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c
@@ -228,7 +228,6 @@ static int modeset_init(struct mdp4_kms *mdp4_kms)
         struct drm_encoder *encoder;
         struct drm_connector *connector;
         struct drm_panel *panel;
-        struct hdmi *hdmi;
         int ret;
 
         /* construct non-private planes: */
@@ -326,11 +325,13 @@ static int modeset_init(struct mdp4_kms *mdp4_kms)
         priv->crtcs[priv->num_crtcs++] = crtc;
         priv->encoders[priv->num_encoders++] = encoder;
 
-        hdmi = hdmi_init(dev, encoder);
-        if (IS_ERR(hdmi)) {
-                ret = PTR_ERR(hdmi);
-                dev_err(dev->dev, "failed to initialize HDMI: %d\n", ret);
-                goto fail;
+        if (priv->hdmi) {
+                /* Construct bridge/connector for HDMI: */
+                ret = hdmi_modeset_init(priv->hdmi, dev, encoder);
+                if (ret) {
+                        dev_err(dev->dev, "failed to initialize HDMI: %d\n", ret);
+                        goto fail;
+                }
         }
 
         return 0;
@@ -381,6 +382,10 @@ struct msm_kms *mdp4_kms_init(struct drm_device *dev)
         if (IS_ERR(mdp4_kms->dsi_pll_vddio))
                 mdp4_kms->dsi_pll_vddio = NULL;
 
+        /* NOTE: driver for this regulator still missing upstream.. use
+         * _get_exclusive() and ignore the error if it does not exist
+         * (and hope that the bootloader left it on for us)
+         */
         mdp4_kms->vdd = devm_regulator_get_exclusive(&pdev->dev, "vdd");
         if (IS_ERR(mdp4_kms->vdd))
                 mdp4_kms->vdd = NULL;
diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h
index 9ff6e7ccfe90..cbd77bc626d5 100644
--- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h
+++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.h
@@ -32,13 +32,6 @@ struct mdp4_kms {
 
         int rev;
 
-        /* Shadow value for MDP4_LAYERMIXER_IN_CFG.. since setup for all
-         * crtcs/encoders is in one shared register, we need to update it
-         * via read/modify/write.  But to avoid getting confused by power-
-         * on-default values after resume, use this shadow value instead:
-         */
-        uint32_t mixer_cfg;
-
         /* mapper-id used to request GEM buffer mapped for scanout: */
         int id;
 
@@ -194,14 +187,6 @@ uint32_t mdp4_get_formats(enum mdp4_pipe pipe_id, uint32_t *pixel_formats,
 
 void mdp4_plane_install_properties(struct drm_plane *plane,
                 struct drm_mode_object *obj);
-void mdp4_plane_set_scanout(struct drm_plane *plane,
-                struct drm_framebuffer *fb);
-int mdp4_plane_mode_set(struct drm_plane *plane,
-                struct drm_crtc *crtc, struct drm_framebuffer *fb,
-                int crtc_x, int crtc_y,
-                unsigned int crtc_w, unsigned int crtc_h,
-                uint32_t src_x, uint32_t src_y,
-                uint32_t src_w, uint32_t src_h);
 enum mdp4_pipe mdp4_plane_pipe(struct drm_plane *plane);
 struct drm_plane *mdp4_plane_init(struct drm_device *dev,
                 enum mdp4_pipe pipe_id, bool private_plane);
@@ -210,8 +195,6 @@ uint32_t mdp4_crtc_vblank(struct drm_crtc *crtc);
 void mdp4_crtc_cancel_pending_flip(struct drm_crtc *crtc, struct drm_file *file);
 void mdp4_crtc_set_config(struct drm_crtc *crtc, uint32_t config);
 void mdp4_crtc_set_intf(struct drm_crtc *crtc, enum mdp4_intf intf, int mixer);
-void mdp4_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane);
-void mdp4_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane);
 struct drm_crtc *mdp4_crtc_init(struct drm_device *dev,
                 struct drm_plane *plane, int id, int ovlp_id,
                 enum mdp4_dma dma_id);
diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_lvds_connector.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_lvds_connector.c
index 310034688c15..4ddc28e1275b 100644
--- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_lvds_connector.c
+++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_lvds_connector.c
@@ -98,6 +98,9 @@ static const struct drm_connector_funcs mdp4_lvds_connector_funcs = {
         .detect = mdp4_lvds_connector_detect,
         .fill_modes = drm_helper_probe_single_connector_modes,
         .destroy = mdp4_lvds_connector_destroy,
+        .reset = drm_atomic_helper_connector_reset,
+        .atomic_duplicate_state = drm_atomic_helper_connector_duplicate_state,
+        .atomic_destroy_state = drm_atomic_helper_connector_destroy_state,
 };
 
 static const struct drm_connector_helper_funcs mdp4_lvds_connector_helper_funcs = {
diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c
index 66f33dba1ebb..1e5ebe83647d 100644
--- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c
+++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c
@@ -31,47 +31,26 @@ struct mdp4_plane {
 };
 #define to_mdp4_plane(x) container_of(x, struct mdp4_plane, base)
 
-static struct mdp4_kms *get_kms(struct drm_plane *plane)
-{
-        struct msm_drm_private *priv = plane->dev->dev_private;
-        return to_mdp4_kms(to_mdp_kms(priv->kms));
-}
-
-static int mdp4_plane_update(struct drm_plane *plane,
+static void mdp4_plane_set_scanout(struct drm_plane *plane,
+                struct drm_framebuffer *fb);
+static int mdp4_plane_mode_set(struct drm_plane *plane,
                 struct drm_crtc *crtc, struct drm_framebuffer *fb,
                 int crtc_x, int crtc_y,
                 unsigned int crtc_w, unsigned int crtc_h,
                 uint32_t src_x, uint32_t src_y,
-                uint32_t src_w, uint32_t src_h)
-{
-        struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane);
-
-        mdp4_plane->enabled = true;
-
-        if (plane->fb)
-                drm_framebuffer_unreference(plane->fb);
-
-        drm_framebuffer_reference(fb);
-
-        return mdp4_plane_mode_set(plane, crtc, fb,
-                        crtc_x, crtc_y, crtc_w, crtc_h,
-                        src_x, src_y, src_w, src_h);
-}
+                uint32_t src_w, uint32_t src_h);
 
-static int mdp4_plane_disable(struct drm_plane *plane)
+static struct mdp4_kms *get_kms(struct drm_plane *plane)
 {
-        struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane);
-        DBG("%s: disable", mdp4_plane->name);
-        if (plane->crtc)
-                mdp4_crtc_detach(plane->crtc, plane);
-        return 0;
+        struct msm_drm_private *priv = plane->dev->dev_private;
+        return to_mdp4_kms(to_mdp_kms(priv->kms));
 }
 
 static void mdp4_plane_destroy(struct drm_plane *plane)
 {
         struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane);
 
-        mdp4_plane_disable(plane);
+        drm_plane_helper_disable(plane);
         drm_plane_cleanup(plane);
 
         kfree(mdp4_plane);
@@ -92,19 +71,75 @@ int mdp4_plane_set_property(struct drm_plane *plane,
 }
 
 static const struct drm_plane_funcs mdp4_plane_funcs = {
-                .update_plane = mdp4_plane_update,
-                .disable_plane = mdp4_plane_disable,
+                .update_plane = drm_atomic_helper_update_plane,
+                .disable_plane = drm_atomic_helper_disable_plane,
                 .destroy = mdp4_plane_destroy,
                 .set_property = mdp4_plane_set_property,
+                .reset = drm_atomic_helper_plane_reset,
+                .atomic_duplicate_state = drm_atomic_helper_plane_duplicate_state,
+                .atomic_destroy_state = drm_atomic_helper_plane_destroy_state,
 };
 
-void mdp4_plane_set_scanout(struct drm_plane *plane,
+static int mdp4_plane_prepare_fb(struct drm_plane *plane,
+                struct drm_framebuffer *fb)
+{
+        struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane);
+        struct mdp4_kms *mdp4_kms = get_kms(plane);
+
+        DBG("%s: prepare: FB[%u]", mdp4_plane->name, fb->base.id);
+        return msm_framebuffer_prepare(fb, mdp4_kms->id);
+}
+
+static void mdp4_plane_cleanup_fb(struct drm_plane *plane,
+                struct drm_framebuffer *fb)
+{
+        struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane);
+        struct mdp4_kms *mdp4_kms = get_kms(plane);
+
+        DBG("%s: cleanup: FB[%u]", mdp4_plane->name, fb->base.id);
+        msm_framebuffer_cleanup(fb, mdp4_kms->id);
+}
+
+
+static int mdp4_plane_atomic_check(struct drm_plane *plane,
+                struct drm_plane_state *state)
+{
+        return 0;
+}
+
+static void mdp4_plane_atomic_update(struct drm_plane *plane,
+                                     struct drm_plane_state *old_state)
+{
+        struct drm_plane_state *state = plane->state;
+        int ret;
+
+        ret = mdp4_plane_mode_set(plane,
+                        state->crtc, state->fb,
+                        state->crtc_x, state->crtc_y,
+                        state->crtc_w, state->crtc_h,
+                        state->src_x,  state->src_y,
+                        state->src_w, state->src_h);
+        /* atomic_check should have ensured that this doesn't fail */
+        WARN_ON(ret < 0);
+}
+
+static const struct drm_plane_helper_funcs mdp4_plane_helper_funcs = {
+                .prepare_fb = mdp4_plane_prepare_fb,
+                .cleanup_fb = mdp4_plane_cleanup_fb,
+                .atomic_check = mdp4_plane_atomic_check,
+                .atomic_update = mdp4_plane_atomic_update,
+};
+
+static void mdp4_plane_set_scanout(struct drm_plane *plane,
                 struct drm_framebuffer *fb)
 {
         struct mdp4_plane *mdp4_plane = to_mdp4_plane(plane);
         struct mdp4_kms *mdp4_kms = get_kms(plane);
         enum mdp4_pipe pipe = mdp4_plane->pipe;
-        uint32_t iova;
+        uint32_t iova = msm_framebuffer_iova(fb, mdp4_kms->id, 0);
+
+        DBG("%s: set_scanout: %08x (%u)", mdp4_plane->name,
+                        iova, fb->pitches[0]);
 
         mdp4_write(mdp4_kms, REG_MDP4_PIPE_SRC_STRIDE_A(pipe),
                         MDP4_PIPE_SRC_STRIDE_A_P0(fb->pitches[0]) |
@@ -114,7 +149,6 @@ void mdp4_plane_set_scanout(struct drm_plane *plane,
                         MDP4_PIPE_SRC_STRIDE_B_P2(fb->pitches[2]) |
                         MDP4_PIPE_SRC_STRIDE_B_P3(fb->pitches[3]));
 
-        msm_gem_get_iova(msm_framebuffer_bo(fb, 0), mdp4_kms->id, &iova);
         mdp4_write(mdp4_kms, REG_MDP4_PIPE_SRCP0_BASE(pipe), iova);
 
         plane->fb = fb;
@@ -122,7 +156,7 @@ void mdp4_plane_set_scanout(struct drm_plane *plane,
 
 #define MDP4_VG_PHASE_STEP_DEFAULT      0x20000000
 
-int mdp4_plane_mode_set(struct drm_plane *plane,
+static int mdp4_plane_mode_set(struct drm_plane *plane,
                 struct drm_crtc *crtc, struct drm_framebuffer *fb,
                 int crtc_x, int crtc_y,
                 unsigned int crtc_w, unsigned int crtc_h,
@@ -137,6 +171,11 @@ int mdp4_plane_mode_set(struct drm_plane *plane,
         uint32_t phasex_step = MDP4_VG_PHASE_STEP_DEFAULT;
         uint32_t phasey_step = MDP4_VG_PHASE_STEP_DEFAULT;
 
+        if (!(crtc && fb)) {
+                DBG("%s: disabled!", mdp4_plane->name);
+                return 0;
+        }
+
         /* src values are in Q16 fixed point, convert to integer: */
         src_x = src_x >> 16;
         src_y = src_y >> 16;
@@ -197,9 +236,6 @@ int mdp4_plane_mode_set(struct drm_plane *plane,
         mdp4_write(mdp4_kms, REG_MDP4_PIPE_PHASEX_STEP(pipe), phasex_step);
         mdp4_write(mdp4_kms, REG_MDP4_PIPE_PHASEY_STEP(pipe), phasey_step);
 
-        /* TODO detach from old crtc (if we had more than one) */
-        mdp4_crtc_attach(crtc, plane);
-
         return 0;
 }
 
@@ -239,9 +275,12 @@ struct drm_plane *mdp4_plane_init(struct drm_device *dev,
                         ARRAY_SIZE(mdp4_plane->formats));
 
         type = private_plane ? DRM_PLANE_TYPE_PRIMARY : DRM_PLANE_TYPE_OVERLAY;
-        drm_universal_plane_init(dev, plane, 0xff, &mdp4_plane_funcs,
-                                 mdp4_plane->formats, mdp4_plane->nformats,
-                                 type);
+        ret = drm_universal_plane_init(dev, plane, 0xff, &mdp4_plane_funcs,
+                                 mdp4_plane->formats, mdp4_plane->nformats, type);
+        if (ret)
+                goto fail;
+
+        drm_plane_helper_add(plane, &mdp4_plane_helper_funcs);
 
         mdp4_plane_install_properties(plane, &plane->base);
 
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5.xml.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5.xml.h
index 67f4f896ba8c..e87ef5512cb0 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5.xml.h
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5.xml.h
@@ -10,14 +10,14 @@ git clone https://github.com/freedreno/envytools.git
 The rules-ng-ng source files this header was generated from are:
 - /home/robclark/src/freedreno/envytools/rnndb/msm.xml                 (    647 bytes, from 2013-11-30 14:45:35)
 - /home/robclark/src/freedreno/envytools/rnndb/freedreno_copyright.xml (   1453 bytes, from 2013-03-31 16:51:27)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  17996 bytes, from 2013-12-01 19:10:31)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1615 bytes, from 2013-11-30 15:00:52)
-- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  22517 bytes, from 2014-06-25 12:55:02)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp4.xml            (  20136 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp_common.xml      (   1940 bytes, from 2014-10-31 16:51:39)
+- /home/robclark/src/freedreno/envytools/rnndb/mdp/mdp5.xml            (  23963 bytes, from 2014-10-31 16:51:46)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/dsi.xml             (  11712 bytes, from 2013-08-17 17:13:43)
 - /home/robclark/src/freedreno/envytools/rnndb/dsi/sfpb.xml            (    344 bytes, from 2013-08-11 19:26:32)
-- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1544 bytes, from 2013-08-16 19:17:05)
+- /home/robclark/src/freedreno/envytools/rnndb/dsi/mmss_cc.xml         (   1686 bytes, from 2014-10-31 16:48:57)
 - /home/robclark/src/freedreno/envytools/rnndb/hdmi/qfprom.xml         (    600 bytes, from 2013-07-05 19:21:12)
-- /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml           (  23613 bytes, from 2014-06-25 12:53:44)
+- /home/robclark/src/freedreno/envytools/rnndb/hdmi/hdmi.xml           (  23613 bytes, from 2014-07-17 15:33:30)
 
 Copyright (C) 2013-2014 by the following authors:
 - Rob Clark <robdclark@gmail.com> (robclark)
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c
new file mode 100644
index 000000000000..b0a44310cf2a
--- /dev/null
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.c
@@ -0,0 +1,207 @@
+/*
+ * Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include "mdp5_kms.h"
+#include "mdp5_cfg.h"
+
+struct mdp5_cfg_handler {
+        int revision;
+        struct mdp5_cfg config;
+};
+
+/* mdp5_cfg must be exposed (used in mdp5.xml.h) */
+const struct mdp5_cfg_hw *mdp5_cfg = NULL;
+
+const struct mdp5_cfg_hw msm8x74_config = {
+        .name = "msm8x74",
+        .smp = {
+                .mmb_count = 22,
+                .mmb_size = 4096,
+        },
+        .ctl = {
+                .count = 5,
+                .base = { 0x00600, 0x00700, 0x00800, 0x00900, 0x00a00 },
+        },
+        .pipe_vig = {
+                .count = 3,
+                .base = { 0x01200, 0x01600, 0x01a00 },
+        },
+        .pipe_rgb = {
+                .count = 3,
+                .base = { 0x01e00, 0x02200, 0x02600 },
+        },
+        .pipe_dma = {
+                .count = 2,
+                .base = { 0x02a00, 0x02e00 },
+        },
+        .lm = {
+                .count = 5,
+                .base = { 0x03200, 0x03600, 0x03a00, 0x03e00, 0x04200 },
+                .nb_stages = 5,
+        },
+        .dspp = {
+                .count = 3,
+                .base = { 0x04600, 0x04a00, 0x04e00 },
+        },
+        .ad = {
+                .count = 2,
+                .base = { 0x13100, 0x13300 }, /* NOTE: no ad in v1.0 */
+        },
+        .intf = {
+                .count = 4,
+                .base = { 0x12500, 0x12700, 0x12900, 0x12b00 },
+        },
+        .max_clk = 200000000,
+};
+
+const struct mdp5_cfg_hw apq8084_config = {
+        .name = "apq8084",
+        .smp = {
+                .mmb_count = 44,
+                .mmb_size = 8192,
+                .reserved_state[0] = GENMASK(7, 0),     /* first 8 MMBs */
+                .reserved[CID_RGB0] = 2,
+                .reserved[CID_RGB1] = 2,
+                .reserved[CID_RGB2] = 2,
+                .reserved[CID_RGB3] = 2,
+        },
+        .ctl = {
+                .count = 5,
+                .base = { 0x00600, 0x00700, 0x00800, 0x00900, 0x00a00 },
+        },
+        .pipe_vig = {
+                .count = 4,
+                .base = { 0x01200, 0x01600, 0x01a00, 0x01e00 },
+        },
+        .pipe_rgb = {
+                .count = 4,
+                .base = { 0x02200, 0x02600, 0x02a00, 0x02e00 },
+        },
+        .pipe_dma = {
+                .count = 2,
+                .base = { 0x03200, 0x03600 },
+        },
+        .lm = {
+                .count = 6,
+                .base = { 0x03a00, 0x03e00, 0x04200, 0x04600, 0x04a00, 0x04e00 },
+                .nb_stages = 5,
+        },
+        .dspp = {
+                .count = 4,
+                .base = { 0x05200, 0x05600, 0x05a00, 0x05e00 },
+
+        },
+        .ad = {
+                .count = 3,
+                .base = { 0x13500, 0x13700, 0x13900 },
+        },
+        .intf = {
+                .count = 5,
+                .base = { 0x12500, 0x12700, 0x12900, 0x12b00, 0x12d00 },
+        },
+        .max_clk = 320000000,
+};
+
+static const struct mdp5_cfg_handler cfg_handlers[] = {
+        { .revision = 0, .config = { .hw = &msm8x74_config } },
+        { .revision = 2, .config = { .hw = &msm8x74_config } },
+        { .revision = 3, .config = { .hw = &apq8084_config } },
+};
+
+
+static struct mdp5_cfg_platform *mdp5_get_config(struct platform_device *dev);
+
+const struct mdp5_cfg_hw *mdp5_cfg_get_hw_config(struct mdp5_cfg_handler *cfg_handler)
+{
+        return cfg_handler->config.hw;
+}
+
+struct mdp5_cfg *mdp5_cfg_get_config(struct mdp5_cfg_handler *cfg_handler)
+{
+        return &cfg_handler->config;
+}
+
+int mdp5_cfg_get_hw_rev(struct mdp5_cfg_handler *cfg_handler)
+{
+        return cfg_handler->revision;
+}
+
+void mdp5_cfg_destroy(struct mdp5_cfg_handler *cfg_handler)
+{
+        kfree(cfg_handler);
+}
+
+struct mdp5_cfg_handler *mdp5_cfg_init(struct mdp5_kms *mdp5_kms,
+                uint32_t major, uint32_t minor)
+{
+        struct drm_device *dev = mdp5_kms->dev;
+        struct platform_device *pdev = dev->platformdev;
+        struct mdp5_cfg_handler *cfg_handler;
+        struct mdp5_cfg_platform *pconfig;
+        int i, ret = 0;
+
+        cfg_handler = kzalloc(sizeof(*cfg_handler), GFP_KERNEL);
+        if (unlikely(!cfg_handler)) {
+                ret = -ENOMEM;
+                goto fail;
+        }
+
+        if (major != 1) {
+                dev_err(dev->dev, "unexpected MDP major version: v%d.%d\n",
+                                major, minor);
+                ret = -ENXIO;
+                goto fail;
+        }
+
+        /* only after mdp5_cfg global pointer's init can we access the hw */
+        for (i = 0; i < ARRAY_SIZE(cfg_handlers); i++) {
+                if (cfg_handlers[i].revision != minor)
+                        continue;
+                mdp5_cfg = cfg_handlers[i].config.hw;
+
+                break;
+        }
+        if (unlikely(!mdp5_cfg)) {
+                dev_err(dev->dev, "unexpected MDP minor revision: v%d.%d\n",
+                                major, minor);
+                ret = -ENXIO;
+                goto fail;
+        }
+
+        cfg_handler->revision = minor;
+        cfg_handler->config.hw = mdp5_cfg;
+
+        pconfig = mdp5_get_config(pdev);
+        memcpy(&cfg_handler->config.platform, pconfig, sizeof(*pconfig));
+
+        DBG("MDP5: %s hw config selected", mdp5_cfg->name);
+
+        return cfg_handler;
+
+fail:
+        if (cfg_handler)
+                mdp5_cfg_destroy(cfg_handler);
+
+        return NULL;
+}
+
+static struct mdp5_cfg_platform *mdp5_get_config(struct platform_device *dev)
+{
+        static struct mdp5_cfg_platform config = {};
+#ifdef CONFIG_OF
+        /* TODO */
+#endif
+        config.iommu = iommu_domain_alloc(&platform_bus_type);
+
+        return &config;
+}
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h
new file mode 100644
index 000000000000..dba4d52cceeb
--- /dev/null
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_cfg.h
@@ -0,0 +1,91 @@
+/*
+ * Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef __MDP5_CFG_H__
+#define __MDP5_CFG_H__
+
+#include "msm_drv.h"
+
+/*
+ * mdp5_cfg
+ *
+ * This module configures the dynamic offsets used by mdp5.xml.h
+ * (initialized in mdp5_cfg.c)
+ */
+extern const struct mdp5_cfg_hw *mdp5_cfg;
+
+#define MAX_CTL                 8
+#define MAX_BASES               8
+#define MAX_SMP_BLOCKS          44
+#define MAX_CLIENTS             32
+
+typedef DECLARE_BITMAP(mdp5_smp_state_t, MAX_SMP_BLOCKS);
+
+#define MDP5_SUB_BLOCK_DEFINITION \
+        int count; \
+        uint32_t base[MAX_BASES]
+
+struct mdp5_sub_block {
+        MDP5_SUB_BLOCK_DEFINITION;
+};
+
+struct mdp5_lm_block {
+        MDP5_SUB_BLOCK_DEFINITION;
+        uint32_t nb_stages;             /* number of stages per blender */
+};
+
+struct mdp5_smp_block {
+        int mmb_count;                  /* number of SMP MMBs */
+        int mmb_size;                   /* MMB: size in bytes */
+        mdp5_smp_state_t reserved_state;/* SMP MMBs statically allocated */
+        int reserved[MAX_CLIENTS];      /* # of MMBs allocated per client */
+};
+
+struct mdp5_cfg_hw {
+        char  *name;
+
+        struct mdp5_smp_block smp;
+        struct mdp5_sub_block ctl;
+        struct mdp5_sub_block pipe_vig;
+        struct mdp5_sub_block pipe_rgb;
+        struct mdp5_sub_block pipe_dma;
+        struct mdp5_lm_block  lm;
+        struct mdp5_sub_block dspp;
+        struct mdp5_sub_block ad;
+        struct mdp5_sub_block intf;
+
+        uint32_t max_clk;
+};
+
+/* platform config data (ie. from DT, or pdata) */
+struct mdp5_cfg_platform {
+        struct iommu_domain *iommu;
+};
+
+struct mdp5_cfg {
+        const struct mdp5_cfg_hw *hw;
+        struct mdp5_cfg_platform platform;
+};
+
+struct mdp5_kms;
+struct mdp5_cfg_handler;
+
+const struct mdp5_cfg_hw *mdp5_cfg_get_hw_config(struct mdp5_cfg_handler *cfg_hnd);
+struct mdp5_cfg *mdp5_cfg_get_config(struct mdp5_cfg_handler *cfg_hnd);
+int mdp5_cfg_get_hw_rev(struct mdp5_cfg_handler *cfg_hnd);
+
+struct mdp5_cfg_handler *mdp5_cfg_init(struct mdp5_kms *mdp5_kms,
+                uint32_t major, uint32_t minor);
+void mdp5_cfg_destroy(struct mdp5_cfg_handler *cfg_hnd);
+
+#endif /* __MDP5_CFG_H__ */
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c
index ebe2e60f3ab1..0e9a2e3a82d7 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c
@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2014 The Linux Foundation. All rights reserved.
  * Copyright (C) 2013 Red Hat
  * Author: Rob Clark <robdclark@gmail.com>
  *
@@ -17,43 +18,35 @@
 
 #include "mdp5_kms.h"
 
+#include <linux/sort.h>
 #include <drm/drm_mode.h>
 #include "drm_crtc.h"
 #include "drm_crtc_helper.h"
 #include "drm_flip_work.h"
 
+#define SSPP_MAX        (SSPP_RGB3 + 1) /* TODO: Add SSPP_MAX in mdp5.xml.h */
+
 struct mdp5_crtc {
         struct drm_crtc base;
         char name[8];
-        struct drm_plane *plane;
-        struct drm_plane *planes[8];
         int id;
         bool enabled;
 
-        /* which mixer/encoder we route output to: */
-        int mixer;
+        /* layer mixer used for this CRTC (+ its lock): */
+#define GET_LM_ID(crtc_id)      ((crtc_id == 3) ? 5 : crtc_id)
+        int lm;
+        spinlock_t lm_lock;     /* protect REG_MDP5_LM_* registers */
+
+        /* CTL used for this CRTC: */
+        struct mdp5_ctl *ctl;
 
         /* if there is a pending flip, these will be non-null: */
         struct drm_pending_vblank_event *event;
-        struct msm_fence_cb pageflip_cb;
 
 #define PENDING_CURSOR 0x1
 #define PENDING_FLIP   0x2
         atomic_t pending;
 
-        /* the fb that we logically (from PoV of KMS API) hold a ref
-         * to.  Which we may not yet be scanning out (we may still
-         * be scanning out previous in case of page_flip while waiting
-         * for gpu rendering to complete:
-         */
-        struct drm_framebuffer *fb;
-
-        /* the fb that we currently hold a scanout ref to: */
-        struct drm_framebuffer *scanout_fb;
-
-        /* for unref'ing framebuffers after scanout completes: */
-        struct drm_flip_work unref_fb_work;
-
         struct mdp_irq vblank;
         struct mdp_irq err;
 };
@@ -73,67 +66,38 @@ static void request_pending(struct drm_crtc *crtc, uint32_t pending)
         mdp_irq_register(&get_kms(crtc)->base, &mdp5_crtc->vblank);
 }
 
-static void crtc_flush(struct drm_crtc *crtc)
-{
-        struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
-        struct mdp5_kms *mdp5_kms = get_kms(crtc);
-        int id = mdp5_crtc->id;
-        uint32_t i, flush = 0;
-
-        for (i = 0; i < ARRAY_SIZE(mdp5_crtc->planes); i++) {
-                struct drm_plane *plane = mdp5_crtc->planes[i];
-                if (plane) {
-                        enum mdp5_pipe pipe = mdp5_plane_pipe(plane);
-                        flush |= pipe2flush(pipe);
-                }
-        }
-        flush |= mixer2flush(mdp5_crtc->id);
-        flush |= MDP5_CTL_FLUSH_CTL;
-
-        DBG("%s: flush=%08x", mdp5_crtc->name, flush);
-
-        mdp5_write(mdp5_kms, REG_MDP5_CTL_FLUSH(id), flush);
-}
+#define mdp5_lm_get_flush(lm)   mdp_ctl_flush_mask_lm(lm)
 
-static void update_fb(struct drm_crtc *crtc, struct drm_framebuffer *new_fb)
+static void crtc_flush(struct drm_crtc *crtc, u32 flush_mask)
 {
         struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
-        struct drm_framebuffer *old_fb = mdp5_crtc->fb;
-
-        /* grab reference to incoming scanout fb: */
-        drm_framebuffer_reference(new_fb);
-        mdp5_crtc->base.primary->fb = new_fb;
-        mdp5_crtc->fb = new_fb;
 
-        if (old_fb)
-                drm_flip_work_queue(&mdp5_crtc->unref_fb_work, old_fb);
+        DBG("%s: flush=%08x", mdp5_crtc->name, flush_mask);
+        mdp5_ctl_commit(mdp5_crtc->ctl, flush_mask);
 }
 
-/* unlike update_fb(), take a ref to the new scanout fb *before* updating
- * plane, then call this.  Needed to ensure we don't unref the buffer that
- * is actually still being scanned out.
- *
- * Note that this whole thing goes away with atomic.. since we can defer
- * calling into driver until rendering is done.
+/*
+ * flush updates, to make sure hw is updated to new scanout fb,
+ * so that we can safely queue unref to current fb (ie. next
+ * vblank we know hw is done w/ previous scanout_fb).
  */
-static void update_scanout(struct drm_crtc *crtc, struct drm_framebuffer *fb)
+static void crtc_flush_all(struct drm_crtc *crtc)
 {
         struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+        struct drm_plane *plane;
+        uint32_t flush_mask = 0;
 
-        /* flush updates, to make sure hw is updated to new scanout fb,
-         * so that we can safely queue unref to current fb (ie. next
-         * vblank we know hw is done w/ previous scanout_fb).
-         */
-        crtc_flush(crtc);
-
-        if (mdp5_crtc->scanout_fb)
-                drm_flip_work_queue(&mdp5_crtc->unref_fb_work,
-                                mdp5_crtc->scanout_fb);
+        /* we could have already released CTL in the disable path: */
+        if (!mdp5_crtc->ctl)
+                return;
 
-        mdp5_crtc->scanout_fb = fb;
+        drm_atomic_crtc_for_each_plane(plane, crtc) {
+                flush_mask |= mdp5_plane_get_flush(plane);
+        }
+        flush_mask |= mdp5_ctl_get_flush(mdp5_crtc->ctl);
+        flush_mask |= mdp5_lm_get_flush(mdp5_crtc->lm);
 
-        /* enable vblank to complete flip: */
-        request_pending(crtc, PENDING_FLIP);
+        crtc_flush(crtc, flush_mask);
 }
 
 /* if file!=NULL, this is preclose potential cancel-flip path */
@@ -142,7 +106,8 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file)
         struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
         struct drm_device *dev = crtc->dev;
         struct drm_pending_vblank_event *event;
-        unsigned long flags, i;
+        struct drm_plane *plane;
+        unsigned long flags;
 
         spin_lock_irqsave(&dev->event_lock, flags);
         event = mdp5_crtc->event;
@@ -153,50 +118,22 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file)
                  */
                 if (!file || (event->base.file_priv == file)) {
                         mdp5_crtc->event = NULL;
+                        DBG("%s: send event: %p", mdp5_crtc->name, event);
                         drm_send_vblank_event(dev, mdp5_crtc->id, event);
                 }
         }
         spin_unlock_irqrestore(&dev->event_lock, flags);
 
-        for (i = 0; i < ARRAY_SIZE(mdp5_crtc->planes); i++) {
-                struct drm_plane *plane = mdp5_crtc->planes[i];
-                if (plane)
-                        mdp5_plane_complete_flip(plane);
+        drm_atomic_crtc_for_each_plane(plane, crtc) {
+                mdp5_plane_complete_flip(plane);
         }
 }
 
-static void pageflip_cb(struct msm_fence_cb *cb)
-{
-        struct mdp5_crtc *mdp5_crtc =
-                container_of(cb, struct mdp5_crtc, pageflip_cb);
-        struct drm_crtc *crtc = &mdp5_crtc->base;
-        struct drm_framebuffer *fb = mdp5_crtc->fb;
-
-        if (!fb)
-                return;
-
-        drm_framebuffer_reference(fb);
-        mdp5_plane_set_scanout(mdp5_crtc->plane, fb);
-        update_scanout(crtc, fb);
-}
-
-static void unref_fb_worker(struct drm_flip_work *work, void *val)
-{
-        struct mdp5_crtc *mdp5_crtc =
-                container_of(work, struct mdp5_crtc, unref_fb_work);
-        struct drm_device *dev = mdp5_crtc->base.dev;
-
-        mutex_lock(&dev->mode_config.mutex);
-        drm_framebuffer_unreference(val);
-        mutex_unlock(&dev->mode_config.mutex);
-}
-
 static void mdp5_crtc_destroy(struct drm_crtc *crtc)
 {
         struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
 
         drm_crtc_cleanup(crtc);
-        drm_flip_work_cleanup(&mdp5_crtc->unref_fb_work);
 
         kfree(mdp5_crtc);
 }
@@ -214,6 +151,8 @@ static void mdp5_crtc_dpms(struct drm_crtc *crtc, int mode)
                         mdp5_enable(mdp5_kms);
                         mdp_irq_register(&mdp5_kms->base, &mdp5_crtc->err);
                 } else {
+                        /* set STAGE_UNUSED for all layers */
+                        mdp5_ctl_blend(mdp5_crtc->ctl, mdp5_crtc->lm, 0x00000000);
                         mdp_irq_unregister(&mdp5_kms->base, &mdp5_crtc->err);
                         mdp5_disable(mdp5_kms);
                 }
@@ -228,54 +167,78 @@ static bool mdp5_crtc_mode_fixup(struct drm_crtc *crtc,
         return true;
 }
 
+/*
+ * blend_setup() - blend all the planes of a CRTC
+ *
+ * When border is enabled, the border color will ALWAYS be the base layer.
+ * Therefore, the first plane (private RGB pipe) will start at STAGE0.
+ * If disabled, the first plane starts at STAGE_BASE.
+ *
+ * Note:
+ * Border is not enabled here because the private plane is exactly
+ * the CRTC resolution.
+ */
 static void blend_setup(struct drm_crtc *crtc)
 {
         struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
         struct mdp5_kms *mdp5_kms = get_kms(crtc);
-        int id = mdp5_crtc->id;
+        struct drm_plane *plane;
+        const struct mdp5_cfg_hw *hw_cfg;
+        uint32_t lm = mdp5_crtc->lm, blend_cfg = 0;
+        unsigned long flags;
+#define blender(stage)  ((stage) - STAGE_BASE)
 
-        /*
-         * Hard-coded setup for now until I figure out how the
-         * layer-mixer works
-         */
+        hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg);
 
-        /* LM[id]: */
-        mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_COLOR_OUT(id),
-                        MDP5_LM_BLEND_COLOR_OUT_STAGE0_FG_ALPHA);
-        mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_OP_MODE(id, 0),
-                        MDP5_LM_BLEND_OP_MODE_FG_ALPHA(FG_CONST) |
-                        MDP5_LM_BLEND_OP_MODE_BG_ALPHA(FG_PIXEL) |
-                        MDP5_LM_BLEND_OP_MODE_BG_INV_ALPHA);
-        mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_FG_ALPHA(id, 0), 0xff);
-        mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_BG_ALPHA(id, 0), 0x00);
-
-        /* NOTE: seems that LM[n] and CTL[m], we do not need n==m.. but
-         * we want to be setting CTL[m].LAYER[n].  Not sure what the
-         * point of having CTL[m].LAYER[o] (for o!=n).. maybe that is
-         * used when chaining up mixers for high resolution displays?
-         */
+        spin_lock_irqsave(&mdp5_crtc->lm_lock, flags);
+
+        /* ctl could be released already when we are shutting down: */
+        if (!mdp5_crtc->ctl)
+                goto out;
 
-        /* CTL[id]: */
-        mdp5_write(mdp5_kms, REG_MDP5_CTL_LAYER_REG(id, 0),
-                        MDP5_CTL_LAYER_REG_RGB0(STAGE0) |
-                        MDP5_CTL_LAYER_REG_BORDER_COLOR);
-        mdp5_write(mdp5_kms, REG_MDP5_CTL_LAYER_REG(id, 1), 0);
-        mdp5_write(mdp5_kms, REG_MDP5_CTL_LAYER_REG(id, 2), 0);
-        mdp5_write(mdp5_kms, REG_MDP5_CTL_LAYER_REG(id, 3), 0);
-        mdp5_write(mdp5_kms, REG_MDP5_CTL_LAYER_REG(id, 4), 0);
+        drm_atomic_crtc_for_each_plane(plane, crtc) {
+                enum mdp_mixer_stage_id stage =
+                        to_mdp5_plane_state(plane->state)->stage;
+
+                /*
+                 * Note: This cannot happen with current implementation but
+                 * we need to check this condition once z property is added
+                 */
+                BUG_ON(stage > hw_cfg->lm.nb_stages);
+
+                /* LM */
+                mdp5_write(mdp5_kms,
+                                REG_MDP5_LM_BLEND_OP_MODE(lm, blender(stage)),
+                                MDP5_LM_BLEND_OP_MODE_FG_ALPHA(FG_CONST) |
+                                MDP5_LM_BLEND_OP_MODE_BG_ALPHA(BG_CONST));
+                mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_FG_ALPHA(lm,
+                                blender(stage)), 0xff);
+                mdp5_write(mdp5_kms, REG_MDP5_LM_BLEND_BG_ALPHA(lm,
+                                blender(stage)), 0x00);
+                /* CTL */
+                blend_cfg |= mdp_ctl_blend_mask(mdp5_plane_pipe(plane), stage);
+                DBG("%s: blending pipe %s on stage=%d", mdp5_crtc->name,
+                                pipe2name(mdp5_plane_pipe(plane)), stage);
+        }
+
+        DBG("%s: lm%d: blend config = 0x%08x", mdp5_crtc->name, lm, blend_cfg);
+        mdp5_ctl_blend(mdp5_crtc->ctl, lm, blend_cfg);
+
+out:
+        spin_unlock_irqrestore(&mdp5_crtc->lm_lock, flags);
 }
 
-static int mdp5_crtc_mode_set(struct drm_crtc *crtc,
-                struct drm_display_mode *mode,
-                struct drm_display_mode *adjusted_mode,
-                int x, int y,
-                struct drm_framebuffer *old_fb)
+static void mdp5_crtc_mode_set_nofb(struct drm_crtc *crtc)
 {
         struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
         struct mdp5_kms *mdp5_kms = get_kms(crtc);
-        int ret;
+        unsigned long flags;
+        struct drm_display_mode *mode;
 
-        mode = adjusted_mode;
+        if (WARN_ON(!crtc->state))
+                return;
+
+        mode = &crtc->state->adjusted_mode;
 
         DBG("%s: set mode: %d:\"%s\" %d %d %d %d %d %d %d %d %d %d 0x%x 0x%x",
                         mdp5_crtc->name, mode->base.id, mode->name,
@@ -286,28 +249,11 @@ static int mdp5_crtc_mode_set(struct drm_crtc *crtc,
                         mode->vsync_end, mode->vtotal,
                         mode->type, mode->flags);
 
-        /* grab extra ref for update_scanout() */
-        drm_framebuffer_reference(crtc->primary->fb);
-
-        ret = mdp5_plane_mode_set(mdp5_crtc->plane, crtc, crtc->primary->fb,
-                        0, 0, mode->hdisplay, mode->vdisplay,
-                        x << 16, y << 16,
-                        mode->hdisplay << 16, mode->vdisplay << 16);
-        if (ret) {
-                drm_framebuffer_unreference(crtc->primary->fb);
-                dev_err(crtc->dev->dev, "%s: failed to set mode on plane: %d\n",
-                                mdp5_crtc->name, ret);
-                return ret;
-        }
-
-        mdp5_write(mdp5_kms, REG_MDP5_LM_OUT_SIZE(mdp5_crtc->id),
+        spin_lock_irqsave(&mdp5_crtc->lm_lock, flags);
+        mdp5_write(mdp5_kms, REG_MDP5_LM_OUT_SIZE(mdp5_crtc->lm),
                         MDP5_LM_OUT_SIZE_WIDTH(mode->hdisplay) |
                         MDP5_LM_OUT_SIZE_HEIGHT(mode->vdisplay));
-
-        update_fb(crtc, crtc->primary->fb);
-        update_scanout(crtc, crtc->primary->fb);
-
-        return 0;
+        spin_unlock_irqrestore(&mdp5_crtc->lm_lock, flags);
 }
 
 static void mdp5_crtc_prepare(struct drm_crtc *crtc)
@@ -321,66 +267,119 @@ static void mdp5_crtc_prepare(struct drm_crtc *crtc)
 
 static void mdp5_crtc_commit(struct drm_crtc *crtc)
 {
+        struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+        DBG("%s", mdp5_crtc->name);
         mdp5_crtc_dpms(crtc, DRM_MODE_DPMS_ON);
-        crtc_flush(crtc);
+        crtc_flush_all(crtc);
         /* drop the ref to mdp clk's that we got in prepare: */
         mdp5_disable(get_kms(crtc));
 }
 
-static int mdp5_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y,
-                struct drm_framebuffer *old_fb)
+static void mdp5_crtc_load_lut(struct drm_crtc *crtc)
+{
+}
+
+struct plane_state {
+        struct drm_plane *plane;
+        struct mdp5_plane_state *state;
+};
+
+static int pstate_cmp(const void *a, const void *b)
+{
+        struct plane_state *pa = (struct plane_state *)a;
+        struct plane_state *pb = (struct plane_state *)b;
+        return pa->state->zpos - pb->state->zpos;
+}
+
+static int mdp5_crtc_atomic_check(struct drm_crtc *crtc,
+                struct drm_crtc_state *state)
 {
         struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
-        struct drm_plane *plane = mdp5_crtc->plane;
-        struct drm_display_mode *mode = &crtc->mode;
-        int ret;
-
-        /* grab extra ref for update_scanout() */
-        drm_framebuffer_reference(crtc->primary->fb);
-
-        ret = mdp5_plane_mode_set(plane, crtc, crtc->primary->fb,
-                        0, 0, mode->hdisplay, mode->vdisplay,
-                        x << 16, y << 16,
-                        mode->hdisplay << 16, mode->vdisplay << 16);
-        if (ret) {
-                drm_framebuffer_unreference(crtc->primary->fb);
-                return ret;
+        struct mdp5_kms *mdp5_kms = get_kms(crtc);
+        struct drm_plane *plane;
+        struct drm_device *dev = crtc->dev;
+        struct plane_state pstates[STAGE3 + 1];
+        int cnt = 0, i;
+
+        DBG("%s: check", mdp5_crtc->name);
+
+        if (mdp5_crtc->event) {
+                dev_err(dev->dev, "already pending flip!\n");
+                return -EBUSY;
         }
 
-        update_fb(crtc, crtc->primary->fb);
-        update_scanout(crtc, crtc->primary->fb);
+        /* request a free CTL, if none is already allocated for this CRTC */
+        if (state->enable && !mdp5_crtc->ctl) {
+                mdp5_crtc->ctl = mdp5_ctlm_request(mdp5_kms->ctlm, crtc);
+                if (WARN_ON(!mdp5_crtc->ctl))
+                        return -EINVAL;
+        }
+
+        /* verify that there are not too many planes attached to crtc
+         * and that we don't have conflicting mixer stages:
+         */
+        drm_atomic_crtc_state_for_each_plane(plane, state) {
+                struct drm_plane_state *pstate;
+
+                if (cnt >= ARRAY_SIZE(pstates)) {
+                        dev_err(dev->dev, "too many planes!\n");
+                        return -EINVAL;
+                }
+
+                pstate = state->state->plane_states[drm_plane_index(plane)];
+
+                /* plane might not have changed, in which case take
+                 * current state:
+                 */
+                if (!pstate)
+                        pstate = plane->state;
+
+                pstates[cnt].plane = plane;
+                pstates[cnt].state = to_mdp5_plane_state(pstate);
+
+                cnt++;
+        }
+
+        sort(pstates, cnt, sizeof(pstates[0]), pstate_cmp, NULL);
+
+        for (i = 0; i < cnt; i++) {
+                pstates[i].state->stage = STAGE_BASE + i;
+                DBG("%s: assign pipe %s on stage=%d", mdp5_crtc->name,
+                                pipe2name(mdp5_plane_pipe(pstates[i].plane)),
+                                pstates[i].state->stage);
+        }
 
         return 0;
 }
 
-static void mdp5_crtc_load_lut(struct drm_crtc *crtc)
+static void mdp5_crtc_atomic_begin(struct drm_crtc *crtc)
 {
+        struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
+        DBG("%s: begin", mdp5_crtc->name);
 }
 
-static int mdp5_crtc_page_flip(struct drm_crtc *crtc,
-                struct drm_framebuffer *new_fb,
-                struct drm_pending_vblank_event *event,
-                uint32_t page_flip_flags)
+static void mdp5_crtc_atomic_flush(struct drm_crtc *crtc)
 {
         struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
         struct drm_device *dev = crtc->dev;
-        struct drm_gem_object *obj;
         unsigned long flags;
 
-        if (mdp5_crtc->event) {
-                dev_err(dev->dev, "already pending flip!\n");
-                return -EBUSY;
-        }
+        DBG("%s: flush", mdp5_crtc->name);
 
-        obj = msm_framebuffer_bo(new_fb, 0);
+        WARN_ON(mdp5_crtc->event);
 
         spin_lock_irqsave(&dev->event_lock, flags);
-        mdp5_crtc->event = event;
+        mdp5_crtc->event = crtc->state->event;
         spin_unlock_irqrestore(&dev->event_lock, flags);
 
-        update_fb(crtc, new_fb);
+        blend_setup(crtc);
+        crtc_flush_all(crtc);
+        request_pending(crtc, PENDING_FLIP);
 
-        return msm_gem_queue_inactive_cb(obj, &mdp5_crtc->pageflip_cb);
+        if (mdp5_crtc->ctl && !crtc->state->enable) {
+                mdp5_ctl_release(mdp5_crtc->ctl);
+                mdp5_crtc->ctl = NULL;
+        }
 }
 
 static int mdp5_crtc_set_property(struct drm_crtc *crtc,
@@ -391,27 +390,33 @@ static int mdp5_crtc_set_property(struct drm_crtc *crtc,
 }
 
 static const struct drm_crtc_funcs mdp5_crtc_funcs = {
-        .set_config = drm_crtc_helper_set_config,
+        .set_config = drm_atomic_helper_set_config,
         .destroy = mdp5_crtc_destroy,
-        .page_flip = mdp5_crtc_page_flip,
+        .page_flip = drm_atomic_helper_page_flip,
         .set_property = mdp5_crtc_set_property,
+        .reset = drm_atomic_helper_crtc_reset,
+        .atomic_duplicate_state = drm_atomic_helper_crtc_duplicate_state,
+        .atomic_destroy_state = drm_atomic_helper_crtc_destroy_state,
 };
 
 static const struct drm_crtc_helper_funcs mdp5_crtc_helper_funcs = {
         .dpms = mdp5_crtc_dpms,
         .mode_fixup = mdp5_crtc_mode_fixup,
-        .mode_set = mdp5_crtc_mode_set,
+        .mode_set_nofb = mdp5_crtc_mode_set_nofb,
+        .mode_set = drm_helper_crtc_mode_set,
+        .mode_set_base = drm_helper_crtc_mode_set_base,
         .prepare = mdp5_crtc_prepare,
         .commit = mdp5_crtc_commit,
-        .mode_set_base = mdp5_crtc_mode_set_base,
         .load_lut = mdp5_crtc_load_lut,
+        .atomic_check = mdp5_crtc_atomic_check,
+        .atomic_begin = mdp5_crtc_atomic_begin,
+        .atomic_flush = mdp5_crtc_atomic_flush,
 };
 
 static void mdp5_crtc_vblank_irq(struct mdp_irq *irq, uint32_t irqstatus)
 {
         struct mdp5_crtc *mdp5_crtc = container_of(irq, struct mdp5_crtc, vblank);
         struct drm_crtc *crtc = &mdp5_crtc->base;
-        struct msm_drm_private *priv = crtc->dev->dev_private;
         unsigned pending;
 
         mdp_irq_unregister(&get_kms(crtc)->base, &mdp5_crtc->vblank);
@@ -420,16 +425,14 @@ static void mdp5_crtc_vblank_irq(struct mdp_irq *irq, uint32_t irqstatus)
 
         if (pending & PENDING_FLIP) {
                 complete_flip(crtc, NULL);
-                drm_flip_work_commit(&mdp5_crtc->unref_fb_work, priv->wq);
         }
 }
 
 static void mdp5_crtc_err_irq(struct mdp_irq *irq, uint32_t irqstatus)
 {
         struct mdp5_crtc *mdp5_crtc = container_of(irq, struct mdp5_crtc, err);
-        struct drm_crtc *crtc = &mdp5_crtc->base;
+
         DBG("%s: error: %08x", mdp5_crtc->name, irqstatus);
-        crtc_flush(crtc);
 }
 
 uint32_t mdp5_crtc_vblank(struct drm_crtc *crtc)
@@ -450,10 +453,9 @@ void mdp5_crtc_set_intf(struct drm_crtc *crtc, int intf,
 {
         struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
         struct mdp5_kms *mdp5_kms = get_kms(crtc);
-        static const enum mdp5_intfnum intfnum[] = {
-                        INTF0, INTF1, INTF2, INTF3,
-        };
+        uint32_t flush_mask = 0;
         uint32_t intf_sel;
+        unsigned long flags;
 
         /* now that we know what irq's we want: */
         mdp5_crtc->err.irqmask = intf2err(intf);
@@ -463,6 +465,7 @@ void mdp5_crtc_set_intf(struct drm_crtc *crtc, int intf,
         if (!mdp5_kms)
                 return;
 
+        spin_lock_irqsave(&mdp5_kms->resource_lock, flags);
         intf_sel = mdp5_read(mdp5_kms, REG_MDP5_DISP_INTF_SEL);
 
         switch (intf) {
@@ -487,45 +490,25 @@ void mdp5_crtc_set_intf(struct drm_crtc *crtc, int intf,
                 break;
         }
 
-        blend_setup(crtc);
+        mdp5_write(mdp5_kms, REG_MDP5_DISP_INTF_SEL, intf_sel);
+        spin_unlock_irqrestore(&mdp5_kms->resource_lock, flags);
 
         DBG("%s: intf_sel=%08x", mdp5_crtc->name, intf_sel);
+        mdp5_ctl_set_intf(mdp5_crtc->ctl, intf);
+        flush_mask |= mdp5_ctl_get_flush(mdp5_crtc->ctl);
+        flush_mask |= mdp5_lm_get_flush(mdp5_crtc->lm);
 
-        mdp5_write(mdp5_kms, REG_MDP5_DISP_INTF_SEL, intf_sel);
-        mdp5_write(mdp5_kms, REG_MDP5_CTL_OP(mdp5_crtc->id),
-                        MDP5_CTL_OP_MODE(MODE_NONE) |
-                        MDP5_CTL_OP_INTF_NUM(intfnum[intf]));
-
-        crtc_flush(crtc);
+        crtc_flush(crtc, flush_mask);
 }
 
-static void set_attach(struct drm_crtc *crtc, enum mdp5_pipe pipe_id,
-                struct drm_plane *plane)
+int mdp5_crtc_get_lm(struct drm_crtc *crtc)
 {
         struct mdp5_crtc *mdp5_crtc = to_mdp5_crtc(crtc);
 
-        BUG_ON(pipe_id >= ARRAY_SIZE(mdp5_crtc->planes));
+        if (WARN_ON(!crtc))
+                return -EINVAL;
 
-        if (mdp5_crtc->planes[pipe_id] == plane)
-                return;
-
-        mdp5_crtc->planes[pipe_id] = plane;
-        blend_setup(crtc);
-        if (mdp5_crtc->enabled && (plane != mdp5_crtc->plane))
-                crtc_flush(crtc);
-}
-
-void mdp5_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane)
-{
-        set_attach(crtc, mdp5_plane_pipe(plane), plane);
-}
-
-void mdp5_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane)
-{
-        /* don't actually detatch our primary plane: */
-        if (to_mdp5_crtc(crtc)->plane == plane)
-                return;
-        set_attach(crtc, mdp5_plane_pipe(plane), NULL);
+        return mdp5_crtc->lm;
 }
 
 /* initialize crtc */
@@ -534,18 +517,17 @@ struct drm_crtc *mdp5_crtc_init(struct drm_device *dev,
 {
         struct drm_crtc *crtc = NULL;
         struct mdp5_crtc *mdp5_crtc;
-        int ret;
 
         mdp5_crtc = kzalloc(sizeof(*mdp5_crtc), GFP_KERNEL);
-        if (!mdp5_crtc) {
-                ret = -ENOMEM;
-                goto fail;
-        }
+        if (!mdp5_crtc)
+                return ERR_PTR(-ENOMEM);
 
         crtc = &mdp5_crtc->base;
 
-        mdp5_crtc->plane = plane;
         mdp5_crtc->id = id;
+        mdp5_crtc->lm = GET_LM_ID(id);
+
+        spin_lock_init(&mdp5_crtc->lm_lock);
 
         mdp5_crtc->vblank.irq = mdp5_crtc_vblank_irq;
         mdp5_crtc->err.irq = mdp5_crtc_err_irq;
@@ -553,23 +535,11 @@ struct drm_crtc *mdp5_crtc_init(struct drm_device *dev,
         snprintf(mdp5_crtc->name, sizeof(mdp5_crtc->name), "%s:%d",
                         pipe2name(mdp5_plane_pipe(plane)), id);
 
-        ret = drm_flip_work_init(&mdp5_crtc->unref_fb_work, 16,
-                        "unref fb", unref_fb_worker);
-        if (ret)
-                goto fail;
-
-        INIT_FENCE_CB(&mdp5_crtc->pageflip_cb, pageflip_cb);
-
         drm_crtc_init_with_planes(dev, crtc, plane, NULL, &mdp5_crtc_funcs);
         drm_crtc_helper_add(crtc, &mdp5_crtc_helper_funcs);
+        plane->crtc = crtc;
 
-        mdp5_plane_install_properties(mdp5_crtc->plane, &crtc->base);
+        mdp5_plane_install_properties(plane, &crtc->base);
 
         return crtc;
-
-fail:
-        if (crtc)
-                mdp5_crtc_destroy(crtc);
-
-        return ERR_PTR(ret);
 }
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c
new file mode 100644
index 000000000000..dea4505ac963
--- /dev/null
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.c
@@ -0,0 +1,322 @@
+/*
+ * Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include "mdp5_kms.h"
+#include "mdp5_ctl.h"
+
+/*
+ * CTL - MDP Control Pool Manager
+ *
+ * Controls are shared between all CRTCs.
+ *
+ * They are intended to be used for data path configuration.
+ * The top level register programming describes the complete data path for
+ * a specific data path ID - REG_MDP5_CTL_*(<id>, ...)
+ *
+ * Hardware capabilities determine the number of concurrent data paths
+ *
+ * In certain use cases (high-resolution dual pipe), one single CTL can be
+ * shared across multiple CRTCs.
+ *
+ * Because the number of CTLs can be less than the number of CRTCs,
+ * CTLs are dynamically allocated from a pool of CTLs, only once a CRTC is
+ * requested by the client (in mdp5_crtc_mode_set()).
+ */
+
+struct mdp5_ctl {
+        struct mdp5_ctl_manager *ctlm;
+
+        u32 id;
+
+        /* whether this CTL has been allocated or not: */
+        bool busy;
+
+        /* memory output connection (@see mdp5_ctl_mode): */
+        u32 mode;
+
+        /* REG_MDP5_CTL_*(<id>) registers access info + lock: */
+        spinlock_t hw_lock;
+        u32 reg_offset;
+
+        /* flush mask used to commit CTL registers */
+        u32 flush_mask;
+
+        bool cursor_on;
+
+        struct drm_crtc *crtc;
+};
+
+struct mdp5_ctl_manager {
+        struct drm_device *dev;
+
+        /* number of CTL / Layer Mixers in this hw config: */
+        u32 nlm;
+        u32 nctl;
+
+        /* pool of CTLs + lock to protect resource allocation (ctls[i].busy) */
+        spinlock_t pool_lock;
+        struct mdp5_ctl ctls[MAX_CTL];
+};
+
+static inline
+struct mdp5_kms *get_kms(struct mdp5_ctl_manager *ctl_mgr)
+{
+        struct msm_drm_private *priv = ctl_mgr->dev->dev_private;
+
+        return to_mdp5_kms(to_mdp_kms(priv->kms));
+}
+
+static inline
+void ctl_write(struct mdp5_ctl *ctl, u32 reg, u32 data)
+{
+        struct mdp5_kms *mdp5_kms = get_kms(ctl->ctlm);
+
+        (void)ctl->reg_offset; /* TODO use this instead of mdp5_write */
+        mdp5_write(mdp5_kms, reg, data);
+}
+
+static inline
+u32 ctl_read(struct mdp5_ctl *ctl, u32 reg)
+{
+        struct mdp5_kms *mdp5_kms = get_kms(ctl->ctlm);
+
+        (void)ctl->reg_offset; /* TODO use this instead of mdp5_write */
+        return mdp5_read(mdp5_kms, reg);
+}
+
+
+int mdp5_ctl_set_intf(struct mdp5_ctl *ctl, enum mdp5_intf intf)
+{
+        unsigned long flags;
+        static const enum mdp5_intfnum intfnum[] = {
+                        INTF0, INTF1, INTF2, INTF3,
+        };
+
+        spin_lock_irqsave(&ctl->hw_lock, flags);
+        ctl_write(ctl, REG_MDP5_CTL_OP(ctl->id),
+                        MDP5_CTL_OP_MODE(ctl->mode) |
+                        MDP5_CTL_OP_INTF_NUM(intfnum[intf]));
+        spin_unlock_irqrestore(&ctl->hw_lock, flags);
+
+        return 0;
+}
+
+int mdp5_ctl_set_cursor(struct mdp5_ctl *ctl, bool enable)
+{
+        struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm;
+        unsigned long flags;
+        u32 blend_cfg;
+        int lm;
+
+        lm = mdp5_crtc_get_lm(ctl->crtc);
+        if (unlikely(WARN_ON(lm < 0))) {
+                dev_err(ctl_mgr->dev->dev, "CTL %d cannot find LM: %d",
+                                ctl->id, lm);
+                return -EINVAL;
+        }
+
+        spin_lock_irqsave(&ctl->hw_lock, flags);
+
+        blend_cfg = ctl_read(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, lm));
+
+        if (enable)
+                blend_cfg |=  MDP5_CTL_LAYER_REG_CURSOR_OUT;
+        else
+                blend_cfg &= ~MDP5_CTL_LAYER_REG_CURSOR_OUT;
+
+        ctl_write(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, lm), blend_cfg);
+
+        spin_unlock_irqrestore(&ctl->hw_lock, flags);
+
+        ctl->cursor_on = enable;
+
+        return 0;
+}
+
+
+int mdp5_ctl_blend(struct mdp5_ctl *ctl, u32 lm, u32 blend_cfg)
+{
+        unsigned long flags;
+
+        if (ctl->cursor_on)
+                blend_cfg |=  MDP5_CTL_LAYER_REG_CURSOR_OUT;
+        else
+                blend_cfg &= ~MDP5_CTL_LAYER_REG_CURSOR_OUT;
+
+        spin_lock_irqsave(&ctl->hw_lock, flags);
+        ctl_write(ctl, REG_MDP5_CTL_LAYER_REG(ctl->id, lm), blend_cfg);
+        spin_unlock_irqrestore(&ctl->hw_lock, flags);
+
+        return 0;
+}
+
+int mdp5_ctl_commit(struct mdp5_ctl *ctl, u32 flush_mask)
+{
+        struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm;
+        unsigned long flags;
+
+        if (flush_mask & MDP5_CTL_FLUSH_CURSOR_DUMMY) {
+                int lm = mdp5_crtc_get_lm(ctl->crtc);
+
+                if (unlikely(WARN_ON(lm < 0))) {
+                        dev_err(ctl_mgr->dev->dev, "CTL %d cannot find LM: %d",
+                                        ctl->id, lm);
+                        return -EINVAL;
+                }
+
+                /* for current targets, cursor bit is the same as LM bit */
+                flush_mask |= mdp_ctl_flush_mask_lm(lm);
+        }
+
+        spin_lock_irqsave(&ctl->hw_lock, flags);
+        ctl_write(ctl, REG_MDP5_CTL_FLUSH(ctl->id), flush_mask);
+        spin_unlock_irqrestore(&ctl->hw_lock, flags);
+
+        return 0;
+}
+
+u32 mdp5_ctl_get_flush(struct mdp5_ctl *ctl)
+{
+        return ctl->flush_mask;
+}
+
+void mdp5_ctl_release(struct mdp5_ctl *ctl)
+{
+        struct mdp5_ctl_manager *ctl_mgr = ctl->ctlm;
+        unsigned long flags;
+
+        if (unlikely(WARN_ON(ctl->id >= MAX_CTL) || !ctl->busy)) {
+                dev_err(ctl_mgr->dev->dev, "CTL %d in bad state (%d)",
+                                ctl->id, ctl->busy);
+                return;
+        }
+
+        spin_lock_irqsave(&ctl_mgr->pool_lock, flags);
+        ctl->busy = false;
+        spin_unlock_irqrestore(&ctl_mgr->pool_lock, flags);
+
+        DBG("CTL %d released", ctl->id);
+}
+
+/*
+ * mdp5_ctl_request() - CTL dynamic allocation
+ *
+ * Note: Current implementation considers that we can only have one CRTC per CTL
+ *
+ * @return first free CTL
+ */
+struct mdp5_ctl *mdp5_ctlm_request(struct mdp5_ctl_manager *ctl_mgr,
+                struct drm_crtc *crtc)
+{
+        struct mdp5_ctl *ctl = NULL;
+        unsigned long flags;
+        int c;
+
+        spin_lock_irqsave(&ctl_mgr->pool_lock, flags);
+
+        for (c = 0; c < ctl_mgr->nctl; c++)
+                if (!ctl_mgr->ctls[c].busy)
+                        break;
+
+        if (unlikely(c >= ctl_mgr->nctl)) {
+                dev_err(ctl_mgr->dev->dev, "No more CTL available!");
+                goto unlock;
+        }
+
+        ctl = &ctl_mgr->ctls[c];
+
+        ctl->crtc = crtc;
+        ctl->busy = true;
+        DBG("CTL %d allocated", ctl->id);
+
+unlock:
+        spin_unlock_irqrestore(&ctl_mgr->pool_lock, flags);
+        return ctl;
+}
+
+void mdp5_ctlm_hw_reset(struct mdp5_ctl_manager *ctl_mgr)
+{
+        unsigned long flags;
+        int c;
+
+        for (c = 0; c < ctl_mgr->nctl; c++) {
+                struct mdp5_ctl *ctl = &ctl_mgr->ctls[c];
+
+                spin_lock_irqsave(&ctl->hw_lock, flags);
+                ctl_write(ctl, REG_MDP5_CTL_OP(ctl->id), 0);
+                spin_unlock_irqrestore(&ctl->hw_lock, flags);
+        }
+}
+
+void mdp5_ctlm_destroy(struct mdp5_ctl_manager *ctl_mgr)
+{
+        kfree(ctl_mgr);
+}
+
+struct mdp5_ctl_manager *mdp5_ctlm_init(struct drm_device *dev,
+                void __iomem *mmio_base, const struct mdp5_cfg_hw *hw_cfg)
+{
+        struct mdp5_ctl_manager *ctl_mgr;
+        const struct mdp5_sub_block *ctl_cfg = &hw_cfg->ctl;
+        unsigned long flags;
+        int c, ret;
+
+        ctl_mgr = kzalloc(sizeof(*ctl_mgr), GFP_KERNEL);
+        if (!ctl_mgr) {
+                dev_err(dev->dev, "failed to allocate CTL manager\n");
+                ret = -ENOMEM;
+                goto fail;
+        }
+
+        if (unlikely(WARN_ON(ctl_cfg->count > MAX_CTL))) {
+                dev_err(dev->dev, "Increase static pool size to at least %d\n",
+                                ctl_cfg->count);
+                ret = -ENOSPC;
+                goto fail;
+        }
+
+        /* initialize the CTL manager: */
+        ctl_mgr->dev = dev;
+        ctl_mgr->nlm = hw_cfg->lm.count;
+        ctl_mgr->nctl = ctl_cfg->count;
+        spin_lock_init(&ctl_mgr->pool_lock);
+
+        /* initialize each CTL of the pool: */
+        spin_lock_irqsave(&ctl_mgr->pool_lock, flags);
+        for (c = 0; c < ctl_mgr->nctl; c++) {
+                struct mdp5_ctl *ctl = &ctl_mgr->ctls[c];
+
+                if (WARN_ON(!ctl_cfg->base[c])) {
+                        dev_err(dev->dev, "CTL_%d: base is null!\n", c);
+                        ret = -EINVAL;
+                        goto fail;
+                }
+                ctl->ctlm = ctl_mgr;
+                ctl->id = c;
+                ctl->mode = MODE_NONE;
+                ctl->reg_offset = ctl_cfg->base[c];
+                ctl->flush_mask = MDP5_CTL_FLUSH_CTL;
+                ctl->busy = false;
+                spin_lock_init(&ctl->hw_lock);
+        }
+        spin_unlock_irqrestore(&ctl_mgr->pool_lock, flags);
+        DBG("Pool of %d CTLs created.", ctl_mgr->nctl);
+
+        return ctl_mgr;
+
+fail:
+        if (ctl_mgr)
+                mdp5_ctlm_destroy(ctl_mgr);
+
+        return ERR_PTR(ret);
+}
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h
new file mode 100644
index 000000000000..1018519b6af2
--- /dev/null
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_ctl.h
@@ -0,0 +1,122 @@
+/*
+ * Copyright (c) 2014 The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef __MDP5_CTL_H__
+#define __MDP5_CTL_H__
+
+#include "msm_drv.h"
+
+/*
+ * CTL Manager prototypes:
+ * mdp5_ctlm_init() returns a ctlm (CTL Manager) handler,
+ * which is then used to call the other mdp5_ctlm_*(ctlm, ...) functions.
+ */
+struct mdp5_ctl_manager;
+struct mdp5_ctl_manager *mdp5_ctlm_init(struct drm_device *dev,
+                void __iomem *mmio_base, const struct mdp5_cfg_hw *hw_cfg);
+void mdp5_ctlm_hw_reset(struct mdp5_ctl_manager *ctlm);
+void mdp5_ctlm_destroy(struct mdp5_ctl_manager *ctlm);
+
+/*
+ * CTL prototypes:
+ * mdp5_ctl_request(ctlm, ...) returns a ctl (CTL resource) handler,
+ * which is then used to call the other mdp5_ctl_*(ctl, ...) functions.
+ */
+struct mdp5_ctl *mdp5_ctlm_request(struct mdp5_ctl_manager *ctlm, struct drm_crtc *crtc);
+
+int mdp5_ctl_set_intf(struct mdp5_ctl *ctl, enum mdp5_intf intf);
+
+int mdp5_ctl_set_cursor(struct mdp5_ctl *ctl, bool enable);
+
+/* @blend_cfg: see LM blender config definition below */
+int mdp5_ctl_blend(struct mdp5_ctl *ctl, u32 lm, u32 blend_cfg);
+
+/* @flush_mask: see CTL flush masks definitions below */
+int mdp5_ctl_commit(struct mdp5_ctl *ctl, u32 flush_mask);
+u32 mdp5_ctl_get_flush(struct mdp5_ctl *ctl);
+
+void mdp5_ctl_release(struct mdp5_ctl *ctl);
+
+/*
+ * blend_cfg (LM blender config):
+ *
+ * The function below allows the caller of mdp5_ctl_blend() to specify how pipes
+ * are being blended according to their stage (z-order), through @blend_cfg arg.
+ */
+static inline u32 mdp_ctl_blend_mask(enum mdp5_pipe pipe,
+                enum mdp_mixer_stage_id stage)
+{
+        switch (pipe) {
+        case SSPP_VIG0: return MDP5_CTL_LAYER_REG_VIG0(stage);
+        case SSPP_VIG1: return MDP5_CTL_LAYER_REG_VIG1(stage);
+        case SSPP_VIG2: return MDP5_CTL_LAYER_REG_VIG2(stage);
+        case SSPP_RGB0: return MDP5_CTL_LAYER_REG_RGB0(stage);
+        case SSPP_RGB1: return MDP5_CTL_LAYER_REG_RGB1(stage);
+        case SSPP_RGB2: return MDP5_CTL_LAYER_REG_RGB2(stage);
+        case SSPP_DMA0: return MDP5_CTL_LAYER_REG_DMA0(stage);
+        case SSPP_DMA1: return MDP5_CTL_LAYER_REG_DMA1(stage);
+        case SSPP_VIG3: return MDP5_CTL_LAYER_REG_VIG3(stage);
+        case SSPP_RGB3: return MDP5_CTL_LAYER_REG_RGB3(stage);
+        default:        return 0;
+        }
+}
+
+/*
+ * flush_mask (CTL flush masks):
+ *
+ * The following functions allow each DRM entity to get and store
+ * their own flush mask.
+ * Once stored, these masks will then be accessed through each DRM's
+ * interface and used by the caller of mdp5_ctl_commit() to specify
+ * which block(s) need to be flushed through @flush_mask parameter.
+ */
+
+#define MDP5_CTL_FLUSH_CURSOR_DUMMY     0x80000000
+
+static inline u32 mdp_ctl_flush_mask_cursor(int cursor_id)
+{
+        /* TODO: use id once multiple cursor support is present */
+        (void)cursor_id;
+
+        return MDP5_CTL_FLUSH_CURSOR_DUMMY;
+}
+
+static inline u32 mdp_ctl_flush_mask_lm(int lm)
+{
+        switch (lm) {
+        case 0:  return MDP5_CTL_FLUSH_LM0;
+        case 1:  return MDP5_CTL_FLUSH_LM1;
+        case 2:  return MDP5_CTL_FLUSH_LM2;
+        case 5:  return MDP5_CTL_FLUSH_LM5;
+        default: return 0;
+        }
+}
+
+static inline u32 mdp_ctl_flush_mask_pipe(enum mdp5_pipe pipe)
+{
+        switch (pipe) {
+        case SSPP_VIG0: return MDP5_CTL_FLUSH_VIG0;
+        case SSPP_VIG1: return MDP5_CTL_FLUSH_VIG1;
+        case SSPP_VIG2: return MDP5_CTL_FLUSH_VIG2;
+        case SSPP_RGB0: return MDP5_CTL_FLUSH_RGB0;
+        case SSPP_RGB1: return MDP5_CTL_FLUSH_RGB1;
+        case SSPP_RGB2: return MDP5_CTL_FLUSH_RGB2;
+        case SSPP_DMA0: return MDP5_CTL_FLUSH_DMA0;
+        case SSPP_DMA1: return MDP5_CTL_FLUSH_DMA1;
+        case SSPP_VIG3: return MDP5_CTL_FLUSH_VIG3;
+        case SSPP_RGB3: return MDP5_CTL_FLUSH_RGB3;
+        default:        return 0;
+        }
+}
+
+#endif /* __MDP5_CTL_H__ */
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c
index edec7bfaa952..0254bfdeb92f 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_encoder.c
@@ -24,6 +24,7 @@ struct mdp5_encoder {
         struct drm_encoder base;
         int intf;
         enum mdp5_intf intf_id;
+        spinlock_t intf_lock;   /* protect REG_MDP5_INTF_* registers */
         bool enabled;
         uint32_t bsc;
 };
@@ -115,6 +116,7 @@ static void mdp5_encoder_dpms(struct drm_encoder *encoder, int mode)
         struct mdp5_kms *mdp5_kms = get_kms(encoder);
         int intf = mdp5_encoder->intf;
         bool enabled = (mode == DRM_MODE_DPMS_ON);
+        unsigned long flags;
 
         DBG("mode=%d", mode);
 
@@ -123,9 +125,24 @@ static void mdp5_encoder_dpms(struct drm_encoder *encoder, int mode)
 
         if (enabled) {
                 bs_set(mdp5_encoder, 1);
+                spin_lock_irqsave(&mdp5_encoder->intf_lock, flags);
                 mdp5_write(mdp5_kms, REG_MDP5_INTF_TIMING_ENGINE_EN(intf), 1);
+                spin_unlock_irqrestore(&mdp5_encoder->intf_lock, flags);
         } else {
+                spin_lock_irqsave(&mdp5_encoder->intf_lock, flags);
                 mdp5_write(mdp5_kms, REG_MDP5_INTF_TIMING_ENGINE_EN(intf), 0);
+                spin_unlock_irqrestore(&mdp5_encoder->intf_lock, flags);
+
+                /*
+                 * Wait for a vsync so we know the ENABLE=0 latched before
+                 * the (connector) source of the vsync's gets disabled,
+                 * otherwise we end up in a funny state if we re-enable
+                 * before the disable latches, which results that some of
+                 * the settings changes for the new modeset (like new
+                 * scanout buffer) don't latch properly..
+                 */
+                mdp_irq_wait(&mdp5_kms->base, intf2vblank(intf));
+
                 bs_set(mdp5_encoder, 0);
         }
 
@@ -150,6 +167,7 @@ static void mdp5_encoder_mode_set(struct drm_encoder *encoder,
         uint32_t display_v_start, display_v_end;
         uint32_t hsync_start_x, hsync_end_x;
         uint32_t format;
+        unsigned long flags;
 
         mode = adjusted_mode;
 
@@ -180,6 +198,8 @@ static void mdp5_encoder_mode_set(struct drm_encoder *encoder,
         display_v_start = (mode->vtotal - mode->vsync_start) * mode->htotal + dtv_hsync_skew;
         display_v_end = vsync_period - ((mode->vsync_start - mode->vdisplay) * mode->htotal) + dtv_hsync_skew - 1;
 
+        spin_lock_irqsave(&mdp5_encoder->intf_lock, flags);
+
         mdp5_write(mdp5_kms, REG_MDP5_INTF_HSYNC_CTL(intf),
                         MDP5_INTF_HSYNC_CTL_PULSEW(mode->hsync_end - mode->hsync_start) |
                         MDP5_INTF_HSYNC_CTL_PERIOD(mode->htotal));
@@ -201,6 +221,8 @@ static void mdp5_encoder_mode_set(struct drm_encoder *encoder,
         mdp5_write(mdp5_kms, REG_MDP5_INTF_ACTIVE_VEND_F0(intf), 0);
         mdp5_write(mdp5_kms, REG_MDP5_INTF_PANEL_FORMAT(intf), format);
         mdp5_write(mdp5_kms, REG_MDP5_INTF_FRAME_LINE_COUNT_EN(intf), 0x3);  /* frame+line? */
+
+        spin_unlock_irqrestore(&mdp5_encoder->intf_lock, flags);
 }
 
 static void mdp5_encoder_prepare(struct drm_encoder *encoder)
@@ -242,6 +264,8 @@ struct drm_encoder *mdp5_encoder_init(struct drm_device *dev, int intf,
         mdp5_encoder->intf_id = intf_id;
         encoder = &mdp5_encoder->base;
 
+        spin_lock_init(&mdp5_encoder->intf_lock);
+
         drm_encoder_init(dev, encoder, &mdp5_encoder_funcs,
                          DRM_MODE_ENCODER_TMDS);
         drm_encoder_helper_add(encoder, &mdp5_encoder_helper_funcs);
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c
index f2b985bc2adf..70ac81edd40f 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_irq.c
@@ -15,6 +15,8 @@
  * this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
+#include <linux/irqdomain.h>
+#include <linux/irq.h>
 
 #include "msm_drv.h"
 #include "mdp5_kms.h"
@@ -88,11 +90,17 @@ irqreturn_t mdp5_irq(struct msm_kms *kms)
 
         VERB("intr=%08x", intr);
 
-        if (intr & MDP5_HW_INTR_STATUS_INTR_MDP)
+        if (intr & MDP5_HW_INTR_STATUS_INTR_MDP) {
                 mdp5_irq_mdp(mdp_kms);
+                intr &= ~MDP5_HW_INTR_STATUS_INTR_MDP;
+        }
 
-        if (intr & MDP5_HW_INTR_STATUS_INTR_HDMI)
-                hdmi_irq(0, mdp5_kms->hdmi);
+        while (intr) {
+                irq_hw_number_t hwirq = fls(intr) - 1;
+                generic_handle_irq(irq_find_mapping(
+                                mdp5_kms->irqcontroller.domain, hwirq));
+                intr &= ~(1 << hwirq);
+        }
 
         return IRQ_HANDLED;
 }
@@ -109,3 +117,82 @@ void mdp5_disable_vblank(struct msm_kms *kms, struct drm_crtc *crtc)
         mdp_update_vblank_mask(to_mdp_kms(kms),
                         mdp5_crtc_vblank(crtc), false);
 }
+
+/*
+ * interrupt-controller implementation, so sub-blocks (hdmi/eDP/dsi/etc)
+ * can register to get their irq's delivered
+ */
+
+#define VALID_IRQS  (MDP5_HW_INTR_STATUS_INTR_DSI0 | \
+                MDP5_HW_INTR_STATUS_INTR_DSI1 | \
+                MDP5_HW_INTR_STATUS_INTR_HDMI | \
+                MDP5_HW_INTR_STATUS_INTR_EDP)
+
+static void mdp5_hw_mask_irq(struct irq_data *irqd)
+{
+        struct mdp5_kms *mdp5_kms = irq_data_get_irq_chip_data(irqd);
+        smp_mb__before_atomic();
+        clear_bit(irqd->hwirq, &mdp5_kms->irqcontroller.enabled_mask);
+        smp_mb__after_atomic();
+}
+
+static void mdp5_hw_unmask_irq(struct irq_data *irqd)
+{
+        struct mdp5_kms *mdp5_kms = irq_data_get_irq_chip_data(irqd);
+        smp_mb__before_atomic();
+        set_bit(irqd->hwirq, &mdp5_kms->irqcontroller.enabled_mask);
+        smp_mb__after_atomic();
+}
+
+static struct irq_chip mdp5_hw_irq_chip = {
+        .name           = "mdp5",
+        .irq_mask       = mdp5_hw_mask_irq,
+        .irq_unmask     = mdp5_hw_unmask_irq,
+};
+
+static int mdp5_hw_irqdomain_map(struct irq_domain *d,
+                unsigned int irq, irq_hw_number_t hwirq)
+{
+        struct mdp5_kms *mdp5_kms = d->host_data;
+
+        if (!(VALID_IRQS & (1 << hwirq)))
+                return -EPERM;
+
+        irq_set_chip_and_handler(irq, &mdp5_hw_irq_chip, handle_level_irq);
+        irq_set_chip_data(irq, mdp5_kms);
+        set_irq_flags(irq, IRQF_VALID);
+
+        return 0;
+}
+
+static struct irq_domain_ops mdp5_hw_irqdomain_ops = {
+        .map = mdp5_hw_irqdomain_map,
+        .xlate = irq_domain_xlate_onecell,
+};
+
+
+int mdp5_irq_domain_init(struct mdp5_kms *mdp5_kms)
+{
+        struct device *dev = mdp5_kms->dev->dev;
+        struct irq_domain *d;
+
+        d = irq_domain_add_linear(dev->of_node, 32,
+                        &mdp5_hw_irqdomain_ops, mdp5_kms);
+        if (!d) {
+                dev_err(dev, "mdp5 irq domain add failed\n");
+                return -ENXIO;
+        }
+
+        mdp5_kms->irqcontroller.enabled_mask = 0;
+        mdp5_kms->irqcontroller.domain = d;
+
+        return 0;
+}
+
+void mdp5_irq_domain_fini(struct mdp5_kms *mdp5_kms)
+{
+        if (mdp5_kms->irqcontroller.domain) {
+                irq_domain_remove(mdp5_kms->irqcontroller.domain);
+                mdp5_kms->irqcontroller.domain = NULL;
+        }
+}
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c
index 31a2c6331a1d..a11f1b80c488 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.c
@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2014, The Linux Foundation. All rights reserved.
  * Copyright (C) 2013 Red Hat
  * Author: Rob Clark <robdclark@gmail.com>
  *
@@ -24,145 +25,11 @@ static const char *iommu_ports[] = {
                 "mdp_0",
 };
 
-static struct mdp5_platform_config *mdp5_get_config(struct platform_device *dev);
-
-const struct mdp5_config *mdp5_cfg;
-
-static const struct mdp5_config msm8x74_config = {
-        .name = "msm8x74",
-        .ctl = {
-                .count = 5,
-                .base = { 0x00600, 0x00700, 0x00800, 0x00900, 0x00a00 },
-        },
-        .pipe_vig = {
-                .count = 3,
-                .base = { 0x01200, 0x01600, 0x01a00 },
-        },
-        .pipe_rgb = {
-                .count = 3,
-                .base = { 0x01e00, 0x02200, 0x02600 },
-        },
-        .pipe_dma = {
-                .count = 2,
-                .base = { 0x02a00, 0x02e00 },
-        },
-        .lm = {
-                .count = 5,
-                .base = { 0x03200, 0x03600, 0x03a00, 0x03e00, 0x04200 },
-        },
-        .dspp = {
-                .count = 3,
-                .base = { 0x04600, 0x04a00, 0x04e00 },
-        },
-        .ad = {
-                .count = 2,
-                .base = { 0x13100, 0x13300 }, /* NOTE: no ad in v1.0 */
-        },
-        .intf = {
-                .count = 4,
-                .base = { 0x12500, 0x12700, 0x12900, 0x12b00 },
-        },
-};
-
-static const struct mdp5_config apq8084_config = {
-        .name = "apq8084",
-        .ctl = {
-                .count = 5,
-                .base = { 0x00600, 0x00700, 0x00800, 0x00900, 0x00a00 },
-        },
-        .pipe_vig = {
-                .count = 4,
-                .base = { 0x01200, 0x01600, 0x01a00, 0x01e00 },
-        },
-        .pipe_rgb = {
-                .count = 4,
-                .base = { 0x02200, 0x02600, 0x02a00, 0x02e00 },
-        },
-        .pipe_dma = {
-                .count = 2,
-                .base = { 0x03200, 0x03600 },
-        },
-        .lm = {
-                .count = 6,
-                .base = { 0x03a00, 0x03e00, 0x04200, 0x04600, 0x04a00, 0x04e00 },
-        },
-        .dspp = {
-                .count = 4,
-                .base = { 0x05200, 0x05600, 0x05a00, 0x05e00 },
-
-        },
-        .ad = {
-                .count = 3,
-                .base = { 0x13500, 0x13700, 0x13900 },
-        },
-        .intf = {
-                .count = 5,
-                .base = { 0x12500, 0x12700, 0x12900, 0x12b00, 0x12d00 },
-        },
-};
-
-struct mdp5_config_entry {
-        int revision;
-        const struct mdp5_config *config;
-};
-
-static const struct mdp5_config_entry mdp5_configs[] = {
-        { .revision = 0, .config = &msm8x74_config },
-        { .revision = 2, .config = &msm8x74_config },
-        { .revision = 3, .config = &apq8084_config },
-};
-
-static int mdp5_select_hw_cfg(struct msm_kms *kms)
-{
-        struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms));
-        struct drm_device *dev = mdp5_kms->dev;
-        uint32_t version, major, minor;
-        int i, ret = 0;
-
-        mdp5_enable(mdp5_kms);
-        version = mdp5_read(mdp5_kms, REG_MDP5_MDP_VERSION);
-        mdp5_disable(mdp5_kms);
-
-        major = FIELD(version, MDP5_MDP_VERSION_MAJOR);
-        minor = FIELD(version, MDP5_MDP_VERSION_MINOR);
-
-        DBG("found MDP5 version v%d.%d", major, minor);
-
-        if (major != 1) {
-                dev_err(dev->dev, "unexpected MDP major version: v%d.%d\n",
-                                major, minor);
-                ret = -ENXIO;
-                goto out;
-        }
-
-        mdp5_kms->rev = minor;
-
-        /* only after mdp5_cfg global pointer's init can we access the hw */
-        for (i = 0; i < ARRAY_SIZE(mdp5_configs); i++) {
-                if (mdp5_configs[i].revision != minor)
-                        continue;
-                mdp5_kms->hw_cfg = mdp5_cfg = mdp5_configs[i].config;
-                break;
-        }
-        if (unlikely(!mdp5_kms->hw_cfg)) {
-                dev_err(dev->dev, "unexpected MDP minor revision: v%d.%d\n",
-                                major, minor);
-                ret = -ENXIO;
-                goto out;
-        }
-
-        DBG("MDP5: %s config selected", mdp5_kms->hw_cfg->name);
-
-        return 0;
-out:
-        return ret;
-}
-
 static int mdp5_hw_init(struct msm_kms *kms)
 {
         struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms));
         struct drm_device *dev = mdp5_kms->dev;
-        int i;
+        unsigned long flags;
 
         pm_runtime_get_sync(dev->dev);
 
@@ -190,10 +57,11 @@ static int mdp5_hw_init(struct msm_kms *kms)
          * care.
          */
 
+        spin_lock_irqsave(&mdp5_kms->resource_lock, flags);
         mdp5_write(mdp5_kms, REG_MDP5_DISP_INTF_SEL, 0);
+        spin_unlock_irqrestore(&mdp5_kms->resource_lock, flags);
 
-        for (i = 0; i < mdp5_kms->hw_cfg->ctl.count; i++)
-                mdp5_write(mdp5_kms, REG_MDP5_CTL_OP(i), 0);
+        mdp5_ctlm_hw_reset(mdp5_kms->ctlm);
 
         pm_runtime_put_sync(dev->dev);
 
@@ -221,10 +89,20 @@ static void mdp5_destroy(struct msm_kms *kms)
         struct mdp5_kms *mdp5_kms = to_mdp5_kms(to_mdp_kms(kms));
         struct msm_mmu *mmu = mdp5_kms->mmu;
 
+        mdp5_irq_domain_fini(mdp5_kms);
+
         if (mmu) {
                 mmu->funcs->detach(mmu, iommu_ports, ARRAY_SIZE(iommu_ports));
                 mmu->funcs->destroy(mmu);
         }
+
+        if (mdp5_kms->ctlm)
+                mdp5_ctlm_destroy(mdp5_kms->ctlm);
+        if (mdp5_kms->smp)
+                mdp5_smp_destroy(mdp5_kms->smp);
+        if (mdp5_kms->cfg)
+                mdp5_cfg_destroy(mdp5_kms->cfg);
+
         kfree(mdp5_kms);
 }
 
@@ -274,17 +152,31 @@ static int modeset_init(struct mdp5_kms *mdp5_kms)
         static const enum mdp5_pipe crtcs[] = {
                         SSPP_RGB0, SSPP_RGB1, SSPP_RGB2, SSPP_RGB3,
         };
+        static const enum mdp5_pipe pub_planes[] = {
+                        SSPP_VIG0, SSPP_VIG1, SSPP_VIG2, SSPP_VIG3,
+        };
         struct drm_device *dev = mdp5_kms->dev;
         struct msm_drm_private *priv = dev->dev_private;
         struct drm_encoder *encoder;
+        const struct mdp5_cfg_hw *hw_cfg;
         int i, ret;
 
-        /* construct CRTCs: */
-        for (i = 0; i < mdp5_kms->hw_cfg->pipe_rgb.count; i++) {
+        hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg);
+
+        /* register our interrupt-controller for hdmi/eDP/dsi/etc
+         * to use for irqs routed through mdp:
+         */
+        ret = mdp5_irq_domain_init(mdp5_kms);
+        if (ret)
+                goto fail;
+
+        /* construct CRTCs and their private planes: */
+        for (i = 0; i < hw_cfg->pipe_rgb.count; i++) {
                 struct drm_plane *plane;
                 struct drm_crtc *crtc;
 
-                plane = mdp5_plane_init(dev, crtcs[i], true);
+                plane = mdp5_plane_init(dev, crtcs[i], true,
+                                hw_cfg->pipe_rgb.base[i]);
                 if (IS_ERR(plane)) {
                         ret = PTR_ERR(plane);
                         dev_err(dev->dev, "failed to construct plane for %s (%d)\n",
@@ -302,6 +194,20 @@ static int modeset_init(struct mdp5_kms *mdp5_kms)
                 priv->crtcs[priv->num_crtcs++] = crtc;
         }
 
+        /* Construct public planes: */
+        for (i = 0; i < hw_cfg->pipe_vig.count; i++) {
+                struct drm_plane *plane;
+
+                plane = mdp5_plane_init(dev, pub_planes[i], false,
+                                hw_cfg->pipe_vig.base[i]);
+                if (IS_ERR(plane)) {
+                        ret = PTR_ERR(plane);
+                        dev_err(dev->dev, "failed to construct %s plane: %d\n",
+                                        pipe2name(pub_planes[i]), ret);
+                        goto fail;
+                }
+        }
+
         /* Construct encoder for HDMI: */
         encoder = mdp5_encoder_init(dev, 3, INTF_HDMI);
         if (IS_ERR(encoder)) {
@@ -324,11 +230,12 @@ static int modeset_init(struct mdp5_kms *mdp5_kms)
         priv->encoders[priv->num_encoders++] = encoder;
 
         /* Construct bridge/connector for HDMI: */
-        mdp5_kms->hdmi = hdmi_init(dev, encoder);
-        if (IS_ERR(mdp5_kms->hdmi)) {
-                ret = PTR_ERR(mdp5_kms->hdmi);
-                dev_err(dev->dev, "failed to initialize HDMI: %d\n", ret);
-                goto fail;
+        if (priv->hdmi) {
+                ret = hdmi_modeset_init(priv->hdmi, dev, encoder);
+                if (ret) {
+                        dev_err(dev->dev, "failed to initialize HDMI: %d\n", ret);
+                        goto fail;
+                }
         }
 
         return 0;
@@ -337,6 +244,21 @@ fail:
         return ret;
 }
 
+static void read_hw_revision(struct mdp5_kms *mdp5_kms,
+                uint32_t *major, uint32_t *minor)
+{
+        uint32_t version;
+
+        mdp5_enable(mdp5_kms);
+        version = mdp5_read(mdp5_kms, REG_MDP5_MDP_VERSION);
+        mdp5_disable(mdp5_kms);
+
+        *major = FIELD(version, MDP5_MDP_VERSION_MAJOR);
+        *minor = FIELD(version, MDP5_MDP_VERSION_MINOR);
+
+        DBG("MDP5 version v%d.%d", *major, *minor);
+}
+
 static int get_clk(struct platform_device *pdev, struct clk **clkp,
                 const char *name)
 {
@@ -353,10 +275,11 @@ static int get_clk(struct platform_device *pdev, struct clk **clkp,
 struct msm_kms *mdp5_kms_init(struct drm_device *dev)
 {
         struct platform_device *pdev = dev->platformdev;
-        struct mdp5_platform_config *config = mdp5_get_config(pdev);
+        struct mdp5_cfg *config;
         struct mdp5_kms *mdp5_kms;
         struct msm_kms *kms = NULL;
         struct msm_mmu *mmu;
+        uint32_t major, minor;
         int i, ret;
 
         mdp5_kms = kzalloc(sizeof(*mdp5_kms), GFP_KERNEL);
@@ -366,12 +289,13 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev)
                 goto fail;
         }
 
+        spin_lock_init(&mdp5_kms->resource_lock);
+
         mdp_kms_init(&mdp5_kms->base, &kms_funcs);
 
         kms = &mdp5_kms->base.base;
 
         mdp5_kms->dev = dev;
-        mdp5_kms->smp_blk_cnt = config->smp_blk_cnt;
 
         mdp5_kms->mmio = msm_ioremap(pdev, "mdp_phys", "MDP5");
         if (IS_ERR(mdp5_kms->mmio)) {
@@ -416,24 +340,52 @@ struct msm_kms *mdp5_kms_init(struct drm_device *dev)
         if (ret)
                 goto fail;
 
-        ret = clk_set_rate(mdp5_kms->src_clk, config->max_clk);
+        /* we need to set a default rate before enabling.  Set a safe
+         * rate first, then figure out hw revision, and then set a
+         * more optimal rate:
+         */
+        clk_set_rate(mdp5_kms->src_clk, 200000000);
+
+        read_hw_revision(mdp5_kms, &major, &minor);
 
-        ret = mdp5_select_hw_cfg(kms);
-        if (ret)
+        mdp5_kms->cfg = mdp5_cfg_init(mdp5_kms, major, minor);
+        if (IS_ERR(mdp5_kms->cfg)) {
+                ret = PTR_ERR(mdp5_kms->cfg);
+                mdp5_kms->cfg = NULL;
                 goto fail;
+        }
+
+        config = mdp5_cfg_get_config(mdp5_kms->cfg);
+
+        /* TODO: compute core clock rate at runtime */
+        clk_set_rate(mdp5_kms->src_clk, config->hw->max_clk);
+
+        mdp5_kms->smp = mdp5_smp_init(mdp5_kms->dev, &config->hw->smp);
+        if (IS_ERR(mdp5_kms->smp)) {
+                ret = PTR_ERR(mdp5_kms->smp);
+                mdp5_kms->smp = NULL;
+                goto fail;
+        }
+
+        mdp5_kms->ctlm = mdp5_ctlm_init(dev, mdp5_kms->mmio, config->hw);
+        if (IS_ERR(mdp5_kms->ctlm)) {
+                ret = PTR_ERR(mdp5_kms->ctlm);
+                mdp5_kms->ctlm = NULL;
+                goto fail;
+        }
 
         /* make sure things are off before attaching iommu (bootloader could
          * have left things on, in which case we'll start getting faults if
          * we don't disable):
          */
         mdp5_enable(mdp5_kms);
-        for (i = 0; i < mdp5_kms->hw_cfg->intf.count; i++)
+        for (i = 0; i < config->hw->intf.count; i++)
                 mdp5_write(mdp5_kms, REG_MDP5_INTF_TIMING_ENGINE_EN(i), 0);
         mdp5_disable(mdp5_kms);
         mdelay(16);
 
-        if (config->iommu) {
-                mmu = msm_iommu_new(&pdev->dev, config->iommu);
+        if (config->platform.iommu) {
+                mmu = msm_iommu_new(&pdev->dev, config->platform.iommu);
                 if (IS_ERR(mmu)) {
                         ret = PTR_ERR(mmu);
                         dev_err(dev->dev, "failed to init iommu: %d\n", ret);
@@ -474,18 +426,3 @@ fail:
                 mdp5_destroy(kms);
         return ERR_PTR(ret);
 }
-
-static struct mdp5_platform_config *mdp5_get_config(struct platform_device *dev)
-{
-        static struct mdp5_platform_config config = {};
-#ifdef CONFIG_OF
-        /* TODO */
-#endif
-        config.iommu = iommu_domain_alloc(&platform_bus_type);
-        /* TODO hard-coded in downstream mdss, but should it be? */
-        config.max_clk = 200000000;
-        /* TODO get from DT: */
-        config.smp_blk_cnt = 22;
-
-        return &config;
-}
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h
index 5bf340dd0f00..dd69c77c0d64 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_kms.h
@@ -21,25 +21,9 @@
 #include "msm_drv.h"
 #include "msm_kms.h"
 #include "mdp/mdp_kms.h"
-/* dynamic offsets used by mdp5.xml.h (initialized in mdp5_kms.c) */
-#define MDP5_MAX_BASES          8
-struct mdp5_sub_block {
-        int     count;
-        uint32_t base[MDP5_MAX_BASES];
-};
-struct mdp5_config {
-        char  *name;
-        struct mdp5_sub_block ctl;
-        struct mdp5_sub_block pipe_vig;
-        struct mdp5_sub_block pipe_rgb;
-        struct mdp5_sub_block pipe_dma;
-        struct mdp5_sub_block lm;
-        struct mdp5_sub_block dspp;
-        struct mdp5_sub_block ad;
-        struct mdp5_sub_block intf;
-};
-extern const struct mdp5_config *mdp5_cfg;
+#include "mdp5_cfg.h"   /* must be included before mdp5.xml.h */
 #include "mdp5.xml.h"
+#include "mdp5_ctl.h"
 #include "mdp5_smp.h"
 
 struct mdp5_kms {
@@ -47,17 +31,14 @@ struct mdp5_kms {
 
         struct drm_device *dev;
 
-        int rev;
-        const struct mdp5_config *hw_cfg;
+        struct mdp5_cfg_handler *cfg;
 
         /* mapper-id used to request GEM buffer mapped for scanout: */
         int id;
         struct msm_mmu *mmu;
 
-        /* for tracking smp allocation amongst pipes: */
-        mdp5_smp_state_t smp_state;
-        struct mdp5_client_smp_state smp_client_state[CID_MAX];
-        int smp_blk_cnt;
+        struct mdp5_smp *smp;
+        struct mdp5_ctl_manager *ctlm;
 
         /* io/register spaces: */
         void __iomem *mmio, *vbif;
@@ -71,18 +52,47 @@ struct mdp5_kms {
         struct clk *lut_clk;
         struct clk *vsync_clk;
 
-        struct hdmi *hdmi;
+        /*
+         * lock to protect access to global resources: ie., following register:
+         *      - REG_MDP5_DISP_INTF_SEL
+         */
+        spinlock_t resource_lock;
 
         struct mdp_irq error_handler;
+
+        struct {
+                volatile unsigned long enabled_mask;
+                struct irq_domain *domain;
+        } irqcontroller;
 };
 #define to_mdp5_kms(x) container_of(x, struct mdp5_kms, base)
 
-/* platform config data (ie. from DT, or pdata) */
-struct mdp5_platform_config {
-        struct iommu_domain *iommu;
-        uint32_t max_clk;
-        int smp_blk_cnt;
+struct mdp5_plane_state {
+        struct drm_plane_state base;
+
+        /* "virtual" zpos.. we calculate actual mixer-stage at runtime
+         * by sorting the attached planes by zpos and then assigning
+         * mixer stage lowest to highest.  Private planes get default
+         * zpos of zero, and public planes a unique value that is
+         * greater than zero.  This way, things work out if a naive
+         * userspace assigns planes to a crtc without setting zpos.
+         */
+        int zpos;
+
+        /* the actual mixer stage, calculated in crtc->atomic_check()
+         * NOTE: this should move to mdp5_crtc_state, when that exists
+         */
+        enum mdp_mixer_stage_id stage;
+
+        /* some additional transactional status to help us know in the
+         * apply path whether we need to update SMP allocation, and
+         * whether current update is still pending:
+         */
+        bool mode_changed : 1;
+        bool pending : 1;
 };
+#define to_mdp5_plane_state(x) \
+                container_of(x, struct mdp5_plane_state, base)
 
 static inline void mdp5_write(struct mdp5_kms *mdp5_kms, u32 reg, u32 data)
 {
@@ -107,23 +117,6 @@ static inline const char *pipe2name(enum mdp5_pipe pipe)
         return names[pipe];
 }
 
-static inline uint32_t pipe2flush(enum mdp5_pipe pipe)
-{
-        switch (pipe) {
-        case SSPP_VIG0: return MDP5_CTL_FLUSH_VIG0;
-        case SSPP_VIG1: return MDP5_CTL_FLUSH_VIG1;
-        case SSPP_VIG2: return MDP5_CTL_FLUSH_VIG2;
-        case SSPP_RGB0: return MDP5_CTL_FLUSH_RGB0;
-        case SSPP_RGB1: return MDP5_CTL_FLUSH_RGB1;
-        case SSPP_RGB2: return MDP5_CTL_FLUSH_RGB2;
-        case SSPP_DMA0: return MDP5_CTL_FLUSH_DMA0;
-        case SSPP_DMA1: return MDP5_CTL_FLUSH_DMA1;
-        case SSPP_VIG3: return MDP5_CTL_FLUSH_VIG3;
-        case SSPP_RGB3: return MDP5_CTL_FLUSH_RGB3;
-        default:        return 0;
-        }
-}
-
 static inline int pipe2nclients(enum mdp5_pipe pipe)
 {
         switch (pipe) {
@@ -137,34 +130,6 @@ static inline int pipe2nclients(enum mdp5_pipe pipe)
         }
 }
 
-static inline enum mdp5_client_id pipe2client(enum mdp5_pipe pipe, int plane)
-{
-        WARN_ON(plane >= pipe2nclients(pipe));
-        switch (pipe) {
-        case SSPP_VIG0: return CID_VIG0_Y + plane;
-        case SSPP_VIG1: return CID_VIG1_Y + plane;
-        case SSPP_VIG2: return CID_VIG2_Y + plane;
-        case SSPP_RGB0: return CID_RGB0;
-        case SSPP_RGB1: return CID_RGB1;
-        case SSPP_RGB2: return CID_RGB2;
-        case SSPP_DMA0: return CID_DMA0_Y + plane;
-        case SSPP_DMA1: return CID_DMA1_Y + plane;
-        case SSPP_VIG3: return CID_VIG3_Y + plane;
-        case SSPP_RGB3: return CID_RGB3;
-        default:        return CID_UNUSED;
-        }
-}
-
-static inline uint32_t mixer2flush(int lm)
-{
-        switch (lm) {
-        case 0:  return MDP5_CTL_FLUSH_LM0;
-        case 1:  return MDP5_CTL_FLUSH_LM1;
-        case 2:  return MDP5_CTL_FLUSH_LM2;
-        default: return 0;
-        }
-}
-
 static inline uint32_t intf2err(int intf)
 {
         switch (intf) {
@@ -197,6 +162,8 @@ void mdp5_irq_uninstall(struct msm_kms *kms);
 irqreturn_t mdp5_irq(struct msm_kms *kms);
 int mdp5_enable_vblank(struct msm_kms *kms, struct drm_crtc *crtc);
 void mdp5_disable_vblank(struct msm_kms *kms, struct drm_crtc *crtc);
+int mdp5_irq_domain_init(struct mdp5_kms *mdp5_kms);
+void mdp5_irq_domain_fini(struct mdp5_kms *mdp5_kms);
 
 static inline
 uint32_t mdp5_get_formats(enum mdp5_pipe pipe, uint32_t *pixel_formats,
@@ -210,26 +177,18 @@ uint32_t mdp5_get_formats(enum mdp5_pipe pipe, uint32_t *pixel_formats,
 
 void mdp5_plane_install_properties(struct drm_plane *plane,
                 struct drm_mode_object *obj);
-void mdp5_plane_set_scanout(struct drm_plane *plane,
-                struct drm_framebuffer *fb);
-int mdp5_plane_mode_set(struct drm_plane *plane,
-                struct drm_crtc *crtc, struct drm_framebuffer *fb,
-                int crtc_x, int crtc_y,
-                unsigned int crtc_w, unsigned int crtc_h,
-                uint32_t src_x, uint32_t src_y,
-                uint32_t src_w, uint32_t src_h);
+uint32_t mdp5_plane_get_flush(struct drm_plane *plane);
 void mdp5_plane_complete_flip(struct drm_plane *plane);
 enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane);
 struct drm_plane *mdp5_plane_init(struct drm_device *dev,
-                enum mdp5_pipe pipe, bool private_plane);
+                enum mdp5_pipe pipe, bool private_plane, uint32_t reg_offset);
 
 uint32_t mdp5_crtc_vblank(struct drm_crtc *crtc);
 
+int mdp5_crtc_get_lm(struct drm_crtc *crtc);
 void mdp5_crtc_cancel_pending_flip(struct drm_crtc *crtc, struct drm_file *file);
 void mdp5_crtc_set_intf(struct drm_crtc *crtc, int intf,
                 enum mdp5_intf intf_id);
-void mdp5_crtc_attach(struct drm_crtc *crtc, struct drm_plane *plane);
-void mdp5_crtc_detach(struct drm_crtc *crtc, struct drm_plane *plane);
 struct drm_crtc *mdp5_crtc_init(struct drm_device *dev,
                 struct drm_plane *plane, int id);
 
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c
index f3daec4412ad..26e5fdea6594 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c
@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2014 The Linux Foundation. All rights reserved.
  * Copyright (C) 2013 Red Hat
  * Author: Rob Clark <robdclark@gmail.com>
  *
@@ -17,6 +18,7 @@
 
 #include "mdp5_kms.h"
 
+#define MAX_PLANE       4
 
 struct mdp5_plane {
         struct drm_plane base;
@@ -24,6 +26,11 @@ struct mdp5_plane {
 
         enum mdp5_pipe pipe;
 
+        spinlock_t pipe_lock;   /* protect REG_MDP5_PIPE_* registers */
+        uint32_t reg_offset;
+
+        uint32_t flush_mask;    /* used to commit pipe registers */
+
         uint32_t nformats;
         uint32_t formats[32];
 
@@ -31,31 +38,24 @@ struct mdp5_plane {
 };
 #define to_mdp5_plane(x) container_of(x, struct mdp5_plane, base)
 
+static int mdp5_plane_mode_set(struct drm_plane *plane,
+                struct drm_crtc *crtc, struct drm_framebuffer *fb,
+                int crtc_x, int crtc_y,
+                unsigned int crtc_w, unsigned int crtc_h,
+                uint32_t src_x, uint32_t src_y,
+                uint32_t src_w, uint32_t src_h);
+static void set_scanout_locked(struct drm_plane *plane,
+                struct drm_framebuffer *fb);
+
 static struct mdp5_kms *get_kms(struct drm_plane *plane)
 {
         struct msm_drm_private *priv = plane->dev->dev_private;
         return to_mdp5_kms(to_mdp_kms(priv->kms));
 }
 
-static int mdp5_plane_update(struct drm_plane *plane,
-                struct drm_crtc *crtc, struct drm_framebuffer *fb,
-                int crtc_x, int crtc_y,
-                unsigned int crtc_w, unsigned int crtc_h,
-                uint32_t src_x, uint32_t src_y,
-                uint32_t src_w, uint32_t src_h)
+static bool plane_enabled(struct drm_plane_state *state)
 {
-        struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
-
-        mdp5_plane->enabled = true;
-
-        if (plane->fb)
-                drm_framebuffer_unreference(plane->fb);
-
-        drm_framebuffer_reference(fb);
-
-        return mdp5_plane_mode_set(plane, crtc, fb,
-                        crtc_x, crtc_y, crtc_w, crtc_h,
-                        src_x, src_y, src_w, src_h);
+        return state->fb && state->crtc;
 }
 
 static int mdp5_plane_disable(struct drm_plane *plane)
@@ -63,21 +63,13 @@ static int mdp5_plane_disable(struct drm_plane *plane)
         struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
         struct mdp5_kms *mdp5_kms = get_kms(plane);
         enum mdp5_pipe pipe = mdp5_plane->pipe;
-        int i;
 
         DBG("%s: disable", mdp5_plane->name);
 
-        /* update our SMP request to zero (release all our blks): */
-        for (i = 0; i < pipe2nclients(pipe); i++)
-                mdp5_smp_request(mdp5_kms, pipe2client(pipe, i), 0);
-
-        /* TODO detaching now will cause us not to get the last
-         * vblank and mdp5_smp_commit().. so other planes will
-         * still see smp blocks previously allocated to us as
-         * in-use..
-         */
-        if (plane->crtc)
-                mdp5_crtc_detach(plane->crtc, plane);
+        if (mdp5_kms) {
+                /* Release the memory we requested earlier from the SMP: */
+                mdp5_smp_release(mdp5_kms->smp, pipe);
+        }
 
         return 0;
 }
@@ -85,11 +77,8 @@ static int mdp5_plane_disable(struct drm_plane *plane)
 static void mdp5_plane_destroy(struct drm_plane *plane)
 {
         struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
-        struct msm_drm_private *priv = plane->dev->dev_private;
-
-        if (priv->kms)
-                mdp5_plane_disable(plane);
 
+        drm_plane_helper_disable(plane);
         drm_plane_cleanup(plane);
 
         kfree(mdp5_plane);
@@ -109,109 +98,186 @@ int mdp5_plane_set_property(struct drm_plane *plane,
         return -EINVAL;
 }
 
+static void mdp5_plane_reset(struct drm_plane *plane)
+{
+        struct mdp5_plane_state *mdp5_state;
+
+        if (plane->state && plane->state->fb)
+                drm_framebuffer_unreference(plane->state->fb);
+
+        kfree(to_mdp5_plane_state(plane->state));
+        mdp5_state = kzalloc(sizeof(*mdp5_state), GFP_KERNEL);
+
+        if (plane->type == DRM_PLANE_TYPE_PRIMARY) {
+                mdp5_state->zpos = 0;
+        } else {
+                mdp5_state->zpos = 1 + drm_plane_index(plane);
+        }
+
+        plane->state = &mdp5_state->base;
+}
+
+static struct drm_plane_state *
+mdp5_plane_duplicate_state(struct drm_plane *plane)
+{
+        struct mdp5_plane_state *mdp5_state;
+
+        if (WARN_ON(!plane->state))
+                return NULL;
+
+        mdp5_state = kmemdup(to_mdp5_plane_state(plane->state),
+                        sizeof(*mdp5_state), GFP_KERNEL);
+
+        if (mdp5_state && mdp5_state->base.fb)
+                drm_framebuffer_reference(mdp5_state->base.fb);
+
+        mdp5_state->mode_changed = false;
+        mdp5_state->pending = false;
+
+        return &mdp5_state->base;
+}
+
+static void mdp5_plane_destroy_state(struct drm_plane *plane,
+                struct drm_plane_state *state)
+{
+        if (state->fb)
+                drm_framebuffer_unreference(state->fb);
+
+        kfree(to_mdp5_plane_state(state));
+}
+
 static const struct drm_plane_funcs mdp5_plane_funcs = {
-                .update_plane = mdp5_plane_update,
-                .disable_plane = mdp5_plane_disable,
+                .update_plane = drm_atomic_helper_update_plane,
+                .disable_plane = drm_atomic_helper_disable_plane,
                 .destroy = mdp5_plane_destroy,
                 .set_property = mdp5_plane_set_property,
+                .reset = mdp5_plane_reset,
+                .atomic_duplicate_state = mdp5_plane_duplicate_state,
+                .atomic_destroy_state = mdp5_plane_destroy_state,
 };
 
-void mdp5_plane_set_scanout(struct drm_plane *plane,
+static int mdp5_plane_prepare_fb(struct drm_plane *plane,
                 struct drm_framebuffer *fb)
 {
         struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
         struct mdp5_kms *mdp5_kms = get_kms(plane);
-        enum mdp5_pipe pipe = mdp5_plane->pipe;
-        uint32_t nplanes = drm_format_num_planes(fb->pixel_format);
-        uint32_t iova[4];
-        int i;
-
-        for (i = 0; i < nplanes; i++) {
-                struct drm_gem_object *bo = msm_framebuffer_bo(fb, i);
-                msm_gem_get_iova(bo, mdp5_kms->id, &iova[i]);
-        }
-        for (; i < 4; i++)
-                iova[i] = 0;
 
-        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_A(pipe),
-                        MDP5_PIPE_SRC_STRIDE_A_P0(fb->pitches[0]) |
-                        MDP5_PIPE_SRC_STRIDE_A_P1(fb->pitches[1]));
-
-        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_B(pipe),
-                        MDP5_PIPE_SRC_STRIDE_B_P2(fb->pitches[2]) |
-                        MDP5_PIPE_SRC_STRIDE_B_P3(fb->pitches[3]));
-
-        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC0_ADDR(pipe), iova[0]);
-        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC1_ADDR(pipe), iova[1]);
-        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC2_ADDR(pipe), iova[2]);
-        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC3_ADDR(pipe), iova[3]);
-
-        plane->fb = fb;
+        DBG("%s: prepare: FB[%u]", mdp5_plane->name, fb->base.id);
+        return msm_framebuffer_prepare(fb, mdp5_kms->id);
 }
 
-/* NOTE: looks like if horizontal decimation is used (if we supported that)
- * then the width used to calculate SMP block requirements is the post-
- * decimated width.  Ie. SMP buffering sits downstream of decimation (which
- * presumably happens during the dma from scanout buffer).
- */
-static int request_smp_blocks(struct drm_plane *plane, uint32_t format,
-                uint32_t nplanes, uint32_t width)
+static void mdp5_plane_cleanup_fb(struct drm_plane *plane,
+                struct drm_framebuffer *fb)
 {
-        struct drm_device *dev = plane->dev;
         struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
         struct mdp5_kms *mdp5_kms = get_kms(plane);
-        enum mdp5_pipe pipe = mdp5_plane->pipe;
-        int i, hsub, nlines, nblks, ret;
 
-        hsub = drm_format_horz_chroma_subsampling(format);
+        DBG("%s: cleanup: FB[%u]", mdp5_plane->name, fb->base.id);
+        msm_framebuffer_cleanup(fb, mdp5_kms->id);
+}
 
-        /* different if BWC (compressed framebuffer?) enabled: */
-        nlines = 2;
+static int mdp5_plane_atomic_check(struct drm_plane *plane,
+                struct drm_plane_state *state)
+{
+        struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
+        struct drm_plane_state *old_state = plane->state;
 
-        for (i = 0, nblks = 0; i < nplanes; i++) {
-                int n, fetch_stride, cpp;
+        DBG("%s: check (%d -> %d)", mdp5_plane->name,
+                        plane_enabled(old_state), plane_enabled(state));
 
-                cpp = drm_format_plane_cpp(format, i);
-                fetch_stride = width * cpp / (i ? hsub : 1);
+        if (plane_enabled(state) && plane_enabled(old_state)) {
+                /* we cannot change SMP block configuration during scanout: */
+                bool full_modeset = false;
+                if (state->fb->pixel_format != old_state->fb->pixel_format) {
+                        DBG("%s: pixel_format change!", mdp5_plane->name);
+                        full_modeset = true;
+                }
+                if (state->src_w != old_state->src_w) {
+                        DBG("%s: src_w change!", mdp5_plane->name);
+                        full_modeset = true;
+                }
+                if (to_mdp5_plane_state(old_state)->pending) {
+                        DBG("%s: still pending!", mdp5_plane->name);
+                        full_modeset = true;
+                }
+                if (full_modeset) {
+                        struct drm_crtc_state *crtc_state =
+                                        drm_atomic_get_crtc_state(state->state, state->crtc);
+                        crtc_state->mode_changed = true;
+                        to_mdp5_plane_state(state)->mode_changed = true;
+                }
+        } else {
+                to_mdp5_plane_state(state)->mode_changed = true;
+        }
 
-                n = DIV_ROUND_UP(fetch_stride * nlines, SMP_BLK_SIZE);
+        return 0;
+}
 
-                /* for hw rev v1.00 */
-                if (mdp5_kms->rev == 0)
-                        n = roundup_pow_of_two(n);
+static void mdp5_plane_atomic_update(struct drm_plane *plane,
+                                     struct drm_plane_state *old_state)
+{
+        struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
+        struct drm_plane_state *state = plane->state;
 
-                DBG("%s[%d]: request %d SMP blocks", mdp5_plane->name, i, n);
-                ret = mdp5_smp_request(mdp5_kms, pipe2client(pipe, i), n);
-                if (ret) {
-                        dev_err(dev->dev, "Could not allocate %d SMP blocks: %d\n",
-                                        n, ret);
-                        return ret;
-                }
+        DBG("%s: update", mdp5_plane->name);
 
-                nblks += n;
+        if (!plane_enabled(state)) {
+                to_mdp5_plane_state(state)->pending = true;
+                mdp5_plane_disable(plane);
+        } else if (to_mdp5_plane_state(state)->mode_changed) {
+                int ret;
+                to_mdp5_plane_state(state)->pending = true;
+                ret = mdp5_plane_mode_set(plane,
+                                state->crtc, state->fb,
+                                state->crtc_x, state->crtc_y,
+                                state->crtc_w, state->crtc_h,
+                                state->src_x,  state->src_y,
+                                state->src_w, state->src_h);
+                /* atomic_check should have ensured that this doesn't fail */
+                WARN_ON(ret < 0);
+        } else {
+                unsigned long flags;
+                spin_lock_irqsave(&mdp5_plane->pipe_lock, flags);
+                set_scanout_locked(plane, state->fb);
+                spin_unlock_irqrestore(&mdp5_plane->pipe_lock, flags);
         }
-
-        /* in success case, return total # of blocks allocated: */
-        return nblks;
 }
 
-static void set_fifo_thresholds(struct drm_plane *plane, int nblks)
+static const struct drm_plane_helper_funcs mdp5_plane_helper_funcs = {
+                .prepare_fb = mdp5_plane_prepare_fb,
+                .cleanup_fb = mdp5_plane_cleanup_fb,
+                .atomic_check = mdp5_plane_atomic_check,
+                .atomic_update = mdp5_plane_atomic_update,
+};
+
+static void set_scanout_locked(struct drm_plane *plane,
+                struct drm_framebuffer *fb)
 {
         struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
         struct mdp5_kms *mdp5_kms = get_kms(plane);
         enum mdp5_pipe pipe = mdp5_plane->pipe;
-        uint32_t val;
 
-        /* 1/4 of SMP pool that is being fetched */
-        val = (nblks * SMP_ENTRIES_PER_BLK) / 4;
+        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_A(pipe),
+                        MDP5_PIPE_SRC_STRIDE_A_P0(fb->pitches[0]) |
+                        MDP5_PIPE_SRC_STRIDE_A_P1(fb->pitches[1]));
 
-        mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_0(pipe), val * 1);
-        mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_1(pipe), val * 2);
-        mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_2(pipe), val * 3);
+        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_STRIDE_B(pipe),
+                        MDP5_PIPE_SRC_STRIDE_B_P2(fb->pitches[2]) |
+                        MDP5_PIPE_SRC_STRIDE_B_P3(fb->pitches[3]));
+
+        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC0_ADDR(pipe),
+                        msm_framebuffer_iova(fb, mdp5_kms->id, 0));
+        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC1_ADDR(pipe),
+                        msm_framebuffer_iova(fb, mdp5_kms->id, 1));
+        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC2_ADDR(pipe),
+                        msm_framebuffer_iova(fb, mdp5_kms->id, 2));
+        mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC3_ADDR(pipe),
+                        msm_framebuffer_iova(fb, mdp5_kms->id, 4));
 
+        plane->fb = fb;
 }
 
-int mdp5_plane_mode_set(struct drm_plane *plane,
+static int mdp5_plane_mode_set(struct drm_plane *plane,
                 struct drm_crtc *crtc, struct drm_framebuffer *fb,
                 int crtc_x, int crtc_y,
                 unsigned int crtc_w, unsigned int crtc_h,
@@ -225,7 +291,8 @@ int mdp5_plane_mode_set(struct drm_plane *plane,
         uint32_t nplanes, config = 0;
         uint32_t phasex_step = 0, phasey_step = 0;
         uint32_t hdecm = 0, vdecm = 0;
-        int i, nblks;
+        unsigned long flags;
+        int ret;
 
         nplanes = drm_format_num_planes(fb->pixel_format);
 
@@ -243,12 +310,11 @@ int mdp5_plane_mode_set(struct drm_plane *plane,
                         fb->base.id, src_x, src_y, src_w, src_h,
                         crtc->base.id, crtc_x, crtc_y, crtc_w, crtc_h);
 
-        /*
-         * Calculate and request required # of smp blocks:
-         */
-        nblks = request_smp_blocks(plane, fb->pixel_format, nplanes, src_w);
-        if (nblks < 0)
-                return nblks;
+        /* Request some memory from the SMP: */
+        ret = mdp5_smp_request(mdp5_kms->smp,
+                        mdp5_plane->pipe, fb->pixel_format, src_w);
+        if (ret)
+                return ret;
 
         /*
          * Currently we update the hw for allocations/requests immediately,
@@ -256,8 +322,7 @@ int mdp5_plane_mode_set(struct drm_plane *plane,
          * would move into atomic->check_plane_state(), while updating the
          * hw would remain here:
          */
-        for (i = 0; i < pipe2nclients(pipe); i++)
-                mdp5_smp_configure(mdp5_kms, pipe2client(pipe, i));
+        mdp5_smp_configure(mdp5_kms->smp, pipe);
 
         if (src_w != crtc_w) {
                 config |= MDP5_PIPE_SCALE_CONFIG_SCALEX_EN;
@@ -269,6 +334,8 @@ int mdp5_plane_mode_set(struct drm_plane *plane,
                 /* TODO calc phasey_step, vdecm */
         }
 
+        spin_lock_irqsave(&mdp5_plane->pipe_lock, flags);
+
         mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_IMG_SIZE(pipe),
                         MDP5_PIPE_SRC_IMG_SIZE_WIDTH(src_w) |
                         MDP5_PIPE_SRC_IMG_SIZE_HEIGHT(src_h));
@@ -289,8 +356,6 @@ int mdp5_plane_mode_set(struct drm_plane *plane,
                         MDP5_PIPE_OUT_XY_X(crtc_x) |
                         MDP5_PIPE_OUT_XY_Y(crtc_y));
 
-        mdp5_plane_set_scanout(plane, fb);
-
         format = to_mdp_format(msm_framebuffer_format(fb));
 
         mdp5_write(mdp5_kms, REG_MDP5_PIPE_SRC_FORMAT(pipe),
@@ -330,22 +395,24 @@ int mdp5_plane_mode_set(struct drm_plane *plane,
                         MDP5_PIPE_SCALE_CONFIG_SCALEX_MAX_FILTER(SCALE_FILTER_NEAREST) |
                         MDP5_PIPE_SCALE_CONFIG_SCALEY_MAX_FILTER(SCALE_FILTER_NEAREST));
 
-        set_fifo_thresholds(plane, nblks);
+        set_scanout_locked(plane, fb);
 
-        /* TODO detach from old crtc (if we had more than one) */
-        mdp5_crtc_attach(crtc, plane);
+        spin_unlock_irqrestore(&mdp5_plane->pipe_lock, flags);
 
-        return 0;
+        return ret;
 }
 
 void mdp5_plane_complete_flip(struct drm_plane *plane)
 {
         struct mdp5_kms *mdp5_kms = get_kms(plane);
-        enum mdp5_pipe pipe = to_mdp5_plane(plane)->pipe;
-        int i;
+        struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
+        enum mdp5_pipe pipe = mdp5_plane->pipe;
+
+        DBG("%s: complete flip", mdp5_plane->name);
 
-        for (i = 0; i < pipe2nclients(pipe); i++)
-                mdp5_smp_commit(mdp5_kms, pipe2client(pipe, i));
+        mdp5_smp_commit(mdp5_kms->smp, pipe);
+
+        to_mdp5_plane_state(plane->state)->pending = false;
 }
 
 enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane)
@@ -354,9 +421,16 @@ enum mdp5_pipe mdp5_plane_pipe(struct drm_plane *plane)
         return mdp5_plane->pipe;
 }
 
+uint32_t mdp5_plane_get_flush(struct drm_plane *plane)
+{
+        struct mdp5_plane *mdp5_plane = to_mdp5_plane(plane);
+
+        return mdp5_plane->flush_mask;
+}
+
 /* initialize plane */
 struct drm_plane *mdp5_plane_init(struct drm_device *dev,
-                enum mdp5_pipe pipe, bool private_plane)
+                enum mdp5_pipe pipe, bool private_plane, uint32_t reg_offset)
 {
         struct drm_plane *plane = NULL;
         struct mdp5_plane *mdp5_plane;
@@ -377,10 +451,18 @@ struct drm_plane *mdp5_plane_init(struct drm_device *dev,
         mdp5_plane->nformats = mdp5_get_formats(pipe, mdp5_plane->formats,
                         ARRAY_SIZE(mdp5_plane->formats));
 
+        mdp5_plane->flush_mask = mdp_ctl_flush_mask_pipe(pipe);
+        mdp5_plane->reg_offset = reg_offset;
+        spin_lock_init(&mdp5_plane->pipe_lock);
+
         type = private_plane ? DRM_PLANE_TYPE_PRIMARY : DRM_PLANE_TYPE_OVERLAY;
-        drm_universal_plane_init(dev, plane, 0xff, &mdp5_plane_funcs,
+        ret = drm_universal_plane_init(dev, plane, 0xff, &mdp5_plane_funcs,
                                  mdp5_plane->formats, mdp5_plane->nformats,
                                  type);
+        if (ret)
+                goto fail;
+
+        drm_plane_helper_add(plane, &mdp5_plane_helper_funcs);
 
         mdp5_plane_install_properties(plane, &plane->base);
 
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c
index 2d0236b963a6..bf551885e019 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c
@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2014, The Linux Foundation. All rights reserved.
  * Copyright (C) 2013 Red Hat
  * Author: Rob Clark <robdclark@gmail.com>
  *
@@ -29,8 +30,11 @@
  * Based on the size of the attached scanout buffer, a certain # of
  * blocks must be allocated to that client out of the shared pool.
  *
- * For each block, it can be either free, or pending/in-use by a
- * client.  The updates happen in three steps:
+ * In some hw, some blocks are statically allocated for certain pipes
+ * and CANNOT be re-allocated (eg: MMB0 and MMB1 both tied to RGB0).
+ *
+ * For each block that can be dynamically allocated, it can be either
+ * free, or pending/in-use by a client. The updates happen in three steps:
  *
  *  1) mdp5_smp_request():
  *     When plane scanout is setup, calculate required number of
@@ -61,21 +65,68 @@
  * inuse and pending state of all clients..
  */
 
-static DEFINE_SPINLOCK(smp_lock);
+struct mdp5_smp {
+        struct drm_device *dev;
+
+        int blk_cnt;
+        int blk_size;
+
+        spinlock_t state_lock;
+        mdp5_smp_state_t state; /* to track smp allocation amongst pipes: */
+
+        struct mdp5_client_smp_state client_state[CID_MAX];
+};
 
+static inline
+struct mdp5_kms *get_kms(struct mdp5_smp *smp)
+{
+        struct msm_drm_private *priv = smp->dev->dev_private;
+
+        return to_mdp5_kms(to_mdp_kms(priv->kms));
+}
+
+static inline enum mdp5_client_id pipe2client(enum mdp5_pipe pipe, int plane)
+{
+        WARN_ON(plane >= pipe2nclients(pipe));
+        switch (pipe) {
+        case SSPP_VIG0: return CID_VIG0_Y + plane;
+        case SSPP_VIG1: return CID_VIG1_Y + plane;
+        case SSPP_VIG2: return CID_VIG2_Y + plane;
+        case SSPP_RGB0: return CID_RGB0;
+        case SSPP_RGB1: return CID_RGB1;
+        case SSPP_RGB2: return CID_RGB2;
+        case SSPP_DMA0: return CID_DMA0_Y + plane;
+        case SSPP_DMA1: return CID_DMA1_Y + plane;
+        case SSPP_VIG3: return CID_VIG3_Y + plane;
+        case SSPP_RGB3: return CID_RGB3;
+        default:        return CID_UNUSED;
+        }
+}
 
 /* step #1: update # of blocks pending for the client: */
-int mdp5_smp_request(struct mdp5_kms *mdp5_kms,
+static int smp_request_block(struct mdp5_smp *smp,
                 enum mdp5_client_id cid, int nblks)
 {
-        struct mdp5_client_smp_state *ps = &mdp5_kms->smp_client_state[cid];
-        int i, ret, avail, cur_nblks, cnt = mdp5_kms->smp_blk_cnt;
+        struct mdp5_kms *mdp5_kms = get_kms(smp);
+        const struct mdp5_cfg_hw *hw_cfg;
+        struct mdp5_client_smp_state *ps = &smp->client_state[cid];
+        int i, ret, avail, cur_nblks, cnt = smp->blk_cnt;
+        int reserved;
         unsigned long flags;
 
-        spin_lock_irqsave(&smp_lock, flags);
+        hw_cfg = mdp5_cfg_get_hw_config(mdp5_kms->cfg);
+        reserved = hw_cfg->smp.reserved[cid];
+
+        spin_lock_irqsave(&smp->state_lock, flags);
 
-        avail = cnt - bitmap_weight(mdp5_kms->smp_state, cnt);
+        nblks -= reserved;
+        if (reserved)
+                DBG("%d MMBs allocated (%d reserved)", nblks, reserved);
+
+        avail = cnt - bitmap_weight(smp->state, cnt);
         if (nblks > avail) {
+                dev_err(mdp5_kms->dev->dev, "out of blks (req=%d > avail=%d)\n",
+                                nblks, avail);
                 ret = -ENOSPC;
                 goto fail;
         }
@@ -84,9 +135,9 @@ int mdp5_smp_request(struct mdp5_kms *mdp5_kms,
         if (nblks > cur_nblks) {
                 /* grow the existing pending reservation: */
                 for (i = cur_nblks; i < nblks; i++) {
-                        int blk = find_first_zero_bit(mdp5_kms->smp_state, cnt);
+                        int blk = find_first_zero_bit(smp->state, cnt);
                         set_bit(blk, ps->pending);
-                        set_bit(blk, mdp5_kms->smp_state);
+                        set_bit(blk, smp->state);
                 }
         } else {
                 /* shrink the existing pending reservation: */
@@ -98,15 +149,88 @@ int mdp5_smp_request(struct mdp5_kms *mdp5_kms,
         }
 
 fail:
-        spin_unlock_irqrestore(&smp_lock, flags);
+        spin_unlock_irqrestore(&smp->state_lock, flags);
+        return 0;
+}
+
+static void set_fifo_thresholds(struct mdp5_smp *smp,
+                enum mdp5_pipe pipe, int nblks)
+{
+        struct mdp5_kms *mdp5_kms = get_kms(smp);
+        u32 smp_entries_per_blk = smp->blk_size / (128 / BITS_PER_BYTE);
+        u32 val;
+
+        /* 1/4 of SMP pool that is being fetched */
+        val = (nblks * smp_entries_per_blk) / 4;
+
+        mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_0(pipe), val * 1);
+        mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_1(pipe), val * 2);
+        mdp5_write(mdp5_kms, REG_MDP5_PIPE_REQPRIO_FIFO_WM_2(pipe), val * 3);
+}
+
+/*
+ * NOTE: looks like if horizontal decimation is used (if we supported that)
+ * then the width used to calculate SMP block requirements is the post-
+ * decimated width.  Ie. SMP buffering sits downstream of decimation (which
+ * presumably happens during the dma from scanout buffer).
+ */
+int mdp5_smp_request(struct mdp5_smp *smp, enum mdp5_pipe pipe, u32 fmt, u32 width)
+{
+        struct mdp5_kms *mdp5_kms = get_kms(smp);
+        struct drm_device *dev = mdp5_kms->dev;
+        int rev = mdp5_cfg_get_hw_rev(mdp5_kms->cfg);
+        int i, hsub, nplanes, nlines, nblks, ret;
+
+        nplanes = drm_format_num_planes(fmt);
+        hsub = drm_format_horz_chroma_subsampling(fmt);
+
+        /* different if BWC (compressed framebuffer?) enabled: */
+        nlines = 2;
+
+        for (i = 0, nblks = 0; i < nplanes; i++) {
+                int n, fetch_stride, cpp;
+
+                cpp = drm_format_plane_cpp(fmt, i);
+                fetch_stride = width * cpp / (i ? hsub : 1);
+
+                n = DIV_ROUND_UP(fetch_stride * nlines, smp->blk_size);
+
+                /* for hw rev v1.00 */
+                if (rev == 0)
+                        n = roundup_pow_of_two(n);
+
+                DBG("%s[%d]: request %d SMP blocks", pipe2name(pipe), i, n);
+                ret = smp_request_block(smp, pipe2client(pipe, i), n);
+                if (ret) {
+                        dev_err(dev->dev, "Cannot allocate %d SMP blocks: %d\n",
+                                        n, ret);
+                        return ret;
+                }
+
+                nblks += n;
+        }
+
+        set_fifo_thresholds(smp, pipe, nblks);
+
         return 0;
 }
 
-static void update_smp_state(struct mdp5_kms *mdp5_kms,
+/* Release SMP blocks for all clients of the pipe */
+void mdp5_smp_release(struct mdp5_smp *smp, enum mdp5_pipe pipe)
+{
+        int i, nblks;
+
+        for (i = 0, nblks = 0; i < pipe2nclients(pipe); i++)
+                smp_request_block(smp, pipe2client(pipe, i), 0);
+        set_fifo_thresholds(smp, pipe, 0);
+}
+
+static void update_smp_state(struct mdp5_smp *smp,
                 enum mdp5_client_id cid, mdp5_smp_state_t *assigned)
 {
-        int cnt = mdp5_kms->smp_blk_cnt;
-        uint32_t blk, val;
+        struct mdp5_kms *mdp5_kms = get_kms(smp);
+        int cnt = smp->blk_cnt;
+        u32 blk, val;
 
         for_each_set_bit(blk, *assigned, cnt) {
                 int idx = blk / 3;
@@ -135,39 +259,80 @@ static void update_smp_state(struct mdp5_kms *mdp5_kms,
 }
 
 /* step #2: configure hw for union(pending, inuse): */
-void mdp5_smp_configure(struct mdp5_kms *mdp5_kms, enum mdp5_client_id cid)
+void mdp5_smp_configure(struct mdp5_smp *smp, enum mdp5_pipe pipe)
 {
-        struct mdp5_client_smp_state *ps = &mdp5_kms->smp_client_state[cid];
-        int cnt = mdp5_kms->smp_blk_cnt;
+        int cnt = smp->blk_cnt;
         mdp5_smp_state_t assigned;
+        int i;
 
-        bitmap_or(assigned, ps->inuse, ps->pending, cnt);
-        update_smp_state(mdp5_kms, cid, &assigned);
+        for (i = 0; i < pipe2nclients(pipe); i++) {
+                enum mdp5_client_id cid = pipe2client(pipe, i);
+                struct mdp5_client_smp_state *ps = &smp->client_state[cid];
+
+                bitmap_or(assigned, ps->inuse, ps->pending, cnt);
+                update_smp_state(smp, cid, &assigned);
+        }
 }
 
 /* step #3: after vblank, copy pending -> inuse: */
-void mdp5_smp_commit(struct mdp5_kms *mdp5_kms, enum mdp5_client_id cid)
+void mdp5_smp_commit(struct mdp5_smp *smp, enum mdp5_pipe pipe)
 {
-        struct mdp5_client_smp_state *ps = &mdp5_kms->smp_client_state[cid];
-        int cnt = mdp5_kms->smp_blk_cnt;
+        int cnt = smp->blk_cnt;
         mdp5_smp_state_t released;
+        int i;
+
+        for (i = 0; i < pipe2nclients(pipe); i++) {
+                enum mdp5_client_id cid = pipe2client(pipe, i);
+                struct mdp5_client_smp_state *ps = &smp->client_state[cid];
+
+                /*
+                 * Figure out if there are any blocks we where previously
+                 * using, which can be released and made available to other
+                 * clients:
+                 */
+                if (bitmap_andnot(released, ps->inuse, ps->pending, cnt)) {
+                        unsigned long flags;
+
+                        spin_lock_irqsave(&smp->state_lock, flags);
+                        /* clear released blocks: */
+                        bitmap_andnot(smp->state, smp->state, released, cnt);
+                        spin_unlock_irqrestore(&smp->state_lock, flags);
 
-        /*
-         * Figure out if there are any blocks we where previously
-         * using, which can be released and made available to other
-         * clients:
-         */
-        if (bitmap_andnot(released, ps->inuse, ps->pending, cnt)) {
-                unsigned long flags;
-
-                spin_lock_irqsave(&smp_lock, flags);
-                /* clear released blocks: */
-                bitmap_andnot(mdp5_kms->smp_state, mdp5_kms->smp_state,
-                                released, cnt);
-                spin_unlock_irqrestore(&smp_lock, flags);
-
-                update_smp_state(mdp5_kms, CID_UNUSED, &released);
+                        update_smp_state(smp, CID_UNUSED, &released);
+                }
+
+                bitmap_copy(ps->inuse, ps->pending, cnt);
         }
+}
+
+void mdp5_smp_destroy(struct mdp5_smp *smp)
+{
+        kfree(smp);
+}
+
+struct mdp5_smp *mdp5_smp_init(struct drm_device *dev, const struct mdp5_smp_block *cfg)
+{
+        struct mdp5_smp *smp = NULL;
+        int ret;
+
+        smp = kzalloc(sizeof(*smp), GFP_KERNEL);
+        if (unlikely(!smp)) {
+                ret = -ENOMEM;
+                goto fail;
+        }
+
+        smp->dev = dev;
+        smp->blk_cnt = cfg->mmb_count;
+        smp->blk_size = cfg->mmb_size;
+
+        /* statically tied MMBs cannot be re-allocated: */
+        bitmap_copy(smp->state, cfg->reserved_state, smp->blk_cnt);
+        spin_lock_init(&smp->state_lock);
+
+        return smp;
+fail:
+        if (smp)
+                mdp5_smp_destroy(smp);
 
-        bitmap_copy(ps->inuse, ps->pending, cnt);
+        return ERR_PTR(ret);
 }
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h
index 0ab739e1a1dd..e47179f63585 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.h
@@ -1,4 +1,5 @@
 /*
+ * Copyright (c) 2014, The Linux Foundation. All rights reserved.
  * Copyright (C) 2013 Red Hat
  * Author: Rob Clark <robdclark@gmail.com>
  *
@@ -20,22 +21,26 @@
 
 #include "msm_drv.h"
 
-#define MAX_SMP_BLOCKS  22
-#define SMP_BLK_SIZE    4096
-#define SMP_ENTRIES_PER_BLK (SMP_BLK_SIZE / 16)
-
-typedef DECLARE_BITMAP(mdp5_smp_state_t, MAX_SMP_BLOCKS);
-
 struct mdp5_client_smp_state {
         mdp5_smp_state_t inuse;
         mdp5_smp_state_t pending;
 };
 
 struct mdp5_kms;
+struct mdp5_smp;
+
+/*
+ * SMP module prototypes:
+ * mdp5_smp_init() returns a SMP @handler,
+ * which is then used to call the other mdp5_smp_*(handler, ...) functions.
+ */
 
-int mdp5_smp_request(struct mdp5_kms *mdp5_kms, enum mdp5_client_id cid, int nblks);
-void mdp5_smp_configure(struct mdp5_kms *mdp5_kms, enum mdp5_client_id cid);
-void mdp5_smp_commit(struct mdp5_kms *mdp5_kms, enum mdp5_client_id cid);
+struct mdp5_smp *mdp5_smp_init(struct drm_device *dev, const struct mdp5_smp_block *cfg);
+void  mdp5_smp_destroy(struct mdp5_smp *smp);
 
+int  mdp5_smp_request(struct mdp5_smp *smp, enum mdp5_pipe pipe, u32 fmt, u32 width);
+void mdp5_smp_configure(struct mdp5_smp *smp, enum mdp5_pipe pipe);
+void mdp5_smp_commit(struct mdp5_smp *smp, enum mdp5_pipe pipe);
+void mdp5_smp_release(struct mdp5_smp *smp, enum mdp5_pipe pipe);
 
 #endif /* __MDP5_SMP_H__ */
diff --git a/drivers/gpu/drm/msm/msm_atomic.c b/drivers/gpu/drm/msm/msm_atomic.c
new file mode 100644
index 000000000000..f0de412e13dc
--- /dev/null
+++ b/drivers/gpu/drm/msm/msm_atomic.c
@@ -0,0 +1,163 @@
+/*
+ * Copyright (C) 2014 Red Hat
+ * Author: Rob Clark <robdclark@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "msm_drv.h"
+#include "msm_kms.h"
+#include "msm_gem.h"
+
+struct msm_commit {
+        struct drm_atomic_state *state;
+        uint32_t fence;
+        struct msm_fence_cb fence_cb;
+};
+
+static void fence_cb(struct msm_fence_cb *cb);
+
+static struct msm_commit *new_commit(struct drm_atomic_state *state)
+{
+        struct msm_commit *c = kzalloc(sizeof(*c), GFP_KERNEL);
+
+        if (!c)
+                return NULL;
+
+        c->state = state;
+        /* TODO we might need a way to indicate to run the cb on a
+         * different wq so wait_for_vblanks() doesn't block retiring
+         * bo's..
+         */
+        INIT_FENCE_CB(&c->fence_cb, fence_cb);
+
+        return c;
+}
+
+/* The (potentially) asynchronous part of the commit.  At this point
+ * nothing can fail short of armageddon.
+ */
+static void complete_commit(struct msm_commit *c)
+{
+        struct drm_atomic_state *state = c->state;
+        struct drm_device *dev = state->dev;
+
+        drm_atomic_helper_commit_pre_planes(dev, state);
+
+        drm_atomic_helper_commit_planes(dev, state);
+
+        drm_atomic_helper_commit_post_planes(dev, state);
+
+        drm_atomic_helper_wait_for_vblanks(dev, state);
+
+        drm_atomic_helper_cleanup_planes(dev, state);
+
+        drm_atomic_state_free(state);
+
+        kfree(c);
+}
+
+static void fence_cb(struct msm_fence_cb *cb)
+{
+        struct msm_commit *c =
+                        container_of(cb, struct msm_commit, fence_cb);
+        complete_commit(c);
+}
+
+static void add_fb(struct msm_commit *c, struct drm_framebuffer *fb)
+{
+        struct drm_gem_object *obj = msm_framebuffer_bo(fb, 0);
+        c->fence = max(c->fence, msm_gem_fence(to_msm_bo(obj), MSM_PREP_READ));
+}
+
+
+/**
+ * drm_atomic_helper_commit - commit validated state object
+ * @dev: DRM device
+ * @state: the driver state object
+ * @async: asynchronous commit
+ *
+ * This function commits a with drm_atomic_helper_check() pre-validated state
+ * object. This can still fail when e.g. the framebuffer reservation fails. For
+ * now this doesn't implement asynchronous commits.
+ *
+ * RETURNS
+ * Zero for success or -errno.
+ */
+int msm_atomic_commit(struct drm_device *dev,
+                struct drm_atomic_state *state, bool async)
+{
+        struct msm_commit *c;
+        int nplanes = dev->mode_config.num_total_plane;
+        int i, ret;
+
+        ret = drm_atomic_helper_prepare_planes(dev, state);
+        if (ret)
+                return ret;
+
+        c = new_commit(state);
+
+        /*
+         * Figure out what fence to wait for:
+         */
+        for (i = 0; i < nplanes; i++) {
+                struct drm_plane *plane = state->planes[i];
+                struct drm_plane_state *new_state = state->plane_states[i];
+
+                if (!plane)
+                        continue;
+
+                if ((plane->state->fb != new_state->fb) && new_state->fb)
+                        add_fb(c, new_state->fb);
+        }
+
+        /*
+         * This is the point of no return - everything below never fails except
+         * when the hw goes bonghits. Which means we can commit the new state on
+         * the software side now.
+         */
+
+        drm_atomic_helper_swap_state(dev, state);
+
+        /*
+         * Everything below can be run asynchronously without the need to grab
+         * any modeset locks at all under one conditions: It must be guaranteed
+         * that the asynchronous work has either been cancelled (if the driver
+         * supports it, which at least requires that the framebuffers get
+         * cleaned up with drm_atomic_helper_cleanup_planes()) or completed
+         * before the new state gets committed on the software side with
+         * drm_atomic_helper_swap_state().
+         *
+         * This scheme allows new atomic state updates to be prepared and
+         * checked in parallel to the asynchronous completion of the previous
+         * update. Which is important since compositors need to figure out the
+         * composition of the next frame right after having submitted the
+         * current layout.
+         */
+
+        if (async) {
+                msm_queue_fence_cb(dev, &c->fence_cb, c->fence);
+                return 0;
+        }
+
+        ret = msm_wait_fence_interruptable(dev, c->fence, NULL);
+        if (ret) {
+                WARN_ON(ret);  // TODO unswap state back?  or??
+                kfree(c);
+                return ret;
+        }
+
+        complete_commit(c);
+
+        return 0;
+}
diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c
index 42e1c48eef28..c795217e1bfc 100644
--- a/drivers/gpu/drm/msm/msm_drv.c
+++ b/drivers/gpu/drm/msm/msm_drv.c
@@ -29,6 +29,8 @@ static void msm_fb_output_poll_changed(struct drm_device *dev)
 static const struct drm_mode_config_funcs mode_config_funcs = {
         .fb_create = msm_framebuffer_create,
         .output_poll_changed = msm_fb_output_poll_changed,
+        .atomic_check = drm_atomic_helper_check,
+        .atomic_commit = msm_atomic_commit,
 };
 
 int msm_register_mmu(struct drm_device *dev, struct msm_mmu *mmu)
@@ -294,6 +296,8 @@ static int msm_load(struct drm_device *dev, unsigned long flags)
                 goto fail;
         }
 
+        drm_mode_config_reset(dev);
+
 #ifdef CONFIG_DRM_MSM_FBDEV
         priv->fbdev = msm_fbdev_init(dev);
 #endif
@@ -619,6 +623,26 @@ int msm_wait_fence_interruptable(struct drm_device *dev, uint32_t fence,
         return ret;
 }
 
+int msm_queue_fence_cb(struct drm_device *dev,
+                struct msm_fence_cb *cb, uint32_t fence)
+{
+        struct msm_drm_private *priv = dev->dev_private;
+        int ret = 0;
+
+        mutex_lock(&dev->struct_mutex);
+        if (!list_empty(&cb->work.entry)) {
+                ret = -EINVAL;
+        } else if (fence > priv->completed_fence) {
+                cb->fence = fence;
+                list_add_tail(&cb->work.entry, &priv->fence_cbs);
+        } else {
+                queue_work(priv->wq, &cb->work);
+        }
+        mutex_unlock(&dev->struct_mutex);
+
+        return ret;
+}
+
 /* called from workqueue */
 void msm_update_fence(struct drm_device *dev, uint32_t fence)
 {
@@ -832,6 +856,7 @@ static struct drm_driver msm_driver = {
         .gem_prime_import_sg_table = msm_gem_prime_import_sg_table,
         .gem_prime_vmap     = msm_gem_prime_vmap,
         .gem_prime_vunmap   = msm_gem_prime_vunmap,
+        .gem_prime_mmap     = msm_gem_prime_mmap,
 #ifdef CONFIG_DEBUG_FS
         .debugfs_init       = msm_debugfs_init,
         .debugfs_cleanup    = msm_debugfs_cleanup,
diff --git a/drivers/gpu/drm/msm/msm_drv.h b/drivers/gpu/drm/msm/msm_drv.h
index 67f9d0a2332c..136303818436 100644
--- a/drivers/gpu/drm/msm/msm_drv.h
+++ b/drivers/gpu/drm/msm/msm_drv.h
@@ -32,15 +32,6 @@
 #include <linux/types.h>
 #include <asm/sizes.h>
 
-
-#if defined(CONFIG_COMPILE_TEST) && !defined(CONFIG_ARCH_QCOM)
-/* stubs we need for compile-test: */
-static inline struct device *msm_iommu_get_ctx(const char *ctx_name)
-{
-        return NULL;
-}
-#endif
-
 #ifndef CONFIG_OF
 #include <mach/board.h>
 #include <mach/socinfo.h>
@@ -48,7 +39,10 @@ static inline struct device *msm_iommu_get_ctx(const char *ctx_name)
 #endif
 
 #include <drm/drmP.h>
+#include <drm/drm_atomic.h>
+#include <drm/drm_atomic_helper.h>
 #include <drm/drm_crtc_helper.h>
+#include <drm/drm_plane_helper.h>
 #include <drm/drm_fb_helper.h>
 #include <drm/msm_drm.h>
 #include <drm/drm_gem.h>
@@ -75,7 +69,12 @@ struct msm_drm_private {
         struct msm_kms *kms;
 
         /* subordinate devices, if present: */
-        struct platform_device *hdmi_pdev, *gpu_pdev;
+        struct platform_device *gpu_pdev;
+
+        /* possibly this should be in the kms component, but it is
+         * shared by both mdp4 and mdp5..
+         */
+        struct hdmi *hdmi;
 
         /* when we have more than one 'msm_gpu' these need to be an array: */
         struct msm_gpu *gpu;
@@ -145,21 +144,29 @@ void __msm_fence_worker(struct work_struct *work);
                 (_cb)->func = _func;                         \
         } while (0)
 
+int msm_atomic_commit(struct drm_device *dev,
+                struct drm_atomic_state *state, bool async);
+
 int msm_register_mmu(struct drm_device *dev, struct msm_mmu *mmu);
 
 int msm_wait_fence_interruptable(struct drm_device *dev, uint32_t fence,
                 struct timespec *timeout);
+int msm_queue_fence_cb(struct drm_device *dev,
+                struct msm_fence_cb *cb, uint32_t fence);
 void msm_update_fence(struct drm_device *dev, uint32_t fence);
 
 int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
                 struct drm_file *file);
 
+int msm_gem_mmap_obj(struct drm_gem_object *obj,
+                        struct vm_area_struct *vma);
 int msm_gem_mmap(struct file *filp, struct vm_area_struct *vma);
 int msm_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf);
 uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj);
 int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id,
                 uint32_t *iova);
 int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint32_t *iova);
+uint32_t msm_gem_iova(struct drm_gem_object *obj, int id);
 struct page **msm_gem_get_pages(struct drm_gem_object *obj);
 void msm_gem_put_pages(struct drm_gem_object *obj);
 void msm_gem_put_iova(struct drm_gem_object *obj, int id);
@@ -170,6 +177,7 @@ int msm_gem_dumb_map_offset(struct drm_file *file, struct drm_device *dev,
 struct sg_table *msm_gem_prime_get_sg_table(struct drm_gem_object *obj);
 void *msm_gem_prime_vmap(struct drm_gem_object *obj);
 void msm_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr);
+int msm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma);
 struct drm_gem_object *msm_gem_prime_import_sg_table(struct drm_device *dev,
                 struct dma_buf_attachment *attach, struct sg_table *sg);
 int msm_gem_prime_pin(struct drm_gem_object *obj);
@@ -192,6 +200,9 @@ struct drm_gem_object *msm_gem_new(struct drm_device *dev,
 struct drm_gem_object *msm_gem_import(struct drm_device *dev,
                 uint32_t size, struct sg_table *sgt);
 
+int msm_framebuffer_prepare(struct drm_framebuffer *fb, int id);
+void msm_framebuffer_cleanup(struct drm_framebuffer *fb, int id);
+uint32_t msm_framebuffer_iova(struct drm_framebuffer *fb, int id, int plane);
 struct drm_gem_object *msm_framebuffer_bo(struct drm_framebuffer *fb, int plane);
 const struct msm_format *msm_framebuffer_format(struct drm_framebuffer *fb);
 struct drm_framebuffer *msm_framebuffer_init(struct drm_device *dev,
@@ -202,8 +213,8 @@ struct drm_framebuffer *msm_framebuffer_create(struct drm_device *dev,
 struct drm_fb_helper *msm_fbdev_init(struct drm_device *dev);
 
 struct hdmi;
-struct hdmi *hdmi_init(struct drm_device *dev, struct drm_encoder *encoder);
-irqreturn_t hdmi_irq(int irq, void *dev_id);
+int hdmi_modeset_init(struct hdmi *hdmi, struct drm_device *dev,
+                struct drm_encoder *encoder);
 void __init hdmi_register(void);
 void __exit hdmi_unregister(void);
 
diff --git a/drivers/gpu/drm/msm/msm_fb.c b/drivers/gpu/drm/msm/msm_fb.c
index 81bafdf19ab3..84dec161d836 100644
--- a/drivers/gpu/drm/msm/msm_fb.c
+++ b/drivers/gpu/drm/msm/msm_fb.c
@@ -24,7 +24,7 @@
 struct msm_framebuffer {
         struct drm_framebuffer base;
         const struct msm_format *format;
-        struct drm_gem_object *planes[2];
+        struct drm_gem_object *planes[3];
 };
 #define to_msm_framebuffer(x) container_of(x, struct msm_framebuffer, base)
 
@@ -87,6 +87,44 @@ void msm_framebuffer_describe(struct drm_framebuffer *fb, struct seq_file *m)
 }
 #endif
 
+/* prepare/pin all the fb's bo's for scanout.  Note that it is not valid
+ * to prepare an fb more multiple different initiator 'id's.  But that
+ * should be fine, since only the scanout (mdpN) side of things needs
+ * this, the gpu doesn't care about fb's.
+ */
+int msm_framebuffer_prepare(struct drm_framebuffer *fb, int id)
+{
+        struct msm_framebuffer *msm_fb = to_msm_framebuffer(fb);
+        int ret, i, n = drm_format_num_planes(fb->pixel_format);
+        uint32_t iova;
+
+        for (i = 0; i < n; i++) {
+                ret = msm_gem_get_iova(msm_fb->planes[i], id, &iova);
+                DBG("FB[%u]: iova[%d]: %08x (%d)", fb->base.id, i, iova, ret);
+                if (ret)
+                        return ret;
+        }
+
+        return 0;
+}
+
+void msm_framebuffer_cleanup(struct drm_framebuffer *fb, int id)
+{
+        struct msm_framebuffer *msm_fb = to_msm_framebuffer(fb);
+        int i, n = drm_format_num_planes(fb->pixel_format);
+
+        for (i = 0; i < n; i++)
+                msm_gem_put_iova(msm_fb->planes[i], id);
+}
+
+uint32_t msm_framebuffer_iova(struct drm_framebuffer *fb, int id, int plane)
+{
+        struct msm_framebuffer *msm_fb = to_msm_framebuffer(fb);
+        if (!msm_fb->planes[plane])
+                return 0;
+        return msm_gem_iova(msm_fb->planes[plane], id);
+}
+
 struct drm_gem_object *msm_framebuffer_bo(struct drm_framebuffer *fb, int plane)
 {
         struct msm_framebuffer *msm_fb = to_msm_framebuffer(fb);
@@ -166,6 +204,11 @@ struct drm_framebuffer *msm_framebuffer_init(struct drm_device *dev,
 
         msm_fb->format = format;
 
+        if (n > ARRAY_SIZE(msm_fb->planes)) {
+                ret = -EINVAL;
+                goto fail;
+        }
+
         for (i = 0; i < n; i++) {
                 unsigned int width = mode_cmd->width / (i ? hsub : 1);
                 unsigned int height = mode_cmd->height / (i ? vsub : 1);
diff --git a/drivers/gpu/drm/msm/msm_fbdev.c b/drivers/gpu/drm/msm/msm_fbdev.c
index ab5bfd2d0ebf..94d55e526b4e 100644
--- a/drivers/gpu/drm/msm/msm_fbdev.c
+++ b/drivers/gpu/drm/msm/msm_fbdev.c
@@ -93,9 +93,6 @@ static int msm_fbdev_create(struct drm_fb_helper *helper,
         uint32_t paddr;
         int ret, size;
 
-        sizes->surface_bpp = 32;
-        sizes->surface_depth = 24;
-
         DBG("create fbdev: %dx%d@%d (%dx%d)", sizes->surface_width,
                         sizes->surface_height, sizes->surface_bpp,
                         sizes->fb_width, sizes->fb_height);
diff --git a/drivers/gpu/drm/msm/msm_gem.c b/drivers/gpu/drm/msm/msm_gem.c
index 4b1b82adabde..4a6f0e49d5b5 100644
--- a/drivers/gpu/drm/msm/msm_gem.c
+++ b/drivers/gpu/drm/msm/msm_gem.c
@@ -309,6 +309,7 @@ int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id,
         return ret;
 }
 
+/* get iova, taking a reference.  Should have a matching put */
 int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint32_t *iova)
 {
         struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -328,6 +329,16 @@ int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint32_t *iova)
         return ret;
 }
 
+/* get iova without taking a reference, used in places where you have
+ * already done a 'msm_gem_get_iova()'.
+ */
+uint32_t msm_gem_iova(struct drm_gem_object *obj, int id)
+{
+        struct msm_gem_object *msm_obj = to_msm_bo(obj);
+        WARN_ON(!msm_obj->domain[id].iova);
+        return msm_obj->domain[id].iova;
+}
+
 void msm_gem_put_iova(struct drm_gem_object *obj, int id)
 {
         // XXX TODO ..
@@ -397,23 +408,10 @@ void *msm_gem_vaddr(struct drm_gem_object *obj)
 int msm_gem_queue_inactive_cb(struct drm_gem_object *obj,
                 struct msm_fence_cb *cb)
 {
-        struct drm_device *dev = obj->dev;
-        struct msm_drm_private *priv = dev->dev_private;
         struct msm_gem_object *msm_obj = to_msm_bo(obj);
-        int ret = 0;
-
-        mutex_lock(&dev->struct_mutex);
-        if (!list_empty(&cb->work.entry)) {
-                ret = -EINVAL;
-        } else if (is_active(msm_obj)) {
-                cb->fence = max(msm_obj->read_fence, msm_obj->write_fence);
-                list_add_tail(&cb->work.entry, &priv->fence_cbs);
-        } else {
-                queue_work(priv->wq, &cb->work);
-        }
-        mutex_unlock(&dev->struct_mutex);
-
-        return ret;
+        uint32_t fence = msm_gem_fence(msm_obj,
+                        MSM_PREP_READ | MSM_PREP_WRITE);
+        return msm_queue_fence_cb(obj->dev, cb, fence);
 }
 
 void msm_gem_move_to_active(struct drm_gem_object *obj,
@@ -452,12 +450,8 @@ int msm_gem_cpu_prep(struct drm_gem_object *obj, uint32_t op,
         int ret = 0;
 
         if (is_active(msm_obj)) {
-                uint32_t fence = 0;
+                uint32_t fence = msm_gem_fence(msm_obj, op);
 
-                if (op & MSM_PREP_READ)
-                        fence = msm_obj->write_fence;
-                if (op & MSM_PREP_WRITE)
-                        fence = max(fence, msm_obj->read_fence);
                 if (op & MSM_PREP_NOSYNC)
                         timeout = NULL;
 
@@ -525,13 +519,11 @@ void msm_gem_free_object(struct drm_gem_object *obj)
         for (id = 0; id < ARRAY_SIZE(msm_obj->domain); id++) {
                 struct msm_mmu *mmu = priv->mmus[id];
                 if (mmu && msm_obj->domain[id].iova) {
-                        uint32_t offset = (uint32_t)mmap_offset(obj);
+                        uint32_t offset = msm_obj->domain[id].iova;
                         mmu->funcs->unmap(mmu, offset, msm_obj->sgt, obj->size);
                 }
         }
 
-        drm_gem_free_mmap_offset(obj);
-
         if (obj->import_attach) {
                 if (msm_obj->vaddr)
                         dma_buf_vunmap(obj->import_attach->dmabuf, msm_obj->vaddr);
diff --git a/drivers/gpu/drm/msm/msm_gem.h b/drivers/gpu/drm/msm/msm_gem.h
index bfb052688f8e..8fbbd0594c46 100644
--- a/drivers/gpu/drm/msm/msm_gem.h
+++ b/drivers/gpu/drm/msm/msm_gem.h
@@ -70,6 +70,19 @@ static inline bool is_active(struct msm_gem_object *msm_obj)
         return msm_obj->gpu != NULL;
 }
 
+static inline uint32_t msm_gem_fence(struct msm_gem_object *msm_obj,
+                uint32_t op)
+{
+        uint32_t fence = 0;
+
+        if (op & MSM_PREP_READ)
+                fence = msm_obj->write_fence;
+        if (op & MSM_PREP_WRITE)
+                fence = max(fence, msm_obj->read_fence);
+
+        return fence;
+}
+
 #define MAX_CMDS 4
 
 /* Created per submit-ioctl, to track bo's and cmdstream bufs, etc,
diff --git a/drivers/gpu/drm/msm/msm_gem_prime.c b/drivers/gpu/drm/msm/msm_gem_prime.c
index ad772fe36115..dd7a7ab603e2 100644
--- a/drivers/gpu/drm/msm/msm_gem_prime.c
+++ b/drivers/gpu/drm/msm/msm_gem_prime.c
@@ -37,6 +37,19 @@ void msm_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr)
         /* TODO msm_gem_vunmap() */
 }
 
+int msm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
+{
+        int ret;
+
+        mutex_lock(&obj->dev->struct_mutex);
+        ret = drm_gem_mmap_obj(obj, obj->size, vma);
+        mutex_unlock(&obj->dev->struct_mutex);
+        if (ret < 0)
+                return ret;
+
+        return msm_gem_mmap_obj(vma->vm_private_data, vma);
+}
+
 struct drm_gem_object *msm_gem_prime_import_sg_table(struct drm_device *dev,
                 struct dma_buf_attachment *attach, struct sg_table *sg)
 {