Merge branch 'staging/gk20a' into staging/master

50 files changed, 1951 insertions, 98 deletions

diff --git a/Documentation/devicetree/bindings/gpu/nvidia,gk20a.txt b/Documentation/devicetree/bindings/gpu/nvidia,gk20a.txt
new file mode 100644
index 00000000000..59abdd0a6b5
--- /dev/null
+++ b/Documentation/devicetree/bindings/gpu/nvidia,gk20a.txt
@@ -0,0 +1,43 @@
+NVIDIA GK20A Graphics Processing Unit
+
+Required properties:
+- compatible: "nvidia,<chip>-<gpu>"
+  Currently recognized values:
+  - nvidia,tegra124-gk20a
+- reg: Physical base address and length of the controller's registers.
+  Must contain two entries:
+  - first entry for bar0
+  - second entry for bar1
+- interrupts: Must contain an entry for each entry in interrupt-names.
+  See ../interrupt-controller/interrupts.txt for details.
+- interrupt-names: Must include the following entries:
+  - stall
+  - nonstall
+- vdd-supply: regulator for supply voltage.
+- clocks: Must contain an entry for each entry in clock-names.
+  See ../clocks/clock-bindings.txt for details.
+- clock-names: Must include the following entries:
+  - gpu
+  - pwr
+- resets: Must contain an entry for each entry in reset-names.
+  See ../reset/reset.txt for details.
+- reset-names: Must include the following entries:
+  - gpu
+
+Example:
+
+	gpu@0,57000000 {
+		compatible = "nvidia,gk20a";
+		reg = <0x0 0x57000000 0x0 0x01000000>,
+		      <0x0 0x58000000 0x0 0x01000000>;
+		interrupts = <GIC_SPI 157 IRQ_TYPE_LEVEL_HIGH>,
+			     <GIC_SPI 158 IRQ_TYPE_LEVEL_HIGH>;
+		interrupt-names = "stall", "nonstall";
+		vdd-supply = <&vdd_gpu>;
+		clocks = <&tegra_car TEGRA124_CLK_GPU>,
+			 <&tegra_car TEGRA124_CLK_PLL_P_OUT5>;
+		clock-names = "gpu", "pll";
+		resets = <&tegra_car 184>;
+		reset-names = "gpu";
+		status = "disabled";
+	};
diff --git a/arch/arm/boot/dts/tegra124-jetson-tk1.dts b/arch/arm/boot/dts/tegra124-jetson-tk1.dts
index e31fb61a81d..15a194d1277 100644
--- a/arch/arm/boot/dts/tegra124-jetson-tk1.dts
+++ b/arch/arm/boot/dts/tegra124-jetson-tk1.dts
@@ -30,6 +30,12 @@
 		};
 	};
 
+	gpu@0,57000000 {
+		status = "okay";
+
+		vdd-supply = <&vdd_gpu>;
+	};
+
 	pinmux: pinmux@0,70000868 {
 		pinctrl-names = "default";
 		pinctrl-0 = <&state_default>;
@@ -1505,7 +1511,7 @@
 					regulator-always-on;
 				};
 
-				sd6 {
+				vdd_gpu: sd6 {
 					regulator-name = "+VDD_GPU_AP";
 					regulator-min-microvolt = <650000>;
 					regulator-max-microvolt = <1200000>;
diff --git a/arch/arm/boot/dts/tegra124-venice2.dts b/arch/arm/boot/dts/tegra124-venice2.dts
index f0bb8424402..86970ee4870 100644
--- a/arch/arm/boot/dts/tegra124-venice2.dts
+++ b/arch/arm/boot/dts/tegra124-venice2.dts
@@ -42,6 +42,12 @@
 		};
 	};
 
+	gpu@0,57000000 {
+		status = "okay";
+
+		vdd-supply = <&vdd_gpu>;
+	};
+
 	pinmux: pinmux@0,70000868 {
 		pinctrl-names = "default";
 		pinctrl-0 = <&pinmux_default>;
@@ -726,7 +732,7 @@
 					regulator-always-on;
 				};
 
-				sd6 {
+				vdd_gpu: sd6 {
 					regulator-name = "+VDD_GPU_AP";
 					regulator-min-microvolt = <650000>;
 					regulator-max-microvolt = <1200000>;
diff --git a/arch/arm/boot/dts/tegra124.dtsi b/arch/arm/boot/dts/tegra124.dtsi
index 011831ffd39..4a40cc54304 100644
--- a/arch/arm/boot/dts/tegra124.dtsi
+++ b/arch/arm/boot/dts/tegra124.dtsi
@@ -103,6 +103,21 @@
 			(GIC_CPU_MASK_SIMPLE(4) | IRQ_TYPE_LEVEL_HIGH)>;
 	};
 
+	gpu@0,57000000 {
+		compatible = "nvidia,gk20a";
+		reg = <0x0 0x57000000 0x0 0x01000000>,
+		      <0x0 0x58000000 0x0 0x01000000>;
+		interrupts = <GIC_SPI 157 IRQ_TYPE_LEVEL_HIGH>,
+			     <GIC_SPI 158 IRQ_TYPE_LEVEL_HIGH>;
+		interrupt-names = "stall", "nonstall";
+		clocks = <&tegra_car TEGRA124_CLK_GPU>,
+			 <&tegra_car TEGRA124_CLK_PLL_P_OUT5>;
+		clock-names = "gpu", "pwr";
+		resets = <&tegra_car 184>;
+		reset-names = "gpu";
+		status = "disabled";
+	};
+
 	timer@0,60005000 {
 		compatible = "nvidia,tegra124-timer", "nvidia,tegra20-timer";
 		reg = <0x0 0x60005000 0x0 0x400>;
diff --git a/arch/arm/configs/tegra_defconfig b/arch/arm/configs/tegra_defconfig
index fb25e2982f6..860165422dd 100644
--- a/arch/arm/configs/tegra_defconfig
+++ b/arch/arm/configs/tegra_defconfig
@@ -90,7 +90,6 @@ CONFIG_RFKILL_INPUT=y
 CONFIG_RFKILL_GPIO=y
 CONFIG_DEVTMPFS=y
 CONFIG_DEVTMPFS_MOUNT=y
-# CONFIG_FIRMWARE_IN_KERNEL is not set
 CONFIG_DMA_CMA=y
 CONFIG_CMA_SIZE_MBYTES=64
 CONFIG_MTD=y
@@ -293,3 +292,11 @@ CONFIG_CRYPTO_TWOFISH=y
 # CONFIG_CRYPTO_ANSI_CPRNG is not set
 CONFIG_CRYPTO_DEV_TEGRA_AES=y
 CONFIG_CRC_CCITT=y
+CONFIG_DRM_NOUVEAU=y
+CONFIG_NOUVEAU_DEBUG=5
+CONFIG_NOUVEAU_DEBUG_DEFAULT=3
+CONFIG_FW_LOADER=y
+CONFIG_FIRMWARE_IN_KERNEL=y
+CONFIG_EXTRA_FIRMWARE="nouveau/nvea_fuc409c nouveau/nvea_fuc409d nouveau/nvea_fuc41ac nouveau/nvea_fuc41ad"
+CONFIG_EXTRA_FIRMWARE_DIR="firmware"
+CONFIG_FW_LOADER_USER_HELPER=y
diff --git a/arch/arm/mach-tegra/tegra.c b/arch/arm/mach-tegra/tegra.c
index 15ac9fcc96b..972fc87ec18 100644
--- a/arch/arm/mach-tegra/tegra.c
+++ b/arch/arm/mach-tegra/tegra.c
@@ -73,6 +73,8 @@ u32 tegra_uart_config[3] = {
 static void __init tegra_init_early(void)
 {
 	of_register_trusted_foundations();
+	/* workaround to avoid FIFO regs freeze on T124 */
+	writel(0x1, IO_ADDRESS(0x70019000 + 0x650));
 	tegra_apb_io_init();
 	tegra_init_fuse();
 	tegra_cpu_reset_handler_init();
diff --git a/drivers/gpu/drm/nouveau/Kconfig b/drivers/gpu/drm/nouveau/Kconfig
index 637c29a3312..b8834ad55eb 100644
--- a/drivers/gpu/drm/nouveau/Kconfig
+++ b/drivers/gpu/drm/nouveau/Kconfig
@@ -25,6 +25,14 @@ config DRM_NOUVEAU
 	help
 	  Choose this option for open-source nVidia support.
 
+config NOUVEAU_PLATFORM_DRIVER
+	bool
+	depends on DRM_NOUVEAU
+	default y if ARCH_TEGRA
+	help
+	 Support for Nouveau platform driver, used for integrated GPUs as found
+	 on NVIDIA Tegra K1.
+
 config NOUVEAU_DEBUG
 	int "Maximum debug level"
 	depends on DRM_NOUVEAU
diff --git a/drivers/gpu/drm/nouveau/Makefile b/drivers/gpu/drm/nouveau/Makefile
index b7d21626477..5d1de264fc9 100644
--- a/drivers/gpu/drm/nouveau/Makefile
+++ b/drivers/gpu/drm/nouveau/Makefile
@@ -64,6 +64,7 @@ nouveau-y += core/subdev/clock/nva3.o
 nouveau-y += core/subdev/clock/nvaa.o
 nouveau-y += core/subdev/clock/nvc0.o
 nouveau-y += core/subdev/clock/nve0.o
+nouveau-y += core/subdev/clock/gk20a.o
 nouveau-y += core/subdev/clock/pllnv04.o
 nouveau-y += core/subdev/clock/pllnva3.o
 nouveau-y += core/subdev/devinit/base.o
@@ -102,6 +103,7 @@ nouveau-y += core/subdev/fb/nvaa.o
 nouveau-y += core/subdev/fb/nvaf.o
 nouveau-y += core/subdev/fb/nvc0.o
 nouveau-y += core/subdev/fb/nve0.o
+nouveau-y += core/subdev/fb/gk20a.o
 nouveau-y += core/subdev/fb/gm107.o
 nouveau-y += core/subdev/fb/ramnv04.o
 nouveau-y += core/subdev/fb/ramnv10.o
@@ -117,6 +119,7 @@ nouveau-y += core/subdev/fb/ramnva3.o
 nouveau-y += core/subdev/fb/ramnvaa.o
 nouveau-y += core/subdev/fb/ramnvc0.o
 nouveau-y += core/subdev/fb/ramnve0.o
+nouveau-y += core/subdev/fb/ramgk20a.o
 nouveau-y += core/subdev/fb/ramgm107.o
 nouveau-y += core/subdev/fb/sddr3.o
 nouveau-y += core/subdev/fb/gddr5.o
@@ -136,10 +139,12 @@ nouveau-y += core/subdev/i2c/nv94.o
 nouveau-y += core/subdev/i2c/nvd0.o
 nouveau-y += core/subdev/ibus/nvc0.o
 nouveau-y += core/subdev/ibus/nve0.o
+nouveau-y += core/subdev/ibus/gk20a.o
 nouveau-y += core/subdev/instmem/base.o
 nouveau-y += core/subdev/instmem/nv04.o
 nouveau-y += core/subdev/instmem/nv40.o
 nouveau-y += core/subdev/instmem/nv50.o
+nouveau-y += core/subdev/instmem/gk20a.o
 nouveau-y += core/subdev/ltcg/gf100.o
 nouveau-y += core/subdev/ltcg/gm107.o
 nouveau-y += core/subdev/mc/base.o
@@ -245,6 +250,7 @@ nouveau-y += core/engine/fifo/nv50.o
 nouveau-y += core/engine/fifo/nv84.o
 nouveau-y += core/engine/fifo/nvc0.o
 nouveau-y += core/engine/fifo/nve0.o
+nouveau-y += core/engine/fifo/gk20a.o
 nouveau-y += core/engine/fifo/nv108.o
 nouveau-y += core/engine/graph/ctxnv40.o
 nouveau-y += core/engine/graph/ctxnv50.o
@@ -255,6 +261,7 @@ nouveau-y += core/engine/graph/ctxnvc8.o
 nouveau-y += core/engine/graph/ctxnvd7.o
 nouveau-y += core/engine/graph/ctxnvd9.o
 nouveau-y += core/engine/graph/ctxnve4.o
+nouveau-y += core/engine/graph/ctxgk20a.o
 nouveau-y += core/engine/graph/ctxnvf0.o
 nouveau-y += core/engine/graph/ctxnv108.o
 nouveau-y += core/engine/graph/ctxgm107.o
@@ -275,6 +282,7 @@ nouveau-y += core/engine/graph/nvc8.o
 nouveau-y += core/engine/graph/nvd7.o
 nouveau-y += core/engine/graph/nvd9.o
 nouveau-y += core/engine/graph/nve4.o
+nouveau-y += core/engine/graph/gk20a.o
 nouveau-y += core/engine/graph/nvf0.o
 nouveau-y += core/engine/graph/nv108.o
 nouveau-y += core/engine/graph/gm107.o
@@ -310,6 +318,7 @@ nouveau-y += nouveau_ttm.o nouveau_sgdma.o nouveau_bo.o nouveau_gem.o
 nouveau-y += nouveau_prime.o nouveau_abi16.o
 nouveau-y += nv04_fence.o nv10_fence.o nv17_fence.o
 nouveau-y += nv50_fence.o nv84_fence.o nvc0_fence.o
+nouveau-y += nouveau_debug.o
 
 # drm/kms
 nouveau-y += nouveau_bios.o nouveau_fbcon.o nouveau_display.o
@@ -325,6 +334,9 @@ nouveau-y += nv50_display.o
 # drm/pm
 nouveau-y += nouveau_hwmon.o nouveau_sysfs.o
 
+# platform driver
+nouveau-$(CONFIG_NOUVEAU_PLATFORM_DRIVER) += nouveau_platform.o
+
 # other random bits
 nouveau-$(CONFIG_COMPAT) += nouveau_ioc32.o
 ifdef CONFIG_X86
diff --git a/drivers/gpu/drm/nouveau/core/engine/device/nve0.c b/drivers/gpu/drm/nouveau/core/engine/device/nve0.c
index 9784cbf8a9d..5b4204f19de 100644
--- a/drivers/gpu/drm/nouveau/core/engine/device/nve0.c
+++ b/drivers/gpu/drm/nouveau/core/engine/device/nve0.c
@@ -156,6 +156,24 @@ nve0_identify(struct nouveau_device *device)
 		device->oclass[NVDEV_ENGINE_PPP    ] = &nvc0_ppp_oclass;
 		device->oclass[NVDEV_ENGINE_PERFMON] = &nve0_perfmon_oclass;
 		break;
+	case 0xea:
+		device->cname = "GK20A";
+		device->oclass[NVDEV_SUBDEV_CLOCK  ] = &gk20a_clock_oclass;
+		device->oclass[NVDEV_SUBDEV_MC     ] =  nvc3_mc_oclass;
+		device->oclass[NVDEV_SUBDEV_BUS    ] =  nvc0_bus_oclass;
+		device->oclass[NVDEV_SUBDEV_TIMER  ] = &gk20a_timer_oclass;
+		device->oclass[NVDEV_SUBDEV_FB     ] =  gk20a_fb_oclass;
+		device->oclass[NVDEV_SUBDEV_IBUS   ] = &gk20a_ibus_oclass;
+		device->oclass[NVDEV_SUBDEV_INSTMEM] = gk20a_instmem_oclass;
+		device->oclass[NVDEV_SUBDEV_VM     ] = &nvc0_vmmgr_oclass;
+		device->oclass[NVDEV_SUBDEV_BAR    ] = &nvc0_bar_oclass;
+		device->oclass[NVDEV_ENGINE_DMAOBJ ] = &nvd0_dmaeng_oclass;
+		device->oclass[NVDEV_ENGINE_FIFO   ] =  gk20a_fifo_oclass;
+		device->oclass[NVDEV_ENGINE_SW     ] =  nvc0_software_oclass;
+		device->oclass[NVDEV_ENGINE_GR     ] =  gk20a_graph_oclass;
+		device->oclass[NVDEV_ENGINE_COPY2  ] = &nve0_copy2_oclass;
+		device->oclass[NVDEV_ENGINE_PERFMON] = &nve0_perfmon_oclass;
+		break;
 	case 0xf0:
 		device->cname = "GK110";
 		device->oclass[NVDEV_SUBDEV_VBIOS  ] = &nouveau_bios_oclass;
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/gk20a.c b/drivers/gpu/drm/nouveau/core/engine/fifo/gk20a.c
new file mode 100644
index 00000000000..327456eae96
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/gk20a.c
@@ -0,0 +1,35 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "nve0.h"
+
+struct nouveau_oclass *
+gk20a_fifo_oclass = &(struct nve0_fifo_impl) {
+	.base.handle = NV_ENGINE(FIFO, 0xea),
+	.base.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nve0_fifo_ctor,
+		.dtor = nve0_fifo_dtor,
+		.init = nve0_fifo_init,
+		.fini = nve0_fifo_fini,
+	},
+	.channels = 128,
+}.base;
diff --git a/drivers/gpu/drm/nouveau/core/engine/fifo/nve0.h b/drivers/gpu/drm/nouveau/core/engine/fifo/nve0.h
index 014344ebee6..e96b32bb1bb 100644
--- a/drivers/gpu/drm/nouveau/core/engine/fifo/nve0.h
+++ b/drivers/gpu/drm/nouveau/core/engine/fifo/nve0.h
@@ -8,6 +8,7 @@ int  nve0_fifo_ctor(struct nouveau_object *, struct nouveau_object *,
 		    struct nouveau_object **);
 void nve0_fifo_dtor(struct nouveau_object *);
 int  nve0_fifo_init(struct nouveau_object *);
+int  nve0_fifo_fini(struct nouveau_object *, bool);
 
 struct nve0_fifo_impl {
 	struct nouveau_oclass base;
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/ctxgk20a.c b/drivers/gpu/drm/nouveau/core/engine/graph/ctxgk20a.c
new file mode 100644
index 00000000000..224ee0287ab
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/ctxgk20a.c
@@ -0,0 +1,53 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "ctxnvc0.h"
+
+static const struct nvc0_graph_pack
+gk20a_grctx_pack_mthd[] = {
+	{ nve4_grctx_init_a097_0, 0xa297 },
+	{ nvc0_grctx_init_902d_0, 0x902d },
+	{}
+};
+
+struct nouveau_oclass *
+gk20a_grctx_oclass = &(struct nvc0_grctx_oclass) {
+	.base.handle = NV_ENGCTX(GR, 0xea),
+	.base.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_graph_context_ctor,
+		.dtor = nvc0_graph_context_dtor,
+		.init = _nouveau_graph_context_init,
+		.fini = _nouveau_graph_context_fini,
+		.rd32 = _nouveau_graph_context_rd32,
+		.wr32 = _nouveau_graph_context_wr32,
+	},
+	.main  = nve4_grctx_generate_main,
+	.mods  = nve4_grctx_generate_mods,
+	.unkn  = nve4_grctx_generate_unkn,
+	.hub   = nve4_grctx_pack_hub,
+	.gpc   = nve4_grctx_pack_gpc,
+	.zcull = nvc0_grctx_pack_zcull,
+	.tpc   = nve4_grctx_pack_tpc,
+	.ppc   = nve4_grctx_pack_ppc,
+	.icmd  = nve4_grctx_pack_icmd,
+	.mthd  = gk20a_grctx_pack_mthd,
+}.base;
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/ctxnvc0.h b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnvc0.h
index 9c815d1f99e..8da8b627b9d 100644
--- a/drivers/gpu/drm/nouveau/core/engine/graph/ctxnvc0.h
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnvc0.h
@@ -69,7 +69,9 @@ extern struct nouveau_oclass *nvd7_grctx_oclass;
 extern struct nouveau_oclass *nvd9_grctx_oclass;
 
 extern struct nouveau_oclass *nve4_grctx_oclass;
+extern struct nouveau_oclass *gk20a_grctx_oclass;
 void nve4_grctx_generate_main(struct nvc0_graph_priv *, struct nvc0_grctx *);
+void nve4_grctx_generate_mods(struct nvc0_graph_priv *, struct nvc0_grctx *);
 void nve4_grctx_generate_unkn(struct nvc0_graph_priv *);
 void nve4_grctx_generate_r418bb8(struct nvc0_graph_priv *);
 
@@ -151,6 +153,13 @@ extern const struct nvc0_graph_init nve4_grctx_init_gpm_0[];
 
 extern const struct nvc0_graph_init nve4_grctx_init_pes_0[];
 
+extern const struct nvc0_graph_pack nve4_grctx_pack_hub[];
+extern const struct nvc0_graph_pack nve4_grctx_pack_gpc[];
+extern const struct nvc0_graph_pack nve4_grctx_pack_tpc[];
+extern const struct nvc0_graph_pack nve4_grctx_pack_ppc[];
+extern const struct nvc0_graph_pack nve4_grctx_pack_icmd[];
+extern const struct nvc0_graph_init nve4_grctx_init_a097_0[];
+
 extern const struct nvc0_graph_pack nvf0_grctx_pack_mthd[];
 
 extern const struct nvc0_graph_init nvf0_grctx_init_pri_0[];
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/ctxnve4.c b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnve4.c
index 49a14b116a5..f722089945d 100644
--- a/drivers/gpu/drm/nouveau/core/engine/graph/ctxnve4.c
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/ctxnve4.c
@@ -272,13 +272,13 @@ nve4_grctx_init_icmd_0[] = {
 	{}
 };
 
-static const struct nvc0_graph_pack
+const struct nvc0_graph_pack
 nve4_grctx_pack_icmd[] = {
 	{ nve4_grctx_init_icmd_0 },
 	{}
 };
 
-static const struct nvc0_graph_init
+const struct nvc0_graph_init
 nve4_grctx_init_a097_0[] = {
 	{ 0x000800,   8, 0x40, 0x00000000 },
 	{ 0x000804,   8, 0x40, 0x00000000 },
@@ -697,7 +697,7 @@ nve4_grctx_init_be_0[] = {
 	{}
 };
 
-static const struct nvc0_graph_pack
+const struct nvc0_graph_pack
 nve4_grctx_pack_hub[] = {
 	{ nvc0_grctx_init_main_0 },
 	{ nve4_grctx_init_fe_0 },
@@ -737,7 +737,7 @@ nve4_grctx_init_gpm_0[] = {
 	{}
 };
 
-static const struct nvc0_graph_pack
+const struct nvc0_graph_pack
 nve4_grctx_pack_gpc[] = {
 	{ nvc0_grctx_init_gpc_unk_0 },
 	{ nvd9_grctx_init_prop_0 },
@@ -802,7 +802,7 @@ nve4_grctx_init_sm_0[] = {
 	{}
 };
 
-static const struct nvc0_graph_pack
+const struct nvc0_graph_pack
 nve4_grctx_pack_tpc[] = {
 	{ nvd7_grctx_init_pe_0 },
 	{ nve4_grctx_init_tex_0 },
@@ -826,7 +826,7 @@ nve4_grctx_init_cbm_0[] = {
 	{}
 };
 
-static const struct nvc0_graph_pack
+const struct nvc0_graph_pack
 nve4_grctx_pack_ppc[] = {
 	{ nve4_grctx_init_pes_0 },
 	{ nve4_grctx_init_cbm_0 },
@@ -838,7 +838,7 @@ nve4_grctx_pack_ppc[] = {
  * PGRAPH context implementation
  ******************************************************************************/
 
-static void
+void
 nve4_grctx_generate_mods(struct nvc0_graph_priv *priv, struct nvc0_grctx *info)
 {
 	u32 magic[GPC_MAX][2];
@@ -957,7 +957,7 @@ nve4_grctx_generate_main(struct nvc0_graph_priv *priv, struct nvc0_grctx *info)
 	struct nvc0_grctx_oclass *oclass = (void *)nv_engine(priv)->cclass;
 	int i;
 
-	nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
+	//nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
 
 	nvc0_graph_mmio(priv, oclass->hub);
 	nvc0_graph_mmio(priv, oclass->gpc);
@@ -991,7 +991,7 @@ nve4_grctx_generate_main(struct nvc0_graph_priv *priv, struct nvc0_grctx *info)
 	nvc0_graph_icmd(priv, oclass->icmd);
 	nv_wr32(priv, 0x404154, 0x00000400);
 	nvc0_graph_mthd(priv, oclass->mthd);
-	nv_mask(priv, 0x000260, 0x00000001, 0x00000001);
+	//nv_mask(priv, 0x000260, 0x00000001, 0x00000001);
 
 	nv_mask(priv, 0x418800, 0x00200000, 0x00200000);
 	nv_mask(priv, 0x41be10, 0x00800000, 0x00800000);
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/gk20a.c b/drivers/gpu/drm/nouveau/core/engine/graph/gk20a.c
new file mode 100644
index 00000000000..83048a56430
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/gk20a.c
@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "nvc0.h"
+#include "ctxnvc0.h"
+
+static struct nouveau_oclass
+gk20a_graph_sclass[] = {
+	{ 0x902d, &nouveau_object_ofuncs },
+	{ 0xa040, &nouveau_object_ofuncs },
+	{ 0xa297, &nouveau_object_ofuncs },
+	{ 0xa0c0, &nouveau_object_ofuncs },
+	{}
+};
+
+struct nouveau_oclass *
+gk20a_graph_oclass = &(struct nvc0_graph_oclass) {
+	.base.handle = NV_ENGINE(GR, 0xea),
+	.base.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = nvc0_graph_ctor,
+		.dtor = nvc0_graph_dtor,
+		.init = nve4_graph_init,
+		.fini = nve4_graph_fini,
+	},
+	.cclass = &gk20a_grctx_oclass,
+	.sclass = gk20a_graph_sclass,
+	.mmio = nve4_graph_pack_mmio,
+}.base;
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.c b/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.c
index f3c7329da0a..7a6cd1e2f29 100644
--- a/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.c
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.c
@@ -894,6 +894,10 @@ nvc0_graph_init_fw(struct nvc0_graph_priv *priv, u32 fuc_base,
 			nv_wr32(priv, fuc_base + 0x0188, i >> 6);
 		nv_wr32(priv, fuc_base + 0x0184, code->data[i]);
 	}
+
+	/* code must be padded to 0x40 words */
+	for (; i & 0x3f; i++)
+		nv_wr32(priv, fuc_base + 0x0184, 0);
 }
 
 static void
@@ -942,17 +946,17 @@ nvc0_graph_init_ctxctl(struct nvc0_graph_priv *priv)
 {
 	struct nvc0_graph_oclass *oclass = (void *)nv_object(priv)->oclass;
 	struct nvc0_grctx_oclass *cclass = (void *)nv_engine(priv)->cclass;
-	u32 r000260;
+	//u32 r000260;
 	int i;
 
 	if (priv->firmware) {
 		/* load fuc microcode */
-		r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
+		//r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
 		nvc0_graph_init_fw(priv, 0x409000, &priv->fuc409c,
 						   &priv->fuc409d);
 		nvc0_graph_init_fw(priv, 0x41a000, &priv->fuc41ac,
 						   &priv->fuc41ad);
-		nv_wr32(priv, 0x000260, r000260);
+		//nv_wr32(priv, 0x000260, r000260);
 
 		/* start both of them running */
 		nv_wr32(priv, 0x409840, 0xffffffff);
@@ -1039,7 +1043,7 @@ nvc0_graph_init_ctxctl(struct nvc0_graph_priv *priv)
 	}
 
 	/* load HUB microcode */
-	r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
+	//r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
 	nv_wr32(priv, 0x4091c0, 0x01000000);
 	for (i = 0; i < oclass->fecs.ucode->data.size / 4; i++)
 		nv_wr32(priv, 0x4091c4, oclass->fecs.ucode->data.data[i]);
@@ -1062,7 +1066,7 @@ nvc0_graph_init_ctxctl(struct nvc0_graph_priv *priv)
 			nv_wr32(priv, 0x41a188, i >> 6);
 		nv_wr32(priv, 0x41a184, oclass->gpccs.ucode->code.data[i]);
 	}
-	nv_wr32(priv, 0x000260, r000260);
+	//nv_wr32(priv, 0x000260, r000260);
 
 	/* load register lists */
 	nvc0_graph_init_csdata(priv, cclass->hub, 0x409000, 0x000, 0x000000);
@@ -1259,10 +1263,14 @@ nvc0_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
 	struct nvc0_graph_oclass *oclass = (void *)bclass;
 	struct nouveau_device *device = nv_device(parent);
 	struct nvc0_graph_priv *priv;
+	bool use_ext_fw, enable;
 	int ret, i;
 
-	ret = nouveau_graph_create(parent, engine, bclass,
-				   (oclass->fecs.ucode != NULL), &priv);
+	use_ext_fw = nouveau_boolopt(device->cfgopt, "NvGrUseFW",
+				     oclass->fecs.ucode == NULL);
+	enable = use_ext_fw || oclass->fecs.ucode != NULL;
+
+	ret = nouveau_graph_create(parent, engine, bclass, enable, &priv);
 	*pobject = nv_object(priv);
 	if (ret)
 		return ret;
@@ -1272,7 +1280,7 @@ nvc0_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
 
 	priv->base.units = nvc0_graph_units;
 
-	if (nouveau_boolopt(device->cfgopt, "NvGrUseFW", false)) {
+	if (use_ext_fw) {
 		nv_info(priv, "using external firmware\n");
 		if (nvc0_graph_ctor_fw(priv, "fuc409c", &priv->fuc409c) ||
 		    nvc0_graph_ctor_fw(priv, "fuc409d", &priv->fuc409d) ||
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.h b/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.h
index 90d44616c87..75203a99d90 100644
--- a/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.h
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nvc0.h
@@ -116,6 +116,7 @@ int  nvc0_graph_ctor(struct nouveau_object *, struct nouveau_object *,
 		     struct nouveau_object **);
 void nvc0_graph_dtor(struct nouveau_object *);
 int  nvc0_graph_init(struct nouveau_object *);
+int  nve4_graph_fini(struct nouveau_object *, bool);
 int  nve4_graph_init(struct nouveau_object *);
 
 extern struct nouveau_oclass nvc0_graph_sclass[];
@@ -217,6 +218,7 @@ extern const struct nvc0_graph_init nve4_graph_init_main_0[];
 extern const struct nvc0_graph_init nve4_graph_init_tpccs_0[];
 extern const struct nvc0_graph_init nve4_graph_init_pe_0[];
 extern const struct nvc0_graph_init nve4_graph_init_be_0[];
+extern const struct nvc0_graph_pack nve4_graph_pack_mmio[];
 
 extern const struct nvc0_graph_init nvf0_graph_init_fe_0[];
 extern const struct nvc0_graph_init nvf0_graph_init_sked_0[];
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nve4.c b/drivers/gpu/drm/nouveau/core/engine/graph/nve4.c
index f7c01121717..51e0c075ad3 100644
--- a/drivers/gpu/drm/nouveau/core/engine/graph/nve4.c
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nve4.c
@@ -151,7 +151,7 @@ nve4_graph_init_be_0[] = {
 	{}
 };
 
-static const struct nvc0_graph_pack
+const struct nvc0_graph_pack
 nve4_graph_pack_mmio[] = {
 	{ nve4_graph_init_main_0 },
 	{ nvc0_graph_init_fe_0 },
@@ -189,7 +189,7 @@ nve4_graph_pack_mmio[] = {
  * PGRAPH engine/subdev functions
  ******************************************************************************/
 
-static int
+int
 nve4_graph_fini(struct nouveau_object *object, bool suspend)
 {
 	struct nvc0_graph_priv *priv = (void *)object;
diff --git a/drivers/gpu/drm/nouveau/core/include/engine/fifo.h b/drivers/gpu/drm/nouveau/core/include/engine/fifo.h
index 26b6b2bb111..b639eb2c74f 100644
--- a/drivers/gpu/drm/nouveau/core/include/engine/fifo.h
+++ b/drivers/gpu/drm/nouveau/core/include/engine/fifo.h
@@ -109,6 +109,7 @@ extern struct nouveau_oclass *nv50_fifo_oclass;
 extern struct nouveau_oclass *nv84_fifo_oclass;
 extern struct nouveau_oclass *nvc0_fifo_oclass;
 extern struct nouveau_oclass *nve0_fifo_oclass;
+extern struct nouveau_oclass *gk20a_fifo_oclass;
 extern struct nouveau_oclass *nv108_fifo_oclass;
 
 void nv04_fifo_intr(struct nouveau_subdev *);
diff --git a/drivers/gpu/drm/nouveau/core/include/engine/graph.h b/drivers/gpu/drm/nouveau/core/include/engine/graph.h
index 871edfdf3d5..8c1d4772da0 100644
--- a/drivers/gpu/drm/nouveau/core/include/engine/graph.h
+++ b/drivers/gpu/drm/nouveau/core/include/engine/graph.h
@@ -68,6 +68,7 @@ extern struct nouveau_oclass *nvc8_graph_oclass;
 extern struct nouveau_oclass *nvd7_graph_oclass;
 extern struct nouveau_oclass *nvd9_graph_oclass;
 extern struct nouveau_oclass *nve4_graph_oclass;
+extern struct nouveau_oclass *gk20a_graph_oclass;
 extern struct nouveau_oclass *nvf0_graph_oclass;
 extern struct nouveau_oclass *nv108_graph_oclass;
 extern struct nouveau_oclass *gm107_graph_oclass;
diff --git a/drivers/gpu/drm/nouveau/core/include/subdev/clock.h b/drivers/gpu/drm/nouveau/core/include/subdev/clock.h
index 8f4ced75444..e330503f101 100644
--- a/drivers/gpu/drm/nouveau/core/include/subdev/clock.h
+++ b/drivers/gpu/drm/nouveau/core/include/subdev/clock.h
@@ -134,6 +134,7 @@ extern struct nouveau_oclass *nvaa_clock_oclass;
 extern struct nouveau_oclass nva3_clock_oclass;
 extern struct nouveau_oclass nvc0_clock_oclass;
 extern struct nouveau_oclass nve0_clock_oclass;
+extern struct nouveau_oclass gk20a_clock_oclass;
 
 int nv04_clock_pll_set(struct nouveau_clock *, u32 type, u32 freq);
 int nv04_clock_pll_calc(struct nouveau_clock *, struct nvbios_pll *,
diff --git a/drivers/gpu/drm/nouveau/core/include/subdev/fb.h b/drivers/gpu/drm/nouveau/core/include/subdev/fb.h
index 58c7ccdebb0..871e73914b2 100644
--- a/drivers/gpu/drm/nouveau/core/include/subdev/fb.h
+++ b/drivers/gpu/drm/nouveau/core/include/subdev/fb.h
@@ -105,6 +105,7 @@ extern struct nouveau_oclass *nvaa_fb_oclass;
 extern struct nouveau_oclass *nvaf_fb_oclass;
 extern struct nouveau_oclass *nvc0_fb_oclass;
 extern struct nouveau_oclass *nve0_fb_oclass;
+extern struct nouveau_oclass *gk20a_fb_oclass;
 extern struct nouveau_oclass *gm107_fb_oclass;
 
 #include <subdev/bios/ramcfg.h>
diff --git a/drivers/gpu/drm/nouveau/core/include/subdev/ibus.h b/drivers/gpu/drm/nouveau/core/include/subdev/ibus.h
index 88814f159d8..31df634c0fd 100644
--- a/drivers/gpu/drm/nouveau/core/include/subdev/ibus.h
+++ b/drivers/gpu/drm/nouveau/core/include/subdev/ibus.h
@@ -30,5 +30,6 @@ nouveau_ibus(void *obj)
 
 extern struct nouveau_oclass nvc0_ibus_oclass;
 extern struct nouveau_oclass nve0_ibus_oclass;
+extern struct nouveau_oclass gk20a_ibus_oclass;
 
 #endif
diff --git a/drivers/gpu/drm/nouveau/core/include/subdev/instmem.h b/drivers/gpu/drm/nouveau/core/include/subdev/instmem.h
index c1df26f3230..6264660bedc 100644
--- a/drivers/gpu/drm/nouveau/core/include/subdev/instmem.h
+++ b/drivers/gpu/drm/nouveau/core/include/subdev/instmem.h
@@ -48,5 +48,6 @@ nouveau_instmem(void *obj)
 extern struct nouveau_oclass *nv04_instmem_oclass;
 extern struct nouveau_oclass *nv40_instmem_oclass;
 extern struct nouveau_oclass *nv50_instmem_oclass;
+extern struct nouveau_oclass *gk20a_instmem_oclass;
 
 #endif
diff --git a/drivers/gpu/drm/nouveau/core/os.h b/drivers/gpu/drm/nouveau/core/os.h
index d0ced94ca54..fc84a258f91 100644
--- a/drivers/gpu/drm/nouveau/core/os.h
+++ b/drivers/gpu/drm/nouveau/core/os.h
@@ -21,6 +21,7 @@
 #include <linux/interrupt.h>
 #include <linux/log2.h>
 #include <linux/pm_runtime.h>
+#include <asm/cacheflush.h>
 
 #include <asm/unaligned.h>
 
@@ -38,4 +39,23 @@
 #endif /* def __BIG_ENDIAN else */
 #endif /* !ioread32_native */
 
+#if defined(__arm__)
+
+static inline void
+nv_cpu_cache_flush_area(void *va, size_t size)
+{
+	phys_addr_t pa = virt_to_phys(va);
+	__cpuc_flush_dcache_area(va, size);
+	outer_flush_range(pa, pa + size);
+}
+
+#else
+
+static inline void
+nv_cpu_cache_flush_area(void *va, size_t size)
+{
+}
+
+#endif /* defined(__arm__) */
+
 #endif
diff --git a/drivers/gpu/drm/nouveau/core/subdev/bar/base.c b/drivers/gpu/drm/nouveau/core/subdev/bar/base.c
index bdf594116f3..73b1ed20c8d 100644
--- a/drivers/gpu/drm/nouveau/core/subdev/bar/base.c
+++ b/drivers/gpu/drm/nouveau/core/subdev/bar/base.c
@@ -118,8 +118,10 @@ nouveau_bar_create_(struct nouveau_object *parent,
 	if (ret)
 		return ret;
 
-	bar->iomem = ioremap(nv_device_resource_start(device, 3),
-			     nv_device_resource_len(device, 3));
+	if (nv_device_resource_len(device, 3) != 0)
+		bar->iomem = ioremap(nv_device_resource_start(device, 3),
+				     nv_device_resource_len(device, 3));
+
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/nouveau/core/subdev/bar/nvc0.c b/drivers/gpu/drm/nouveau/core/subdev/bar/nvc0.c
index 3f30db62e65..ca8139b9ab2 100644
--- a/drivers/gpu/drm/nouveau/core/subdev/bar/nvc0.c
+++ b/drivers/gpu/drm/nouveau/core/subdev/bar/nvc0.c
@@ -30,14 +30,16 @@
 
 #include "priv.h"
 
+struct nvc0_bar_priv_vm {
+	struct nouveau_gpuobj *mem;
+	struct nouveau_gpuobj *pgd;
+	struct nouveau_vm *vm;
+};
+
 struct nvc0_bar_priv {
 	struct nouveau_bar base;
 	spinlock_t lock;
-	struct {
-		struct nouveau_gpuobj *mem;
-		struct nouveau_gpuobj *pgd;
-		struct nouveau_vm *vm;
-	} bar[2];
+	struct nvc0_bar_priv_vm bar[2];
 };
 
 static int
@@ -79,87 +81,87 @@ nvc0_bar_unmap(struct nouveau_bar *bar, struct nouveau_vma *vma)
 }
 
 static int
-nvc0_bar_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
-	      struct nouveau_oclass *oclass, void *data, u32 size,
-	      struct nouveau_object **pobject)
+nvc0_bar_init_vm(struct nvc0_bar_priv *priv, struct nvc0_bar_priv_vm *bar_vm,
+		 int bar_nr)
 {
-	struct nouveau_device *device = nv_device(parent);
-	struct nvc0_bar_priv *priv;
-	struct nouveau_gpuobj *mem;
+	struct nouveau_device *device = nv_device(&priv->base);
 	struct nouveau_vm *vm;
+	resource_size_t bar_len;
 	int ret;
 
-	ret = nouveau_bar_create(parent, engine, oclass, &priv);
-	*pobject = nv_object(priv);
-	if (ret)
-		return ret;
-
-	/* BAR3 */
 	ret = nouveau_gpuobj_new(nv_object(priv), NULL, 0x1000, 0, 0,
-				&priv->bar[0].mem);
-	mem = priv->bar[0].mem;
+				&bar_vm->mem);
 	if (ret)
 		return ret;
 
 	ret = nouveau_gpuobj_new(nv_object(priv), NULL, 0x8000, 0, 0,
-				&priv->bar[0].pgd);
+				&bar_vm->pgd);
 	if (ret)
 		return ret;
 
-	ret = nouveau_vm_new(device, 0, nv_device_resource_len(device, 3), 0, &vm);
+	bar_len = nv_device_resource_len(device, bar_nr);
+
+	ret = nouveau_vm_new(device, 0, bar_len, 0, &vm);
 	if (ret)
 		return ret;
 
 	atomic_inc(&vm->engref[NVDEV_SUBDEV_BAR]);
 
-	ret = nouveau_gpuobj_new(nv_object(priv), NULL,
-				 (nv_device_resource_len(device, 3) >> 12) * 8,
-				 0x1000, NVOBJ_FLAG_ZERO_ALLOC,
-				 &vm->pgt[0].obj[0]);
-	vm->pgt[0].refcount[0] = 1;
-	if (ret)
-		return ret;
+	/*
+	 * Bootstrap page table lookup.
+	 */
+	if (bar_nr == 3) {
+		ret = nouveau_gpuobj_new(nv_object(priv), NULL,
+					 (bar_len >> 12) * 8, 0x1000,
+					 NVOBJ_FLAG_ZERO_ALLOC,
+					&vm->pgt[0].obj[0]);
+		vm->pgt[0].refcount[0] = 1;
+		if (ret)
+			return ret;
+	}
 
-	ret = nouveau_vm_ref(vm, &priv->bar[0].vm, priv->bar[0].pgd);
+	ret = nouveau_vm_ref(vm, &bar_vm->vm, bar_vm->pgd);
 	nouveau_vm_ref(NULL, &vm, NULL);
 	if (ret)
 		return ret;
 
-	nv_wo32(mem, 0x0200, lower_32_bits(priv->bar[0].pgd->addr));
-	nv_wo32(mem, 0x0204, upper_32_bits(priv->bar[0].pgd->addr));
-	nv_wo32(mem, 0x0208, lower_32_bits(nv_device_resource_len(device, 3) - 1));
-	nv_wo32(mem, 0x020c, upper_32_bits(nv_device_resource_len(device, 3) - 1));
+	nv_wo32(bar_vm->mem, 0x0200, lower_32_bits(bar_vm->pgd->addr));
+	nv_wo32(bar_vm->mem, 0x0204, upper_32_bits(bar_vm->pgd->addr));
+	nv_wo32(bar_vm->mem, 0x0208, lower_32_bits(bar_len - 1));
+	nv_wo32(bar_vm->mem, 0x020c, upper_32_bits(bar_len - 1));
 
-	/* BAR1 */
-	ret = nouveau_gpuobj_new(nv_object(priv), NULL, 0x1000, 0, 0,
-				&priv->bar[1].mem);
-	mem = priv->bar[1].mem;
-	if (ret)
-		return ret;
+	return 0;
+}
 
-	ret = nouveau_gpuobj_new(nv_object(priv), NULL, 0x8000, 0, 0,
-				&priv->bar[1].pgd);
-	if (ret)
-		return ret;
+static int
+nvc0_bar_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	      struct nouveau_oclass *oclass, void *data, u32 size,
+	      struct nouveau_object **pobject)
+{
+	struct nouveau_device *device = nv_device(parent);
+	struct nvc0_bar_priv *priv;
+	bool has_bar3 = nv_device_resource_len(device, 3) != 0;
+	int ret;
 
-	ret = nouveau_vm_new(device, 0, nv_device_resource_len(device, 1), 0, &vm);
+	ret = nouveau_bar_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
 	if (ret)
 		return ret;
 
-	atomic_inc(&vm->engref[NVDEV_SUBDEV_BAR]);
+	/* BAR3 */
+	if (has_bar3) {
+		ret = nvc0_bar_init_vm(priv, &priv->bar[0], 3);
+		if (ret)
+			return ret;
+		priv->base.alloc = nouveau_bar_alloc;
+		priv->base.kmap = nvc0_bar_kmap;
+	}
 
-	ret = nouveau_vm_ref(vm, &priv->bar[1].vm, priv->bar[1].pgd);
-	nouveau_vm_ref(NULL, &vm, NULL);
+	/* BAR1 */
+	ret = nvc0_bar_init_vm(priv, &priv->bar[1], 1);
 	if (ret)
 		return ret;
 
-	nv_wo32(mem, 0x0200, lower_32_bits(priv->bar[1].pgd->addr));
-	nv_wo32(mem, 0x0204, upper_32_bits(priv->bar[1].pgd->addr));
-	nv_wo32(mem, 0x0208, lower_32_bits(nv_device_resource_len(device, 1) - 1));
-	nv_wo32(mem, 0x020c, upper_32_bits(nv_device_resource_len(device, 1) - 1));
-
-	priv->base.alloc = nouveau_bar_alloc;
-	priv->base.kmap = nvc0_bar_kmap;
 	priv->base.umap = nvc0_bar_umap;
 	priv->base.unmap = nvc0_bar_unmap;
 	priv->base.flush = nv84_bar_flush;
@@ -201,7 +203,9 @@ nvc0_bar_init(struct nouveau_object *object)
 	nv_mask(priv, 0x100c80, 0x00000001, 0x00000000);
 
 	nv_wr32(priv, 0x001704, 0x80000000 | priv->bar[1].mem->addr >> 12);
-	nv_wr32(priv, 0x001714, 0xc0000000 | priv->bar[0].mem->addr >> 12);
+	if (priv->bar[0].mem)
+		nv_wr32(priv, 0x001714,
+			0xc0000000 | priv->bar[0].mem->addr >> 12);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/nouveau/core/subdev/clock/gk20a.c b/drivers/gpu/drm/nouveau/core/subdev/clock/gk20a.c
new file mode 100644
index 00000000000..ac80513ddea
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/subdev/clock/gk20a.c
@@ -0,0 +1,496 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Shamelessly ripped off from ChromeOS's gk20a/clk_pllg.c
+ *
+ */
+
+#define MHZ (1000 * 1000)
+
+#define MASK(w)	((1 << w) - 1)
+
+#define SYS_GPCPLL_CFG_BASE			0x00137000
+#define GPC_BCASE_GPCPLL_CFG_BASE	0x00132800
+
+#define GPCPLL_CFG		(SYS_GPCPLL_CFG_BASE + 0)
+#define GPCPLL_CFG_ENABLE	BIT(0)
+#define GPCPLL_CFG_IDDQ		BIT(1)
+#define GPCPLL_CFG_LOCK_DET_OFF	BIT(4)
+#define GPCPLL_CFG_LOCK		BIT(17)
+
+#define GPCPLL_COEFF		(SYS_GPCPLL_CFG_BASE + 4)
+#define GPCPLL_COEFF_M_SHIFT	0
+#define GPCPLL_COEFF_N_SHIFT	8
+#define GPCPLL_COEFF_P_SHIFT	16
+
+#define GPCPLL_CFG2			(SYS_GPCPLL_CFG_BASE + 0xc)
+#define GPCPLL_CFG2_SETUP2_SHIFT	16
+#define GPCPLL_CFG2_PLL_STEPA_SHIFT	24
+
+#define GPCPLL_CFG3			(SYS_GPCPLL_CFG_BASE + 0x18)
+#define GPCPLL_CFG3_PLL_STEPB_SHIFT		16
+
+#define GPCPLL_NDIV_SLOWDOWN	(SYS_GPCPLL_CFG_BASE + 0x1c)
+#define GPCPLL_NDIV_SLOWDOWN_NDIV_LO_SHIFT	0
+#define GPCPLL_NDIV_SLOWDOWN_NDIV_MID_SHIFT	8
+#define GPCPLL_NDIV_SLOWDOWN_STEP_SIZE_LO2MID_SHIFT	16
+#define GPCPLL_NDIV_SLOWDOWN_SLOWDOWN_USING_PLL_SHIFT	22
+#define GPCPLL_NDIV_SLOWDOWN_EN_DYNRAMP_SHIFT	31
+
+#define SEL_VCO		(SYS_GPCPLL_CFG_BASE + 0x100)
+#define SEL_VCO_GPC2CLK_OUT_SHIFT	0
+
+#define GPC2CLK_OUT	(SYS_GPCPLL_CFG_BASE + 0x250)
+#define GPC2CLK_OUT_SDIV14_INDIV4_WIDTH 1
+#define GPC2CLK_OUT_SDIV14_INDIV4_SHIFT 31
+#define GPC2CLK_OUT_SDIV14_INDIV4_MODE 1
+#define GPC2CLK_OUT_VCODIV_WIDTH 6
+#define GPC2CLK_OUT_VCODIV_SHIFT 8
+#define GPC2CLK_OUT_VCODIV1 0
+#define GPC2CLK_OUT_VCODIV_MASK (MASK(GPC2CLK_OUT_VCODIV_WIDTH) << \
+		GPC2CLK_OUT_VCODIV_SHIFT)
+#define	GPC2CLK_OUT_BYPDIV_WIDTH 6
+#define GPC2CLK_OUT_BYPDIV_SHIFT 0
+#define GPC2CLK_OUT_BYPDIV31 0x3c
+#define GPC2CLK_OUT_INIT_MASK	((MASK(GPC2CLK_OUT_SDIV14_INDIV4_WIDTH) << \
+		GPC2CLK_OUT_SDIV14_INDIV4_SHIFT)\
+		| (MASK(GPC2CLK_OUT_VCODIV_WIDTH) << GPC2CLK_OUT_VCODIV_SHIFT)\
+		| (MASK(GPC2CLK_OUT_BYPDIV_WIDTH) << GPC2CLK_OUT_BYPDIV_SHIFT))
+#define GPC2CLK_OUT_INIT_VAL	((GPC2CLK_OUT_SDIV14_INDIV4_MODE << \
+		GPC2CLK_OUT_SDIV14_INDIV4_SHIFT) \
+		| (GPC2CLK_OUT_VCODIV1 << GPC2CLK_OUT_VCODIV_SHIFT) \
+		| (GPC2CLK_OUT_BYPDIV31 << GPC2CLK_OUT_BYPDIV_SHIFT))
+
+#define GPC_BCAST_NDIV_SLOWDOWN_DEBUG	(GPC_BCASE_GPCPLL_CFG_BASE + 0xa0)
+#define GPC_BCAST_NDIV_SLOWDOWN_DEBUG_PLL_DYNRAMP_DONE_SYNCED_SHIFT	24
+#define GPC_BCAST_NDIV_SLOWDOWN_DEBUG_PLL_DYNRAMP_DONE_SYNCED_MASK \
+	    (0x1 << GPC_BCAST_NDIV_SLOWDOWN_DEBUG_PLL_DYNRAMP_DONE_SYNCED_SHIFT)
+
+#include <linux/types.h>
+
+static u8 pl_to_div[] = {
+/* PL:   0, 1, 2, 3, 4, 5, 6,  7,  8,  9, 10, 11, 12, 13, 14 */
+/* p: */ 1, 2, 3, 4, 5, 6, 8, 10, 12, 16, 12, 16, 20, 24, 32,
+};
+
+struct gk20a_clk_pllg_params {
+	u32 min_freq, max_freq; /* MHz */
+	u32 min_vco, max_vco;   /* MHz */
+	u32 min_u, max_u;       /* MHz */
+	u32 min_m, max_m;
+	u32 min_n, max_n;
+	u32 min_pl, max_pl;
+};
+
+const struct gk20a_clk_pllg_params gk20a_pllg_params = {
+	.min_freq = 144, .max_freq = 2064,
+	.min_vco = 1000, .max_vco = 2064,
+	.min_u = 12, .max_u = 38,
+	.min_m = 1, .max_m = 255,
+	.min_n = 8, .max_n = 255,
+	.min_pl = 1, .max_pl = 32,
+};
+
+#include <subdev/clock.h>
+#include <subdev/timer.h>
+
+#include <nouveau_platform.h>
+
+#include <linux/clk.h>
+#include <linux/delay.h>
+
+struct gk20a_clock_priv {
+	struct nouveau_clock base;
+	const struct gk20a_clk_pllg_params *params;
+	u32 m, n, pl;
+	unsigned long parent_rate;
+	bool enabled;
+	bool init;
+};
+#define to_gk20a_clock(base) container_of(base, struct gk20a_clock_priv, base)
+
+static unsigned long
+gk20a_pllg_calc_rate(struct gk20a_clock_priv *priv)
+{
+	unsigned long rate;
+	int divider;
+
+	rate = priv->parent_rate * priv->n;
+	divider = priv->m * pl_to_div[priv->pl];
+	do_div(rate, divider);
+
+	return rate;
+}
+
+static int
+gk20a_pllg_calc_mnp(struct gk20a_clock_priv *priv, unsigned long rate)
+{
+	unsigned int target_clk_f, ref_clk_f, target_freq;
+	unsigned int min_vco_f, max_vco_f;
+	u32 low_pl, high_pl, best_pl;
+	unsigned int target_vco_f, vco_f;
+	u32 best_m, best_n;
+	unsigned int u_f;
+	u32 m, n, n2;
+	u32 delta, lwv, best_delta = ~0;
+	int pl;
+
+	target_clk_f = rate * 2 / MHZ;
+	ref_clk_f = priv->parent_rate / MHZ;
+
+	max_vco_f = priv->params->max_vco;
+	min_vco_f = priv->params->min_vco;
+	best_m = priv->params->max_m;
+	best_n = priv->params->min_n;
+	best_pl = priv->params->min_pl;
+
+	target_vco_f = target_clk_f + target_clk_f / 50;
+	if (max_vco_f < target_vco_f)
+		max_vco_f = target_vco_f;
+
+	/* min_pl <= high_pl <= max_pl */
+	high_pl = (max_vco_f + target_vco_f - 1) / target_vco_f;
+	high_pl = min(high_pl, priv->params->max_pl);
+	high_pl = max(high_pl, priv->params->min_pl);
+
+	/* min_pl <= low_pl <= max_pl */
+	low_pl = min_vco_f / target_vco_f;
+	low_pl = min(low_pl, priv->params->max_pl);
+	low_pl = max(low_pl, priv->params->min_pl);
+
+	/* Find Indices of high_pl and low_pl */
+	for (pl = 0; pl < ARRAY_SIZE(pl_to_div) - 1; pl++) {
+		if (pl_to_div[pl] >= low_pl) {
+			low_pl = pl;
+			break;
+		}
+	}
+	for (pl = 0; pl < ARRAY_SIZE(pl_to_div) - 1; pl++) {
+		if (pl_to_div[pl] >= high_pl) {
+			high_pl = pl;
+			break;
+		}
+	}
+
+	/* Select lowest possible VCO */
+	for (pl = low_pl; pl <= high_pl; pl++) {
+		target_vco_f = target_clk_f * pl_to_div[pl];
+		for (m = priv->params->min_m; m <= priv->params->max_m; m++) {
+			u_f = ref_clk_f / m;
+
+			if (u_f < priv->params->min_u)
+				break;
+			if (u_f > priv->params->max_u)
+				continue;
+
+			n = (target_vco_f * m) / ref_clk_f;
+			n2 = ((target_vco_f * m) + (ref_clk_f - 1)) / ref_clk_f;
+
+			if (n > priv->params->max_n)
+				break;
+
+			for (; n <= n2; n++) {
+				if (n < priv->params->min_n)
+					continue;
+				if (n > priv->params->max_n)
+					break;
+
+				vco_f = ref_clk_f * n / m;
+
+				if (vco_f >= min_vco_f && vco_f <= max_vco_f) {
+					lwv = (vco_f + (pl_to_div[pl] / 2))
+						/ pl_to_div[pl];
+					delta = abs(lwv - target_clk_f);
+
+					if (delta < best_delta) {
+						best_delta = delta;
+						best_m = m;
+						best_n = n;
+						best_pl = pl;
+
+						if (best_delta == 0)
+							goto found_match;
+					}
+				}
+			}
+		}
+	}
+
+found_match:
+	WARN_ON(best_delta == ~0);
+
+	if (best_delta != 0)
+		nv_debug(priv, "no best match for target @ %dMHz on gpc_pll",
+			 target_clk_f);
+
+	priv->m = best_m;
+	priv->n = best_n;
+	priv->pl = best_pl;
+
+	target_freq = gk20a_pllg_calc_rate(priv) / MHZ;
+
+	nv_debug(priv, "actual target freq %d MHz, M %d, N %d, PL %d(div%d)\n",
+		 target_freq, priv->m, priv->n, priv->pl, pl_to_div[priv->pl]);
+
+	return 0;
+}
+
+static int
+gk20a_pllg_slide(struct gk20a_clock_priv *priv, u32 n)
+{
+	u32 val;
+	int ramp_timeout;
+
+	/* get old coefficients */
+	val = nv_rd32(priv, GPCPLL_COEFF);
+	/* do nothing if NDIV is the same */
+	if (n == ((val >> GPCPLL_COEFF_N_SHIFT) & 0xff))
+		return 0;
+
+	/* setup */
+	nv_mask(priv, GPCPLL_CFG2, 0xff << GPCPLL_CFG2_PLL_STEPA_SHIFT,
+		0x2b << GPCPLL_CFG2_PLL_STEPA_SHIFT);
+	nv_mask(priv, GPCPLL_CFG3, 0xff << GPCPLL_CFG3_PLL_STEPB_SHIFT,
+		0xb << GPCPLL_CFG3_PLL_STEPB_SHIFT);
+
+	/* pll slowdown mode */
+	nv_mask(priv, GPCPLL_NDIV_SLOWDOWN,
+		BIT(GPCPLL_NDIV_SLOWDOWN_SLOWDOWN_USING_PLL_SHIFT),
+		BIT(GPCPLL_NDIV_SLOWDOWN_SLOWDOWN_USING_PLL_SHIFT));
+
+	/* new ndiv ready for ramp */
+	val = nv_rd32(priv, GPCPLL_COEFF);
+	val &= ~(0xff << GPCPLL_COEFF_N_SHIFT);
+	val |= (n & 0xff) << GPCPLL_COEFF_N_SHIFT;
+	udelay(1);
+	nv_wr32(priv, GPCPLL_COEFF, val);
+
+	/* dynamic ramp to new ndiv */
+	val = nv_rd32(priv, GPCPLL_NDIV_SLOWDOWN);
+	val |= 0x1 << GPCPLL_NDIV_SLOWDOWN_EN_DYNRAMP_SHIFT;
+	udelay(1);
+	nv_wr32(priv, GPCPLL_NDIV_SLOWDOWN, val);
+
+	for (ramp_timeout = 500; ramp_timeout > 0; ramp_timeout--) {
+		udelay(1);
+		val = nv_rd32(priv, GPC_BCAST_NDIV_SLOWDOWN_DEBUG);
+		if (val & GPC_BCAST_NDIV_SLOWDOWN_DEBUG_PLL_DYNRAMP_DONE_SYNCED_MASK)
+			break;
+	}
+
+	/* exit slowdown mode */
+	nv_mask(priv, GPCPLL_NDIV_SLOWDOWN,
+		BIT(GPCPLL_NDIV_SLOWDOWN_SLOWDOWN_USING_PLL_SHIFT) |
+		BIT(GPCPLL_NDIV_SLOWDOWN_EN_DYNRAMP_SHIFT), 0);
+	nv_rd32(priv, GPCPLL_NDIV_SLOWDOWN);
+
+	if (ramp_timeout <= 0) {
+		nv_error(priv, "gpcpll dynamic ramp timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+static void
+_gk20a_pllg_disable(struct gk20a_clock_priv *priv)
+{
+	/* disable PLL */
+	nv_mask(priv, GPCPLL_CFG, GPCPLL_CFG_ENABLE, 0);
+	nv_rd32(priv, GPCPLL_CFG);
+}
+
+
+static int
+_gk20a_pllg_program_mnp(struct gk20a_clock_priv *priv, bool allow_slide)
+{
+	u32 val, cfg;
+	u32 m_old, pl_old, n_lo;
+
+	/* get old coefficients */
+	val = nv_rd32(priv, GPCPLL_COEFF);
+	m_old = (val >> GPCPLL_COEFF_M_SHIFT) & 0xff;
+	pl_old = (val >> GPCPLL_COEFF_P_SHIFT) & 0xff;
+
+	/* do NDIV slide if there is no change in M and PL */
+	cfg = nv_rd32(priv, GPCPLL_CFG);
+	if (allow_slide && priv->m == m_old && priv->pl == pl_old &&
+	    (cfg & GPCPLL_CFG_ENABLE)) {
+		return gk20a_pllg_slide(priv, priv->n);
+	}
+
+	/* slide down to NDIV_LO */
+	n_lo = DIV_ROUND_UP(m_old * priv->params->min_vco,
+			    priv->parent_rate / MHZ);
+	if (allow_slide && (cfg & GPCPLL_CFG_ENABLE)) {
+		int ret = gk20a_pllg_slide(priv, n_lo);
+		if (ret)
+			return ret;
+	}
+
+	/* split FO-to-bypass jump in halfs by setting out divider 1:2 */
+	nv_mask(priv, GPC2CLK_OUT, GPC2CLK_OUT_VCODIV_MASK,
+		0x2 << GPC2CLK_OUT_VCODIV_SHIFT);
+
+	/* put PLL in bypass before programming it */
+	val = nv_rd32(priv, SEL_VCO);
+	val &= ~(0x1 << SEL_VCO_GPC2CLK_OUT_SHIFT);
+	udelay(2);
+	nv_wr32(priv, SEL_VCO, val);
+
+	/* get out from IDDQ */
+	val = nv_rd32(priv, GPCPLL_CFG);
+	if (val & GPCPLL_CFG_IDDQ) {
+		val &= ~GPCPLL_CFG_IDDQ;
+		nv_wr32(priv, GPCPLL_CFG, val);
+		nv_rd32(priv, GPCPLL_CFG);
+		udelay(2);
+	}
+
+	_gk20a_pllg_disable(priv);
+
+	nv_debug(priv, "%s: m=%d n=%d pl=%d\n", __func__, priv->m, priv->n,
+		 priv->pl);
+
+	n_lo = DIV_ROUND_UP(priv->m * priv->params->min_vco,
+			    priv->parent_rate / MHZ);
+	val = priv->m << GPCPLL_COEFF_M_SHIFT;
+	val |= (allow_slide ? n_lo : priv->n) << GPCPLL_COEFF_N_SHIFT;
+	val |= priv->pl << GPCPLL_COEFF_P_SHIFT;
+	nv_wr32(priv, GPCPLL_COEFF, val);
+
+	/* enable PLL */
+	nv_mask(priv, GPCPLL_CFG, GPCPLL_CFG_ENABLE, GPCPLL_CFG_ENABLE);
+
+	val = nv_rd32(priv, GPCPLL_CFG);
+	if (val & GPCPLL_CFG_LOCK_DET_OFF) {
+		val &= ~GPCPLL_CFG_LOCK_DET_OFF;
+		nv_wr32(priv, GPCPLL_CFG, val);
+	}
+
+	if (!nouveau_timer_wait_eq(priv, 300000, GPCPLL_CFG, GPCPLL_CFG_LOCK,
+				   GPCPLL_CFG_LOCK)) {
+		nv_error(priv, "%s: timeout waiting for pllg lock\n", __func__);
+		return -ETIMEDOUT;
+	}
+
+	/* switch to VCO mode */
+	nv_mask(priv, SEL_VCO, 0, BIT(SEL_VCO_GPC2CLK_OUT_SHIFT));
+
+	/* restore out divider 1:1 */
+	val = nv_rd32(priv, GPC2CLK_OUT);
+	val &= ~GPC2CLK_OUT_VCODIV_MASK;
+	udelay(2);
+	nv_wr32(priv, GPC2CLK_OUT, val);
+
+	/* slide up to new NDIV */
+	return allow_slide ? gk20a_pllg_slide(priv, priv->n) : 0;
+}
+
+static int
+gk20a_pllg_program_mnp(struct gk20a_clock_priv *priv)
+{
+	int err;
+
+	err = _gk20a_pllg_program_mnp(priv, true);
+	if (err)
+		err = _gk20a_pllg_program_mnp(priv, false);
+
+	return err;
+}
+
+static void
+gk20a_pllg_init(struct gk20a_clock_priv *priv)
+{
+	nv_mask(priv, GPC2CLK_OUT, GPC2CLK_OUT_INIT_MASK, GPC2CLK_OUT_INIT_VAL);
+
+	priv->init = true;
+}
+
+static int
+gk20a_pllg_enable(struct gk20a_clock_priv *priv)
+{
+	int err;
+
+	if (!priv->init)
+		gk20a_pllg_init(priv);
+
+	err = gk20a_pllg_program_mnp(priv);
+	if (!err)
+		priv->enabled = true;
+
+	return err;
+}
+
+static int
+gk20a_clock_init(struct nouveau_object *object)
+{
+	struct gk20a_clock_priv *priv = (void *)object;
+	int ret;
+
+	ret = gk20a_pllg_calc_mnp(priv, 72 * MHZ);
+	if (ret) {
+		nv_error(priv, "cannot compute clock parameters\n");
+		return ret;
+	}
+
+	ret = gk20a_pllg_enable(priv);
+	if (ret) {
+		nv_error(priv, "cannot initialize PLLG\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+static int
+gk20a_clock_ctor(struct nouveau_object *parent,  struct nouveau_object *engine,
+		 struct nouveau_oclass *oclass, void *data, u32 size,
+		 struct nouveau_object **pobject)
+{
+	struct gk20a_clock_priv *priv;
+	struct nouveau_platform_gpu *plat = to_platform_gpu(nv_device(parent));
+	int ret;
+
+	ret = nouveau_subdev_create(parent, engine, oclass, 0, "CLK", "clock",
+				    &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	priv->params = &gk20a_pllg_params;
+	priv->parent_rate = clk_get_rate(plat->clk);
+
+	nv_info(priv, "parent clk rate: %ld\n", priv->parent_rate);
+
+	return 0;
+}
+
+struct nouveau_oclass
+gk20a_clock_oclass = {
+	.handle = NV_SUBDEV(CLOCK, 0xea),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = gk20a_clock_ctor,
+		.dtor = _nouveau_subdev_dtor,
+		.init = gk20a_clock_init,
+		.fini = _nouveau_subdev_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/gk20a.c b/drivers/gpu/drm/nouveau/core/subdev/fb/gk20a.c
new file mode 100644
index 00000000000..a16024a7477
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/subdev/fb/gk20a.c
@@ -0,0 +1,56 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "nvc0.h"
+
+struct gk20a_fb_priv {
+	struct nouveau_fb base;
+};
+
+static int
+gk20a_fb_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	     struct nouveau_oclass *oclass, void *data, u32 size,
+	     struct nouveau_object **pobject)
+{
+	struct gk20a_fb_priv *priv;
+	int ret;
+
+	ret = nouveau_fb_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+struct nouveau_oclass *
+gk20a_fb_oclass = &(struct nouveau_fb_impl) {
+	.base.handle = NV_SUBDEV(FB, 0xea),
+	.base.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = gk20a_fb_ctor,
+		.dtor = _nouveau_fb_dtor,
+		.init = _nouveau_fb_init,
+		.fini = _nouveau_fb_fini,
+	},
+	.memtype = nvc0_fb_memtype_valid,
+	.ram = &gk20a_ram_oclass,
+}.base;
diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/priv.h b/drivers/gpu/drm/nouveau/core/subdev/fb/priv.h
index da74c889aed..82273f832e4 100644
--- a/drivers/gpu/drm/nouveau/core/subdev/fb/priv.h
+++ b/drivers/gpu/drm/nouveau/core/subdev/fb/priv.h
@@ -32,6 +32,7 @@ extern struct nouveau_oclass nva3_ram_oclass;
 extern struct nouveau_oclass nvaa_ram_oclass;
 extern struct nouveau_oclass nvc0_ram_oclass;
 extern struct nouveau_oclass nve0_ram_oclass;
+extern struct nouveau_oclass gk20a_ram_oclass;
 extern struct nouveau_oclass gm107_ram_oclass;
 
 int nouveau_sddr3_calc(struct nouveau_ram *ram);
diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c b/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c
new file mode 100644
index 00000000000..1b8b91dc5ba
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c
@@ -0,0 +1,150 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "priv.h"
+#include "ramgk20a.h"
+
+#include <subdev/fb.h>
+
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/dma-mapping.h>
+
+static void
+gk20a_ram_put(struct nouveau_fb *pfb, struct nouveau_mem **pmem)
+{
+	struct device *dev = nv_device_base(nv_device(pfb));
+	struct gk20a_mem *mem = to_gk20a_mem(*pmem);
+
+	*pmem = NULL;
+	if (unlikely(mem == NULL))
+		return;
+
+	if (likely(mem->cpuaddr))
+		dma_free_coherent(dev, mem->base.size << PAGE_SHIFT,
+				  mem->cpuaddr, mem->handle);
+
+	kfree(mem->base.pages);
+	kfree(mem);
+}
+
+static int
+gk20a_ram_get(struct nouveau_fb *pfb, u64 size, u32 align, u32 ncmin,
+	     u32 memtype, struct nouveau_mem **pmem)
+{
+	struct device *dev = nv_device_base(nv_device(pfb));
+	struct gk20a_mem *mem;
+	u32 type = memtype & 0xff;
+	u32 npages, order;
+	int i;
+
+	nv_debug(pfb, "%s: size: %llx align: %x, ncmin: %x\n", __func__, size,
+		 align, ncmin);
+
+	npages = size >> PAGE_SHIFT;
+	if (npages == 0)
+		npages = 1;
+
+	if (align == 0)
+		align = PAGE_SIZE;
+	align >>= PAGE_SHIFT;
+
+	/* round alignment to the next power of 2, if needed */
+	order = fls(align);
+	if ((align & (align - 1)) == 0)
+		order--;
+	align = BIT(order);
+
+	/* ensure returned address is correctly aligned */
+	npages = max(align, npages);
+
+	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
+	if (!mem)
+		return -ENOMEM;
+
+	mem->base.size = npages;
+	mem->base.memtype = type;
+
+	mem->base.pages = kzalloc(sizeof(dma_addr_t) * npages, GFP_KERNEL);
+	if (!mem->base.pages) {
+		kfree(mem);
+		return -ENOMEM;
+	}
+
+	*pmem = &mem->base;
+
+	mem->cpuaddr = dma_alloc_coherent(dev, npages << PAGE_SHIFT,
+					  &mem->handle, GFP_KERNEL);
+	if (!mem->cpuaddr) {
+		nv_error(pfb, "%s: cannot allocate memory!\n", __func__);
+		gk20a_ram_put(pfb, pmem);
+		return -ENOMEM;
+	}
+
+	align <<= PAGE_SHIFT;
+
+	/* alignment check */
+	if (unlikely(mem->handle & (align - 1)))
+		nv_warn(pfb, "memory not aligned as requested: %pad (0x%x)\n",
+			&mem->handle, align);
+
+	nv_debug(pfb, "alloc size: 0x%x, align: 0x%x, paddr: %pad, vaddr: %p\n",
+		 npages << PAGE_SHIFT, align, &mem->handle, mem->cpuaddr);
+
+	for (i = 0; i < npages; i++)
+		mem->base.pages[i] = mem->handle + (PAGE_SIZE * i);
+
+	mem->base.offset = (u64)mem->base.pages[0];
+
+	return 0;
+}
+
+static int
+gk20a_ram_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	      struct nouveau_oclass *oclass, void *data, u32 datasize,
+	      struct nouveau_object **pobject)
+{
+	struct nouveau_ram *ram;
+	int ret;
+
+	ret = nouveau_ram_create(parent, engine, oclass, &ram);
+	*pobject = nv_object(ram);
+	if (ret)
+		return ret;
+	ram->type = NV_MEM_TYPE_STOLEN;
+	ram->size = get_num_physpages() << PAGE_SHIFT;
+
+	ram->get = gk20a_ram_get;
+	ram->put = gk20a_ram_put;
+
+	return 0;
+}
+
+struct nouveau_oclass
+gk20a_ram_oclass = {
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = gk20a_ram_ctor,
+		.dtor = _nouveau_ram_dtor,
+		.init = _nouveau_ram_init,
+		.fini = _nouveau_ram_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.h b/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.h
new file mode 100644
index 00000000000..105ad258169
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.h
@@ -0,0 +1,38 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef __NVKM_FB_RAMGK20A_H__
+#define __NVKM_FB_RAMGK20A_H__
+
+#include <linux/types.h>
+#include <subdev/fb.h>
+
+struct gk20a_mem {
+	struct nouveau_mem base;
+	void *cpuaddr;
+	dma_addr_t handle;
+};
+
+#define to_gk20a_mem(m) container_of(m, struct gk20a_mem, base)
+
+#endif
diff --git a/drivers/gpu/drm/nouveau/core/subdev/ibus/gk20a.c b/drivers/gpu/drm/nouveau/core/subdev/ibus/gk20a.c
new file mode 100644
index 00000000000..245f0ebaa6a
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/subdev/ibus/gk20a.c
@@ -0,0 +1,103 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <subdev/ibus.h>
+#include <subdev/timer.h>
+
+struct gk20a_ibus_priv {
+	struct nouveau_ibus base;
+};
+
+static void
+gk20a_ibus_init_priv_ring(struct gk20a_ibus_priv *priv)
+{
+	nv_mask(priv, 0x137250, 0x3f, 0);
+
+	nv_mask(priv, 0x000200, 0x20, 0);
+	usleep_range(20, 30);
+	nv_mask(priv, 0x000200, 0x20, 0x20);
+
+	nv_wr32(priv, 0x12004c, 0x4);
+	nv_wr32(priv, 0x122204, 0x2);
+	nv_rd32(priv, 0x122204);
+}
+
+static void
+gk20a_ibus_intr(struct nouveau_subdev *subdev)
+{
+	struct gk20a_ibus_priv *priv = (void *)subdev;
+	u32 status0 = nv_rd32(priv, 0x120058);
+
+	if (status0 & 0x7) {
+		nv_debug(priv, "resetting priv ring\n");
+		gk20a_ibus_init_priv_ring(priv);
+	}
+
+	/* Acknowledge interrupt */
+	nv_mask(priv, 0x12004c, 0x2, 0x2);
+
+	if (!nv_wait(subdev, 0x12004c, 0x3f, 0x00))
+		nv_warn(priv, "timeout waiting for ringmaster ack\n");
+}
+
+static int
+gk20a_ibus_init(struct nouveau_object *object)
+{
+	struct gk20a_ibus_priv *priv = (void *)object;
+	int ret;
+
+	ret = _nouveau_ibus_init(object);
+	if (ret)
+		return ret;
+
+	gk20a_ibus_init_priv_ring(priv);
+
+	return 0;
+}
+
+static int
+gk20a_ibus_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+	       struct nouveau_oclass *oclass, void *data, u32 size,
+	       struct nouveau_object **pobject)
+{
+	struct gk20a_ibus_priv *priv;
+	int ret;
+
+	ret = nouveau_ibus_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	nv_subdev(priv)->intr = gk20a_ibus_intr;
+	return 0;
+}
+
+struct nouveau_oclass
+gk20a_ibus_oclass = {
+	.handle = NV_SUBDEV(IBUS, 0xea),
+	.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = gk20a_ibus_ctor,
+		.dtor = _nouveau_ibus_dtor,
+		.init = gk20a_ibus_init,
+		.fini = _nouveau_ibus_fini,
+	},
+};
diff --git a/drivers/gpu/drm/nouveau/core/subdev/instmem/gk20a.c b/drivers/gpu/drm/nouveau/core/subdev/instmem/gk20a.c
new file mode 100644
index 00000000000..0bcbba09eb0
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/subdev/instmem/gk20a.c
@@ -0,0 +1,138 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <subdev/fb.h>
+
+#include "priv.h"
+#include "core/subdev/fb/ramgk20a.h"
+
+struct gk20a_instmem_priv {
+	struct nouveau_instmem base;
+};
+
+struct gk20a_instobj_priv {
+	struct nouveau_instobj base;
+	struct nouveau_mem *mem;
+};
+
+static u32
+gk20a_instobj_rd32(struct nouveau_object *object, u64 offset)
+{
+	struct gk20a_instobj_priv *node = (void *)object;
+	struct gk20a_mem *mem = to_gk20a_mem(node->mem);
+
+	return ((u32 *)mem->cpuaddr)[offset / 4];
+}
+
+static void
+gk20a_instobj_wr32(struct nouveau_object *object, u64 offset, u32 data)
+{
+	struct gk20a_instobj_priv *node = (void *)object;
+	struct gk20a_mem *mem = to_gk20a_mem(node->mem);
+
+	((u32 *)mem->cpuaddr)[offset / 4] = data;
+}
+
+static void
+gk20a_instobj_dtor(struct nouveau_object *object)
+{
+	struct gk20a_instobj_priv *node = (void *)object;
+	struct nouveau_fb *pfb = nouveau_fb(object);
+
+	pfb->ram->put(pfb, &node->mem);
+	nouveau_instobj_destroy(&node->base);
+}
+
+static int
+gk20a_instobj_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		  struct nouveau_oclass *oclass, void *data, u32 size,
+		  struct nouveau_object **pobject)
+{
+	struct nouveau_fb *pfb = nouveau_fb(parent);
+	struct nouveau_instobj_args *args = data;
+	struct gk20a_instobj_priv *node;
+	int ret;
+
+	args->size  = max((args->size  + 4095) & ~4095, (u32)4096);
+	args->align = max((args->align + 4095) & ~4095, (u32)4096);
+
+	ret = nouveau_instobj_create(parent, engine, oclass, &node);
+	*pobject = nv_object(node);
+	if (ret)
+		return ret;
+
+	ret = pfb->ram->get(pfb, args->size, args->align, 0, 0x800, &node->mem);
+	if (ret)
+		return ret;
+
+	node->base.addr = node->mem->offset;
+	node->base.size = node->mem->size << 12;
+	node->mem->page_shift = 12;
+	return 0;
+}
+
+static struct nouveau_instobj_impl
+gk20a_instobj_oclass = {
+	.base.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = gk20a_instobj_ctor,
+		.dtor = gk20a_instobj_dtor,
+		.init = _nouveau_instobj_init,
+		.fini = _nouveau_instobj_fini,
+		.rd32 = gk20a_instobj_rd32,
+		.wr32 = gk20a_instobj_wr32,
+	},
+};
+
+static int
+gk20a_instmem_fini(struct nouveau_object *object, bool suspend)
+{
+	struct gk20a_instmem_priv *priv = (void *)object;
+	return nouveau_instmem_fini(&priv->base, suspend);
+}
+
+static int
+gk20a_instmem_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
+		  struct nouveau_oclass *oclass, void *data, u32 size,
+		  struct nouveau_object **pobject)
+{
+	struct gk20a_instmem_priv *priv;
+	int ret;
+
+	ret = nouveau_instmem_create(parent, engine, oclass, &priv);
+	*pobject = nv_object(priv);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+struct nouveau_oclass *
+gk20a_instmem_oclass = &(struct nouveau_instmem_impl) {
+	.base.handle = NV_SUBDEV(INSTMEM, 0xea),
+	.base.ofuncs = &(struct nouveau_ofuncs) {
+		.ctor = gk20a_instmem_ctor,
+		.dtor = _nouveau_instmem_dtor,
+		.init = _nouveau_instmem_init,
+		.fini = gk20a_instmem_fini,
+	},
+	.instobj = &gk20a_instobj_oclass.base,
+}.base;
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index b6dc85c614b..f00ae18003f 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -33,6 +33,7 @@
 #include <subdev/fb.h>
 #include <subdev/vm.h>
 #include <subdev/bar.h>
+#include <subdev/timer.h>
 
 #include "nouveau_drm.h"
 #include "nouveau_dma.h"
@@ -407,6 +408,8 @@ nouveau_bo_validate(struct nouveau_bo *nvbo, bool interruptible,
 {
 	int ret;
 
+	nouveau_bo_sync_for_device(nvbo);
+
 	ret = ttm_bo_validate(&nvbo->bo, &nvbo->placement,
 			      interruptible, no_wait_gpu);
 	if (ret)
@@ -435,8 +438,10 @@ nouveau_bo_wr16(struct nouveau_bo *nvbo, unsigned index, u16 val)
 	mem = &mem[index];
 	if (is_iomem)
 		iowrite16_native(val, (void __force __iomem *)mem);
-	else
+	else {
 		*mem = val;
+		nv_cpu_cache_flush_area(mem, 2);
+	}
 }
 
 u32
@@ -459,8 +464,10 @@ nouveau_bo_wr32(struct nouveau_bo *nvbo, unsigned index, u32 val)
 	mem = &mem[index];
 	if (is_iomem)
 		iowrite32_native(val, (void __force __iomem *)mem);
-	else
+	else {
 		*mem = val;
+		nv_cpu_cache_flush_area(mem, 4);
+	}
 }
 
 static struct ttm_tt *
@@ -487,6 +494,38 @@ nouveau_bo_invalidate_caches(struct ttm_bo_device *bdev, uint32_t flags)
 	return 0;
 }
 
+void
+nouveau_bo_sync_for_cpu(struct nouveau_bo *nvbo)
+{
+	struct nouveau_device *device;
+	struct ttm_tt *ttm = nvbo->bo.ttm;
+
+	device = nouveau_dev(nouveau_bdev(ttm->bdev)->dev);
+
+	if (nvbo->bo.ttm && nvbo->bo.ttm->caching_state == tt_cached)
+		ttm_dma_tt_cache_sync_for_cpu((struct ttm_dma_tt *)nvbo->bo.ttm,
+					      nv_device_base(device));
+}
+
+void
+nouveau_bo_sync_for_device(struct nouveau_bo *nvbo)
+{
+	struct ttm_tt *ttm = nvbo->bo.ttm;
+
+	if (ttm && ttm->caching_state == tt_cached) {
+		struct nouveau_device *device;
+
+		device = nouveau_dev(nouveau_bdev(ttm->bdev)->dev);
+
+		ttm_dma_tt_cache_sync_for_device((struct ttm_dma_tt *)ttm,
+						 nv_device_base(device));
+
+		nv_wr32(device, 0x70004, 0x00000001);
+		if (!nv_wait(device, 0x070004, 0x00000001, 0x00000000))
+			nv_warn(device, "L2 invalidate timeout\n");
+	}
+}
+
 static int
 nouveau_bo_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
 			 struct ttm_mem_type_manager *man)
@@ -511,7 +550,11 @@ nouveau_bo_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
 			     TTM_MEMTYPE_FLAG_MAPPABLE;
 		man->available_caching = TTM_PL_FLAG_UNCACHED |
 					 TTM_PL_FLAG_WC;
+#if defined(__arm__)
+		man->default_caching = TTM_PL_FLAG_UNCACHED;
+#else
 		man->default_caching = TTM_PL_FLAG_WC;
+#endif
 		break;
 	case TTM_PL_TT:
 		if (nv_device(drm->device)->card_type >= NV_50)
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.h b/drivers/gpu/drm/nouveau/nouveau_bo.h
index ff17c1f432f..50f721eac65 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.h
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.h
@@ -89,6 +89,9 @@ int  nouveau_bo_vma_add(struct nouveau_bo *, struct nouveau_vm *,
 			struct nouveau_vma *);
 void nouveau_bo_vma_del(struct nouveau_bo *, struct nouveau_vma *);
 
+void nouveau_bo_sync_for_cpu(struct nouveau_bo *);
+void nouveau_bo_sync_for_device(struct nouveau_bo *);
+
 /* TODO: submit equivalent to TTM generic API upstream? */
 static inline void __iomem *
 nvbo_kmap_obj_iovirtual(struct nouveau_bo *nvbo)
diff --git a/drivers/gpu/drm/nouveau/nouveau_debug.c b/drivers/gpu/drm/nouveau/nouveau_debug.c
new file mode 100644
index 00000000000..92b5f5a2bb7
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_debug.c
@@ -0,0 +1,185 @@
+
+#include <linux/highmem.h>
+#include <linux/printk.h>
+#include <core/object.h>
+#include <core/gpuobj.h>
+#include <engine/fifo.h>
+#include <subdev/vm.h>
+#include <nouveau_drm.h>
+#include <nouveau_chan.h>
+#include <nouveau_bo.h>
+
+static u32
+pfn_for_vaddr(u32 pgtpfn, u64 vaddr)
+{
+	u32 pde = vaddr >> 26;
+	u32 pte = (vaddr & 0x3ffffff) >> 12;
+	u32 ptepfn;
+	struct page *p = pfn_to_page(pgtpfn);
+	u32 *pgt = kmap(p);
+	u32 *pet;
+	u32 ret;
+	ptepfn = pgt[pde * 2 + 1] >> 4;
+	kunmap(p);
+
+	p = pfn_to_page(ptepfn);
+	pet = kmap(p);
+	ret = pet[pte * 2] >> 4;
+	kunmap(p);
+
+	return ret;
+}
+
+void
+dump_pte(u32 ptepfn, u32 base)
+{
+	struct page *p = pfn_to_page(ptepfn);
+	u32 *pte = kmap(p);
+	int k;
+
+	for (k = 0; k < 512; k++) {
+		u32 mappfn = pte[0] >> 4;
+		bool ptevol = pte[1] & 0x1;
+		u32 pteflags = pte[1] & (0x7 << 1);
+		if (pte[0] & 1) {
+			printk("    %08x -> %08x VOL=%d FLAGS=%x\n", base * 0x4000000 + k * 0x1000, mappfn << 12, ptevol, pteflags);
+		}
+		pte += 2;
+	}
+
+	kunmap(p);
+}
+
+void
+dump_pgt(u32 pgtpfn)
+{
+	struct page *p = pfn_to_page(pgtpfn);
+	u32 *pgt = kmap(p);
+	int j;
+
+	for (j = 0; j < 128; j++) {
+		int flags = pgt[1] & 0xf;
+		int ptepfn = pgt[1] >> 4;
+		/* PTE valid */
+		if ((flags & 0x3) != 0) {
+			dump_pte(ptepfn, j);
+		}
+		pgt += 2;
+	}
+
+	kunmap(p);
+}
+
+void
+dump_ramuser(struct nouveau_channel *ch)
+{
+	struct nouveau_fifo_chan *chan = (struct nouveau_fifo_chan *) ch->object;
+	/* Read the values from the snoop area to get up-to-date information */
+	u32 *ramuser = chan->user;
+	u32 get, get_hi, put, put_hi;
+	/* Read the low fields first as specified in the doc */
+	get = ramuser[17];
+	rmb();
+	get_hi = ramuser[24];
+	rmb();
+	put = ramuser[16];
+	rmb();
+	put_hi = ramuser[19];
+	rmb();
+	printk("    PUT: %08x%08x\n", put_hi, put);
+	printk("    GET: %08x%08x\n", get_hi, get);
+	printk("    GP_PUT: %08x\n", ramuser[35]);
+	printk("    GP_GET: %08x\n", ramuser[34]);
+}
+
+void dump_pb(u32 pbpfn, u32 start, u32 len)
+{
+	struct page *p = pfn_to_page(pbpfn);
+	u32 *pb = kmap(p) + start;
+	u32 i;
+
+	for (i = start; i < start + len; i++)
+		printk("%08x:  %08x\n", (pbpfn << PAGE_SHIFT) + i * 4, pb[i]);
+
+	kunmap(p);
+}
+
+void dump_gpentries(struct nouveau_channel *ch, u32 gppfn, u32 pgtpfn)
+{
+	struct nouveau_fifo_chan *chan = (struct nouveau_fifo_chan *) ch->object;
+	struct page *p = pfn_to_page(gppfn);
+	u32 *gp = kmap(p);
+	u32 *ramuser = chan->user;
+	u32 gp0, gp1;
+	u64 addr;
+	u32 len;
+	u32 gp_get, gp_put, i;
+
+	gp_get = ramuser[34];
+	gp_put = ramuser[35];
+
+	for (i = gp_get; i <= gp_put; i++) {
+		gp0 = gp[i * 2];
+		gp1 = gp[i * 2 + 1];
+		addr = ((u64)gp0) | (((u64)(gp1 & 0xff)) << 32);
+		len = ((gp1 & 0x7ffffc00) >> 10);
+		printk("    addr: %llx len: %x\n", addr, len);
+		if (addr != 0)
+			dump_pb(pfn_for_vaddr(pgtpfn, addr), addr & (~PAGE_MASK), len);
+	}
+
+	kunmap(p);
+}
+
+void dump_ramin(struct nouveau_channel *ch, u32 raminpfn)
+{
+	struct page *p = pfn_to_page(raminpfn);
+	u32 *ramin = kmap(p);
+	u32 pgtpfn;
+	u64 gpbase;
+
+	pgtpfn = (ramin[128] & (~0x3)) >> PAGE_SHIFT;
+
+	gpbase = (((u64)(ramin[19] & 0xff)) << 32) | (ramin[18] & 0xfffffff8);
+
+	printk("  GP_BASE: %llx\n", gpbase);
+	printk("   phys: %x\n", pfn_for_vaddr(pgtpfn, gpbase));
+	printk("  GP ENTRIES:\n");
+	dump_gpentries(ch, pfn_for_vaddr(pgtpfn, gpbase), pgtpfn);
+	printk("  USERD: %08x%08x\n", ramin[3], ramin[2]);
+	dump_ramuser(ch);
+	printk("  PGT: %02x%08x %02x%08x\n", ramin[129], ramin[128], ramin[131], ramin[130]);
+	dump_pgt(pgtpfn);
+	kunmap(p);
+}
+
+void dump_channel(struct nouveau_channel *chan)
+{
+	struct nouveau_device *device = nv_device(chan->drm->device);
+	struct nouveau_fifo_chan *fchan = (struct nouveau_fifo_chan *) chan->object;
+
+	u32 raminpfn = nv_rd32(device, 0x800000 + (fchan->chid * 8)) & 0xfffff;
+	printk("Channel 0x%x, RAMIN: %x\n", fchan->chid, raminpfn << PAGE_SHIFT);
+
+	dump_ramin(chan, raminpfn);
+}
+
+/*
+void
+dump_runlist(struct nouveau_object *priv, u32 nchan, struct nouveau_gpuobj *cur)
+{
+	int i;
+
+	for (i = 0; i < nchan; i++) {
+		u32 val = nv_ro32(cur, i * 8);
+		u32 chst1 = nv_rd32(priv, 0x800000 + (val * 8));
+		u32 chst2 = nv_rd32(priv, 0x800004 + (val * 8));
+		u32 raminpfn = chst1 & 0xfffff;
+		if ((chst1 & 0x80000000) && val) {
+			printk("  channel %x %08x %08x\n", val, chst1, chst2);
+			printk("    RAMIN: %08x\n", raminpfn << PAGE_SHIFT);
+			dump_ramin(raminpfn);
+		}
+	}
+}
+*/
\ No newline at end of file
diff --git a/drivers/gpu/drm/nouveau/nouveau_debug.h b/drivers/gpu/drm/nouveau/nouveau_debug.h
new file mode 100644
index 00000000000..b5a3a420926
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_debug.h
@@ -0,0 +1,11 @@
+
+struct nouveau_object;
+struct nouveau_gpuobj;
+struct nouveau_channel;
+
+void dump_pte(u32 ptepfn, u32 base);
+void dump_pgt(u32 pgtpfn);
+void dump_ramuser(struct nouveau_channel *chan);
+void dump_ramin(struct nouveau_channel *chan, u32 raminpfn);
+void dump_channel(struct nouveau_channel *chan);
+/*void dump_runlist(struct nouveau_object *priv, u32 nchan, struct nouveau_gpuobj *cur);*/
\ No newline at end of file
diff --git a/drivers/gpu/drm/nouveau/nouveau_drm.c b/drivers/gpu/drm/nouveau/nouveau_drm.c
index e8ae68a9aaf..e838fb8624c 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_drm.c
@@ -172,6 +172,11 @@ nouveau_accel_init(struct nouveau_drm *drm)
 		return;
 	}
 
+	if (device->chipset == 0xea) {
+		/* gk20a does not have CE0/CE1 */
+		arg0 = NVE0_CHANNEL_IND_ENGINE_GR;
+		arg1 = 1;
+	} else
 	if (device->card_type >= NV_E0) {
 		ret = nouveau_channel_new(drm, &drm->client, NVDRM_DEVICE,
 					  NVDRM_CHAN + 1,
@@ -359,6 +364,9 @@ nouveau_drm_load(struct drm_device *dev, unsigned long flags)
 
 	dev->dev_private = drm;
 	drm->dev = dev;
+	if (dev->platformdev)
+		platform_set_drvdata(dev->platformdev, dev);
+
 	nouveau_client(drm)->debug = nouveau_dbgopt(nouveau_debug, "DRM");
 
 	INIT_LIST_HEAD(&drm->clients);
@@ -1004,23 +1012,19 @@ nouveau_drm_pci_driver = {
 	.driver.pm = &nouveau_pm_ops,
 };
 
-int nouveau_drm_platform_probe(struct platform_device *pdev)
+int nouveau_drm_platform_device_create(struct platform_device *pdev, int length,
+				       void **pobject)
 {
-	struct nouveau_device *device;
-	int ret;
-
-	ret = nouveau_device_create(pdev, NOUVEAU_BUS_PLATFORM,
-				    nouveau_platform_name(pdev),
-				    dev_name(&pdev->dev), nouveau_config,
-				    nouveau_debug, &device);
-
-	ret = drm_platform_init(&driver, pdev);
-	if (ret) {
-		nouveau_object_ref(NULL, (struct nouveau_object **)&device);
-		return ret;
-	}
+	return nouveau_device_create_(pdev, NOUVEAU_BUS_PLATFORM,
+				      nouveau_platform_name(pdev),
+				      dev_name(&pdev->dev),
+				      nouveau_config, nouveau_debug, length,
+				      pobject);
+}
 
-	return ret;
+int nouveau_drm_platform_device_init(struct platform_device *pdev)
+{
+	return drm_platform_init(&driver, pdev);
 }
 
 static int __init
@@ -1036,6 +1040,8 @@ nouveau_drm_init(void)
 	if (!nouveau_modeset)
 		return 0;
 
+	nouveau_platform_driver_init();
+
 	nouveau_register_dsm_handler();
 	return drm_pci_init(&driver, &nouveau_drm_pci_driver);
 }
@@ -1048,6 +1054,8 @@ nouveau_drm_exit(void)
 
 	drm_pci_exit(&driver, &nouveau_drm_pci_driver);
 	nouveau_unregister_dsm_handler();
+
+	nouveau_platform_driver_exit();
 }
 
 module_init(nouveau_drm_init);
diff --git a/drivers/gpu/drm/nouveau/nouveau_drm.h b/drivers/gpu/drm/nouveau/nouveau_drm.h
index 7efbafaf7c1..178f8cfb148 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drm.h
+++ b/drivers/gpu/drm/nouveau/nouveau_drm.h
@@ -157,6 +157,27 @@ nouveau_dev(struct drm_device *dev)
 int nouveau_pmops_suspend(struct device *);
 int nouveau_pmops_resume(struct device *);
 
+int nouveau_drm_platform_device_create(struct platform_device *, int, void **);
+int nouveau_drm_platform_device_init(struct platform_device *);
+
+#if IS_ENABLED(CONFIG_NOUVEAU_PLATFORM_DRIVER)
+
+int nouveau_platform_driver_init(void);
+void nouveau_platform_driver_exit(void);
+
+#else
+
+static inline int nouveau_platform_driver_init(void)
+{
+	return 0;
+}
+
+static inline void nouveau_platform_driver_exit(void)
+{
+}
+
+#endif
+
 #define NV_FATAL(cli, fmt, args...) nv_fatal((cli), fmt, ##args)
 #define NV_ERROR(cli, fmt, args...) nv_error((cli), fmt, ##args)
 #define NV_WARN(cli, fmt, args...) nv_warn((cli), fmt, ##args)
diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c
index c90c0dc0afe..b7e42fdc963 100644
--- a/drivers/gpu/drm/nouveau/nouveau_gem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_gem.c
@@ -897,7 +897,13 @@ nouveau_gem_ioctl_cpu_prep(struct drm_device *dev, void *data,
 	ret = ttm_bo_wait(&nvbo->bo, true, true, no_wait);
 	spin_unlock(&nvbo->bo.bdev->fence_lock);
 	drm_gem_object_unreference_unlocked(gem);
-	return ret;
+
+	if (ret)
+		return ret;
+
+	nouveau_bo_sync_for_cpu(nvbo);
+
+	return 0;
 }
 
 int
diff --git a/drivers/gpu/drm/nouveau/nouveau_platform.c b/drivers/gpu/drm/nouveau/nouveau_platform.c
new file mode 100644
index 00000000000..fc720376554
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_platform.c
@@ -0,0 +1,193 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <linux/clk.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/of.h>
+#include <linux/reset.h>
+#include <linux/regulator/consumer.h>
+#include <linux/tegra-powergate.h>
+
+#include "engine/device.h"
+#include "nouveau_drm.h"
+#include "nouveau_platform.h"
+
+static int nouveau_platform_power_up(struct nouveau_platform_gpu *gpu)
+{
+	int err;
+
+	err = regulator_enable(gpu->vdd);
+	if (err)
+		goto err_power;
+
+	err = clk_prepare_enable(gpu->clk);
+	if (err)
+		goto err_clk;
+	err = clk_prepare_enable(gpu->clk_pwr);
+	if (err)
+		goto err_clk_pwr;
+	clk_set_rate(gpu->clk_pwr, 204000000);
+	udelay(10);
+
+	reset_control_assert(gpu->rst);
+	udelay(10);
+
+	err = tegra_powergate_remove_clamping(TEGRA_POWERGATE_3D);
+	if (err)
+		goto err_clamp;
+	udelay(10);
+
+	reset_control_deassert(gpu->rst);
+	udelay(10);
+
+	return 0;
+
+err_clamp:
+	clk_disable_unprepare(gpu->clk_pwr);
+err_clk_pwr:
+	clk_disable_unprepare(gpu->clk);
+err_clk:
+	regulator_disable(gpu->vdd);
+err_power:
+	return err;
+}
+
+static int nouveau_platform_power_down(struct nouveau_platform_gpu *gpu)
+{
+	int err;
+
+	reset_control_assert(gpu->rst);
+	udelay(10);
+
+	clk_disable_unprepare(gpu->clk_pwr);
+	clk_disable_unprepare(gpu->clk);
+	udelay(10);
+
+	err = regulator_disable(gpu->vdd);
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static int nouveau_platform_probe(struct platform_device *pdev)
+{
+	struct nouveau_platform_gpu *gpu;
+	struct regulator *vdd;
+	struct reset_control *rst;
+	struct clk *clk, *pwr;
+	int err;
+
+	/*
+	 * get the resources we need before we allocate the device' memory
+	 * in case we need to return -EPROBE_DEFER
+	 */
+	vdd = devm_regulator_get(&pdev->dev, "vdd");
+	if (IS_ERR(vdd))
+		return PTR_ERR(vdd);
+
+	rst = devm_reset_control_get(&pdev->dev, "gpu");
+	if (IS_ERR(rst))
+		return PTR_ERR(rst);
+
+	clk = devm_clk_get(&pdev->dev, "gpu");
+	if (IS_ERR(clk))
+		return PTR_ERR(clk);
+
+	pwr = devm_clk_get(&pdev->dev, "pwr");
+	if (IS_ERR(pwr))
+		return PTR_ERR(pwr);
+
+	err = nouveau_drm_platform_device_create(pdev, sizeof(*gpu),
+						 (void **)&gpu);
+	if (err)
+		return err;
+
+	gpu->vdd = vdd;
+	gpu->rst = rst;
+	gpu->clk = clk;
+	gpu->clk_pwr = pwr;
+
+	err = nouveau_platform_power_up(gpu);
+	if (err)
+		goto err_probe;
+
+	err = nouveau_drm_platform_device_init(pdev);
+	if (err)
+		goto err_probe;
+
+	return 0;
+
+err_probe:
+	nouveau_object_ref(NULL, (struct nouveau_object **)&gpu->device);
+	return err;
+}
+
+static int nouveau_platform_remove(struct platform_device *pdev)
+{
+	struct drm_device *drm_dev = platform_get_drvdata(pdev);
+	struct nouveau_device *device = nouveau_dev(drm_dev);
+	struct nouveau_platform_gpu *gpu = to_platform_gpu(device);
+	int err;
+
+	drm_dev->irq_enabled = false;
+	drm_put_dev(drm_dev);
+
+	err = nouveau_platform_power_down(gpu);
+	if (err)
+		return err;
+
+	nouveau_object_ref(NULL, (struct nouveau_object **)&device);
+	nouveau_object_debug();
+
+	return 0;
+}
+
+#if IS_ENABLED(CONFIG_OF)
+static const struct of_device_id nouveau_platform_match[] = {
+	{ .compatible = "nvidia,gk20a" },
+	{ }
+};
+
+MODULE_DEVICE_TABLE(of, nouveau_platform_match);
+#endif
+
+struct platform_driver nouveau_platform_driver = {
+	.driver = {
+		.name = "nouveau",
+		.of_match_table = of_match_ptr(nouveau_platform_match),
+	},
+	.probe = nouveau_platform_probe,
+	.remove = nouveau_platform_remove,
+};
+
+int __init nouveau_platform_driver_init(void)
+{
+	return platform_driver_register(&nouveau_platform_driver);
+}
+
+void __exit nouveau_platform_driver_exit(void)
+{
+	platform_driver_unregister(&nouveau_platform_driver);
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_platform.h b/drivers/gpu/drm/nouveau/nouveau_platform.h
new file mode 100644
index 00000000000..847b3671e60
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nouveau_platform.h
@@ -0,0 +1,44 @@
+/*
+ * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef __NOUVEAU_PLATFORM_H__
+#define __NOUVEAU_PLATFORM_H__
+
+#include "core/device.h"
+
+struct reset_control;
+struct clk;
+struct regulator;
+
+struct nouveau_platform_gpu {
+	struct nouveau_device device;
+
+	struct reset_control *rst;
+	struct clk *clk;
+	struct clk *clk_pwr;
+
+	struct regulator *vdd;
+};
+#define to_platform_gpu(d) container_of(d, struct nouveau_platform_gpu, device)
+
+
+#endif
diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 1df856f7856..30e5d90cb7b 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -500,7 +500,7 @@ pgprot_t ttm_io_prot(uint32_t caching_flags, pgprot_t tmp)
 			pgprot_val(tmp) |= _PAGE_GUARDED;
 	}
 #endif
-#if defined(__ia64__)
+#if defined(__ia64__) || defined(__arm__)
 	if (caching_flags & TTM_PL_FLAG_WC)
 		tmp = pgprot_writecombine(tmp);
 	else
diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c
index 75f31909004..66c16ad35f7 100644
--- a/drivers/gpu/drm/ttm/ttm_tt.c
+++ b/drivers/gpu/drm/ttm/ttm_tt.c
@@ -38,6 +38,7 @@
 #include <linux/swap.h>
 #include <linux/slab.h>
 #include <linux/export.h>
+#include <linux/dma-mapping.h>
 #include <drm/drm_cache.h>
 #include <drm/drm_mem_util.h>
 #include <drm/ttm/ttm_module.h>
@@ -248,6 +249,30 @@ void ttm_dma_tt_fini(struct ttm_dma_tt *ttm_dma)
 }
 EXPORT_SYMBOL(ttm_dma_tt_fini);
 
+void ttm_dma_tt_cache_sync_for_device(struct ttm_dma_tt *ttm_dma,
+				      struct device *dev)
+{
+	unsigned long i;
+
+	for (i = 0; i < ttm_dma->ttm.num_pages; i++) {
+		dma_sync_single_for_device(dev, ttm_dma->dma_address[i],
+					   PAGE_SIZE, DMA_TO_DEVICE);
+	}
+}
+EXPORT_SYMBOL(ttm_dma_tt_cache_sync_for_device);
+
+void ttm_dma_tt_cache_sync_for_cpu(struct ttm_dma_tt *ttm_dma,
+				   struct device *dev)
+{
+	unsigned long i;
+
+	for (i = 0; i < ttm_dma->ttm.num_pages; i++) {
+		dma_sync_single_for_cpu(dev, ttm_dma->dma_address[i],
+					PAGE_SIZE, DMA_FROM_DEVICE);
+	}
+}
+EXPORT_SYMBOL(ttm_dma_tt_cache_sync_for_cpu);
+
 void ttm_tt_unbind(struct ttm_tt *ttm)
 {
 	int ret;
diff --git a/firmware/nouveau/nvea_fuc409c b/firmware/nouveau/nvea_fuc409c
new file mode 100644
index 00000000000..176bcc184e2
--- /dev/null
+++ b/firmware/nouveau/nvea_fuc409c
diff --git a/firmware/nouveau/nvea_fuc409d b/firmware/nouveau/nvea_fuc409d
new file mode 100644
index 00000000000..e6997c0252c
--- /dev/null
+++ b/firmware/nouveau/nvea_fuc409d
diff --git a/firmware/nouveau/nvea_fuc41ac b/firmware/nouveau/nvea_fuc41ac
new file mode 100644
index 00000000000..143a923b953
--- /dev/null
+++ b/firmware/nouveau/nvea_fuc41ac
diff --git a/firmware/nouveau/nvea_fuc41ad b/firmware/nouveau/nvea_fuc41ad
new file mode 100644
index 00000000000..b14ead08bdf
--- /dev/null
+++ b/firmware/nouveau/nvea_fuc41ad
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index a5183da3ef9..52fb709568f 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -41,6 +41,7 @@
 #include <linux/fs.h>
 #include <linux/spinlock.h>
 #include <linux/reservation.h>
+#include <linux/device.h>
 
 struct ttm_backend_func {
 	/**
@@ -690,6 +691,33 @@ extern int ttm_tt_swapout(struct ttm_tt *ttm,
  */
 extern void ttm_tt_unpopulate(struct ttm_tt *ttm);
 
+/**
+ * ttm_dma_tt_cache_sync_for_device:
+ *
+ * @ttm A struct ttm_tt of the type returned by ttm_dma_tt_init.
+ * @dev A struct device representing the device to which to sync.
+ *
+ * This function will flush the CPU caches on arches where snooping in the
+ * TT is not available. On fully coherent arches this will turn into an (almost)
+ * noop. This makes sure that data written by the CPU is visible to the device.
+ */
+extern void ttm_dma_tt_cache_sync_for_device(struct ttm_dma_tt *ttm_dma,
+					     struct device *dev);
+
+/**
+ * ttm_dma_tt_cache_sync_for_cpu:
+ *
+ * @ttm A struct ttm_tt of the type returned by ttm_dma_tt_init.
+ * @dev A struct device representing the device from which to sync.
+ *
+ * This function will invalidate the CPU caches on arches where snooping in the
+ * TT is not available. On fully coherent arches this will turn into an (almost)
+ * noop. This makes sure that the CPU does not read any stale cached or
+ * prefetched data.
+ */
+extern void ttm_dma_tt_cache_sync_for_cpu(struct ttm_dma_tt *ttm_dma,
+					  struct device *dev);
+
 /*
  * ttm_bo.c
  */