Merge tag 'drm-for-v4.15' of git://people.freedesktop.org/~airlied/linux

Pull drm updates from Dave Airlie:
 "This is the main drm pull request for v4.15.

  Core:
   - Atomic object lifetime fixes
   - Atomic iterator improvements
   - Sparse/smatch fixes
   - Legacy kms ioctls to be interruptible
   - EDID override improvements
   - fb/gem helper cleanups
   - Simple outreachy patches
   - Documentation improvements
   - Fix dma-buf rcu races
   - DRM mode object leasing for improving VR use cases.
   - vgaarb improvements for non-x86 platforms.

  New driver:
   - tve200: Faraday Technology TVE200 block.

     This "TV Encoder" encodes a ITU-T BT.656 stream and can be found in
     the StorLink SL3516 (later Cortina Systems CS3516) as well as the
     Grain Media GM8180.

  New bridges:
   - SiI9234 support

  New panels:
   - S6E63J0X03, OTM8009A, Seiko 43WVF1G, 7" rpi touch panel, Toshiba
     LT089AC19000, Innolux AT043TN24

  i915:
   - Remove Coffeelake from alpha support
   - Cannonlake workarounds
   - Infoframe refactoring for DisplayPort
   - VBT updates
   - DisplayPort vswing/emph/buffer translation refactoring
   - CCS fixes
   - Restore GPU clock boost on missed vblanks
   - Scatter list updates for userptr allocations
   - Gen9+ transition watermarks
   - Display IPC (Isochronous Priority Control)
   - Private PAT management
   - GVT: improved error handling and pci config sanitizing
   - Execlist refactoring
   - Transparent Huge Page support
   - User defined priorities support
   - HuC/GuC firmware refactoring
   - DP MST fixes
   - eDP power sequencing fixes
   - Use RCU instead of stop_machine
   - PSR state tracking support
   - Eviction fixes
   - BDW DP aux channel timeout fixes
   - LSPCON fixes
   - Cannonlake PLL fixes

  amdgpu:
   - Per VM BO support
   - Powerplay cleanups
   - CI powerplay support
   - PASID mgr for kfd
   - SR-IOV fixes
   - initial GPU reset for vega10
   - Prime mmap support
   - TTM updates
   - Clock query interface for Raven
   - Fence to handle ioctl
   - UVD encode ring support on Polaris
   - Transparent huge page DMA support
   - Compute LRU pipe tweaks
   - BO flag to allow buffers to opt out of implicit sync
   - CTX priority setting API
   - VRAM lost infrastructure plumbing

  qxl:
   - fix flicker since atomic rework

  amdkfd:
   - Further improvements from internal AMD tree
   - Usermode events
   - Drop radeon support

  nouveau:
   - Pascal temperature sensor support
   - Improved BAR2 handling
   - MMU rework to support Pascal MMU

  exynos:
   - Improved HDMI/mixer support
   - HDMI audio interface support

  tegra:
   - Prep work for tegra186
   - Cleanup/fixes

  msm:
   - Preemption support for a5xx
   - Display fixes for 8x96 (snapdragon 820)
   - Async cursor plane fixes
   - FW loading rework
   - GPU debugging improvements

  vc4:
   - Prep for DSI panels
   - fix T-format tiling scanout
   - New madvise ioctl

  Rockchip:
   - LVDS support

  omapdrm:
   - omap4 HDMI CEC support

  etnaviv:
   - GPU performance counters groundwork

  sun4i:
   - refactor driver load + TCON backend
   - HDMI improvements
   - A31 support
   - Misc fixes

  udl:
   - Probe/EDID read fixes.

  tilcdc:
   - Misc fixes.

  pl111:
   - Support more variants

  adv7511:
   - Improve EDID handling.
   - HDMI CEC support

  sii8620:
   - Add remote control support"

* tag 'drm-for-v4.15' of git://people.freedesktop.org/~airlied/linux: (1480 commits)
  drm/rockchip: analogix_dp: Use mutex rather than spinlock
  drm/mode_object: fix documentation for object lookups.
  drm/i915: Reorder context-close to avoid calling i915_vma_close() under RCU
  drm/i915: Move init_clock_gating() back to where it was
  drm/i915: Prune the reservation shared fence array
  drm/i915: Idle the GPU before shinking everything
  drm/i915: Lock llist_del_first() vs llist_del_all()
  drm/i915: Calculate ironlake intermediate watermarks correctly, v2.
  drm/i915: Disable lazy PPGTT page table optimization for vGPU
  drm/i915/execlists: Remove the priority "optimisation"
  drm/i915: Filter out spurious execlists context-switch interrupts
  drm/amdgpu: use irq-safe lock for kiq->ring_lock
  drm/amdgpu: bypass lru touch for KIQ ring submission
  drm/amdgpu: Potential uninitialized variable in amdgpu_vm_update_directories()
  drm/amdgpu: potential uninitialized variable in amdgpu_vce_ring_parse_cs()
  drm/amd/powerplay: initialize a variable before using it
  drm/amd/powerplay: suppress KASAN out of bounds warning in vega10_populate_all_memory_levels
  drm/amd/amdgpu: fix evicted VRAM bo adjudgement condition
  drm/vblank: Tune drm_crtc_accurate_vblank_count() WARN down to a debug
  drm/rockchip: add CONFIG_OF dependency for lvds
  ...

Documentation/admin-guide/kernel-parameters.txt

@@ -857,7 +857,7 @@
 			The filter can be disabled or changed to another
 			driver later using sysfs.
 
-	drm_kms_helper.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
+	drm.edid_firmware=[<connector>:]<file>[,[<connector>:]<file>]
 			Broken monitors, graphic adapters, KVMs and EDIDless
 			panels may send no or incorrect EDID data sets.
 			This parameter allows to specify an EDID data sets

Documentation/devicetree/bindings/display/bridge/adi,adv7511.txt

@@ -68,6 +68,8 @@ Optional properties:
 - adi,disable-timing-generator: Only for ADV7533. Disables the internal timing
   generator. The chip will rely on the sync signals in the DSI data lanes,
   rather than generate its own timings for HDMI output.
+- clocks: from common clock binding: reference to the CEC clock.
+- clock-names: from common clock binding: must be "cec".
 
 Required nodes:
 
@@ -89,6 +91,8 @@ Example
 		reg = <39>;
 		interrupt-parent = <&gpio3>;
 		interrupts = <29 IRQ_TYPE_EDGE_FALLING>;
+		clocks = <&cec_clock>;
+		clock-names = "cec";
 
 		adi,input-depth = <8>;
 		adi,input-colorspace = "rgb";

Documentation/devicetree/bindings/display/bridge/sii9234.txt

@@ -0,0 +1,49 @@
+Silicon Image SiI9234 HDMI/MHL bridge bindings
+
+Required properties:
+	- compatible : "sil,sii9234".
+	- reg : I2C address for TPI interface, use 0x39
+	- avcc33-supply : MHL/USB Switch Supply Voltage (3.3V)
+	- iovcc18-supply : I/O Supply Voltage (1.8V)
+	- avcc12-supply : TMDS Analog Supply Voltage (1.2V)
+	- cvcc12-supply : Digital Core Supply Voltage (1.2V)
+	- interrupts, interrupt-parent: interrupt specifier of INT pin
+	- reset-gpios: gpio specifier of RESET pin (active low)
+	- video interfaces: Device node can contain two video interface port
+			    nodes for HDMI encoder and connector according to [1].
+			    - port@0 - MHL to HDMI
+			    - port@1 - MHL to connector
+
+[1]: Documentation/devicetree/bindings/media/video-interfaces.txt
+
+
+Example:
+	sii9234@39 {
+		compatible = "sil,sii9234";
+		reg = <0x39>;
+		avcc33-supply = <&vcc33mhl>;
+		iovcc18-supply = <&vcc18mhl>;
+		avcc12-supply = <&vsil12>;
+		cvcc12-supply = <&vsil12>;
+		reset-gpios = <&gpf3 4 GPIO_ACTIVE_LOW>;
+		interrupt-parent = <&gpf3>;
+		interrupts = <5 IRQ_TYPE_LEVEL_HIGH>;
+
+		ports {
+			#address-cells = <1>;
+			#size-cells = <0>;
+
+			port@0 {
+				reg = <0>;
+				mhl_to_hdmi: endpoint {
+					remote-endpoint = <&hdmi_to_mhl>;
+				};
+			};
+			port@1 {
+				reg = <1>;
+				mhl_to_connector: endpoint {
+					remote-endpoint = <&connector_to_mhl>;
+				};
+			};
+		};
+	};

Documentation/devicetree/bindings/display/faraday,tve200.txt

@@ -0,0 +1,54 @@
+* Faraday TV Encoder TVE200
+
+Required properties:
+
+- compatible: must be one of:
+	"faraday,tve200"
+	"cortina,gemini-tvc", "faraday,tve200"
+
+- reg: base address and size of the control registers block
+
+- interrupts: contains an interrupt specifier for the interrupt
+	line from the TVE200
+
+- clock-names: should contain "PCLK" for the clock line clocking the
+	silicon and "TVE" for the 27MHz clock to the video driver
+
+- clocks: contains phandle and clock specifier pairs for the entries
+	in the clock-names property. See
+	Documentation/devicetree/bindings/clock/clock-bindings.txt
+
+Optional properties:
+
+- resets: contains the reset line phandle for the block
+
+Required sub-nodes:
+
+- port: describes LCD panel signals, following the common binding
+	for video transmitter interfaces; see
+	Documentation/devicetree/bindings/media/video-interfaces.txt
+	This port should have the properties:
+	reg = <0>;
+	It should have one endpoint connected to a remote endpoint where
+	the display is connected.
+
+Example:
+
+display-controller@6a000000 {
+	#address-cells = <1>;
+	#size-cells = <0>;
+	compatible = "faraday,tve200";
+	reg = <0x6a000000 0x1000>;
+	interrupts = <13 IRQ_TYPE_EDGE_RISING>;
+	resets = <&syscon GEMINI_RESET_TVC>;
+	clocks = <&syscon GEMINI_CLK_GATE_TVC>,
+		 <&syscon GEMINI_CLK_TVC>;
+	clock-names = "PCLK", "TVE";
+
+	port@0 {
+		reg = <0>;
+		display_out: endpoint {
+			remote-endpoint = <&panel_in>;
+		};
+	};
+};

Documentation/devicetree/bindings/display/msm/dsi.txt

@@ -13,16 +13,16 @@ Required properties:
 - power-domains: Should be <&mmcc MDSS_GDSC>.
 - clocks: Phandles to device clocks.
 - clock-names: the following clocks are required:
-  * "mdp_core_clk"
-  * "iface_clk"
-  * "bus_clk"
-  * "core_mmss_clk"
-  * "byte_clk"
-  * "pixel_clk"
-  * "core_clk"
+  * "mdp_core"
+  * "iface"
+  * "bus"
+  * "core_mmss"
+  * "byte"
+  * "pixel"
+  * "core"
   For DSIv2, we need an additional clock:
-   * "src_clk"
-- assigned-clocks: Parents of "byte_clk" and "pixel_clk" for the given platform.
+   * "src"
+- assigned-clocks: Parents of "byte" and "pixel" for the given platform.
 - assigned-clock-parents: The Byte clock and Pixel clock PLL outputs provided
   by a DSI PHY block. See [1] for details on clock bindings.
 - vdd-supply: phandle to vdd regulator device node
@@ -101,7 +101,7 @@ Required properties:
 - power-domains: Should be <&mmcc MDSS_GDSC>.
 - clocks: Phandles to device clocks. See [1] for details on clock bindings.
 - clock-names: the following clocks are required:
-  * "iface_clk"
+  * "iface"
 - vddio-supply: phandle to vdd-io regulator device node
 
 Optional properties:
@@ -123,13 +123,13 @@ Example:
 		reg = <0xfd922800 0x200>;
 		power-domains = <&mmcc MDSS_GDSC>;
 		clock-names =
-			"bus_clk",
-			"byte_clk",
-			"core_clk",
-			"core_mmss_clk",
-			"iface_clk",
-			"mdp_core_clk",
-			"pixel_clk";
+			"bus",
+			"byte",
+			"core",
+			"core_mmss",
+			"iface",
+			"mdp_core",
+			"pixel";
 		clocks =
 			<&mmcc MDSS_AXI_CLK>,
 			<&mmcc MDSS_BYTE0_CLK>,
@@ -207,7 +207,7 @@ Example:
 		reg =   <0xfd922a00 0xd4>,
 			<0xfd922b00 0x2b0>,
 			<0xfd922d80 0x7b>;
-		clock-names = "iface_clk";
+		clock-names = "iface";
 		clocks = <&mmcc MDSS_AHB_CLK>;
 		#clock-cells = <1>;
 		vddio-supply = <&pma8084_l12>;

Documentation/devicetree/bindings/display/msm/edp.txt

@@ -12,11 +12,11 @@ Required properties:
 - clocks: device clocks
   See Documentation/devicetree/bindings/clock/clock-bindings.txt for details.
 - clock-names: the following clocks are required:
-  * "core_clk"
-  * "iface_clk"
-  * "mdp_core_clk"
-  * "pixel_clk"
-  * "link_clk"
+  * "core"
+  * "iface"
+  * "mdp_core"
+  * "pixel"
+  * "link"
 - #clock-cells: The value should be 1.
 - vdda-supply: phandle to vdda regulator device node
 - lvl-vdd-supply: phandle to regulator device node which is used to supply power
@@ -41,11 +41,11 @@ Example:
 			interrupts = <12 0>;
 			power-domains = <&mmcc MDSS_GDSC>;
 			clock-names =
-				"core_clk",
-				"pixel_clk",
-				"iface_clk",
-				"link_clk",
-				"mdp_core_clk";
+				"core",
+				"pixel",
+				"iface",
+				"link",
+				"mdp_core";
 			clocks =
 				<&mmcc MDSS_EDPAUX_CLK>,
 				<&mmcc MDSS_EDPPIXEL_CLK>,

Documentation/devicetree/bindings/display/msm/hdmi.txt

@@ -64,9 +64,9 @@ Example:
 		interrupts = <GIC_SPI 79 0>;
 		power-domains = <&mmcc MDSS_GDSC>;
 		clock-names =
-		    "core_clk",
-		    "master_iface_clk",
-		    "slave_iface_clk";
+		    "core",
+		    "master_iface",
+		    "slave_iface";
 		clocks =
 		    <&mmcc HDMI_APP_CLK>,
 		    <&mmcc HDMI_M_AHB_CLK>,
@@ -92,7 +92,7 @@ Example:
 		      <0x4a00500 0x100>;
 		#phy-cells = <0>;
 		power-domains = <&mmcc MDSS_GDSC>;
-		clock-names = "slave_iface_clk";
+		clock-names = "slave_iface";
 		clocks = <&mmcc HDMI_S_AHB_CLK>;
 		core-vdda-supply = <&pm8921_hdmi_mvs>;
 	};

Documentation/devicetree/bindings/display/msm/mdp5.txt

@@ -22,16 +22,16 @@ Required properties:
   Documentation/devicetree/bindings/power/power_domain.txt
 - clocks: device clocks. See ../clocks/clock-bindings.txt for details.
 - clock-names: the following clocks are required.
-  * "iface_clk"
-  * "bus_clk"
-  * "vsync_clk"
+  * "iface"
+  * "bus"
+  * "vsync"
 - #address-cells: number of address cells for the MDSS children. Should be 1.
 - #size-cells: Should be 1.
 - ranges: parent bus address space is the same as the child bus address space.
 
 Optional properties:
 - clock-names: the following clocks are optional:
-  * "lut_clk"
+  * "lut"
 
 MDP5:
 Required properties:
@@ -45,10 +45,10 @@ Required properties:
   through MDP block
 - clocks: device clocks. See ../clocks/clock-bindings.txt for details.
 - clock-names: the following clocks are required.
--   * "bus_clk"
--   * "iface_clk"
--   * "core_clk"
--   * "vsync_clk"
+-   * "bus"
+-   * "iface"
+-   * "core"
+-   * "vsync"
 - ports: contains the list of output ports from MDP. These connect to interfaces
   that are external to the MDP hardware, such as HDMI, DSI, EDP etc (LVDS is a
   special case since it is a part of the MDP block itself).
@@ -77,7 +77,7 @@ Required properties:
 
 Optional properties:
 - clock-names: the following clocks are optional:
-  * "lut_clk"
+  * "lut"
 
 Example:
 
@@ -95,9 +95,9 @@ Example:
 		clocks = <&gcc GCC_MDSS_AHB_CLK>,
 			 <&gcc GCC_MDSS_AXI_CLK>,
 			 <&gcc GCC_MDSS_VSYNC_CLK>;
-		clock-names = "iface_clk",
-			      "bus_clk",
-			      "vsync_clk"
+		clock-names = "iface",
+			      "bus",
+			      "vsync"
 
 		interrupts = <0 72 0>;
 
@@ -120,10 +120,10 @@ Example:
 				 <&gcc GCC_MDSS_AXI_CLK>,
 				 <&gcc GCC_MDSS_MDP_CLK>,
 				 <&gcc GCC_MDSS_VSYNC_CLK>;
-			clock-names = "iface_clk",
-				      "bus_clk",
-				      "core_clk",
-				      "vsync_clk";
+			clock-names = "iface",
+				      "bus",
+				      "core",
+				      "vsync";
 
 			ports {
 				#address-cells = <1>;

Documentation/devicetree/bindings/display/panel/orisetech,otm8009a.txt

@@ -0,0 +1,21 @@
+Orise Tech OTM8009A 3.97" 480x800 TFT LCD panel (MIPI-DSI video mode)
+
+The Orise Tech OTM8009A is a 3.97" 480x800 TFT LCD panel connected using
+a MIPI-DSI video interface. Its backlight is managed through the DSI link.
+
+Required properties:
+  - compatible: "orisetech,otm8009a"
+  - reg: the virtual channel number of a DSI peripheral
+
+Optional properties:
+  - reset-gpios: a GPIO spec for the reset pin (active low).
+
+Example:
+&dsi {
+	...
+	panel@0 {
+		compatible = "orisetech,otm8009a";
+		reg = <0>;
+		reset-gpios = <&gpioh 7 GPIO_ACTIVE_LOW>;
+	};
+};

Documentation/devicetree/bindings/display/panel/raspberrypi,7inch-touchscreen.txt

@@ -0,0 +1,49 @@
+This binding covers the official 7" (800x480) Raspberry Pi touchscreen
+panel.
+
+This DSI panel contains:
+
+- TC358762 DSI->DPI bridge
+- Atmel microcontroller on I2C for power sequencing the DSI bridge and
+  controlling backlight
+- Touchscreen controller on I2C for touch input
+
+and this binding covers the DSI display parts but not its touch input.
+
+Required properties:
+- compatible:	Must be "raspberrypi,7inch-touchscreen-panel"
+- reg:		Must be "45"
+- port:		See panel-common.txt
+
+Example:
+
+dsi1: dsi@7e700000 {
+	#address-cells = <1>;
+	#size-cells = <0>;
+	<...>
+
+	port {
+		dsi_out_port: endpoint {
+			remote-endpoint = <&panel_dsi_port>;
+		};
+	};
+};
+
+i2c_dsi: i2c {
+	compatible = "i2c-gpio";
+	#address-cells = <1>;
+	#size-cells = <0>;
+	gpios = <&gpio 28 0
+		 &gpio 29 0>;
+
+	lcd@45 {
+		compatible = "raspberrypi,7inch-touchscreen-panel";
+		reg = <0x45>;
+
+		port {
+			panel_dsi_port: endpoint {
+				remote-endpoint = <&dsi_out_port>;
+			};
+		};
+	};
+};

Documentation/devicetree/bindings/display/panel/samsung,s6e63j0x03.txt

@@ -0,0 +1,24 @@
+Samsung S6E63J0X03 1.63" 320x320 AMOLED panel (interface: MIPI-DSI command mode)
+
+Required properties:
+  - compatible: "samsung,s6e63j0x03"
+  - reg: the virtual channel number of a DSI peripheral
+  - vdd3-supply: I/O voltage supply
+  - vci-supply: voltage supply for analog circuits
+  - reset-gpios: a GPIO spec for the reset pin (active low)
+  - te-gpios: a GPIO spec for the tearing effect synchronization signal
+    gpio pin (active high)
+
+Example:
+&dsi {
+	...
+
+	panel@0 {
+		compatible = "samsung,s6e63j0x03";
+		reg = <0>;
+		vdd3-supply = <&ldo16_reg>;
+		vci-supply = <&ldo20_reg>;
+		reset-gpios = <&gpe0 1 GPIO_ACTIVE_LOW>;
+		te-gpios = <&gpx0 6 GPIO_ACTIVE_HIGH>;
+	};
+};

Documentation/devicetree/bindings/display/panel/seiko,43wvf1g.txt

@@ -0,0 +1,23 @@
+Seiko Instruments Inc. 4.3" WVGA (800 x RGB x 480) TFT with Touch-Panel
+
+Required properties:
+- compatible: should be "sii,43wvf1g".
+- "dvdd-supply": 3v3 digital regulator.
+- "avdd-supply": 5v analog regulator.
+
+Optional properties:
+- backlight: phandle for the backlight control.
+
+Example:
+
+	panel {
+		compatible = "sii,43wvf1g";
+		backlight = <&backlight_display>;
+		dvdd-supply = <&reg_lcd_3v3>;
+		avdd-supply = <&reg_lcd_5v>;
+		port {
+			panel_in: endpoint {
+				remote-endpoint = <&display_out>;
+			};
+		};
+	};

Documentation/devicetree/bindings/display/panel/toshiba,lt089ac29000.txt

@@ -0,0 +1,8 @@
+Toshiba 8.9" WXGA (1280x768) TFT LCD panel
+
+Required properties:
+- compatible: should be "toshiba,lt089ac29000.txt"
+- power-supply: as specified in the base binding
+
+This binding is compatible with the simple-panel binding, which is specified
+in simple-panel.txt in this directory.

Documentation/devicetree/bindings/display/rockchip/rockchip-lvds.txt

@@ -0,0 +1,99 @@
+Rockchip RK3288 LVDS interface
+================================
+
+Required properties:
+- compatible: matching the soc type, one of
+	- "rockchip,rk3288-lvds";
+
+- reg: physical base address of the controller and length
+	of memory mapped region.
+- clocks: must include clock specifiers corresponding to entries in the
+	clock-names property.
+- clock-names: must contain "pclk_lvds"
+
+- avdd1v0-supply: regulator phandle for 1.0V analog power
+- avdd1v8-supply: regulator phandle for 1.8V analog power
+- avdd3v3-supply: regulator phandle for 3.3V analog power
+
+- rockchip,grf: phandle to the general register files syscon
+- rockchip,output: "rgb", "lvds" or "duallvds", This describes the output interface
+
+Optional properties:
+- pinctrl-names: must contain a "lcdc" entry.
+- pinctrl-0: pin control group to be used for this controller.
+
+Required nodes:
+
+The lvds has two video ports as described by
+	Documentation/devicetree/bindings/media/video-interfaces.txt
+Their connections are modeled using the OF graph bindings specified in
+	Documentation/devicetree/bindings/graph.txt.
+
+- video port 0 for the VOP input, the remote endpoint maybe vopb or vopl
+- video port 1 for either a panel or subsequent encoder
+
+the lvds panel described by
+	Documentation/devicetree/bindings/display/panel/simple-panel.txt
+
+Panel required properties:
+- ports for remote LVDS output
+
+Panel optional properties:
+- data-mapping: should be "vesa-24","jeida-24" or "jeida-18".
+This describes decribed by:
+	Documentation/devicetree/bindings/display/panel/panel-lvds.txt
+
+Example:
+
+lvds_panel: lvds-panel {
+	compatible = "auo,b101ean01";
+	enable-gpios = <&gpio7 21 GPIO_ACTIVE_HIGH>;
+	data-mapping = "jeida-24";
+
+	ports {
+		panel_in_lvds: endpoint {
+			remote-endpoint = <&lvds_out_panel>;
+		};
+	};
+};
+
+For Rockchip RK3288:
+
+	lvds: lvds@ff96c000 {
+		compatible = "rockchip,rk3288-lvds";
+		rockchip,grf = <&grf>;
+		reg = <0xff96c000 0x4000>;
+		clocks = <&cru PCLK_LVDS_PHY>;
+		clock-names = "pclk_lvds";
+		pinctrl-names = "lcdc";
+		pinctrl-0 = <&lcdc_ctl>;
+		avdd1v0-supply = <&vdd10_lcd>;
+		avdd1v8-supply = <&vcc18_lcd>;
+		avdd3v3-supply = <&vcca_33>;
+		rockchip,output = "rgb";
+		ports {
+			#address-cells = <1>;
+			#size-cells = <0>;
+
+			lvds_in: port@0 {
+				reg = <0>;
+
+				lvds_in_vopb: endpoint@0 {
+					reg = <0>;
+					remote-endpoint = <&vopb_out_lvds>;
+				};
+				lvds_in_vopl: endpoint@1 {
+					reg = <1>;
+					remote-endpoint = <&vopl_out_lvds>;
+				};
+			};
+
+			lvds_out: port@1 {
+				reg = <1>;
+
+				lvds_out_panel: endpoint {
+					remote-endpoint = <&panel_in_lvds>;
+				};
+			};
+		};
+	};

Documentation/devicetree/bindings/display/sunxi/sun4i-drm.txt

@@ -40,15 +40,19 @@ CEC. It is one end of the pipeline.
 
 Required properties:
   - compatible: value must be one of:
+    * allwinner,sun4i-a10-hdmi
     * allwinner,sun5i-a10s-hdmi
+    * allwinner,sun6i-a31-hdmi
   - reg: base address and size of memory-mapped region
   - interrupts: interrupt associated to this IP
   - clocks: phandles to the clocks feeding the HDMI encoder
     * ahb: the HDMI interface clock
     * mod: the HDMI module clock
+    * ddc: the HDMI ddc clock (A31 only)
     * pll-0: the first video PLL
     * pll-1: the second video PLL
   - clock-names: the clock names mentioned above
+  - resets: phandle to the reset control for the HDMI encoder (A31 only)
   - dmas: phandles to the DMA channels used by the HDMI encoder
     * ddc-tx: The channel for DDC transmission
     * ddc-rx: The channel for DDC reception
@@ -83,9 +87,11 @@ The TCON acts as a timing controller for RGB, LVDS and TV interfaces.
 
 Required properties:
  - compatible: value must be either:
+   * allwinner,sun4i-a10-tcon
    * allwinner,sun5i-a13-tcon
    * allwinner,sun6i-a31-tcon
    * allwinner,sun6i-a31s-tcon
+   * allwinner,sun7i-a20-tcon
    * allwinner,sun8i-a33-tcon
    * allwinner,sun8i-v3s-tcon
  - reg: base address and size of memory-mapped region
@@ -150,8 +156,10 @@ system.
 
 Required properties:
   - compatible: value must be one of:
+    * allwinner,sun4i-a10-display-backend
     * allwinner,sun5i-a13-display-backend
     * allwinner,sun6i-a31-display-backend
+    * allwinner,sun7i-a20-display-backend
     * allwinner,sun8i-a33-display-backend
   - reg: base address and size of the memory-mapped region.
   - interrupts: interrupt associated to this IP
@@ -182,8 +190,10 @@ deinterlacing and color space conversion.
 
 Required properties:
   - compatible: value must be one of:
+    * allwinner,sun4i-a10-display-frontend
     * allwinner,sun5i-a13-display-frontend
     * allwinner,sun6i-a31-display-frontend
+    * allwinner,sun7i-a20-display-frontend
     * allwinner,sun8i-a33-display-frontend
   - reg: base address and size of the memory-mapped region.
   - interrupts: interrupt associated to this IP
@@ -228,10 +238,12 @@ extra node.
 
 Required properties:
   - compatible: value must be one of:
+    * allwinner,sun4i-a10-display-engine
     * allwinner,sun5i-a10s-display-engine
     * allwinner,sun5i-a13-display-engine
     * allwinner,sun6i-a31-display-engine
     * allwinner,sun6i-a31s-display-engine
+    * allwinner,sun7i-a20-display-engine
     * allwinner,sun8i-a33-display-engine
     * allwinner,sun8i-v3s-display-engine
 

Documentation/devicetree/bindings/display/tegra/nvidia,tegra20-host1x.txt

@@ -3,6 +3,10 @@ NVIDIA Tegra host1x
 Required properties:
 - compatible: "nvidia,tegra<chip>-host1x"
 - reg: Physical base address and length of the controller's registers.
+  For pre-Tegra186, one entry describing the whole register area.
+  For Tegra186, one entry for each entry in reg-names:
+    "vm" - VM region assigned to Linux
+    "hypervisor" - Hypervisor region (only if Linux acts as hypervisor)
 - interrupts: The interrupt outputs from the controller.
 - #address-cells: The number of cells used to represent physical base addresses
   in the host1x address space. Should be 1.

Documentation/devicetree/bindings/vendor-prefixes.txt

@@ -254,6 +254,7 @@ opencores	OpenCores.org
 openrisc	OpenRISC.io
 option	Option NV
 ORCL	Oracle Corporation
+orisetech	Orise Technology
 ortustech	Ortus Technology Co., Ltd.
 ovti	OmniVision Technologies
 oxsemi	Oxford Semiconductor, Ltd.

Documentation/gpu/drm-uapi.rst

@@ -168,6 +168,61 @@ IOCTL Support on Device Nodes
 .. kernel-doc:: drivers/gpu/drm/drm_ioctl.c
    :doc: driver specific ioctls
 
+Recommended IOCTL Return Values
+-------------------------------
+
+In theory a driver's IOCTL callback is only allowed to return very few error
+codes. In practice it's good to abuse a few more. This section documents common
+practice within the DRM subsystem:
+
+ENOENT:
+        Strictly this should only be used when a file doesn't exist e.g. when
+        calling the open() syscall. We reuse that to signal any kind of object
+        lookup failure, e.g. for unknown GEM buffer object handles, unknown KMS
+        object handles and similar cases.
+
+ENOSPC:
+        Some drivers use this to differentiate "out of kernel memory" from "out
+        of VRAM". Sometimes also applies to other limited gpu resources used for
+        rendering (e.g. when you have a special limited compression buffer).
+        Sometimes resource allocation/reservation issues in command submission
+        IOCTLs are also signalled through EDEADLK.
+
+        Simply running out of kernel/system memory is signalled through ENOMEM.
+
+EPERM/EACCESS:
+        Returned for an operation that is valid, but needs more privileges.
+        E.g. root-only or much more common, DRM master-only operations return
+        this when when called by unpriviledged clients. There's no clear
+        difference between EACCESS and EPERM.
+
+ENODEV:
+        Feature (like PRIME, modesetting, GEM) is not supported by the driver.
+
+ENXIO:
+        Remote failure, either a hardware transaction (like i2c), but also used
+        when the exporting driver of a shared dma-buf or fence doesn't support a
+        feature needed.
+
+EINTR:
+        DRM drivers assume that userspace restarts all IOCTLs. Any DRM IOCTL can
+        return EINTR and in such a case should be restarted with the IOCTL
+        parameters left unchanged.
+
+EIO:
+        The GPU died and couldn't be resurrected through a reset. Modesetting
+        hardware failures are signalled through the "link status" connector
+        property.
+
+EINVAL:
+        Catch-all for anything that is an invalid argument combination which
+        cannot work.
+
+IOCTL also use other error codes like ETIME, EFAULT, EBUSY, ENOTTY but their
+usage is in line with the common meanings. The above list tries to just document
+DRM specific patterns. Note that ENOTTY has the slightly unintuitive meaning of
+"this IOCTL does not exist", and is used exactly as such in DRM.
+
 .. kernel-doc:: include/drm/drm_ioctl.h
    :internal:
 

Documentation/gpu/index.rst

@@ -15,6 +15,7 @@ Linux GPU Driver Developer's Guide
    pl111
    tegra
    tinydrm
+   tve200
    vc4
    vga-switcheroo
    vgaarbiter

Documentation/gpu/todo.rst

@@ -75,17 +75,6 @@ helpers.
 
 Contact: Ville Syrjälä, Daniel Vetter, driver maintainers
 
-Implement deferred fbdev setup in the helper
---------------------------------------------
-
-Many (especially embedded drivers) want to delay fbdev setup until there's a
-real screen plugged in. This is to avoid the dreaded fallback to the low-res
-fbdev default. Many drivers have a hacked-up (and often broken) version of this,
-better to do it once in the shared helpers. Thierry has a patch series, but that
-one needs to be rebased and final polish applied.
-
-Contact: Thierry Reding, Daniel Vetter, driver maintainers
-
 Convert early atomic drivers to async commit helpers
 ----------------------------------------------------
 
@@ -138,6 +127,8 @@ interfaces to fix these issues:
   the acquire context explicitly on stack and then also pass it down into
   drivers explicitly so that the legacy-on-atomic functions can use them.
 
+  Except for some driver code this is done.
+
 * A bunch of the vtable hooks are now in the wrong place: DRM has a split
   between core vfunc tables (named ``drm_foo_funcs``), which are used to
   implement the userspace ABI. And then there's the optional hooks for the
@@ -151,6 +142,8 @@ interfaces to fix these issues:
   connector at runtime. That's almost all of them, and would allow us to get
   rid of a lot of ``best_encoder`` boilerplate in drivers.
 
+  This was almost done, but new drivers added a few more cases again.
+
 Contact: Daniel Vetter
 
 Get rid of dev->struct_mutex from GEM drivers
@@ -177,14 +170,19 @@ following drivers still use ``struct_mutex``: ``msm``, ``omapdrm`` and
 
 Contact: Daniel Vetter, respective driver maintainers
 
-Core refactorings
-=================
+Convert instances of dev_info/dev_err/dev_warn to their DRM_DEV_* equivalent
+----------------------------------------------------------------------------
 
-Use new IDR deletion interface to clean up drm_gem_handle_delete()
-------------------------------------------------------------------
+For drivers which could have multiple instances, it is necessary to
+differentiate between which is which in the logs. Since DRM_INFO/WARN/ERROR
+don't do this, drivers used dev_info/warn/err to make this differentiation. We
+now have DRM_DEV_* variants of the drm print macros, so we can start to convert
+those drivers back to using drm-formwatted specific log messages.
 
-See the "This is gross" comment -- apparently the IDR system now can return an
-error code instead of oopsing.
+Contact: Sean Paul, Maintainer of the driver you plan to convert
+
+Core refactorings
+=================
 
 Clean up the DRM header mess
 ----------------------------
@@ -306,6 +304,18 @@ There's a bunch of issues with it:
 
 Contact: Daniel Vetter
 
+KMS cleanups
+------------
+
+Some of these date from the very introduction of KMS in 2008 ...
+
+- drm_mode_config.crtc_idr is misnamed, since it contains all KMS object. Should
+  be renamed to drm_mode_config.object_idr.
+
+- drm_display_mode doesn't need to be derived from drm_mode_object. That's
+  leftovers from older (never merged into upstream) KMS designs where modes
+  where set using their ID, including support to add/remove modes.
+
 Better Testing
 ==============
 
@@ -353,7 +363,16 @@ those drivers as simple as possible, so lots of room for refactoring:
 - backlight helpers, probably best to put them into a new drm_backlight.c.
   This is because drivers/video is de-facto unmaintained. We could also
   move drivers/video/backlight to drivers/gpu/backlight and take it all
-  over within drm-misc, but that's more work.
+  over within drm-misc, but that's more work. Backlight helpers require a fair
+  bit of reworking and refactoring. A simple example is the enabling of a backlight.
+  Tinydrm has helpers for this. It would be good if other drivers can also use the
+  helper. However, there are various cases we need to consider i.e different
+  drivers seem to have different ways of enabling/disabling a backlight.
+  We also need to consider the backlight drivers (like gpio_backlight). The situation
+  is further complicated by the fact that the backlight is tied to fbdev
+  via fb_notifier_callback() which has complicated logic. For further details, refer
+  to the following discussion thread:
+  https://groups.google.com/forum/#!topic/outreachy-kernel/8rBe30lwtdA
 
 - spi helpers, probably best put into spi core/helper code. Thierry said
   the spi maintainer is fast&reactive, so shouldn't be a big issue.

Documentation/gpu/tve200.rst

@@ -0,0 +1,6 @@
+==================================
+ drm/tve200 Faraday TV Encoder 200
+==================================
+
+.. kernel-doc:: drivers/gpu/drm/tve200/tve200_drv.c
+   :doc: Faraday TV Encoder 200

MAINTAINERS

@@ -754,8 +754,6 @@ F:	drivers/gpu/drm/amd/amdkfd/
 F:	drivers/gpu/drm/amd/include/cik_structs.h
 F:	drivers/gpu/drm/amd/include/kgd_kfd_interface.h
 F:	drivers/gpu/drm/amd/include/vi_structs.h
-F:	drivers/gpu/drm/radeon/radeon_kfd.c
-F:	drivers/gpu/drm/radeon/radeon_kfd.h
 F:	include/uapi/linux/kfd_ioctl.h
 
 AMD SEATTLE DEVICE TREE SUPPORT
@@ -4400,6 +4398,12 @@ T:	git git://anongit.freedesktop.org/drm/drm-misc
 S:	Maintained
 F:	drivers/gpu/drm/bochs/
 
+DRM DRIVER FOR FARADAY TVE200 TV ENCODER
+M:	Linus Walleij <linus.walleij@linaro.org>
+T:	git git://anongit.freedesktop.org/drm/drm-misc
+S:	Maintained
+F:	drivers/gpu/drm/tve200/
+
 DRM DRIVER FOR INTEL I810 VIDEO CARDS
 S:	Orphan / Obsolete
 F:	drivers/gpu/drm/i810/
@@ -4543,7 +4547,7 @@ L:	dri-devel@lists.freedesktop.org
 S:	Supported
 F:	drivers/gpu/drm/sun4i/
 F:	Documentation/devicetree/bindings/display/sunxi/sun4i-drm.txt
-T:	git git://git.kernel.org/pub/scm/linux/kernel/git/mripard/linux.git
+T:	git git://anongit.freedesktop.org/drm/drm-misc
 
 DRM DRIVERS FOR AMLOGIC SOCS
 M:	Neil Armstrong <narmstrong@baylibre.com>
@@ -4727,7 +4731,7 @@ T:	git git://anongit.freedesktop.org/drm/drm-misc
 DRM PANEL DRIVERS
 M:	Thierry Reding <thierry.reding@gmail.com>
 L:	dri-devel@lists.freedesktop.org
-T:	git git://anongit.freedesktop.org/tegra/linux.git
+T:	git git://anongit.freedesktop.org/drm/drm-misc
 S:	Maintained
 F:	drivers/gpu/drm/drm_panel.c
 F:	drivers/gpu/drm/panel/
@@ -5495,6 +5499,7 @@ F:	drivers/net/wan/sdla.c
 
 FRAMEBUFFER LAYER
 M:	Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
+L:	dri-devel@lists.freedesktop.org
 L:	linux-fbdev@vger.kernel.org
 T:	git git://github.com/bzolnier/linux.git
 Q:	http://patchwork.kernel.org/project/linux-fbdev/list/

arch/powerpc/kernel/pci-common.c

@@ -1740,15 +1740,3 @@ static void fixup_hide_host_resource_fsl(struct pci_dev *dev)
 }
 DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_MOTOROLA, PCI_ANY_ID, fixup_hide_host_resource_fsl);
 DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_FREESCALE, PCI_ANY_ID, fixup_hide_host_resource_fsl);
-
-static void fixup_vga(struct pci_dev *pdev)
-{
-	u16 cmd;
-
-	pci_read_config_word(pdev, PCI_COMMAND, &cmd);
-	if ((cmd & (PCI_COMMAND_IO | PCI_COMMAND_MEMORY)) || !vga_default_device())
-		vga_set_default_device(pdev);
-
-}
-DECLARE_PCI_FIXUP_CLASS_FINAL(PCI_ANY_ID, PCI_ANY_ID,
-			      PCI_CLASS_DISPLAY_VGA, 8, fixup_vga);

drivers/dma-buf/dma-buf.c

@@ -625,7 +625,7 @@ EXPORT_SYMBOL_GPL(dma_buf_detach);
 struct sg_table *dma_buf_map_attachment(struct dma_buf_attachment *attach,
 					enum dma_data_direction direction)
 {
-	struct sg_table *sg_table = ERR_PTR(-EINVAL);
+	struct sg_table *sg_table;
 
 	might_sleep();
 

drivers/dma-buf/reservation.c

@@ -266,8 +266,7 @@ EXPORT_SYMBOL(reservation_object_add_excl_fence);
 * @dst: the destination reservation object
 * @src: the source reservation object
 *
-* Copy all fences from src to dst. Both src->lock as well as dst-lock must be
-* held.
+* Copy all fences from src to dst. dst-lock must be held.
 */
 int reservation_object_copy_fences(struct reservation_object *dst,
 				   struct reservation_object *src)
@@ -277,33 +276,62 @@ int reservation_object_copy_fences(struct reservation_object *dst,
 	size_t size;
 	unsigned i;
 
-	src_list = reservation_object_get_list(src);
+	rcu_read_lock();
+	src_list = rcu_dereference(src->fence);
 
+retry:
 	if (src_list) {
-		size = offsetof(typeof(*src_list),
-				shared[src_list->shared_count]);
+		unsigned shared_count = src_list->shared_count;
+
+		size = offsetof(typeof(*src_list), shared[shared_count]);
+		rcu_read_unlock();
+
 		dst_list = kmalloc(size, GFP_KERNEL);
 		if (!dst_list)
 			return -ENOMEM;
 
-		dst_list->shared_count = src_list->shared_count;
-		dst_list->shared_max = src_list->shared_count;
-		for (i = 0; i < src_list->shared_count; ++i)
-			dst_list->shared[i] =
-				dma_fence_get(src_list->shared[i]);
+		rcu_read_lock();
+		src_list = rcu_dereference(src->fence);
+		if (!src_list || src_list->shared_count > shared_count) {
+			kfree(dst_list);
+			goto retry;
+		}
+
+		dst_list->shared_count = 0;
+		dst_list->shared_max = shared_count;
+		for (i = 0; i < src_list->shared_count; ++i) {
+			struct dma_fence *fence;
+
+			fence = rcu_dereference(src_list->shared[i]);
+			if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT,
+				     &fence->flags))
+				continue;
+
+			if (!dma_fence_get_rcu(fence)) {
+				kfree(dst_list);
+				src_list = rcu_dereference(src->fence);
+				goto retry;
+			}
+
+			if (dma_fence_is_signaled(fence)) {
+				dma_fence_put(fence);
+				continue;
+			}
+
+			dst_list->shared[dst_list->shared_count++] = fence;
+		}
 	} else {
 		dst_list = NULL;
 	}
 
+	new = dma_fence_get_rcu_safe(&src->fence_excl);
+	rcu_read_unlock();
+
 	kfree(dst->staged);
 	dst->staged = NULL;
 
 	src_list = reservation_object_get_list(dst);
-
 	old = reservation_object_get_excl(dst);
-	new = reservation_object_get_excl(src);
-
-	dma_fence_get(new);
 
 	preempt_disable();
 	write_seqcount_begin(&dst->seq);

drivers/dma-buf/sw_sync.c

@@ -321,8 +321,16 @@ static int sw_sync_debugfs_open(struct inode *inode, struct file *file)
 static int sw_sync_debugfs_release(struct inode *inode, struct file *file)
 {
 	struct sync_timeline *obj = file->private_data;
+	struct sync_pt *pt, *next;
+
+	spin_lock_irq(&obj->lock);
+
+	list_for_each_entry_safe(pt, next, &obj->pt_list, link) {
+		dma_fence_set_error(&pt->base, -ENOENT);
+		dma_fence_signal_locked(&pt->base);
+	}
 
-	smp_wmb();
+	spin_unlock_irq(&obj->lock);
 
 	sync_timeline_put(obj);
 	return 0;

drivers/gpu/drm/Kconfig

@@ -110,7 +110,7 @@ config DRM_FBDEV_OVERALLOC
 
 config DRM_LOAD_EDID_FIRMWARE
 	bool "Allow to specify an EDID data set instead of probing for it"
-	depends on DRM_KMS_HELPER
+	depends on DRM
 	help
 	  Say Y here, if you want to use EDID data to be loaded from the
 	  /lib/firmware directory or one of the provided built-in
@@ -184,6 +184,7 @@ config DRM_AMDGPU
 	select BACKLIGHT_CLASS_DEVICE
 	select BACKLIGHT_LCD_SUPPORT
 	select INTERVAL_TREE
+	select CHASH
 	help
 	  Choose this option if you have a recent AMD Radeon graphics card.
 
@@ -191,6 +192,8 @@ config DRM_AMDGPU
 
 source "drivers/gpu/drm/amd/amdgpu/Kconfig"
 
+source "drivers/gpu/drm/amd/lib/Kconfig"
+
 source "drivers/gpu/drm/nouveau/Kconfig"
 
 source "drivers/gpu/drm/i915/Kconfig"
@@ -278,6 +281,8 @@ source "drivers/gpu/drm/tinydrm/Kconfig"
 
 source "drivers/gpu/drm/pl111/Kconfig"
 
+source "drivers/gpu/drm/tve200/Kconfig"
+
 # Keep legacy drivers last
 
 menuconfig DRM_LEGACY

drivers/gpu/drm/Makefile

@@ -18,7 +18,7 @@ drm-y       :=	drm_auth.o drm_bufs.o drm_cache.o \
 		drm_encoder.o drm_mode_object.o drm_property.o \
 		drm_plane.o drm_color_mgmt.o drm_print.o \
 		drm_dumb_buffers.o drm_mode_config.o drm_vblank.o \
-		drm_syncobj.o
+		drm_syncobj.o drm_lease.o
 
 drm-$(CONFIG_DRM_LIB_RANDOM) += lib/drm_random.o
 drm-$(CONFIG_DRM_VM) += drm_vm.o
@@ -29,6 +29,7 @@ drm-$(CONFIG_DRM_PANEL) += drm_panel.o
 drm-$(CONFIG_OF) += drm_of.o
 drm-$(CONFIG_AGP) += drm_agpsupport.o
 drm-$(CONFIG_DEBUG_FS) += drm_debugfs.o drm_debugfs_crc.o
+drm-$(CONFIG_DRM_LOAD_EDID_FIRMWARE) += drm_edid_load.o
 
 drm_kms_helper-y := drm_crtc_helper.o drm_dp_helper.o drm_probe_helper.o \
 		drm_plane_helper.o drm_dp_mst_topology.o drm_atomic_helper.o \
@@ -37,7 +38,6 @@ drm_kms_helper-y := drm_crtc_helper.o drm_dp_helper.o drm_probe_helper.o \
 		drm_scdc_helper.o drm_gem_framebuffer_helper.o
 
 drm_kms_helper-$(CONFIG_DRM_PANEL_BRIDGE) += bridge/panel.o
-drm_kms_helper-$(CONFIG_DRM_LOAD_EDID_FIRMWARE) += drm_edid_load.o
 drm_kms_helper-$(CONFIG_DRM_FBDEV_EMULATION) += drm_fb_helper.o
 drm_kms_helper-$(CONFIG_DRM_KMS_CMA_HELPER) += drm_fb_cma_helper.o
 drm_kms_helper-$(CONFIG_DRM_DP_AUX_CHARDEV) += drm_dp_aux_dev.o
@@ -45,14 +45,13 @@ drm_kms_helper-$(CONFIG_DRM_DP_AUX_CHARDEV) += drm_dp_aux_dev.o
 obj-$(CONFIG_DRM_KMS_HELPER) += drm_kms_helper.o
 obj-$(CONFIG_DRM_DEBUG_MM_SELFTEST) += selftests/
 
-CFLAGS_drm_trace_points.o := -I$(src)
-
 obj-$(CONFIG_DRM)	+= drm.o
 obj-$(CONFIG_DRM_MIPI_DSI) += drm_mipi_dsi.o
 obj-$(CONFIG_DRM_ARM)	+= arm/
 obj-$(CONFIG_DRM_TTM)	+= ttm/
 obj-$(CONFIG_DRM_TDFX)	+= tdfx/
 obj-$(CONFIG_DRM_R128)	+= r128/
+obj-y			+= amd/lib/
 obj-$(CONFIG_HSA_AMD) += amd/amdkfd/
 obj-$(CONFIG_DRM_RADEON)+= radeon/
 obj-$(CONFIG_DRM_AMDGPU)+= amd/amdgpu/
@@ -101,3 +100,4 @@ obj-$(CONFIG_DRM_ZTE)	+= zte/
 obj-$(CONFIG_DRM_MXSFB)	+= mxsfb/
 obj-$(CONFIG_DRM_TINYDRM) += tinydrm/
 obj-$(CONFIG_DRM_PL111) += pl111/
+obj-$(CONFIG_DRM_TVE200) += tve200/

drivers/gpu/drm/amd/amdgpu/Makefile

@@ -26,7 +26,7 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
 	amdgpu_prime.o amdgpu_vm.o amdgpu_ib.o amdgpu_pll.o \
 	amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \
 	amdgpu_gtt_mgr.o amdgpu_vram_mgr.o amdgpu_virt.o amdgpu_atomfirmware.o \
-	amdgpu_queue_mgr.o amdgpu_vf_error.o
+	amdgpu_queue_mgr.o amdgpu_vf_error.o amdgpu_sched.o
 
 # add asic specific block
 amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o kv_smc.o kv_dpm.o \
@@ -134,5 +134,3 @@ include $(FULL_AMD_PATH)/powerplay/Makefile
 amdgpu-y += $(AMD_POWERPLAY_FILES)
 
 obj-$(CONFIG_DRM_AMDGPU)+= amdgpu.o
-
-CFLAGS_amdgpu_trace_points.o := -I$(src)

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -65,6 +65,7 @@
 #include "amdgpu_uvd.h"
 #include "amdgpu_vce.h"
 #include "amdgpu_vcn.h"
+#include "amdgpu_mn.h"
 
 #include "gpu_scheduler.h"
 #include "amdgpu_virt.h"
@@ -91,7 +92,7 @@ extern int amdgpu_dpm;
 extern int amdgpu_fw_load_type;
 extern int amdgpu_aspm;
 extern int amdgpu_runtime_pm;
-extern unsigned amdgpu_ip_block_mask;
+extern uint amdgpu_ip_block_mask;
 extern int amdgpu_bapm;
 extern int amdgpu_deep_color;
 extern int amdgpu_vm_size;
@@ -104,14 +105,14 @@ extern int amdgpu_sched_jobs;
 extern int amdgpu_sched_hw_submission;
 extern int amdgpu_no_evict;
 extern int amdgpu_direct_gma_size;
-extern unsigned amdgpu_pcie_gen_cap;
-extern unsigned amdgpu_pcie_lane_cap;
-extern unsigned amdgpu_cg_mask;
-extern unsigned amdgpu_pg_mask;
-extern unsigned amdgpu_sdma_phase_quantum;
+extern uint amdgpu_pcie_gen_cap;
+extern uint amdgpu_pcie_lane_cap;
+extern uint amdgpu_cg_mask;
+extern uint amdgpu_pg_mask;
+extern uint amdgpu_sdma_phase_quantum;
 extern char *amdgpu_disable_cu;
 extern char *amdgpu_virtual_display;
-extern unsigned amdgpu_pp_feature_mask;
+extern uint amdgpu_pp_feature_mask;
 extern int amdgpu_vram_page_split;
 extern int amdgpu_ngg;
 extern int amdgpu_prim_buf_per_se;
@@ -120,6 +121,7 @@ extern int amdgpu_cntl_sb_buf_per_se;
 extern int amdgpu_param_buf_per_se;
 extern int amdgpu_job_hang_limit;
 extern int amdgpu_lbpw;
+extern int amdgpu_compute_multipipe;
 
 #ifdef CONFIG_DRM_AMDGPU_SI
 extern int amdgpu_si_support;
@@ -178,6 +180,7 @@ struct amdgpu_cs_parser;
 struct amdgpu_job;
 struct amdgpu_irq_src;
 struct amdgpu_fpriv;
+struct amdgpu_bo_va_mapping;
 
 enum amdgpu_cp_irq {
 	AMDGPU_CP_IRQ_GFX_EOP = 0,
@@ -292,14 +295,25 @@ struct amdgpu_buffer_funcs {
 
 /* provided by hw blocks that can write ptes, e.g., sdma */
 struct amdgpu_vm_pte_funcs {
+	/* number of dw to reserve per operation */
+	unsigned	copy_pte_num_dw;
+
 	/* copy pte entries from GART */
 	void (*copy_pte)(struct amdgpu_ib *ib,
 			 uint64_t pe, uint64_t src,
 			 unsigned count);
+
 	/* write pte one entry at a time with addr mapping */
 	void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe,
 			  uint64_t value, unsigned count,
 			  uint32_t incr);
+
+	/* maximum nums of PTEs/PDEs in a single operation */
+	uint32_t	set_max_nums_pte_pde;
+
+	/* number of dw to reserve per operation */
+	unsigned	set_pte_pde_num_dw;
+
 	/* for linear pte/pde updates without addr mapping */
 	void (*set_pte_pde)(struct amdgpu_ib *ib,
 			    uint64_t pe,
@@ -332,6 +346,7 @@ struct amdgpu_gart_funcs {
 struct amdgpu_ih_funcs {
 	/* ring read/write ptr handling, called from interrupt context */
 	u32 (*get_wptr)(struct amdgpu_device *adev);
+	bool (*prescreen_iv)(struct amdgpu_device *adev);
 	void (*decode_iv)(struct amdgpu_device *adev,
 			  struct amdgpu_iv_entry *entry);
 	void (*set_rptr)(struct amdgpu_device *adev);
@@ -399,6 +414,7 @@ void amdgpu_gem_prime_unpin(struct drm_gem_object *obj);
 struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *);
 void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj);
 void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr);
+int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma);
 int amdgpu_gem_debugfs_init(struct amdgpu_device *adev);
 
 /* sub-allocation manager, it has to be protected by another lock.
@@ -455,9 +471,10 @@ struct amdgpu_sa_bo {
  */
 void amdgpu_gem_force_release(struct amdgpu_device *adev);
 int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
-				int alignment, u32 initial_domain,
-				u64 flags, bool kernel,
-				struct drm_gem_object **obj);
+			     int alignment, u32 initial_domain,
+			     u64 flags, bool kernel,
+			     struct reservation_object *resv,
+			     struct drm_gem_object **obj);
 
 int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 			    struct drm_device *dev,
@@ -715,10 +732,14 @@ struct amdgpu_ctx {
 	struct amdgpu_device    *adev;
 	struct amdgpu_queue_mgr queue_mgr;
 	unsigned		reset_counter;
+	uint32_t		vram_lost_counter;
 	spinlock_t		ring_lock;
 	struct dma_fence	**fences;
 	struct amdgpu_ctx_ring	rings[AMDGPU_MAX_RINGS];
-	bool preamble_presented;
+	bool			preamble_presented;
+	enum amd_sched_priority init_priority;
+	enum amd_sched_priority override_priority;
+	struct mutex            lock;
 };
 
 struct amdgpu_ctx_mgr {
@@ -731,17 +752,22 @@ struct amdgpu_ctx_mgr {
 struct amdgpu_ctx *amdgpu_ctx_get(struct amdgpu_fpriv *fpriv, uint32_t id);
 int amdgpu_ctx_put(struct amdgpu_ctx *ctx);
 
-uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
-			      struct dma_fence *fence);
+int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
+			      struct dma_fence *fence, uint64_t *seq);
 struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,
 				   struct amdgpu_ring *ring, uint64_t seq);
+void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
+				  enum amd_sched_priority priority);
 
 int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
 		     struct drm_file *filp);
 
+int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, unsigned ring_id);
+
 void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr);
 void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr);
 
+
 /*
  * file private structure
  */
@@ -753,7 +779,6 @@ struct amdgpu_fpriv {
 	struct mutex		bo_list_lock;
 	struct idr		bo_list_handles;
 	struct amdgpu_ctx_mgr	ctx_mgr;
-	u32			vram_lost_counter;
 };
 
 /*
@@ -854,7 +879,7 @@ struct amdgpu_mec {
 struct amdgpu_kiq {
 	u64			eop_gpu_addr;
 	struct amdgpu_bo	*eop_obj;
-	struct mutex		ring_mutex;
+	spinlock_t              ring_lock;
 	struct amdgpu_ring	ring;
 	struct amdgpu_irq_src	irq;
 };
@@ -1014,11 +1039,14 @@ struct amdgpu_gfx {
 	/* reset mask */
 	uint32_t                        grbm_soft_reset;
 	uint32_t                        srbm_soft_reset;
-	bool                            in_reset;
 	/* s3/s4 mask */
 	bool                            in_suspend;
 	/* NGG */
 	struct amdgpu_ngg		ngg;
+
+	/* pipe reservation */
+	struct mutex			pipe_reserve_mutex;
+	DECLARE_BITMAP			(pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 };
 
 int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
@@ -1056,6 +1084,7 @@ struct amdgpu_cs_parser {
 	/* buffer objects */
 	struct ww_acquire_ctx		ticket;
 	struct amdgpu_bo_list		*bo_list;
+	struct amdgpu_mn		*mn;
 	struct amdgpu_bo_list_entry	vm_pd;
 	struct list_head		validated;
 	struct dma_fence		*fence;
@@ -1096,6 +1125,7 @@ struct amdgpu_job {
 	uint32_t		gds_base, gds_size;
 	uint32_t		gws_base, gws_size;
 	uint32_t		oa_base, oa_size;
+	uint32_t		vram_lost_counter;
 
 	/* user fence handling */
 	uint64_t		uf_addr;
@@ -1121,7 +1151,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,
 /*
  * Writeback
  */
-#define AMDGPU_MAX_WB 1024	/* Reserve at most 1024 WB slots for amdgpu-owned rings. */
+#define AMDGPU_MAX_WB 512	/* Reserve at most 512 WB slots for amdgpu-owned rings. */
 
 struct amdgpu_wb {
 	struct amdgpu_bo	*wb_obj;
@@ -1183,6 +1213,9 @@ struct amdgpu_firmware {
 
 	/* gpu info firmware data pointer */
 	const struct firmware *gpu_info_fw;
+
+	void *fw_buf_ptr;
+	uint64_t fw_buf_mc;
 };
 
 /*
@@ -1196,20 +1229,6 @@ void amdgpu_benchmark(struct amdgpu_device *adev, int test_number);
  */
 void amdgpu_test_moves(struct amdgpu_device *adev);
 
-/*
- * MMU Notifier
- */
-#if defined(CONFIG_MMU_NOTIFIER)
-int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr);
-void amdgpu_mn_unregister(struct amdgpu_bo *bo);
-#else
-static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
-{
-	return -ENODEV;
-}
-static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {}
-#endif
-
 /*
  * Debugfs
  */
@@ -1305,6 +1324,8 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp);
 int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
+int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data,
+				    struct drm_file *filp);
 int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
 int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data,
 				struct drm_file *filp);
@@ -1370,6 +1391,18 @@ struct amdgpu_atcs {
 	struct amdgpu_atcs_functions functions;
 };
 
+/*
+ * Firmware VRAM reservation
+ */
+struct amdgpu_fw_vram_usage {
+	u64 start_offset;
+	u64 size;
+	struct amdgpu_bo *reserved_bo;
+	void *va;
+};
+
+int amdgpu_fw_reserve_vram_init(struct amdgpu_device *adev);
+
 /*
  * CGS
  */
@@ -1519,7 +1552,6 @@ struct amdgpu_device {
 
 	/* powerplay */
 	struct amd_powerplay		powerplay;
-	bool				pp_enabled;
 	bool				pp_force_state_enabled;
 
 	/* dpm */
@@ -1575,6 +1607,8 @@ struct amdgpu_device {
 	struct delayed_work     late_init_work;
 
 	struct amdgpu_virt	virt;
+	/* firmware VRAM reservation */
+	struct amdgpu_fw_vram_usage fw_vram_usage;
 
 	/* link all shadow bo */
 	struct list_head                shadow_list;
@@ -1592,6 +1626,7 @@ struct amdgpu_device {
 
 	/* record last mm index being written through WREG32*/
 	unsigned long last_mm_index;
+	bool                            in_sriov_reset;
 };
 
 static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev)
@@ -1759,6 +1794,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
 #define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r))
 #define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o))
 #define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev))
+#define amdgpu_ih_prescreen_iv(adev) (adev)->irq.ih_funcs->prescreen_iv((adev))
 #define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv))
 #define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev))
 #define amdgpu_display_vblank_get_counter(adev, crtc) (adev)->mode_info.funcs->vblank_get_counter((adev), (crtc))
@@ -1791,18 +1827,6 @@ void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes,
 				  u64 num_vis_bytes);
 void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain);
 bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo);
-int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages);
-int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
-				     uint32_t flags);
-bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm);
-struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm);
-bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
-				  unsigned long end);
-bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
-				       int *last_invalidated);
-bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm);
-uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm,
-				 struct ttm_mem_reg *mem);
 void amdgpu_vram_location(struct amdgpu_device *adev, struct amdgpu_mc *mc, u64 base);
 void amdgpu_gart_location(struct amdgpu_device *adev, struct amdgpu_mc *mc);
 void amdgpu_ttm_set_active_vram_size(struct amdgpu_device *adev, u64 size);
@@ -1836,8 +1860,6 @@ static inline bool amdgpu_has_atpx(void) { return false; }
 extern const struct drm_ioctl_desc amdgpu_ioctls_kms[];
 extern const int amdgpu_max_kms_ioctl;
 
-bool amdgpu_kms_vram_lost(struct amdgpu_device *adev,
-			  struct amdgpu_fpriv *fpriv);
 int amdgpu_driver_load_kms(struct drm_device *dev, unsigned long flags);
 void amdgpu_driver_unload_kms(struct drm_device *dev);
 void amdgpu_driver_lastclose_kms(struct drm_device *dev);
@@ -1885,10 +1907,9 @@ static inline int amdgpu_acpi_init(struct amdgpu_device *adev) { return 0; }
 static inline void amdgpu_acpi_fini(struct amdgpu_device *adev) { }
 #endif
 
-struct amdgpu_bo_va_mapping *
-amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
-		       uint64_t addr, struct amdgpu_bo **bo);
-int amdgpu_cs_sysvm_access_required(struct amdgpu_cs_parser *parser);
+int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
+			   uint64_t addr, struct amdgpu_bo **bo,
+			   struct amdgpu_bo_va_mapping **mapping);
 
 #include "amdgpu_object.h"
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c

@@ -35,41 +35,50 @@
 
 #include "acp_gfx_if.h"
 
-#define ACP_TILE_ON_MASK                0x03
-#define ACP_TILE_OFF_MASK               0x02
-#define ACP_TILE_ON_RETAIN_REG_MASK     0x1f
-#define ACP_TILE_OFF_RETAIN_REG_MASK    0x20
-
-#define ACP_TILE_P1_MASK                0x3e
-#define ACP_TILE_P2_MASK                0x3d
-#define ACP_TILE_DSP0_MASK              0x3b
-#define ACP_TILE_DSP1_MASK              0x37
-
-#define ACP_TILE_DSP2_MASK              0x2f
-
-#define ACP_DMA_REGS_END		0x146c0
-#define ACP_I2S_PLAY_REGS_START		0x14840
-#define ACP_I2S_PLAY_REGS_END		0x148b4
-#define ACP_I2S_CAP_REGS_START		0x148b8
-#define ACP_I2S_CAP_REGS_END		0x1496c
-
-#define ACP_I2S_COMP1_CAP_REG_OFFSET	0xac
-#define ACP_I2S_COMP2_CAP_REG_OFFSET	0xa8
-#define ACP_I2S_COMP1_PLAY_REG_OFFSET	0x6c
-#define ACP_I2S_COMP2_PLAY_REG_OFFSET	0x68
-
-#define mmACP_PGFSM_RETAIN_REG		0x51c9
-#define mmACP_PGFSM_CONFIG_REG		0x51ca
-#define mmACP_PGFSM_READ_REG_0		0x51cc
-
-#define mmACP_MEM_SHUT_DOWN_REQ_LO	0x51f8
-#define mmACP_MEM_SHUT_DOWN_REQ_HI	0x51f9
-#define mmACP_MEM_SHUT_DOWN_STS_LO	0x51fa
-#define mmACP_MEM_SHUT_DOWN_STS_HI	0x51fb
-
-#define ACP_TIMEOUT_LOOP		0x000000FF
-#define ACP_DEVS			3
-#define ACP_SRC_ID			162
+#define ACP_TILE_ON_MASK                	0x03
+#define ACP_TILE_OFF_MASK               	0x02
+#define ACP_TILE_ON_RETAIN_REG_MASK     	0x1f
+#define ACP_TILE_OFF_RETAIN_REG_MASK    	0x20
+
+#define ACP_TILE_P1_MASK                	0x3e
+#define ACP_TILE_P2_MASK                	0x3d
+#define ACP_TILE_DSP0_MASK              	0x3b
+#define ACP_TILE_DSP1_MASK              	0x37
+
+#define ACP_TILE_DSP2_MASK              	0x2f
+
+#define ACP_DMA_REGS_END			0x146c0
+#define ACP_I2S_PLAY_REGS_START			0x14840
+#define ACP_I2S_PLAY_REGS_END			0x148b4
+#define ACP_I2S_CAP_REGS_START			0x148b8
+#define ACP_I2S_CAP_REGS_END			0x1496c
+
+#define ACP_I2S_COMP1_CAP_REG_OFFSET		0xac
+#define ACP_I2S_COMP2_CAP_REG_OFFSET		0xa8
+#define ACP_I2S_COMP1_PLAY_REG_OFFSET		0x6c
+#define ACP_I2S_COMP2_PLAY_REG_OFFSET		0x68
+
+#define mmACP_PGFSM_RETAIN_REG			0x51c9
+#define mmACP_PGFSM_CONFIG_REG			0x51ca
+#define mmACP_PGFSM_READ_REG_0			0x51cc
+
+#define mmACP_MEM_SHUT_DOWN_REQ_LO		0x51f8
+#define mmACP_MEM_SHUT_DOWN_REQ_HI		0x51f9
+#define mmACP_MEM_SHUT_DOWN_STS_LO		0x51fa
+#define mmACP_MEM_SHUT_DOWN_STS_HI		0x51fb
+
+#define mmACP_CONTROL				0x5131
+#define mmACP_STATUS				0x5133
+#define mmACP_SOFT_RESET			0x5134
+#define ACP_CONTROL__ClkEn_MASK 		0x1
+#define ACP_SOFT_RESET__SoftResetAud_MASK 	0x100
+#define ACP_SOFT_RESET__SoftResetAudDone_MASK	0x1000000
+#define ACP_CLOCK_EN_TIME_OUT_VALUE		0x000000FF
+#define ACP_SOFT_RESET_DONE_TIME_OUT_VALUE	0x000000FF
+
+#define ACP_TIMEOUT_LOOP			0x000000FF
+#define ACP_DEVS				3
+#define ACP_SRC_ID				162
 
 enum {
 	ACP_TILE_P1 = 0,
@@ -260,6 +269,8 @@ static int acp_hw_init(void *handle)
 {
 	int r, i;
 	uint64_t acp_base;
+	u32 val = 0;
+	u32 count = 0;
 	struct device *dev;
 	struct i2s_platform_data *i2s_pdata;
 
@@ -402,6 +413,46 @@ static int acp_hw_init(void *handle)
 		}
 	}
 
+	/* Assert Soft reset of ACP */
+	val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET);
+
+	val |= ACP_SOFT_RESET__SoftResetAud_MASK;
+	cgs_write_register(adev->acp.cgs_device, mmACP_SOFT_RESET, val);
+
+	count = ACP_SOFT_RESET_DONE_TIME_OUT_VALUE;
+	while (true) {
+		val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET);
+		if (ACP_SOFT_RESET__SoftResetAudDone_MASK ==
+		    (val & ACP_SOFT_RESET__SoftResetAudDone_MASK))
+			break;
+		if (--count == 0) {
+			dev_err(&adev->pdev->dev, "Failed to reset ACP\n");
+			return -ETIMEDOUT;
+		}
+		udelay(100);
+	}
+	/* Enable clock to ACP and wait until the clock is enabled */
+	val = cgs_read_register(adev->acp.cgs_device, mmACP_CONTROL);
+	val = val | ACP_CONTROL__ClkEn_MASK;
+	cgs_write_register(adev->acp.cgs_device, mmACP_CONTROL, val);
+
+	count = ACP_CLOCK_EN_TIME_OUT_VALUE;
+
+	while (true) {
+		val = cgs_read_register(adev->acp.cgs_device, mmACP_STATUS);
+		if (val & (u32) 0x1)
+			break;
+		if (--count == 0) {
+			dev_err(&adev->pdev->dev, "Failed to reset ACP\n");
+			return -ETIMEDOUT;
+		}
+		udelay(100);
+	}
+	/* Deassert the SOFT RESET flags */
+	val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET);
+	val &= ~ACP_SOFT_RESET__SoftResetAud_MASK;
+	cgs_write_register(adev->acp.cgs_device, mmACP_SOFT_RESET, val);
+
 	return 0;
 }
 
@@ -414,6 +465,8 @@ static int acp_hw_init(void *handle)
 static int acp_hw_fini(void *handle)
 {
 	int i, ret;
+	u32 val = 0;
+	u32 count = 0;
 	struct device *dev;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
@@ -421,6 +474,42 @@ static int acp_hw_fini(void *handle)
 	if (!adev->acp.acp_cell)
 		return 0;
 
+	/* Assert Soft reset of ACP */
+	val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET);
+
+	val |= ACP_SOFT_RESET__SoftResetAud_MASK;
+	cgs_write_register(adev->acp.cgs_device, mmACP_SOFT_RESET, val);
+
+	count = ACP_SOFT_RESET_DONE_TIME_OUT_VALUE;
+	while (true) {
+		val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET);
+		if (ACP_SOFT_RESET__SoftResetAudDone_MASK ==
+		    (val & ACP_SOFT_RESET__SoftResetAudDone_MASK))
+			break;
+		if (--count == 0) {
+			dev_err(&adev->pdev->dev, "Failed to reset ACP\n");
+			return -ETIMEDOUT;
+		}
+		udelay(100);
+	}
+	/* Disable ACP clock */
+	val = cgs_read_register(adev->acp.cgs_device, mmACP_CONTROL);
+	val &= ~ACP_CONTROL__ClkEn_MASK;
+	cgs_write_register(adev->acp.cgs_device, mmACP_CONTROL, val);
+
+	count = ACP_CLOCK_EN_TIME_OUT_VALUE;
+
+	while (true) {
+		val = cgs_read_register(adev->acp.cgs_device, mmACP_STATUS);
+		if (val & (u32) 0x1)
+			break;
+		if (--count == 0) {
+			dev_err(&adev->pdev->dev, "Failed to reset ACP\n");
+			return -ETIMEDOUT;
+		}
+		udelay(100);
+	}
+
 	if (adev->acp.acp_genpd) {
 		for (i = 0; i < ACP_DEVS ; i++) {
 			dev = get_mfd_cell_dev(adev->acp.acp_cell[i].name, i);

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c

@@ -169,6 +169,8 @@ static const struct kfd2kgd_calls kfd2kgd = {
 	.get_vmem_size = get_vmem_size,
 	.get_gpu_clock_counter = get_gpu_clock_counter,
 	.get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
+	.alloc_pasid = amdgpu_vm_alloc_pasid,
+	.free_pasid = amdgpu_vm_free_pasid,
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
 	.init_pipeline = kgd_init_pipeline,
@@ -336,6 +338,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	struct cik_mqd *m;
 	uint32_t *mqd_hqd;
 	uint32_t reg, wptr_val, data;
+	bool valid_wptr = false;
 
 	m = get_mqd(mqd);
 
@@ -354,7 +357,14 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 			     CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
 	WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, data);
 
-	if (read_user_wptr(mm, wptr, wptr_val))
+	/* read_user_ptr may take the mm->mmap_sem.
+	 * release srbm_mutex to avoid circular dependency between
+	 * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex.
+	 */
+	release_queue(kgd);
+	valid_wptr = read_user_wptr(mm, wptr, wptr_val);
+	acquire_queue(kgd, pipe_id, queue_id);
+	if (valid_wptr)
 		WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask);
 
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c

@@ -128,6 +128,8 @@ static const struct kfd2kgd_calls kfd2kgd = {
 	.get_vmem_size = get_vmem_size,
 	.get_gpu_clock_counter = get_gpu_clock_counter,
 	.get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
+	.alloc_pasid = amdgpu_vm_alloc_pasid,
+	.free_pasid = amdgpu_vm_free_pasid,
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
 	.init_pipeline = kgd_init_pipeline,
@@ -290,6 +292,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	struct vi_mqd *m;
 	uint32_t *mqd_hqd;
 	uint32_t reg, wptr_val, data;
+	bool valid_wptr = false;
 
 	m = get_mqd(mqd);
 
@@ -337,7 +340,14 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 			     CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
 	WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, data);
 
-	if (read_user_wptr(mm, wptr, wptr_val))
+	/* read_user_ptr may take the mm->mmap_sem.
+	 * release srbm_mutex to avoid circular dependency between
+	 * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex.
+	 */
+	release_queue(kgd);
+	valid_wptr = read_user_wptr(mm, wptr, wptr_val);
+	acquire_queue(kgd, pipe_id, queue_id);
+	if (valid_wptr)
 		WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask);
 
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);

drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c

@@ -1766,34 +1766,32 @@ bool amdgpu_atombios_scratch_need_asic_init(struct amdgpu_device *adev)
 		return true;
 }
 
-/* Atom needs data in little endian format
- * so swap as appropriate when copying data to
- * or from atom. Note that atom operates on
- * dw units.
+/* Atom needs data in little endian format so swap as appropriate when copying
+ * data to or from atom. Note that atom operates on dw units.
+ *
+ * Use to_le=true when sending data to atom and provide at least
+ * ALIGN(num_bytes,4) bytes in the dst buffer.
+ *
+ * Use to_le=false when receiving data from atom and provide ALIGN(num_bytes,4)
+ * byes in the src buffer.
  */
 void amdgpu_atombios_copy_swap(u8 *dst, u8 *src, u8 num_bytes, bool to_le)
 {
 #ifdef __BIG_ENDIAN
-	u8 src_tmp[20], dst_tmp[20]; /* used for byteswapping */
-	u32 *dst32, *src32;
+	u32 src_tmp[5], dst_tmp[5];
 	int i;
+	u8 align_num_bytes = ALIGN(num_bytes, 4);
 
-	memcpy(src_tmp, src, num_bytes);
-	src32 = (u32 *)src_tmp;
-	dst32 = (u32 *)dst_tmp;
 	if (to_le) {
-		for (i = 0; i < ((num_bytes + 3) / 4); i++)
-			dst32[i] = cpu_to_le32(src32[i]);
-		memcpy(dst, dst_tmp, num_bytes);
+		memcpy(src_tmp, src, num_bytes);
+		for (i = 0; i < align_num_bytes / 4; i++)
+			dst_tmp[i] = cpu_to_le32(src_tmp[i]);
+		memcpy(dst, dst_tmp, align_num_bytes);
 	} else {
-		u8 dws = num_bytes & ~3;
-		for (i = 0; i < ((num_bytes + 3) / 4); i++)
-			dst32[i] = le32_to_cpu(src32[i]);
-		memcpy(dst, dst_tmp, dws);
-		if (num_bytes % 4) {
-			for (i = 0; i < (num_bytes % 4); i++)
-				dst[dws+i] = dst_tmp[dws+i];
-		}
+		memcpy(src_tmp, src, align_num_bytes);
+		for (i = 0; i < align_num_bytes / 4; i++)
+			dst_tmp[i] = le32_to_cpu(src_tmp[i]);
+		memcpy(dst, dst_tmp, num_bytes);
 	}
 #else
 	memcpy(dst, src, num_bytes);
@@ -1807,6 +1805,8 @@ int amdgpu_atombios_allocate_fb_scratch(struct amdgpu_device *adev)
 	uint16_t data_offset;
 	int usage_bytes = 0;
 	struct _ATOM_VRAM_USAGE_BY_FIRMWARE *firmware_usage;
+	u64 start_addr;
+	u64 size;
 
 	if (amdgpu_atom_parse_data_header(ctx, index, NULL, NULL, NULL, &data_offset)) {
 		firmware_usage = (struct _ATOM_VRAM_USAGE_BY_FIRMWARE *)(ctx->bios + data_offset);
@@ -1815,7 +1815,21 @@ int amdgpu_atombios_allocate_fb_scratch(struct amdgpu_device *adev)
 			  le32_to_cpu(firmware_usage->asFirmwareVramReserveInfo[0].ulStartAddrUsedByFirmware),
 			  le16_to_cpu(firmware_usage->asFirmwareVramReserveInfo[0].usFirmwareUseInKb));
 
-		usage_bytes = le16_to_cpu(firmware_usage->asFirmwareVramReserveInfo[0].usFirmwareUseInKb) * 1024;
+		start_addr = firmware_usage->asFirmwareVramReserveInfo[0].ulStartAddrUsedByFirmware;
+		size = firmware_usage->asFirmwareVramReserveInfo[0].usFirmwareUseInKb;
+
+		if ((uint32_t)(start_addr & ATOM_VRAM_OPERATION_FLAGS_MASK) ==
+			(uint32_t)(ATOM_VRAM_BLOCK_SRIOV_MSG_SHARE_RESERVATION <<
+			ATOM_VRAM_OPERATION_FLAGS_SHIFT)) {
+			/* Firmware request VRAM reservation for SR-IOV */
+			adev->fw_vram_usage.start_offset = (start_addr &
+				(~ATOM_VRAM_OPERATION_FLAGS_MASK)) << 10;
+			adev->fw_vram_usage.size = size << 10;
+			/* Use the default scratch size */
+			usage_bytes = 0;
+		} else {
+			usage_bytes = le16_to_cpu(firmware_usage->asFirmwareVramReserveInfo[0].usFirmwareUseInKb) * 1024;
+		}
 	}
 	ctx->scratch_size_bytes = 0;
 	if (usage_bytes == 0)

drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c

@@ -71,19 +71,33 @@ int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev)
 	struct atom_context *ctx = adev->mode_info.atom_context;
 	int index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
 						vram_usagebyfirmware);
+	struct vram_usagebyfirmware_v2_1 *	firmware_usage;
+	uint32_t start_addr, size;
 	uint16_t data_offset;
 	int usage_bytes = 0;
 
 	if (amdgpu_atom_parse_data_header(ctx, index, NULL, NULL, NULL, &data_offset)) {
-		struct vram_usagebyfirmware_v2_1 *firmware_usage =
-			(struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset);
-
+		firmware_usage = (struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset);
 		DRM_DEBUG("atom firmware requested %08x %dkb fw %dkb drv\n",
 			  le32_to_cpu(firmware_usage->start_address_in_kb),
 			  le16_to_cpu(firmware_usage->used_by_firmware_in_kb),
 			  le16_to_cpu(firmware_usage->used_by_driver_in_kb));
 
-		usage_bytes = le16_to_cpu(firmware_usage->used_by_driver_in_kb) * 1024;
+		start_addr = le32_to_cpu(firmware_usage->start_address_in_kb);
+		size = le16_to_cpu(firmware_usage->used_by_firmware_in_kb);
+
+		if ((uint32_t)(start_addr & ATOM_VRAM_OPERATION_FLAGS_MASK) ==
+			(uint32_t)(ATOM_VRAM_BLOCK_SRIOV_MSG_SHARE_RESERVATION <<
+			ATOM_VRAM_OPERATION_FLAGS_SHIFT)) {
+			/* Firmware request VRAM reservation for SR-IOV */
+			adev->fw_vram_usage.start_offset = (start_addr &
+				(~ATOM_VRAM_OPERATION_FLAGS_MASK)) << 10;
+			adev->fw_vram_usage.size = size << 10;
+			/* Use the default scratch size */
+			usage_bytes = 0;
+		} else {
+			usage_bytes = le16_to_cpu(firmware_usage->used_by_driver_in_kb) << 10;
+		}
 	}
 	ctx->scratch_size_bytes = 0;
 	if (usage_bytes == 0)

drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c

@@ -42,10 +42,31 @@ struct amdgpu_cgs_device {
 	struct amdgpu_device *adev =					\
 		((struct amdgpu_cgs_device *)cgs_device)->adev
 
+static void *amdgpu_cgs_register_pp_handle(struct cgs_device *cgs_device,
+			int (*call_back_func)(struct amd_pp_init *, void **))
+{
+	CGS_FUNC_ADEV;
+	struct amd_pp_init pp_init;
+	struct amd_powerplay *amd_pp;
+
+	if (call_back_func == NULL)
+		return NULL;
+
+	amd_pp = &(adev->powerplay);
+	pp_init.chip_family = adev->family;
+	pp_init.chip_id = adev->asic_type;
+	pp_init.pm_en = (amdgpu_dpm != 0 && !amdgpu_sriov_vf(adev)) ? true : false;
+	pp_init.feature_mask = amdgpu_pp_feature_mask;
+	pp_init.device = cgs_device;
+	if (call_back_func(&pp_init, &(amd_pp->pp_handle)))
+		return NULL;
+
+	return adev->powerplay.pp_handle;
+}
+
 static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device,
 				    enum cgs_gpu_mem_type type,
 				    uint64_t size, uint64_t align,
-				    uint64_t min_offset, uint64_t max_offset,
 				    cgs_handle_t *handle)
 {
 	CGS_FUNC_ADEV;
@@ -53,13 +74,6 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device,
 	int ret = 0;
 	uint32_t domain = 0;
 	struct amdgpu_bo *obj;
-	struct ttm_placement placement;
-	struct ttm_place place;
-
-	if (min_offset > max_offset) {
-		BUG_ON(1);
-		return -EINVAL;
-	}
 
 	/* fail if the alignment is not a power of 2 */
 	if (((align != 1) && (align & (align - 1)))
@@ -73,41 +87,19 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device,
 		flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
 			AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		domain = AMDGPU_GEM_DOMAIN_VRAM;
-		if (max_offset > adev->mc.real_vram_size)
-			return -EINVAL;
-		place.fpfn = min_offset >> PAGE_SHIFT;
-		place.lpfn = max_offset >> PAGE_SHIFT;
-		place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED |
-			TTM_PL_FLAG_VRAM;
 		break;
 	case CGS_GPU_MEM_TYPE__INVISIBLE_CONTIG_FB:
 	case CGS_GPU_MEM_TYPE__INVISIBLE_FB:
 		flags = AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
 			AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		domain = AMDGPU_GEM_DOMAIN_VRAM;
-		if (adev->mc.visible_vram_size < adev->mc.real_vram_size) {
-			place.fpfn =
-				max(min_offset, adev->mc.visible_vram_size) >> PAGE_SHIFT;
-			place.lpfn =
-				min(max_offset, adev->mc.real_vram_size) >> PAGE_SHIFT;
-			place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED |
-				TTM_PL_FLAG_VRAM;
-		}
-
 		break;
 	case CGS_GPU_MEM_TYPE__GART_CACHEABLE:
 		domain = AMDGPU_GEM_DOMAIN_GTT;
-		place.fpfn = min_offset >> PAGE_SHIFT;
-		place.lpfn = max_offset >> PAGE_SHIFT;
-		place.flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_TT;
 		break;
 	case CGS_GPU_MEM_TYPE__GART_WRITECOMBINE:
 		flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC;
 		domain = AMDGPU_GEM_DOMAIN_GTT;
-		place.fpfn = min_offset >> PAGE_SHIFT;
-		place.lpfn = max_offset >> PAGE_SHIFT;
-		place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_TT |
-			TTM_PL_FLAG_UNCACHED;
 		break;
 	default:
 		return -EINVAL;
@@ -116,15 +108,8 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device,
 
 	*handle = 0;
 
-	placement.placement = &place;
-	placement.num_placement = 1;
-	placement.busy_placement = &place;
-	placement.num_busy_placement = 1;
-
-	ret = amdgpu_bo_create_restricted(adev, size, PAGE_SIZE,
-					  true, domain, flags,
-					  NULL, &placement, NULL,
-					  0, &obj);
+	ret = amdgpu_bo_create(adev, size, align, true, domain, flags,
+			       NULL, NULL, 0, &obj);
 	if (ret) {
 		DRM_ERROR("(%d) bo create failed\n", ret);
 		return ret;
@@ -155,19 +140,14 @@ static int amdgpu_cgs_gmap_gpu_mem(struct cgs_device *cgs_device, cgs_handle_t h
 				   uint64_t *mcaddr)
 {
 	int r;
-	u64 min_offset, max_offset;
 	struct amdgpu_bo *obj = (struct amdgpu_bo *)handle;
 
 	WARN_ON_ONCE(obj->placement.num_placement > 1);
 
-	min_offset = obj->placements[0].fpfn << PAGE_SHIFT;
-	max_offset = obj->placements[0].lpfn << PAGE_SHIFT;
-
 	r = amdgpu_bo_reserve(obj, true);
 	if (unlikely(r != 0))
 		return r;
-	r = amdgpu_bo_pin_restricted(obj, obj->preferred_domains,
-				     min_offset, max_offset, mcaddr);
+	r = amdgpu_bo_pin(obj, obj->preferred_domains, mcaddr);
 	amdgpu_bo_unreserve(obj);
 	return r;
 }
@@ -675,6 +655,85 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 
 		if (!adev->pm.fw) {
 			switch (adev->asic_type) {
+			case CHIP_TAHITI:
+				strcpy(fw_name, "radeon/tahiti_smc.bin");
+				break;
+			case CHIP_PITCAIRN:
+				if ((adev->pdev->revision == 0x81) &&
+				    ((adev->pdev->device == 0x6810) ||
+				    (adev->pdev->device == 0x6811))) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/pitcairn_k_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/pitcairn_smc.bin");
+				}
+				break;
+			case CHIP_VERDE:
+				if (((adev->pdev->device == 0x6820) &&
+					((adev->pdev->revision == 0x81) ||
+					(adev->pdev->revision == 0x83))) ||
+				    ((adev->pdev->device == 0x6821) &&
+					((adev->pdev->revision == 0x83) ||
+					(adev->pdev->revision == 0x87))) ||
+				    ((adev->pdev->revision == 0x87) &&
+					((adev->pdev->device == 0x6823) ||
+					(adev->pdev->device == 0x682b)))) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/verde_k_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/verde_smc.bin");
+				}
+				break;
+			case CHIP_OLAND:
+				if (((adev->pdev->revision == 0x81) &&
+					((adev->pdev->device == 0x6600) ||
+					(adev->pdev->device == 0x6604) ||
+					(adev->pdev->device == 0x6605) ||
+					(adev->pdev->device == 0x6610))) ||
+				    ((adev->pdev->revision == 0x83) &&
+					(adev->pdev->device == 0x6610))) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/oland_k_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/oland_smc.bin");
+				}
+				break;
+			case CHIP_HAINAN:
+				if (((adev->pdev->revision == 0x81) &&
+					(adev->pdev->device == 0x6660)) ||
+				    ((adev->pdev->revision == 0x83) &&
+					((adev->pdev->device == 0x6660) ||
+					(adev->pdev->device == 0x6663) ||
+					(adev->pdev->device == 0x6665) ||
+					 (adev->pdev->device == 0x6667)))) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/hainan_k_smc.bin");
+				} else if ((adev->pdev->revision == 0xc3) &&
+					 (adev->pdev->device == 0x6665)) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/banks_k_2_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/hainan_smc.bin");
+				}
+				break;
+			case CHIP_BONAIRE:
+				if ((adev->pdev->revision == 0x80) ||
+					(adev->pdev->revision == 0x81) ||
+					(adev->pdev->device == 0x665f)) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/bonaire_k_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/bonaire_smc.bin");
+				}
+				break;
+			case CHIP_HAWAII:
+				if (adev->pdev->revision == 0x80) {
+					info->is_kicker = true;
+					strcpy(fw_name, "radeon/hawaii_k_smc.bin");
+				} else {
+					strcpy(fw_name, "radeon/hawaii_smc.bin");
+				}
+				break;
 			case CHIP_TOPAZ:
 				if (((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0x81)) ||
 				    ((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0x83)) ||
@@ -838,6 +897,9 @@ static int amdgpu_cgs_query_system_info(struct cgs_device *cgs_device,
 	case CGS_SYSTEM_INFO_PCIE_SUB_SYS_VENDOR_ID:
 		sys_info->value = adev->pdev->subsystem_vendor;
 		break;
+	case CGS_SYSTEM_INFO_PCIE_BUS_DEVFN:
+		sys_info->value = adev->pdev->devfn;
+		break;
 	default:
 		return -ENODEV;
 	}
@@ -1139,6 +1201,7 @@ static const struct cgs_ops amdgpu_cgs_ops = {
 	.is_virtualization_enabled = amdgpu_cgs_is_virtualization_enabled,
 	.enter_safe_mode = amdgpu_cgs_enter_safe_mode,
 	.lock_grbm_idx = amdgpu_cgs_lock_grbm_idx,
+	.register_pp_handle = amdgpu_cgs_register_pp_handle,
 };
 
 static const struct cgs_os_ops amdgpu_cgs_os_ops = {

drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c

@@ -231,7 +231,7 @@ amdgpu_connector_update_scratch_regs(struct drm_connector *connector,
 		if (connector->encoder_ids[i] == 0)
 			break;
 
-		encoder = drm_encoder_find(connector->dev,
+		encoder = drm_encoder_find(connector->dev, NULL,
 					connector->encoder_ids[i]);
 		if (!encoder)
 			continue;
@@ -256,7 +256,7 @@ amdgpu_connector_find_encoder(struct drm_connector *connector,
 	for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) {
 		if (connector->encoder_ids[i] == 0)
 			break;
-		encoder = drm_encoder_find(connector->dev,
+		encoder = drm_encoder_find(connector->dev, NULL,
 					connector->encoder_ids[i]);
 		if (!encoder)
 			continue;
@@ -346,10 +346,8 @@ static void amdgpu_connector_free_edid(struct drm_connector *connector)
 {
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
 
-	if (amdgpu_connector->edid) {
-		kfree(amdgpu_connector->edid);
-		amdgpu_connector->edid = NULL;
-	}
+	kfree(amdgpu_connector->edid);
+	amdgpu_connector->edid = NULL;
 }
 
 static int amdgpu_connector_ddc_get_modes(struct drm_connector *connector)
@@ -374,7 +372,7 @@ amdgpu_connector_best_single_encoder(struct drm_connector *connector)
 
 	/* pick the encoder ids */
 	if (enc_id)
-		return drm_encoder_find(connector->dev, enc_id);
+		return drm_encoder_find(connector->dev, NULL, enc_id);
 	return NULL;
 }
 
@@ -1079,7 +1077,7 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force)
 			if (connector->encoder_ids[i] == 0)
 				break;
 
-			encoder = drm_encoder_find(connector->dev, connector->encoder_ids[i]);
+			encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]);
 			if (!encoder)
 				continue;
 
@@ -1136,7 +1134,7 @@ amdgpu_connector_dvi_encoder(struct drm_connector *connector)
 		if (connector->encoder_ids[i] == 0)
 			break;
 
-		encoder = drm_encoder_find(connector->dev, connector->encoder_ids[i]);
+		encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]);
 		if (!encoder)
 			continue;
 
@@ -1155,7 +1153,7 @@ amdgpu_connector_dvi_encoder(struct drm_connector *connector)
 	/* then check use digitial */
 	/* pick the first one */
 	if (enc_id)
-		return drm_encoder_find(connector->dev, enc_id);
+		return drm_encoder_find(connector->dev, NULL, enc_id);
 	return NULL;
 }
 
@@ -1296,7 +1294,7 @@ u16 amdgpu_connector_encoder_get_dp_bridge_encoder_id(struct drm_connector *conn
 		if (connector->encoder_ids[i] == 0)
 			break;
 
-		encoder = drm_encoder_find(connector->dev,
+		encoder = drm_encoder_find(connector->dev, NULL,
 					connector->encoder_ids[i]);
 		if (!encoder)
 			continue;
@@ -1325,7 +1323,7 @@ static bool amdgpu_connector_encoder_is_hbr2(struct drm_connector *connector)
 	for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) {
 		if (connector->encoder_ids[i] == 0)
 			break;
-		encoder = drm_encoder_find(connector->dev,
+		encoder = drm_encoder_find(connector->dev, NULL,
 					connector->encoder_ids[i]);
 		if (!encoder)
 			continue;

drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

@@ -25,6 +25,7 @@
  *    Jerome Glisse <glisse@freedesktop.org>
  */
 #include <linux/pagemap.h>
+#include <linux/sync_file.h>
 #include <drm/drmP.h>
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_syncobj.h>
@@ -89,12 +90,14 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
 		goto free_chunk;
 	}
 
+	mutex_lock(&p->ctx->lock);
+
 	/* get chunks */
 	chunk_array_user = u64_to_user_ptr(cs->in.chunks);
 	if (copy_from_user(chunk_array, chunk_array_user,
 			   sizeof(uint64_t)*cs->in.num_chunks)) {
 		ret = -EFAULT;
-		goto put_ctx;
+		goto free_chunk;
 	}
 
 	p->nchunks = cs->in.num_chunks;
@@ -102,7 +105,7 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
 			    GFP_KERNEL);
 	if (!p->chunks) {
 		ret = -ENOMEM;
-		goto put_ctx;
+		goto free_chunk;
 	}
 
 	for (i = 0; i < p->nchunks; i++) {
@@ -169,6 +172,11 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
 	if (ret)
 		goto free_all_kdata;
 
+	if (p->ctx->vram_lost_counter != p->job->vram_lost_counter) {
+		ret = -ECANCELED;
+		goto free_all_kdata;
+	}
+
 	if (p->uf_entry.robj)
 		p->job->uf_addr = uf_offset;
 	kfree(chunk_array);
@@ -182,8 +190,6 @@ free_partial_kdata:
 	kfree(p->chunks);
 	p->chunks = NULL;
 	p->nchunks = 0;
-put_ctx:
-	amdgpu_ctx_put(p->ctx);
 free_chunk:
 	kfree(chunk_array);
 
@@ -473,11 +479,16 @@ static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p,
 			return -EPERM;
 
 		/* Check if we have user pages and nobody bound the BO already */
-		if (lobj->user_pages && bo->tbo.ttm->state != tt_bound) {
-			size_t size = sizeof(struct page *);
-
-			size *= bo->tbo.ttm->num_pages;
-			memcpy(bo->tbo.ttm->pages, lobj->user_pages, size);
+		if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) &&
+		    lobj->user_pages) {
+			amdgpu_ttm_placement_from_domain(bo,
+							 AMDGPU_GEM_DOMAIN_CPU);
+			r = ttm_bo_validate(&bo->tbo, &bo->placement, true,
+					    false);
+			if (r)
+				return r;
+			amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm,
+						     lobj->user_pages);
 			binding_userptr = true;
 		}
 
@@ -502,7 +513,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	struct amdgpu_bo_list_entry *e;
 	struct list_head duplicates;
-	bool need_mmap_lock = false;
 	unsigned i, tries = 10;
 	int r;
 
@@ -510,9 +520,9 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 
 	p->bo_list = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle);
 	if (p->bo_list) {
-		need_mmap_lock = p->bo_list->first_userptr !=
-			p->bo_list->num_entries;
 		amdgpu_bo_list_get_list(p->bo_list, &p->validated);
+		if (p->bo_list->first_userptr != p->bo_list->num_entries)
+			p->mn = amdgpu_mn_get(p->adev);
 	}
 
 	INIT_LIST_HEAD(&duplicates);
@@ -521,9 +531,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 	if (p->uf_entry.robj)
 		list_add(&p->uf_entry.tv.head, &p->validated);
 
-	if (need_mmap_lock)
-		down_read(&current->mm->mmap_sem);
-
 	while (1) {
 		struct list_head need_pages;
 		unsigned i;
@@ -543,22 +550,24 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 		INIT_LIST_HEAD(&need_pages);
 		for (i = p->bo_list->first_userptr;
 		     i < p->bo_list->num_entries; ++i) {
+			struct amdgpu_bo *bo;
 
 			e = &p->bo_list->array[i];
+			bo = e->robj;
 
-			if (amdgpu_ttm_tt_userptr_invalidated(e->robj->tbo.ttm,
+			if (amdgpu_ttm_tt_userptr_invalidated(bo->tbo.ttm,
 				 &e->user_invalidated) && e->user_pages) {
 
 				/* We acquired a page array, but somebody
 				 * invalidated it. Free it and try again
 				 */
 				release_pages(e->user_pages,
-					      e->robj->tbo.ttm->num_pages);
+					      bo->tbo.ttm->num_pages);
 				kvfree(e->user_pages);
 				e->user_pages = NULL;
 			}
 
-			if (e->robj->tbo.ttm->state != tt_bound &&
+			if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) &&
 			    !e->user_pages) {
 				list_del(&e->tv.head);
 				list_add(&e->tv.head, &need_pages);
@@ -635,9 +644,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 
 	amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved,
 				     p->bytes_moved_vis);
-	fpriv->vm.last_eviction_counter =
-		atomic64_read(&p->adev->num_evictions);
-
 	if (p->bo_list) {
 		struct amdgpu_bo *gds = p->bo_list->gds_obj;
 		struct amdgpu_bo *gws = p->bo_list->gws_obj;
@@ -678,9 +684,6 @@ error_validate:
 
 error_free_pages:
 
-	if (need_mmap_lock)
-		up_read(&current->mm->mmap_sem);
-
 	if (p->bo_list) {
 		for (i = p->bo_list->first_userptr;
 		     i < p->bo_list->num_entries; ++i) {
@@ -705,7 +708,8 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p)
 
 	list_for_each_entry(e, &p->validated, tv.head) {
 		struct reservation_object *resv = e->robj->tbo.resv;
-		r = amdgpu_sync_resv(p->adev, &p->job->sync, resv, p->filp);
+		r = amdgpu_sync_resv(p->adev, &p->job->sync, resv, p->filp,
+				     amdgpu_bo_explicit_sync(e->robj));
 
 		if (r)
 			return r;
@@ -726,11 +730,7 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error,
 {
 	unsigned i;
 
-	if (!error)
-		ttm_eu_fence_buffer_objects(&parser->ticket,
-					    &parser->validated,
-					    parser->fence);
-	else if (backoff)
+	if (error && backoff)
 		ttm_eu_backoff_reservation(&parser->ticket,
 					   &parser->validated);
 
@@ -740,8 +740,10 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error,
 
 	dma_fence_put(parser->fence);
 
-	if (parser->ctx)
+	if (parser->ctx) {
+		mutex_unlock(&parser->ctx->lock);
 		amdgpu_ctx_put(parser->ctx);
+	}
 	if (parser->bo_list)
 		amdgpu_bo_list_put(parser->bo_list);
 
@@ -766,10 +768,6 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p)
 	if (r)
 		return r;
 
-	r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_dir_update);
-	if (r)
-		return r;
-
 	r = amdgpu_vm_clear_freed(adev, vm, NULL);
 	if (r)
 		return r;
@@ -823,7 +821,13 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p)
 
 	}
 
-	r = amdgpu_vm_clear_moved(adev, vm, &p->job->sync);
+	r = amdgpu_vm_handle_moved(adev, vm);
+	if (r)
+		return r;
+
+	r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_update);
+	if (r)
+		return r;
 
 	if (amdgpu_vm_debug && p->bo_list) {
 		/* Invalidate all BOs to test for userspace bugs */
@@ -833,7 +837,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p)
 			if (!bo)
 				continue;
 
-			amdgpu_vm_bo_invalidate(adev, bo);
+			amdgpu_vm_bo_invalidate(adev, bo, false);
 		}
 	}
 
@@ -846,19 +850,63 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev,
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_ring *ring = p->job->ring;
-	int i, r;
+	int r;
 
 	/* Only for UVD/VCE VM emulation */
-	if (ring->funcs->parse_cs) {
-		for (i = 0; i < p->job->num_ibs; i++) {
-			r = amdgpu_ring_parse_cs(ring, p, i);
+	if (p->job->ring->funcs->parse_cs) {
+		unsigned i, j;
+
+		for (i = 0, j = 0; i < p->nchunks && j < p->job->num_ibs; i++) {
+			struct drm_amdgpu_cs_chunk_ib *chunk_ib;
+			struct amdgpu_bo_va_mapping *m;
+			struct amdgpu_bo *aobj = NULL;
+			struct amdgpu_cs_chunk *chunk;
+			struct amdgpu_ib *ib;
+			uint64_t offset;
+			uint8_t *kptr;
+
+			chunk = &p->chunks[i];
+			ib = &p->job->ibs[j];
+			chunk_ib = chunk->kdata;
+
+			if (chunk->chunk_id != AMDGPU_CHUNK_ID_IB)
+				continue;
+
+			r = amdgpu_cs_find_mapping(p, chunk_ib->va_start,
+						   &aobj, &m);
+			if (r) {
+				DRM_ERROR("IB va_start is invalid\n");
+				return r;
+			}
+
+			if ((chunk_ib->va_start + chunk_ib->ib_bytes) >
+			    (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) {
+				DRM_ERROR("IB va_start+ib_bytes is invalid\n");
+				return -EINVAL;
+			}
+
+			/* the IB should be reserved at this point */
+			r = amdgpu_bo_kmap(aobj, (void **)&kptr);
+			if (r) {
+				return r;
+			}
+
+			offset = m->start * AMDGPU_GPU_PAGE_SIZE;
+			kptr += chunk_ib->va_start - offset;
+
+			memcpy(ib->ptr, kptr, chunk_ib->ib_bytes);
+			amdgpu_bo_kunmap(aobj);
+
+			r = amdgpu_ring_parse_cs(ring, p, j);
 			if (r)
 				return r;
+
+			j++;
 		}
 	}
 
 	if (p->job->vm) {
-		p->job->vm_pd_addr = amdgpu_bo_gpu_offset(vm->root.bo);
+		p->job->vm_pd_addr = amdgpu_bo_gpu_offset(vm->root.base.bo);
 
 		r = amdgpu_bo_vm_update_pte(p);
 		if (r)
@@ -920,54 +968,18 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev,
 
 		parser->job->ring = ring;
 
-		if (ring->funcs->parse_cs) {
-			struct amdgpu_bo_va_mapping *m;
-			struct amdgpu_bo *aobj = NULL;
-			uint64_t offset;
-			uint8_t *kptr;
-
-			m = amdgpu_cs_find_mapping(parser, chunk_ib->va_start,
-						   &aobj);
-			if (!aobj) {
-				DRM_ERROR("IB va_start is invalid\n");
-				return -EINVAL;
-			}
-
-			if ((chunk_ib->va_start + chunk_ib->ib_bytes) >
-			    (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) {
-				DRM_ERROR("IB va_start+ib_bytes is invalid\n");
-				return -EINVAL;
-			}
-
-			/* the IB should be reserved at this point */
-			r = amdgpu_bo_kmap(aobj, (void **)&kptr);
-			if (r) {
-				return r;
-			}
-
-			offset = m->start * AMDGPU_GPU_PAGE_SIZE;
-			kptr += chunk_ib->va_start - offset;
-
-			r =  amdgpu_ib_get(adev, vm, chunk_ib->ib_bytes, ib);
-			if (r) {
-				DRM_ERROR("Failed to get ib !\n");
-				return r;
-			}
-
-			memcpy(ib->ptr, kptr, chunk_ib->ib_bytes);
-			amdgpu_bo_kunmap(aobj);
-		} else {
-			r =  amdgpu_ib_get(adev, vm, 0, ib);
-			if (r) {
-				DRM_ERROR("Failed to get ib !\n");
-				return r;
-			}
-
+		r =  amdgpu_ib_get(adev, vm,
+					ring->funcs->parse_cs ? chunk_ib->ib_bytes : 0,
+					ib);
+		if (r) {
+			DRM_ERROR("Failed to get ib !\n");
+			return r;
 		}
 
 		ib->gpu_addr = chunk_ib->va_start;
 		ib->length_dw = chunk_ib->ib_bytes / 4;
 		ib->flags = chunk_ib->flags;
+
 		j++;
 	}
 
@@ -977,7 +989,7 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev,
 	    parser->job->ring->funcs->type == AMDGPU_RING_TYPE_VCE))
 		return -EINVAL;
 
-	return 0;
+	return amdgpu_ctx_wait_prev_fence(parser->ctx, parser->job->ring->idx);
 }
 
 static int amdgpu_cs_process_fence_dep(struct amdgpu_cs_parser *p,
@@ -1131,14 +1143,31 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	struct amdgpu_ring *ring = p->job->ring;
 	struct amd_sched_entity *entity = &p->ctx->rings[ring->idx].entity;
 	struct amdgpu_job *job;
+	unsigned i;
+	uint64_t seq;
+
 	int r;
 
+	amdgpu_mn_lock(p->mn);
+	if (p->bo_list) {
+		for (i = p->bo_list->first_userptr;
+		     i < p->bo_list->num_entries; ++i) {
+			struct amdgpu_bo *bo = p->bo_list->array[i].robj;
+
+			if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm)) {
+				amdgpu_mn_unlock(p->mn);
+				return -ERESTARTSYS;
+			}
+		}
+	}
+
 	job = p->job;
 	p->job = NULL;
 
 	r = amd_sched_job_init(&job->base, &ring->sched, entity, p->filp);
 	if (r) {
 		amdgpu_job_free(job);
+		amdgpu_mn_unlock(p->mn);
 		return r;
 	}
 
@@ -1146,21 +1175,36 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	job->fence_ctx = entity->fence_context;
 	p->fence = dma_fence_get(&job->base.s_fence->finished);
 
+	r = amdgpu_ctx_add_fence(p->ctx, ring, p->fence, &seq);
+	if (r) {
+		dma_fence_put(p->fence);
+		dma_fence_put(&job->base.s_fence->finished);
+		amdgpu_job_free(job);
+		amdgpu_mn_unlock(p->mn);
+		return r;
+	}
+
 	amdgpu_cs_post_dependencies(p);
 
-	cs->out.handle = amdgpu_ctx_add_fence(p->ctx, ring, p->fence);
-	job->uf_sequence = cs->out.handle;
+	cs->out.handle = seq;
+	job->uf_sequence = seq;
+
 	amdgpu_job_free_resources(job);
+	amdgpu_ring_priority_get(job->ring,
+				 amd_sched_get_job_priority(&job->base));
 
 	trace_amdgpu_cs_ioctl(job);
 	amd_sched_entity_push_job(&job->base);
+
+	ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence);
+	amdgpu_mn_unlock(p->mn);
+
 	return 0;
 }
 
 int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 {
 	struct amdgpu_device *adev = dev->dev_private;
-	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	union drm_amdgpu_cs *cs = data;
 	struct amdgpu_cs_parser parser = {};
 	bool reserved_buffers = false;
@@ -1168,8 +1212,6 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 
 	if (!adev->accel_working)
 		return -EBUSY;
-	if (amdgpu_kms_vram_lost(adev, fpriv))
-		return -ENODEV;
 
 	parser.adev = adev;
 	parser.filp = filp;
@@ -1180,6 +1222,10 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		goto out;
 	}
 
+	r = amdgpu_cs_ib_fill(adev, &parser);
+	if (r)
+		goto out;
+
 	r = amdgpu_cs_parser_bos(&parser, data);
 	if (r) {
 		if (r == -ENOMEM)
@@ -1190,9 +1236,6 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	}
 
 	reserved_buffers = true;
-	r = amdgpu_cs_ib_fill(adev, &parser);
-	if (r)
-		goto out;
 
 	r = amdgpu_cs_dependencies(adev, &parser);
 	if (r) {
@@ -1228,16 +1271,12 @@ int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data,
 {
 	union drm_amdgpu_wait_cs *wait = data;
 	struct amdgpu_device *adev = dev->dev_private;
-	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	unsigned long timeout = amdgpu_gem_timeout(wait->in.timeout);
 	struct amdgpu_ring *ring = NULL;
 	struct amdgpu_ctx *ctx;
 	struct dma_fence *fence;
 	long r;
 
-	if (amdgpu_kms_vram_lost(adev, fpriv))
-		return -ENODEV;
-
 	ctx = amdgpu_ctx_get(filp->driver_priv, wait->in.ctx_id);
 	if (ctx == NULL)
 		return -EINVAL;
@@ -1255,6 +1294,8 @@ int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data,
 		r = PTR_ERR(fence);
 	else if (fence) {
 		r = dma_fence_wait_timeout(fence, true, timeout);
+		if (r > 0 && fence->error)
+			r = fence->error;
 		dma_fence_put(fence);
 	} else
 		r = 1;
@@ -1302,6 +1343,62 @@ static struct dma_fence *amdgpu_cs_get_fence(struct amdgpu_device *adev,
 	return fence;
 }
 
+int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data,
+				    struct drm_file *filp)
+{
+	struct amdgpu_device *adev = dev->dev_private;
+	union drm_amdgpu_fence_to_handle *info = data;
+	struct dma_fence *fence;
+	struct drm_syncobj *syncobj;
+	struct sync_file *sync_file;
+	int fd, r;
+
+	fence = amdgpu_cs_get_fence(adev, filp, &info->in.fence);
+	if (IS_ERR(fence))
+		return PTR_ERR(fence);
+
+	switch (info->in.what) {
+	case AMDGPU_FENCE_TO_HANDLE_GET_SYNCOBJ:
+		r = drm_syncobj_create(&syncobj, 0, fence);
+		dma_fence_put(fence);
+		if (r)
+			return r;
+		r = drm_syncobj_get_handle(filp, syncobj, &info->out.handle);
+		drm_syncobj_put(syncobj);
+		return r;
+
+	case AMDGPU_FENCE_TO_HANDLE_GET_SYNCOBJ_FD:
+		r = drm_syncobj_create(&syncobj, 0, fence);
+		dma_fence_put(fence);
+		if (r)
+			return r;
+		r = drm_syncobj_get_fd(syncobj, (int*)&info->out.handle);
+		drm_syncobj_put(syncobj);
+		return r;
+
+	case AMDGPU_FENCE_TO_HANDLE_GET_SYNC_FILE_FD:
+		fd = get_unused_fd_flags(O_CLOEXEC);
+		if (fd < 0) {
+			dma_fence_put(fence);
+			return fd;
+		}
+
+		sync_file = sync_file_create(fence);
+		dma_fence_put(fence);
+		if (!sync_file) {
+			put_unused_fd(fd);
+			return -ENOMEM;
+		}
+
+		fd_install(fd, sync_file->file);
+		info->out.handle = fd;
+		return 0;
+
+	default:
+		return -EINVAL;
+	}
+}
+
 /**
  * amdgpu_cs_wait_all_fence - wait on all fences to signal
  *
@@ -1336,6 +1433,9 @@ static int amdgpu_cs_wait_all_fences(struct amdgpu_device *adev,
 
 		if (r == 0)
 			break;
+
+		if (fence->error)
+			return fence->error;
 	}
 
 	memset(wait, 0, sizeof(*wait));
@@ -1381,6 +1481,7 @@ static int amdgpu_cs_wait_any_fence(struct amdgpu_device *adev,
 			array[i] = fence;
 		} else { /* NULL, the fence has been already signaled */
 			r = 1;
+			first = i;
 			goto out;
 		}
 	}
@@ -1395,7 +1496,7 @@ out:
 	wait->out.status = (r > 0);
 	wait->out.first_signaled = first;
 	/* set return value 0 to indicate success */
-	r = 0;
+	r = array[first]->error;
 
 err_free_fence_array:
 	for (i = 0; i < fence_count; i++)
@@ -1416,15 +1517,12 @@ int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data,
 				struct drm_file *filp)
 {
 	struct amdgpu_device *adev = dev->dev_private;
-	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	union drm_amdgpu_wait_fences *wait = data;
 	uint32_t fence_count = wait->in.fence_count;
 	struct drm_amdgpu_fence *fences_user;
 	struct drm_amdgpu_fence *fences;
 	int r;
 
-	if (amdgpu_kms_vram_lost(adev, fpriv))
-		return -ENODEV;
 	/* Get the fences from userspace */
 	fences = kmalloc_array(fence_count, sizeof(struct drm_amdgpu_fence),
 			GFP_KERNEL);
@@ -1460,78 +1558,36 @@ err_free_fences:
  * virtual memory address. Returns allocation structure when found, NULL
  * otherwise.
  */
-struct amdgpu_bo_va_mapping *
-amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
-		       uint64_t addr, struct amdgpu_bo **bo)
+int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
+			   uint64_t addr, struct amdgpu_bo **bo,
+			   struct amdgpu_bo_va_mapping **map)
 {
+	struct amdgpu_fpriv *fpriv = parser->filp->driver_priv;
+	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_bo_va_mapping *mapping;
-	unsigned i;
-
-	if (!parser->bo_list)
-		return NULL;
-
-	addr /= AMDGPU_GPU_PAGE_SIZE;
-
-	for (i = 0; i < parser->bo_list->num_entries; i++) {
-		struct amdgpu_bo_list_entry *lobj;
-
-		lobj = &parser->bo_list->array[i];
-		if (!lobj->bo_va)
-			continue;
-
-		list_for_each_entry(mapping, &lobj->bo_va->valids, list) {
-			if (mapping->start > addr ||
-			    addr > mapping->last)
-				continue;
-
-			*bo = lobj->bo_va->base.bo;
-			return mapping;
-		}
-
-		list_for_each_entry(mapping, &lobj->bo_va->invalids, list) {
-			if (mapping->start > addr ||
-			    addr > mapping->last)
-				continue;
-
-			*bo = lobj->bo_va->base.bo;
-			return mapping;
-		}
-	}
-
-	return NULL;
-}
-
-/**
- * amdgpu_cs_sysvm_access_required - make BOs accessible by the system VM
- *
- * @parser: command submission parser context
- *
- * Helper for UVD/VCE VM emulation, make sure BOs are accessible by the system VM.
- */
-int amdgpu_cs_sysvm_access_required(struct amdgpu_cs_parser *parser)
-{
-	unsigned i;
 	int r;
 
-	if (!parser->bo_list)
-		return 0;
+	addr /= AMDGPU_GPU_PAGE_SIZE;
 
-	for (i = 0; i < parser->bo_list->num_entries; i++) {
-		struct amdgpu_bo *bo = parser->bo_list->array[i].robj;
+	mapping = amdgpu_vm_bo_lookup_mapping(vm, addr);
+	if (!mapping || !mapping->bo_va || !mapping->bo_va->base.bo)
+		return -EINVAL;
 
-		r = amdgpu_ttm_bind(&bo->tbo, &bo->tbo.mem);
-		if (unlikely(r))
-			return r;
+	*bo = mapping->bo_va->base.bo;
+	*map = mapping;
 
-		if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
-			continue;
+	/* Double check that the BO is reserved by this CS */
+	if (READ_ONCE((*bo)->tbo.resv->lock.ctx) != &parser->ticket)
+		return -EINVAL;
 
-		bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
-		amdgpu_ttm_placement_from_domain(bo, bo->allowed_domains);
-		r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false);
-		if (unlikely(r))
+	if (!((*bo)->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)) {
+		(*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+		amdgpu_ttm_placement_from_domain(*bo, (*bo)->allowed_domains);
+		r = ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, false,
+				    false);
+		if (r)
 			return r;
 	}
 
-	return 0;
+	return amdgpu_ttm_bind(&(*bo)->tbo, &(*bo)->tbo.mem);
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c

@@ -23,13 +23,41 @@
  */
 
 #include <drm/drmP.h>
+#include <drm/drm_auth.h>
 #include "amdgpu.h"
+#include "amdgpu_sched.h"
 
-static int amdgpu_ctx_init(struct amdgpu_device *adev, struct amdgpu_ctx *ctx)
+static int amdgpu_ctx_priority_permit(struct drm_file *filp,
+				      enum amd_sched_priority priority)
+{
+	/* NORMAL and below are accessible by everyone */
+	if (priority <= AMD_SCHED_PRIORITY_NORMAL)
+		return 0;
+
+	if (capable(CAP_SYS_NICE))
+		return 0;
+
+	if (drm_is_current_master(filp))
+		return 0;
+
+	return -EACCES;
+}
+
+static int amdgpu_ctx_init(struct amdgpu_device *adev,
+			   enum amd_sched_priority priority,
+			   struct drm_file *filp,
+			   struct amdgpu_ctx *ctx)
 {
 	unsigned i, j;
 	int r;
 
+	if (priority < 0 || priority >= AMD_SCHED_PRIORITY_MAX)
+		return -EINVAL;
+
+	r = amdgpu_ctx_priority_permit(filp, priority);
+	if (r)
+		return r;
+
 	memset(ctx, 0, sizeof(*ctx));
 	ctx->adev = adev;
 	kref_init(&ctx->refcount);
@@ -39,19 +67,24 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, struct amdgpu_ctx *ctx)
 	if (!ctx->fences)
 		return -ENOMEM;
 
+	mutex_init(&ctx->lock);
+
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 		ctx->rings[i].sequence = 1;
 		ctx->rings[i].fences = &ctx->fences[amdgpu_sched_jobs * i];
 	}
 
 	ctx->reset_counter = atomic_read(&adev->gpu_reset_counter);
+	ctx->vram_lost_counter = atomic_read(&adev->vram_lost_counter);
+	ctx->init_priority = priority;
+	ctx->override_priority = AMD_SCHED_PRIORITY_UNSET;
 
 	/* create context entity for each ring */
 	for (i = 0; i < adev->num_rings; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
 		struct amd_sched_rq *rq;
 
-		rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL];
+		rq = &ring->sched.sched_rq[priority];
 
 		if (ring == &adev->gfx.kiq.ring)
 			continue;
@@ -96,10 +129,14 @@ static void amdgpu_ctx_fini(struct amdgpu_ctx *ctx)
 				      &ctx->rings[i].entity);
 
 	amdgpu_queue_mgr_fini(adev, &ctx->queue_mgr);
+
+	mutex_destroy(&ctx->lock);
 }
 
 static int amdgpu_ctx_alloc(struct amdgpu_device *adev,
 			    struct amdgpu_fpriv *fpriv,
+			    struct drm_file *filp,
+			    enum amd_sched_priority priority,
 			    uint32_t *id)
 {
 	struct amdgpu_ctx_mgr *mgr = &fpriv->ctx_mgr;
@@ -117,8 +154,9 @@ static int amdgpu_ctx_alloc(struct amdgpu_device *adev,
 		kfree(ctx);
 		return r;
 	}
+
 	*id = (uint32_t)r;
-	r = amdgpu_ctx_init(adev, ctx);
+	r = amdgpu_ctx_init(adev, priority, filp, ctx);
 	if (r) {
 		idr_remove(&mgr->ctx_handles, *id);
 		*id = 0;
@@ -193,6 +231,7 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
 {
 	int r;
 	uint32_t id;
+	enum amd_sched_priority priority;
 
 	union drm_amdgpu_ctx *args = data;
 	struct amdgpu_device *adev = dev->dev_private;
@@ -200,10 +239,16 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
 
 	r = 0;
 	id = args->in.ctx_id;
+	priority = amdgpu_to_sched_priority(args->in.priority);
+
+	/* For backwards compatibility reasons, we need to accept
+	 * ioctls with garbage in the priority field */
+	if (priority == AMD_SCHED_PRIORITY_INVALID)
+		priority = AMD_SCHED_PRIORITY_NORMAL;
 
 	switch (args->in.op) {
 	case AMDGPU_CTX_OP_ALLOC_CTX:
-		r = amdgpu_ctx_alloc(adev, fpriv, &id);
+		r = amdgpu_ctx_alloc(adev, fpriv, filp, priority, &id);
 		args->out.alloc.ctx_id = id;
 		break;
 	case AMDGPU_CTX_OP_FREE_CTX:
@@ -246,8 +291,8 @@ int amdgpu_ctx_put(struct amdgpu_ctx *ctx)
 	return 0;
 }
 
-uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
-			      struct dma_fence *fence)
+int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
+			      struct dma_fence *fence, uint64_t* handler)
 {
 	struct amdgpu_ctx_ring *cring = & ctx->rings[ring->idx];
 	uint64_t seq = cring->sequence;
@@ -256,12 +301,8 @@ uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
 
 	idx = seq & (amdgpu_sched_jobs - 1);
 	other = cring->fences[idx];
-	if (other) {
-		signed long r;
-		r = dma_fence_wait_timeout(other, false, MAX_SCHEDULE_TIMEOUT);
-		if (r < 0)
-			DRM_ERROR("Error (%ld) waiting for fence!\n", r);
-	}
+	if (other)
+		BUG_ON(!dma_fence_is_signaled(other));
 
 	dma_fence_get(fence);
 
@@ -271,8 +312,10 @@ uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
 	spin_unlock(&ctx->ring_lock);
 
 	dma_fence_put(other);
+	if (handler)
+		*handler = seq;
 
-	return seq;
+	return 0;
 }
 
 struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,
@@ -303,6 +346,51 @@ struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,
 	return fence;
 }
 
+void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
+				  enum amd_sched_priority priority)
+{
+	int i;
+	struct amdgpu_device *adev = ctx->adev;
+	struct amd_sched_rq *rq;
+	struct amd_sched_entity *entity;
+	struct amdgpu_ring *ring;
+	enum amd_sched_priority ctx_prio;
+
+	ctx->override_priority = priority;
+
+	ctx_prio = (ctx->override_priority == AMD_SCHED_PRIORITY_UNSET) ?
+			ctx->init_priority : ctx->override_priority;
+
+	for (i = 0; i < adev->num_rings; i++) {
+		ring = adev->rings[i];
+		entity = &ctx->rings[i].entity;
+		rq = &ring->sched.sched_rq[ctx_prio];
+
+		if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
+			continue;
+
+		amd_sched_entity_set_rq(entity, rq);
+	}
+}
+
+int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, unsigned ring_id)
+{
+	struct amdgpu_ctx_ring *cring = &ctx->rings[ring_id];
+	unsigned idx = cring->sequence & (amdgpu_sched_jobs - 1);
+	struct dma_fence *other = cring->fences[idx];
+
+	if (other) {
+		signed long r;
+		r = dma_fence_wait_timeout(other, false, MAX_SCHEDULE_TIMEOUT);
+		if (r < 0) {
+			DRM_ERROR("Error (%ld) waiting for fence!\n", r);
+			return r;
+		}
+	}
+
+	return 0;
+}
+
 void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr)
 {
 	mutex_init(&mgr->lock);

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

@@ -56,6 +56,7 @@
 #include "amdgpu_vf_error.h"
 
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_pm.h"
 
 MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin");
 MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin");
@@ -65,6 +66,7 @@ MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin");
 static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev);
 static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev);
 static int amdgpu_debugfs_test_ib_ring_init(struct amdgpu_device *adev);
+static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev);
 
 static const char *amdgpu_asic_name[] = {
 	"TAHITI",
@@ -107,10 +109,8 @@ uint32_t amdgpu_mm_rreg(struct amdgpu_device *adev, uint32_t reg,
 {
 	uint32_t ret;
 
-	if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev)) {
-		BUG_ON(in_interrupt());
+	if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev))
 		return amdgpu_virt_kiq_rreg(adev, reg);
-	}
 
 	if ((reg * 4) < adev->rmmio_size && !(acc_flags & AMDGPU_REGS_IDX))
 		ret = readl(((void __iomem *)adev->rmmio) + (reg * 4));
@@ -135,10 +135,8 @@ void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v,
 		adev->last_mm_index = v;
 	}
 
-	if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev)) {
-		BUG_ON(in_interrupt());
+	if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev))
 		return amdgpu_virt_kiq_wreg(adev, reg, v);
-	}
 
 	if ((reg * 4) < adev->rmmio_size && !(acc_flags & AMDGPU_REGS_IDX))
 		writel(v, ((void __iomem *)adev->rmmio) + (reg * 4));
@@ -402,6 +400,15 @@ void amdgpu_pci_config_reset(struct amdgpu_device *adev)
  */
 static int amdgpu_doorbell_init(struct amdgpu_device *adev)
 {
+	/* No doorbell on SI hardware generation */
+	if (adev->asic_type < CHIP_BONAIRE) {
+		adev->doorbell.base = 0;
+		adev->doorbell.size = 0;
+		adev->doorbell.num_doorbells = 0;
+		adev->doorbell.ptr = NULL;
+		return 0;
+	}
+
 	/* doorbell bar mapping */
 	adev->doorbell.base = pci_resource_start(adev->pdev, 2);
 	adev->doorbell.size = pci_resource_len(adev->pdev, 2);
@@ -539,7 +546,7 @@ int amdgpu_wb_get(struct amdgpu_device *adev, u32 *wb)
 
 	if (offset < adev->wb.num_wb) {
 		__set_bit(offset, adev->wb.used);
-		*wb = offset * 8; /* convert to dw offset */
+		*wb = offset << 3; /* convert to dw offset */
 		return 0;
 	} else {
 		return -EINVAL;
@@ -557,7 +564,7 @@ int amdgpu_wb_get(struct amdgpu_device *adev, u32 *wb)
 void amdgpu_wb_free(struct amdgpu_device *adev, u32 wb)
 {
 	if (wb < adev->wb.num_wb)
-		__clear_bit(wb, adev->wb.used);
+		__clear_bit(wb >> 3, adev->wb.used);
 }
 
 /**
@@ -647,42 +654,96 @@ void amdgpu_gart_location(struct amdgpu_device *adev, struct amdgpu_mc *mc)
 }
 
 /*
- * GPU helpers function.
+ * Firmware Reservation functions
  */
 /**
- * amdgpu_need_post - check if the hw need post or not
+ * amdgpu_fw_reserve_vram_fini - free fw reserved vram
  *
  * @adev: amdgpu_device pointer
  *
- * Check if the asic has been initialized (all asics) at driver startup
- * or post is needed if  hw reset is performed.
- * Returns true if need or false if not.
+ * free fw reserved vram if it has been reserved.
  */
-bool amdgpu_need_post(struct amdgpu_device *adev)
+void amdgpu_fw_reserve_vram_fini(struct amdgpu_device *adev)
 {
-	uint32_t reg;
+	amdgpu_bo_free_kernel(&adev->fw_vram_usage.reserved_bo,
+		NULL, &adev->fw_vram_usage.va);
+}
 
-	if (adev->has_hw_reset) {
-		adev->has_hw_reset = false;
-		return true;
-	}
+/**
+ * amdgpu_fw_reserve_vram_init - create bo vram reservation from fw
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * create bo vram reservation from fw.
+ */
+int amdgpu_fw_reserve_vram_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+	u64 gpu_addr;
+	u64 vram_size = adev->mc.visible_vram_size;
 
-	/* bios scratch used on CIK+ */
-	if (adev->asic_type >= CHIP_BONAIRE)
-		return amdgpu_atombios_scratch_need_asic_init(adev);
+	adev->fw_vram_usage.va = NULL;
+	adev->fw_vram_usage.reserved_bo = NULL;
 
-	/* check MEM_SIZE for older asics */
-	reg = amdgpu_asic_get_config_memsize(adev);
+	if (adev->fw_vram_usage.size > 0 &&
+		adev->fw_vram_usage.size <= vram_size) {
 
-	if ((reg != 0) && (reg != 0xffffffff))
-		return false;
+		r = amdgpu_bo_create(adev, adev->fw_vram_usage.size,
+			PAGE_SIZE, true, 0,
+			AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
+			AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, NULL, NULL, 0,
+			&adev->fw_vram_usage.reserved_bo);
+		if (r)
+			goto error_create;
 
-	return true;
+		r = amdgpu_bo_reserve(adev->fw_vram_usage.reserved_bo, false);
+		if (r)
+			goto error_reserve;
+		r = amdgpu_bo_pin_restricted(adev->fw_vram_usage.reserved_bo,
+			AMDGPU_GEM_DOMAIN_VRAM,
+			adev->fw_vram_usage.start_offset,
+			(adev->fw_vram_usage.start_offset +
+			adev->fw_vram_usage.size), &gpu_addr);
+		if (r)
+			goto error_pin;
+		r = amdgpu_bo_kmap(adev->fw_vram_usage.reserved_bo,
+			&adev->fw_vram_usage.va);
+		if (r)
+			goto error_kmap;
+
+		amdgpu_bo_unreserve(adev->fw_vram_usage.reserved_bo);
+	}
+	return r;
 
+error_kmap:
+	amdgpu_bo_unpin(adev->fw_vram_usage.reserved_bo);
+error_pin:
+	amdgpu_bo_unreserve(adev->fw_vram_usage.reserved_bo);
+error_reserve:
+	amdgpu_bo_unref(&adev->fw_vram_usage.reserved_bo);
+error_create:
+	adev->fw_vram_usage.va = NULL;
+	adev->fw_vram_usage.reserved_bo = NULL;
+	return r;
 }
 
-static bool amdgpu_vpost_needed(struct amdgpu_device *adev)
+
+/*
+ * GPU helpers function.
+ */
+/**
+ * amdgpu_need_post - check if the hw need post or not
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Check if the asic has been initialized (all asics) at driver startup
+ * or post is needed if  hw reset is performed.
+ * Returns true if need or false if not.
+ */
+bool amdgpu_need_post(struct amdgpu_device *adev)
 {
+	uint32_t reg;
+
 	if (amdgpu_sriov_vf(adev))
 		return false;
 
@@ -705,7 +766,23 @@ static bool amdgpu_vpost_needed(struct amdgpu_device *adev)
 				return true;
 		}
 	}
-	return amdgpu_need_post(adev);
+
+	if (adev->has_hw_reset) {
+		adev->has_hw_reset = false;
+		return true;
+	}
+
+	/* bios scratch used on CIK+ */
+	if (adev->asic_type >= CHIP_BONAIRE)
+		return amdgpu_atombios_scratch_need_asic_init(adev);
+
+	/* check MEM_SIZE for older asics */
+	reg = amdgpu_asic_get_config_memsize(adev);
+
+	if ((reg != 0) && (reg != 0xffffffff))
+		return false;
+
+	return true;
 }
 
 /**
@@ -887,6 +964,20 @@ static uint32_t cail_ioreg_read(struct card_info *info, uint32_t reg)
 	return r;
 }
 
+static ssize_t amdgpu_atombios_get_vbios_version(struct device *dev,
+						 struct device_attribute *attr,
+						 char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = ddev->dev_private;
+	struct atom_context *ctx = adev->mode_info.atom_context;
+
+	return snprintf(buf, PAGE_SIZE, "%s\n", ctx->vbios_version);
+}
+
+static DEVICE_ATTR(vbios_version, 0444, amdgpu_atombios_get_vbios_version,
+		   NULL);
+
 /**
  * amdgpu_atombios_fini - free the driver info and callbacks for atombios
  *
@@ -906,6 +997,7 @@ static void amdgpu_atombios_fini(struct amdgpu_device *adev)
 	adev->mode_info.atom_context = NULL;
 	kfree(adev->mode_info.atom_card_info);
 	adev->mode_info.atom_card_info = NULL;
+	device_remove_file(adev->dev, &dev_attr_vbios_version);
 }
 
 /**
@@ -922,6 +1014,7 @@ static int amdgpu_atombios_init(struct amdgpu_device *adev)
 {
 	struct card_info *atom_card_info =
 	    kzalloc(sizeof(struct card_info), GFP_KERNEL);
+	int ret;
 
 	if (!atom_card_info)
 		return -ENOMEM;
@@ -958,6 +1051,13 @@ static int amdgpu_atombios_init(struct amdgpu_device *adev)
 		amdgpu_atombios_scratch_regs_init(adev);
 		amdgpu_atombios_allocate_fb_scratch(adev);
 	}
+
+	ret = device_create_file(adev->dev, &dev_attr_vbios_version);
+	if (ret) {
+		DRM_ERROR("Failed to create device file for VBIOS version\n");
+		return ret;
+	}
+
 	return 0;
 }
 
@@ -1757,10 +1857,8 @@ static int amdgpu_fini(struct amdgpu_device *adev)
 		adev->ip_blocks[i].status.late_initialized = false;
 	}
 
-	if (amdgpu_sriov_vf(adev)) {
-		amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL);
+	if (amdgpu_sriov_vf(adev))
 		amdgpu_virt_release_full_gpu(adev, false);
-	}
 
 	return 0;
 }
@@ -1848,6 +1946,7 @@ static int amdgpu_sriov_reinit_late(struct amdgpu_device *adev)
 
 	static enum amd_ip_block_type ip_order[] = {
 		AMD_IP_BLOCK_TYPE_SMC,
+		AMD_IP_BLOCK_TYPE_PSP,
 		AMD_IP_BLOCK_TYPE_DCE,
 		AMD_IP_BLOCK_TYPE_GFX,
 		AMD_IP_BLOCK_TYPE_SDMA,
@@ -1933,12 +2032,17 @@ static int amdgpu_resume(struct amdgpu_device *adev)
 
 static void amdgpu_device_detect_sriov_bios(struct amdgpu_device *adev)
 {
-	if (adev->is_atom_fw) {
-		if (amdgpu_atomfirmware_gpu_supports_virtualization(adev))
-			adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS;
-	} else {
-		if (amdgpu_atombios_has_gpu_virtualization_table(adev))
-			adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS;
+	if (amdgpu_sriov_vf(adev)) {
+		if (adev->is_atom_fw) {
+			if (amdgpu_atomfirmware_gpu_supports_virtualization(adev))
+				adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS;
+		} else {
+			if (amdgpu_atombios_has_gpu_virtualization_table(adev))
+				adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS;
+		}
+
+		if (!(adev->virt.caps & AMDGPU_SRIOV_CAPS_SRIOV_VBIOS))
+			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_NO_VBIOS, 0, 0);
 	}
 }
 
@@ -1979,6 +2083,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	adev->vm_manager.vm_pte_num_rings = 0;
 	adev->gart.gart_funcs = NULL;
 	adev->fence_context = dma_fence_context_alloc(AMDGPU_MAX_RINGS);
+	bitmap_zero(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	adev->smc_rreg = &amdgpu_invalid_rreg;
 	adev->smc_wreg = &amdgpu_invalid_wreg;
@@ -2007,8 +2112,10 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	mutex_init(&adev->pm.mutex);
 	mutex_init(&adev->gfx.gpu_clock_mutex);
 	mutex_init(&adev->srbm_mutex);
+	mutex_init(&adev->gfx.pipe_reserve_mutex);
 	mutex_init(&adev->grbm_idx_mutex);
 	mutex_init(&adev->mn_lock);
+	mutex_init(&adev->virt.vf_errors.lock);
 	hash_init(adev->mn_hash);
 
 	amdgpu_check_arguments(adev);
@@ -2051,9 +2158,8 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	DRM_INFO("register mmio base: 0x%08X\n", (uint32_t)adev->rmmio_base);
 	DRM_INFO("register mmio size: %u\n", (unsigned)adev->rmmio_size);
 
-	if (adev->asic_type >= CHIP_BONAIRE)
-		/* doorbell bar mapping */
-		amdgpu_doorbell_init(adev);
+	/* doorbell bar mapping */
+	amdgpu_doorbell_init(adev);
 
 	/* io port mapping */
 	for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
@@ -2095,7 +2201,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_atombios_init(adev);
 	if (r) {
 		dev_err(adev->dev, "amdgpu_atombios_init failed\n");
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0);
 		goto failed;
 	}
 
@@ -2103,10 +2209,9 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	amdgpu_device_detect_sriov_bios(adev);
 
 	/* Post card if necessary */
-	if (amdgpu_vpost_needed(adev)) {
+	if (amdgpu_need_post(adev)) {
 		if (!adev->bios) {
 			dev_err(adev->dev, "no vBIOS found\n");
-			amdgpu_vf_error_put(AMDGIM_ERROR_VF_NO_VBIOS, 0, 0);
 			r = -EINVAL;
 			goto failed;
 		}
@@ -2114,7 +2219,6 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		r = amdgpu_atom_asic_init(adev->mode_info.atom_context);
 		if (r) {
 			dev_err(adev->dev, "gpu post error!\n");
-			amdgpu_vf_error_put(AMDGIM_ERROR_VF_GPU_POST_ERROR, 0, 0);
 			goto failed;
 		}
 	} else {
@@ -2126,7 +2230,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		r = amdgpu_atomfirmware_get_clock_info(adev);
 		if (r) {
 			dev_err(adev->dev, "amdgpu_atomfirmware_get_clock_info failed\n");
-			amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
+			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
 			goto failed;
 		}
 	} else {
@@ -2134,7 +2238,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		r = amdgpu_atombios_get_clock_info(adev);
 		if (r) {
 			dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n");
-			amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
+			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
 			goto failed;
 		}
 		/* init i2c buses */
@@ -2145,7 +2249,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_fence_driver_init(adev);
 	if (r) {
 		dev_err(adev->dev, "amdgpu_fence_driver_init failed\n");
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0);
 		goto failed;
 	}
 
@@ -2155,7 +2259,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_init(adev);
 	if (r) {
 		dev_err(adev->dev, "amdgpu_init failed\n");
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0);
 		amdgpu_fini(adev);
 		goto failed;
 	}
@@ -2175,7 +2279,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_ib_pool_init(adev);
 	if (r) {
 		dev_err(adev->dev, "IB initialization failed (%d).\n", r);
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r);
 		goto failed;
 	}
 
@@ -2183,8 +2287,15 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	if (r)
 		DRM_ERROR("ib ring test failed (%d).\n", r);
 
+	if (amdgpu_sriov_vf(adev))
+		amdgpu_virt_init_data_exchange(adev);
+
 	amdgpu_fbdev_init(adev);
 
+	r = amdgpu_pm_sysfs_init(adev);
+	if (r)
+		DRM_ERROR("registering pm debugfs failed (%d).\n", r);
+
 	r = amdgpu_gem_debugfs_init(adev);
 	if (r)
 		DRM_ERROR("registering gem debugfs failed (%d).\n", r);
@@ -2201,6 +2312,10 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	if (r)
 		DRM_ERROR("registering firmware debugfs failed (%d).\n", r);
 
+	r = amdgpu_debugfs_vbios_dump_init(adev);
+	if (r)
+		DRM_ERROR("Creating vbios dump debugfs failed (%d).\n", r);
+
 	if ((amdgpu_testing & 1)) {
 		if (adev->accel_working)
 			amdgpu_test_moves(adev);
@@ -2220,7 +2335,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_late_init(adev);
 	if (r) {
 		dev_err(adev->dev, "amdgpu_late_init failed\n");
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r);
 		goto failed;
 	}
 
@@ -2252,6 +2367,7 @@ void amdgpu_device_fini(struct amdgpu_device *adev)
 	/* evict vram memory */
 	amdgpu_bo_evict_vram(adev);
 	amdgpu_ib_pool_fini(adev);
+	amdgpu_fw_reserve_vram_fini(adev);
 	amdgpu_fence_driver_fini(adev);
 	amdgpu_fbdev_fini(adev);
 	r = amdgpu_fini(adev);
@@ -2276,8 +2392,8 @@ void amdgpu_device_fini(struct amdgpu_device *adev)
 	adev->rio_mem = NULL;
 	iounmap(adev->rmmio);
 	adev->rmmio = NULL;
-	if (adev->asic_type >= CHIP_BONAIRE)
-		amdgpu_doorbell_fini(adev);
+	amdgpu_doorbell_fini(adev);
+	amdgpu_pm_sysfs_fini(adev);
 	amdgpu_debugfs_regs_cleanup(adev);
 }
 
@@ -2504,6 +2620,9 @@ static bool amdgpu_check_soft_reset(struct amdgpu_device *adev)
 	int i;
 	bool asic_hang = false;
 
+	if (amdgpu_sriov_vf(adev))
+		return true;
+
 	for (i = 0; i < adev->num_ip_blocks; i++) {
 		if (!adev->ip_blocks[i].status.valid)
 			continue;
@@ -2546,7 +2665,8 @@ static bool amdgpu_need_full_reset(struct amdgpu_device *adev)
 		if ((adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) ||
 		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) ||
 		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_ACP) ||
-		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE)) {
+		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) ||
+		     adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP) {
 			if (adev->ip_blocks[i].status.hang) {
 				DRM_INFO("Some block need full reset!\n");
 				return true;
@@ -2654,7 +2774,7 @@ int amdgpu_sriov_gpu_reset(struct amdgpu_device *adev, struct amdgpu_job *job)
 
 	mutex_lock(&adev->virt.lock_reset);
 	atomic_inc(&adev->gpu_reset_counter);
-	adev->gfx.in_reset = true;
+	adev->in_sriov_reset = true;
 
 	/* block TTM */
 	resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev);
@@ -2765,7 +2885,7 @@ give_up_reset:
 		dev_info(adev->dev, "GPU reset successed!\n");
 	}
 
-	adev->gfx.in_reset = false;
+	adev->in_sriov_reset = false;
 	mutex_unlock(&adev->virt.lock_reset);
 	return r;
 }
@@ -2902,7 +3022,6 @@ out:
 		}
 	} else {
 		dev_err(adev->dev, "asic resume failed (%d).\n", r);
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_ASIC_RESUME_FAIL, 0, r);
 		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 			if (adev->rings[i] && adev->rings[i]->sched.thread) {
 				kthread_unpark(adev->rings[i]->sched.thread);
@@ -2916,7 +3035,6 @@ out:
 	if (r) {
 		/* bad news, how to tell it to userspace ? */
 		dev_info(adev->dev, "GPU reset failed\n");
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r);
 	}
 	else {
 		dev_info(adev->dev, "GPU reset successed!\n");
@@ -3463,10 +3581,7 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf,
 
 	valuesize = sizeof(values);
 	if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->read_sensor)
-		r = adev->powerplay.pp_funcs->read_sensor(adev->powerplay.pp_handle, idx, &values[0], &valuesize);
-	else if (adev->pm.funcs && adev->pm.funcs->read_sensor)
-		r = adev->pm.funcs->read_sensor(adev, idx, &values[0],
-						&valuesize);
+		r = amdgpu_dpm_read_sensor(adev, idx, &values[0], &valuesize);
 	else
 		return -EINVAL;
 
@@ -3754,6 +3869,28 @@ int amdgpu_debugfs_init(struct drm_minor *minor)
 {
 	return 0;
 }
+
+static int amdgpu_debugfs_get_vbios_dump(struct seq_file *m, void *data)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+
+	seq_write(m, adev->bios, adev->bios_size);
+	return 0;
+}
+
+static const struct drm_info_list amdgpu_vbios_dump_list[] = {
+		{"amdgpu_vbios",
+		 amdgpu_debugfs_get_vbios_dump,
+		 0, NULL},
+};
+
+static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev)
+{
+	return amdgpu_debugfs_add_files(adev,
+					amdgpu_vbios_dump_list, 1);
+}
 #else
 static int amdgpu_debugfs_test_ib_ring_init(struct amdgpu_device *adev)
 {
@@ -3763,5 +3900,9 @@ static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
 {
 	return 0;
 }
+static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev)
+{
+	return 0;
+}
 static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev) { }
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c

@@ -960,8 +960,10 @@ u8 amdgpu_encode_pci_lane_width(u32 lanes)
 }
 
 struct amd_vce_state*
-amdgpu_get_vce_clock_state(struct amdgpu_device *adev, unsigned idx)
+amdgpu_get_vce_clock_state(void *handle, u32 idx)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
 	if (idx < adev->pm.dpm.num_of_vce_states)
 		return &adev->pm.dpm.vce_states[idx];
 

drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h

@@ -241,179 +241,125 @@ enum amdgpu_pcie_gen {
 	AMDGPU_PCIE_GEN_INVALID = 0xffff
 };
 
-struct amdgpu_dpm_funcs {
-	int (*get_temperature)(struct amdgpu_device *adev);
-	int (*pre_set_power_state)(struct amdgpu_device *adev);
-	int (*set_power_state)(struct amdgpu_device *adev);
-	void (*post_set_power_state)(struct amdgpu_device *adev);
-	void (*display_configuration_changed)(struct amdgpu_device *adev);
-	u32 (*get_sclk)(struct amdgpu_device *adev, bool low);
-	u32 (*get_mclk)(struct amdgpu_device *adev, bool low);
-	void (*print_power_state)(struct amdgpu_device *adev, struct amdgpu_ps *ps);
-	void (*debugfs_print_current_performance_level)(struct amdgpu_device *adev, struct seq_file *m);
-	int (*force_performance_level)(struct amdgpu_device *adev, enum amd_dpm_forced_level level);
-	bool (*vblank_too_short)(struct amdgpu_device *adev);
-	void (*powergate_uvd)(struct amdgpu_device *adev, bool gate);
-	void (*powergate_vce)(struct amdgpu_device *adev, bool gate);
-	void (*enable_bapm)(struct amdgpu_device *adev, bool enable);
-	void (*set_fan_control_mode)(struct amdgpu_device *adev, u32 mode);
-	u32 (*get_fan_control_mode)(struct amdgpu_device *adev);
-	int (*set_fan_speed_percent)(struct amdgpu_device *adev, u32 speed);
-	int (*get_fan_speed_percent)(struct amdgpu_device *adev, u32 *speed);
-	int (*force_clock_level)(struct amdgpu_device *adev, enum pp_clock_type type, uint32_t mask);
-	int (*print_clock_levels)(struct amdgpu_device *adev, enum pp_clock_type type, char *buf);
-	int (*get_sclk_od)(struct amdgpu_device *adev);
-	int (*set_sclk_od)(struct amdgpu_device *adev, uint32_t value);
-	int (*get_mclk_od)(struct amdgpu_device *adev);
-	int (*set_mclk_od)(struct amdgpu_device *adev, uint32_t value);
-	int (*check_state_equal)(struct amdgpu_device *adev,
-				struct amdgpu_ps *cps,
-				struct amdgpu_ps *rps,
-				bool *equal);
-	int (*read_sensor)(struct amdgpu_device *adev, int idx, void *value,
-			   int *size);
-
-	struct amd_vce_state* (*get_vce_clock_state)(struct amdgpu_device *adev, unsigned idx);
-	int (*reset_power_profile_state)(struct amdgpu_device *adev,
-			struct amd_pp_profile *request);
-	int (*get_power_profile_state)(struct amdgpu_device *adev,
-			struct amd_pp_profile *query);
-	int (*set_power_profile_state)(struct amdgpu_device *adev,
-			struct amd_pp_profile *request);
-	int (*switch_power_profile)(struct amdgpu_device *adev,
-			enum amd_pp_profile_type type);
-};
+#define amdgpu_dpm_pre_set_power_state(adev) \
+		((adev)->powerplay.pp_funcs->pre_set_power_state((adev)->powerplay.pp_handle))
+
+#define amdgpu_dpm_set_power_state(adev) \
+		((adev)->powerplay.pp_funcs->set_power_state((adev)->powerplay.pp_handle))
+
+#define amdgpu_dpm_post_set_power_state(adev) \
+		((adev)->powerplay.pp_funcs->post_set_power_state((adev)->powerplay.pp_handle))
+
+#define amdgpu_dpm_display_configuration_changed(adev) \
+		((adev)->powerplay.pp_funcs->display_configuration_changed((adev)->powerplay.pp_handle))
 
-#define amdgpu_dpm_pre_set_power_state(adev) (adev)->pm.funcs->pre_set_power_state((adev))
-#define amdgpu_dpm_set_power_state(adev) (adev)->pm.funcs->set_power_state((adev))
-#define amdgpu_dpm_post_set_power_state(adev) (adev)->pm.funcs->post_set_power_state((adev))
-#define amdgpu_dpm_display_configuration_changed(adev) (adev)->pm.funcs->display_configuration_changed((adev))
-#define amdgpu_dpm_print_power_state(adev, ps) (adev)->pm.funcs->print_power_state((adev), (ps))
-#define amdgpu_dpm_vblank_too_short(adev) (adev)->pm.funcs->vblank_too_short((adev))
-#define amdgpu_dpm_enable_bapm(adev, e) (adev)->pm.funcs->enable_bapm((adev), (e))
+#define amdgpu_dpm_print_power_state(adev, ps) \
+		((adev)->powerplay.pp_funcs->print_power_state((adev)->powerplay.pp_handle, (ps)))
+
+#define amdgpu_dpm_vblank_too_short(adev) \
+		((adev)->powerplay.pp_funcs->vblank_too_short((adev)->powerplay.pp_handle))
+
+#define amdgpu_dpm_enable_bapm(adev, e) \
+		((adev)->powerplay.pp_funcs->enable_bapm((adev)->powerplay.pp_handle, (e)))
 
 #define amdgpu_dpm_read_sensor(adev, idx, value, size) \
-	((adev)->pp_enabled ? \
-		(adev)->powerplay.pp_funcs->read_sensor(adev->powerplay.pp_handle, (idx), (value), (size)) : \
-		(adev)->pm.funcs->read_sensor((adev), (idx), (value), (size)))
+		((adev)->powerplay.pp_funcs->read_sensor((adev)->powerplay.pp_handle, (idx), (value), (size)))
 
 #define amdgpu_dpm_get_temperature(adev) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_temperature((adev)->powerplay.pp_handle) : \
-	      (adev)->pm.funcs->get_temperature((adev)))
+		((adev)->powerplay.pp_funcs->get_temperature((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_set_fan_control_mode(adev, m) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->set_fan_control_mode((adev)->powerplay.pp_handle, (m)) : \
-	      (adev)->pm.funcs->set_fan_control_mode((adev), (m)))
+		((adev)->powerplay.pp_funcs->set_fan_control_mode((adev)->powerplay.pp_handle, (m)))
 
 #define amdgpu_dpm_get_fan_control_mode(adev) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_fan_control_mode((adev)->powerplay.pp_handle) : \
-	      (adev)->pm.funcs->get_fan_control_mode((adev)))
+		((adev)->powerplay.pp_funcs->get_fan_control_mode((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_set_fan_speed_percent(adev, s) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->set_fan_speed_percent((adev)->powerplay.pp_handle, (s)) : \
-	      (adev)->pm.funcs->set_fan_speed_percent((adev), (s)))
+		((adev)->powerplay.pp_funcs->set_fan_speed_percent((adev)->powerplay.pp_handle, (s)))
 
 #define amdgpu_dpm_get_fan_speed_percent(adev, s) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_fan_speed_percent((adev)->powerplay.pp_handle, (s)) : \
-	      (adev)->pm.funcs->get_fan_speed_percent((adev), (s)))
+		((adev)->powerplay.pp_funcs->get_fan_speed_percent((adev)->powerplay.pp_handle, (s)))
 
 #define amdgpu_dpm_get_fan_speed_rpm(adev, s) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_fan_speed_rpm((adev)->powerplay.pp_handle, (s)) : \
-	      -EINVAL)
+		((adev)->powerplay.pp_funcs->get_fan_speed_rpm)((adev)->powerplay.pp_handle, (s))
 
 #define amdgpu_dpm_get_sclk(adev, l) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (l)) : \
-		(adev)->pm.funcs->get_sclk((adev), (l)))
+		((adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (l)))
 
 #define amdgpu_dpm_get_mclk(adev, l)  \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->get_mclk((adev)->powerplay.pp_handle, (l)) : \
-	      (adev)->pm.funcs->get_mclk((adev), (l)))
-
+		((adev)->powerplay.pp_funcs->get_mclk((adev)->powerplay.pp_handle, (l)))
 
 #define amdgpu_dpm_force_performance_level(adev, l) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->force_performance_level((adev)->powerplay.pp_handle, (l)) : \
-	      (adev)->pm.funcs->force_performance_level((adev), (l)))
+		((adev)->powerplay.pp_funcs->force_performance_level((adev)->powerplay.pp_handle, (l)))
 
 #define amdgpu_dpm_powergate_uvd(adev, g) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->powergate_uvd((adev)->powerplay.pp_handle, (g)) : \
-	      (adev)->pm.funcs->powergate_uvd((adev), (g)))
+		((adev)->powerplay.pp_funcs->powergate_uvd((adev)->powerplay.pp_handle, (g)))
 
 #define amdgpu_dpm_powergate_vce(adev, g) \
-	((adev)->pp_enabled ?						\
-	      (adev)->powerplay.pp_funcs->powergate_vce((adev)->powerplay.pp_handle, (g)) : \
-	      (adev)->pm.funcs->powergate_vce((adev), (g)))
+		((adev)->powerplay.pp_funcs->powergate_vce((adev)->powerplay.pp_handle, (g)))
 
 #define amdgpu_dpm_get_current_power_state(adev) \
-	(adev)->powerplay.pp_funcs->get_current_power_state((adev)->powerplay.pp_handle)
+		((adev)->powerplay.pp_funcs->get_current_power_state((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_get_pp_num_states(adev, data) \
-	(adev)->powerplay.pp_funcs->get_pp_num_states((adev)->powerplay.pp_handle, data)
+		((adev)->powerplay.pp_funcs->get_pp_num_states((adev)->powerplay.pp_handle, data))
 
 #define amdgpu_dpm_get_pp_table(adev, table) \
-	(adev)->powerplay.pp_funcs->get_pp_table((adev)->powerplay.pp_handle, table)
+		((adev)->powerplay.pp_funcs->get_pp_table((adev)->powerplay.pp_handle, table))
 
 #define amdgpu_dpm_set_pp_table(adev, buf, size) \
-	(adev)->powerplay.pp_funcs->set_pp_table((adev)->powerplay.pp_handle, buf, size)
+		((adev)->powerplay.pp_funcs->set_pp_table((adev)->powerplay.pp_handle, buf, size))
 
 #define amdgpu_dpm_print_clock_levels(adev, type, buf) \
-	(adev)->powerplay.pp_funcs->print_clock_levels((adev)->powerplay.pp_handle, type, buf)
+		((adev)->powerplay.pp_funcs->print_clock_levels((adev)->powerplay.pp_handle, type, buf))
 
 #define amdgpu_dpm_force_clock_level(adev, type, level) \
-		(adev)->powerplay.pp_funcs->force_clock_level((adev)->powerplay.pp_handle, type, level)
+		((adev)->powerplay.pp_funcs->force_clock_level((adev)->powerplay.pp_handle, type, level))
 
 #define amdgpu_dpm_get_sclk_od(adev) \
-	(adev)->powerplay.pp_funcs->get_sclk_od((adev)->powerplay.pp_handle)
+		((adev)->powerplay.pp_funcs->get_sclk_od((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_set_sclk_od(adev, value) \
-	(adev)->powerplay.pp_funcs->set_sclk_od((adev)->powerplay.pp_handle, value)
+		((adev)->powerplay.pp_funcs->set_sclk_od((adev)->powerplay.pp_handle, value))
 
 #define amdgpu_dpm_get_mclk_od(adev) \
-	((adev)->powerplay.pp_funcs->get_mclk_od((adev)->powerplay.pp_handle))
+		((adev)->powerplay.pp_funcs->get_mclk_od((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_set_mclk_od(adev, value) \
-	((adev)->powerplay.pp_funcs->set_mclk_od((adev)->powerplay.pp_handle, value))
+		((adev)->powerplay.pp_funcs->set_mclk_od((adev)->powerplay.pp_handle, value))
 
-#define amdgpu_dpm_dispatch_task(adev, event_id, input, output)		\
-	(adev)->powerplay.pp_funcs->dispatch_tasks((adev)->powerplay.pp_handle, (event_id), (input), (output))
+#define amdgpu_dpm_dispatch_task(adev, task_id, input, output)		\
+		((adev)->powerplay.pp_funcs->dispatch_tasks)((adev)->powerplay.pp_handle, (task_id), (input), (output))
 
-#define amgdpu_dpm_check_state_equal(adev, cps, rps, equal) (adev)->pm.funcs->check_state_equal((adev), (cps),(rps),(equal))
+#define amdgpu_dpm_check_state_equal(adev, cps, rps, equal) \
+		((adev)->powerplay.pp_funcs->check_state_equal((adev)->powerplay.pp_handle, (cps), (rps), (equal)))
 
 #define amdgpu_dpm_get_vce_clock_state(adev, i)				\
-	((adev)->pp_enabled ?						\
-	 (adev)->powerplay.pp_funcs->get_vce_clock_state((adev)->powerplay.pp_handle, (i)) : \
-	 (adev)->pm.funcs->get_vce_clock_state((adev), (i)))
+		((adev)->powerplay.pp_funcs->get_vce_clock_state((adev)->powerplay.pp_handle, (i)))
 
-#define amdgpu_dpm_get_performance_level(adev) \
-	((adev)->pp_enabled ?						\
-	(adev)->powerplay.pp_funcs->get_performance_level((adev)->powerplay.pp_handle) : \
-	(adev)->pm.dpm.forced_level)
+#define amdgpu_dpm_get_performance_level(adev)				\
+		((adev)->powerplay.pp_funcs->get_performance_level((adev)->powerplay.pp_handle))
 
 #define amdgpu_dpm_reset_power_profile_state(adev, request) \
-	((adev)->powerplay.pp_funcs->reset_power_profile_state(\
+		((adev)->powerplay.pp_funcs->reset_power_profile_state(\
 			(adev)->powerplay.pp_handle, request))
 
 #define amdgpu_dpm_get_power_profile_state(adev, query) \
-	((adev)->powerplay.pp_funcs->get_power_profile_state(\
+		((adev)->powerplay.pp_funcs->get_power_profile_state(\
 			(adev)->powerplay.pp_handle, query))
 
 #define amdgpu_dpm_set_power_profile_state(adev, request) \
-	((adev)->powerplay.pp_funcs->set_power_profile_state(\
+		((adev)->powerplay.pp_funcs->set_power_profile_state(\
 			(adev)->powerplay.pp_handle, request))
 
 #define amdgpu_dpm_switch_power_profile(adev, type) \
-	((adev)->powerplay.pp_funcs->switch_power_profile(\
+		((adev)->powerplay.pp_funcs->switch_power_profile(\
 			(adev)->powerplay.pp_handle, type))
 
+#define amdgpu_dpm_set_clockgating_by_smu(adev, msg_id) \
+		((adev)->powerplay.pp_funcs->set_clockgating_by_smu(\
+			(adev)->powerplay.pp_handle, msg_id))
+
 struct amdgpu_dpm {
 	struct amdgpu_ps        *ps;
 	/* number of valid power states */
@@ -485,7 +431,6 @@ struct amdgpu_pm {
 	struct amdgpu_dpm       dpm;
 	const struct firmware	*fw;	/* SMC firmware */
 	uint32_t                fw_version;
-	const struct amdgpu_dpm_funcs *funcs;
 	uint32_t                pcie_gen_mask;
 	uint32_t                pcie_mlw_mask;
 	struct amd_pp_display_configuration pm_display_cfg;/* set by DAL */
@@ -551,6 +496,6 @@ u16 amdgpu_get_pcie_lane_support(struct amdgpu_device *adev,
 u8 amdgpu_encode_pci_lane_width(u32 lanes);
 
 struct amd_vce_state*
-amdgpu_get_vce_clock_state(struct amdgpu_device *adev, unsigned idx);
+amdgpu_get_vce_clock_state(void *handle, u32 idx);
 
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -69,9 +69,13 @@
  * - 3.17.0 - Add AMDGPU_NUM_VRAM_CPU_PAGE_FAULTS.
  * - 3.18.0 - Export gpu always on cu bitmap
  * - 3.19.0 - Add support for UVD MJPEG decode
+ * - 3.20.0 - Add support for local BOs
+ * - 3.21.0 - Add DRM_AMDGPU_FENCE_TO_HANDLE ioctl
+ * - 3.22.0 - Add DRM_AMDGPU_SCHED ioctl
+ * - 3.23.0 - Add query for VRAM lost counter
  */
 #define KMS_DRIVER_MAJOR	3
-#define KMS_DRIVER_MINOR	19
+#define KMS_DRIVER_MINOR	23
 #define KMS_DRIVER_PATCHLEVEL	0
 
 int amdgpu_vram_limit = 0;
@@ -91,7 +95,7 @@ int amdgpu_dpm = -1;
 int amdgpu_fw_load_type = -1;
 int amdgpu_aspm = -1;
 int amdgpu_runtime_pm = -1;
-unsigned amdgpu_ip_block_mask = 0xffffffff;
+uint amdgpu_ip_block_mask = 0xffffffff;
 int amdgpu_bapm = -1;
 int amdgpu_deep_color = 0;
 int amdgpu_vm_size = -1;
@@ -106,14 +110,14 @@ int amdgpu_sched_jobs = 32;
 int amdgpu_sched_hw_submission = 2;
 int amdgpu_no_evict = 0;
 int amdgpu_direct_gma_size = 0;
-unsigned amdgpu_pcie_gen_cap = 0;
-unsigned amdgpu_pcie_lane_cap = 0;
-unsigned amdgpu_cg_mask = 0xffffffff;
-unsigned amdgpu_pg_mask = 0xffffffff;
-unsigned amdgpu_sdma_phase_quantum = 32;
+uint amdgpu_pcie_gen_cap = 0;
+uint amdgpu_pcie_lane_cap = 0;
+uint amdgpu_cg_mask = 0xffffffff;
+uint amdgpu_pg_mask = 0xffffffff;
+uint amdgpu_sdma_phase_quantum = 32;
 char *amdgpu_disable_cu = NULL;
 char *amdgpu_virtual_display = NULL;
-unsigned amdgpu_pp_feature_mask = 0xffffffff;
+uint amdgpu_pp_feature_mask = 0xffffffff;
 int amdgpu_ngg = 0;
 int amdgpu_prim_buf_per_se = 0;
 int amdgpu_pos_buf_per_se = 0;
@@ -121,6 +125,7 @@ int amdgpu_cntl_sb_buf_per_se = 0;
 int amdgpu_param_buf_per_se = 0;
 int amdgpu_job_hang_limit = 0;
 int amdgpu_lbpw = -1;
+int amdgpu_compute_multipipe = -1;
 
 MODULE_PARM_DESC(vramlimit, "Restrict VRAM for testing, in megabytes");
 module_param_named(vramlimit, amdgpu_vram_limit, int, 0600);
@@ -264,6 +269,9 @@ module_param_named(job_hang_limit, amdgpu_job_hang_limit, int ,0444);
 MODULE_PARM_DESC(lbpw, "Load Balancing Per Watt (LBPW) support (1 = enable, 0 = disable, -1 = auto)");
 module_param_named(lbpw, amdgpu_lbpw, int, 0444);
 
+MODULE_PARM_DESC(compute_multipipe, "Force compute queues to be spread across pipes (1 = enable, 0 = disable, -1 = auto)");
+module_param_named(compute_multipipe, amdgpu_compute_multipipe, int, 0444);
+
 #ifdef CONFIG_DRM_AMDGPU_SI
 
 #if defined(CONFIG_DRM_RADEON) || defined(CONFIG_DRM_RADEON_MODULE)
@@ -608,6 +616,8 @@ amdgpu_pci_remove(struct pci_dev *pdev)
 
 	drm_dev_unregister(dev);
 	drm_dev_unref(dev);
+	pci_disable_device(pdev);
+	pci_set_drvdata(pdev, NULL);
 }
 
 static void
@@ -852,6 +862,7 @@ static struct drm_driver kms_driver = {
 	.gem_prime_import_sg_table = amdgpu_gem_prime_import_sg_table,
 	.gem_prime_vmap = amdgpu_gem_prime_vmap,
 	.gem_prime_vunmap = amdgpu_gem_prime_vunmap,
+	.gem_prime_mmap = amdgpu_gem_prime_mmap,
 
 	.name = DRIVER_NAME,
 	.desc = DRIVER_DESC,

drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c

@@ -149,7 +149,7 @@ static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev,
 				       AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
 				       AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
 				       AMDGPU_GEM_CREATE_VRAM_CLEARED,
-				       true, &gobj);
+				       true, NULL, &gobj);
 	if (ret) {
 		pr_err("failed to allocate framebuffer (%d)\n", aligned_size);
 		return -ENOMEM;
@@ -303,10 +303,10 @@ static int amdgpu_fbdev_destroy(struct drm_device *dev, struct amdgpu_fbdev *rfb
 	if (rfb->obj) {
 		amdgpufb_destroy_pinned_object(rfb->obj);
 		rfb->obj = NULL;
+		drm_framebuffer_unregister_private(&rfb->base);
+		drm_framebuffer_cleanup(&rfb->base);
 	}
 	drm_fb_helper_fini(&rfbdev->helper);
-	drm_framebuffer_unregister_private(&rfb->base);
-	drm_framebuffer_cleanup(&rfb->base);
 
 	return 0;
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c

@@ -168,6 +168,32 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f)
 	return 0;
 }
 
+/**
+ * amdgpu_fence_emit_polling - emit a fence on the requeste ring
+ *
+ * @ring: ring the fence is associated with
+ * @s: resulting sequence number
+ *
+ * Emits a fence command on the requested ring (all asics).
+ * Used For polling fence.
+ * Returns 0 on success, -ENOMEM on failure.
+ */
+int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s)
+{
+	uint32_t seq;
+
+	if (!s)
+		return -EINVAL;
+
+	seq = ++ring->fence_drv.sync_seq;
+	amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr,
+			       seq, AMDGPU_FENCE_FLAG_INT);
+
+	*s = seq;
+
+	return 0;
+}
+
 /**
  * amdgpu_fence_schedule_fallback - schedule fallback check
  *
@@ -281,6 +307,30 @@ int amdgpu_fence_wait_empty(struct amdgpu_ring *ring)
 	return r;
 }
 
+/**
+ * amdgpu_fence_wait_polling - busy wait for givn sequence number
+ *
+ * @ring: ring index the fence is associated with
+ * @wait_seq: sequence number to wait
+ * @timeout: the timeout for waiting in usecs
+ *
+ * Wait for all fences on the requested ring to signal (all asics).
+ * Returns left time if no timeout, 0 or minus if timeout.
+ */
+signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring,
+				      uint32_t wait_seq,
+				      signed long timeout)
+{
+	uint32_t seq;
+
+	do {
+		seq = amdgpu_fence_read(ring);
+		udelay(5);
+		timeout -= 5;
+	} while ((int32_t)(wait_seq - seq) > 0 && timeout > 0);
+
+	return timeout > 0 ? timeout : 0;
+}
 /**
  * amdgpu_fence_count_emitted - get the count of emitted fences
  *
@@ -641,6 +691,19 @@ static int amdgpu_debugfs_fence_info(struct seq_file *m, void *data)
 			   atomic_read(&ring->fence_drv.last_seq));
 		seq_printf(m, "Last emitted        0x%08x\n",
 			   ring->fence_drv.sync_seq);
+
+		if (ring->funcs->type != AMDGPU_RING_TYPE_GFX)
+			continue;
+
+		/* set in CP_VMID_PREEMPT and preemption occurred */
+		seq_printf(m, "Last preempted      0x%08x\n",
+			   le32_to_cpu(*(ring->fence_drv.cpu_addr + 2)));
+		/* set in CP_VMID_RESET and reset occurred */
+		seq_printf(m, "Last reset          0x%08x\n",
+			   le32_to_cpu(*(ring->fence_drv.cpu_addr + 4)));
+		/* Both preemption and reset occurred */
+		seq_printf(m, "Last both           0x%08x\n",
+			   le32_to_cpu(*(ring->fence_drv.cpu_addr + 6)));
 	}
 	return 0;
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c

@@ -332,12 +332,13 @@ int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset,
 		adev->gart.pages[p] = pagelist[i];
 #endif
 
-	if (adev->gart.ptr) {
-		r = amdgpu_gart_map(adev, offset, pages, dma_addr, flags,
-			    adev->gart.ptr);
-		if (r)
-			return r;
-	}
+	if (!adev->gart.ptr)
+		return 0;
+
+	r = amdgpu_gart_map(adev, offset, pages, dma_addr, flags,
+		    adev->gart.ptr);
+	if (r)
+		return r;
 
 	mb();
 	amdgpu_gart_flush_gpu_tlb(adev, 0);

drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c

@@ -44,11 +44,12 @@ void amdgpu_gem_object_free(struct drm_gem_object *gobj)
 }
 
 int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
-				int alignment, u32 initial_domain,
-				u64 flags, bool kernel,
-				struct drm_gem_object **obj)
+			     int alignment, u32 initial_domain,
+			     u64 flags, bool kernel,
+			     struct reservation_object *resv,
+			     struct drm_gem_object **obj)
 {
-	struct amdgpu_bo *robj;
+	struct amdgpu_bo *bo;
 	int r;
 
 	*obj = NULL;
@@ -59,7 +60,7 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 
 retry:
 	r = amdgpu_bo_create(adev, size, alignment, kernel, initial_domain,
-			     flags, NULL, NULL, 0, &robj);
+			     flags, NULL, resv, 0, &bo);
 	if (r) {
 		if (r != -ERESTARTSYS) {
 			if (initial_domain == AMDGPU_GEM_DOMAIN_VRAM) {
@@ -71,7 +72,7 @@ retry:
 		}
 		return r;
 	}
-	*obj = &robj->gem_base;
+	*obj = &bo->gem_base;
 
 	return 0;
 }
@@ -112,7 +113,17 @@ int amdgpu_gem_object_open(struct drm_gem_object *obj,
 	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
 	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_bo_va *bo_va;
+	struct mm_struct *mm;
 	int r;
+
+	mm = amdgpu_ttm_tt_get_usermm(abo->tbo.ttm);
+	if (mm && mm != current->mm)
+		return -EPERM;
+
+	if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID &&
+	    abo->tbo.resv != vm->root.base.bo->tbo.resv)
+		return -EPERM;
+
 	r = amdgpu_bo_reserve(abo, false);
 	if (r)
 		return r;
@@ -127,35 +138,6 @@ int amdgpu_gem_object_open(struct drm_gem_object *obj,
 	return 0;
 }
 
-static int amdgpu_gem_vm_check(void *param, struct amdgpu_bo *bo)
-{
-	/* if anything is swapped out don't swap it in here,
-	   just abort and wait for the next CS */
-	if (!amdgpu_bo_gpu_accessible(bo))
-		return -ERESTARTSYS;
-
-	if (bo->shadow && !amdgpu_bo_gpu_accessible(bo->shadow))
-		return -ERESTARTSYS;
-
-	return 0;
-}
-
-static bool amdgpu_gem_vm_ready(struct amdgpu_device *adev,
-				struct amdgpu_vm *vm,
-				struct list_head *list)
-{
-	struct ttm_validate_buffer *entry;
-
-	list_for_each_entry(entry, list, head) {
-		struct amdgpu_bo *bo =
-			container_of(entry->bo, struct amdgpu_bo, tbo);
-		if (amdgpu_gem_vm_check(NULL, bo))
-			return false;
-	}
-
-	return !amdgpu_vm_validate_pt_bos(adev, vm, amdgpu_gem_vm_check, NULL);
-}
-
 void amdgpu_gem_object_close(struct drm_gem_object *obj,
 			     struct drm_file *file_priv)
 {
@@ -165,13 +147,14 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj,
 	struct amdgpu_vm *vm = &fpriv->vm;
 
 	struct amdgpu_bo_list_entry vm_pd;
-	struct list_head list;
+	struct list_head list, duplicates;
 	struct ttm_validate_buffer tv;
 	struct ww_acquire_ctx ticket;
 	struct amdgpu_bo_va *bo_va;
 	int r;
 
 	INIT_LIST_HEAD(&list);
+	INIT_LIST_HEAD(&duplicates);
 
 	tv.bo = &bo->tbo;
 	tv.shared = true;
@@ -179,7 +162,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj,
 
 	amdgpu_vm_get_pd_bo(vm, &list, &vm_pd);
 
-	r = ttm_eu_reserve_buffers(&ticket, &list, false, NULL);
+	r = ttm_eu_reserve_buffers(&ticket, &list, false, &duplicates);
 	if (r) {
 		dev_err(adev->dev, "leaking bo va because "
 			"we fail to reserve bo (%d)\n", r);
@@ -189,7 +172,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj,
 	if (bo_va && --bo_va->ref_count == 0) {
 		amdgpu_vm_bo_rmv(adev, bo_va);
 
-		if (amdgpu_gem_vm_ready(adev, vm, &list)) {
+		if (amdgpu_vm_ready(vm)) {
 			struct dma_fence *fence = NULL;
 
 			r = amdgpu_vm_clear_freed(adev, vm, &fence);
@@ -214,18 +197,24 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 			    struct drm_file *filp)
 {
 	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_vm *vm = &fpriv->vm;
 	union drm_amdgpu_gem_create *args = data;
+	uint64_t flags = args->in.domain_flags;
 	uint64_t size = args->in.bo_size;
+	struct reservation_object *resv = NULL;
 	struct drm_gem_object *gobj;
 	uint32_t handle;
-	bool kernel = false;
 	int r;
 
 	/* reject invalid gem flags */
-	if (args->in.domain_flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
-				      AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
-				      AMDGPU_GEM_CREATE_CPU_GTT_USWC |
-				      AMDGPU_GEM_CREATE_VRAM_CLEARED))
+	if (flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
+		      AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
+		      AMDGPU_GEM_CREATE_CPU_GTT_USWC |
+		      AMDGPU_GEM_CREATE_VRAM_CLEARED |
+		      AMDGPU_GEM_CREATE_VM_ALWAYS_VALID |
+		      AMDGPU_GEM_CREATE_EXPLICIT_SYNC))
+
 		return -EINVAL;
 
 	/* reject invalid gem domains */
@@ -240,7 +229,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 	/* create a gem object to contain this object in */
 	if (args->in.domains & (AMDGPU_GEM_DOMAIN_GDS |
 	    AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) {
-		kernel = true;
+		flags |= AMDGPU_GEM_CREATE_NO_CPU_ACCESS;
 		if (args->in.domains == AMDGPU_GEM_DOMAIN_GDS)
 			size = size << AMDGPU_GDS_SHIFT;
 		else if (args->in.domains == AMDGPU_GEM_DOMAIN_GWS)
@@ -252,10 +241,25 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 	}
 	size = roundup(size, PAGE_SIZE);
 
+	if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) {
+		r = amdgpu_bo_reserve(vm->root.base.bo, false);
+		if (r)
+			return r;
+
+		resv = vm->root.base.bo->tbo.resv;
+	}
+
 	r = amdgpu_gem_object_create(adev, size, args->in.alignment,
 				     (u32)(0xffffffff & args->in.domains),
-				     args->in.domain_flags,
-				     kernel, &gobj);
+				     flags, false, resv, &gobj);
+	if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) {
+		if (!r) {
+			struct amdgpu_bo *abo = gem_to_amdgpu_bo(gobj);
+
+			abo->parent = amdgpu_bo_ref(vm->root.base.bo);
+		}
+		amdgpu_bo_unreserve(vm->root.base.bo);
+	}
 	if (r)
 		return r;
 
@@ -297,9 +301,8 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 	}
 
 	/* create a gem object to contain this object in */
-	r = amdgpu_gem_object_create(adev, args->size, 0,
-				     AMDGPU_GEM_DOMAIN_CPU, 0,
-				     0, &gobj);
+	r = amdgpu_gem_object_create(adev, args->size, 0, AMDGPU_GEM_DOMAIN_CPU,
+				     0, 0, NULL, &gobj);
 	if (r)
 		return r;
 
@@ -317,8 +320,6 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 	}
 
 	if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE) {
-		down_read(&current->mm->mmap_sem);
-
 		r = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm,
 						 bo->tbo.ttm->pages);
 		if (r)
@@ -333,8 +334,6 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 		amdgpu_bo_unreserve(bo);
 		if (r)
 			goto free_pages;
-
-		up_read(&current->mm->mmap_sem);
 	}
 
 	r = drm_gem_handle_create(filp, gobj, &handle);
@@ -511,10 +510,10 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 				    struct list_head *list,
 				    uint32_t operation)
 {
-	int r = -ERESTARTSYS;
+	int r;
 
-	if (!amdgpu_gem_vm_ready(adev, vm, list))
-		goto error;
+	if (!amdgpu_vm_ready(vm))
+		return;
 
 	r = amdgpu_vm_update_directories(adev, vm);
 	if (r)
@@ -551,7 +550,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	struct amdgpu_bo_list_entry vm_pd;
 	struct ttm_validate_buffer tv;
 	struct ww_acquire_ctx ticket;
-	struct list_head list;
+	struct list_head list, duplicates;
 	uint64_t va_flags;
 	int r = 0;
 
@@ -580,13 +579,9 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 			args->operation);
 		return -EINVAL;
 	}
-	if ((args->operation == AMDGPU_VA_OP_MAP) ||
-	    (args->operation == AMDGPU_VA_OP_REPLACE)) {
-		if (amdgpu_kms_vram_lost(adev, fpriv))
-			return -ENODEV;
-	}
 
 	INIT_LIST_HEAD(&list);
+	INIT_LIST_HEAD(&duplicates);
 	if ((args->operation != AMDGPU_VA_OP_CLEAR) &&
 	    !(args->flags & AMDGPU_VM_PAGE_PRT)) {
 		gobj = drm_gem_object_lookup(filp, args->handle);
@@ -603,7 +598,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 
 	amdgpu_vm_get_pd_bo(&fpriv->vm, &list, &vm_pd);
 
-	r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL);
+	r = ttm_eu_reserve_buffers(&ticket, &list, true, &duplicates);
 	if (r)
 		goto error_unref;
 
@@ -669,6 +664,7 @@ error_unref:
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp)
 {
+	struct amdgpu_device *adev = dev->dev_private;
 	struct drm_amdgpu_gem_op *args = data;
 	struct drm_gem_object *gobj;
 	struct amdgpu_bo *robj;
@@ -716,6 +712,9 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 		if (robj->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM)
 			robj->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT;
 
+		if (robj->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
+			amdgpu_vm_bo_invalidate(adev, robj, true);
+
 		amdgpu_bo_unreserve(robj);
 		break;
 	default:
@@ -745,8 +744,7 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 	r = amdgpu_gem_object_create(adev, args->size, 0,
 				     AMDGPU_GEM_DOMAIN_VRAM,
 				     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-				     ttm_bo_type_device,
-				     &gobj);
+				     false, NULL, &gobj);
 	if (r)
 		return -ENOMEM;
 

drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c

@@ -109,9 +109,26 @@ void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_s
 	}
 }
 
+static bool amdgpu_gfx_is_multipipe_capable(struct amdgpu_device *adev)
+{
+	if (amdgpu_compute_multipipe != -1) {
+		DRM_INFO("amdgpu: forcing compute pipe policy %d\n",
+			 amdgpu_compute_multipipe);
+		return amdgpu_compute_multipipe == 1;
+	}
+
+	/* FIXME: spreading the queues across pipes causes perf regressions
+	 * on POLARIS11 compute workloads */
+	if (adev->asic_type == CHIP_POLARIS11)
+		return false;
+
+	return adev->gfx.mec.num_mec > 1;
+}
+
 void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
 {
 	int i, queue, pipe, mec;
+	bool multipipe_policy = amdgpu_gfx_is_multipipe_capable(adev);
 
 	/* policy for amdgpu compute queue ownership */
 	for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) {
@@ -125,8 +142,7 @@ void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
 		if (mec >= adev->gfx.mec.num_mec)
 			break;
 
-		/* FIXME: spreading the queues across pipes causes perf regressions */
-		if (0) {
+		if (multipipe_policy) {
 			/* policy: amdgpu owns the first two queues of the first MEC */
 			if (mec == 0 && queue < 2)
 				set_bit(i, adev->gfx.mec.queue_bitmap);
@@ -185,7 +201,7 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
 	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
 	int r = 0;
 
-	mutex_init(&kiq->ring_mutex);
+	spin_lock_init(&kiq->ring_lock);
 
 	r = amdgpu_wb_get(adev, &adev->virt.reg_val_offs);
 	if (r)
@@ -260,8 +276,13 @@ int amdgpu_gfx_compute_mqd_sw_init(struct amdgpu_device *adev,
 	/* create MQD for KIQ */
 	ring = &adev->gfx.kiq.ring;
 	if (!ring->mqd_obj) {
+		/* originaly the KIQ MQD is put in GTT domain, but for SRIOV VRAM domain is a must
+		 * otherwise hypervisor trigger SAVE_VF fail after driver unloaded which mean MQD
+		 * deallocated and gart_unbind, to strict diverage we decide to use VRAM domain for
+		 * KIQ MQD no matter SRIOV or Bare-metal
+		 */
 		r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
-					    AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj,
+					    AMDGPU_GEM_DOMAIN_VRAM, &ring->mqd_obj,
 					    &ring->mqd_gpu_addr, &ring->mqd_ptr);
 		if (r) {
 			dev_warn(adev->dev, "failed to create ring mqd ob (%d)", r);

drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c

@@ -169,7 +169,8 @@ static int amdgpu_gtt_mgr_new(struct ttm_mem_type_manager *man,
 	int r;
 
 	spin_lock(&mgr->lock);
-	if (atomic64_read(&mgr->available) < mem->num_pages) {
+	if ((&tbo->mem == mem || tbo->mem.mem_type != TTM_PL_TT) &&
+	    atomic64_read(&mgr->available) < mem->num_pages) {
 		spin_unlock(&mgr->lock);
 		return 0;
 	}
@@ -244,8 +245,9 @@ static void amdgpu_gtt_mgr_del(struct ttm_mem_type_manager *man,
 uint64_t amdgpu_gtt_mgr_usage(struct ttm_mem_type_manager *man)
 {
 	struct amdgpu_gtt_mgr *mgr = man->priv;
+	s64 result = man->size - atomic64_read(&mgr->available);
 
-	return (u64)(man->size - atomic64_read(&mgr->available)) * PAGE_SIZE;
+	return (result > 0 ? result : 0) * PAGE_SIZE;
 }
 
 /**
@@ -265,7 +267,7 @@ static void amdgpu_gtt_mgr_debug(struct ttm_mem_type_manager *man,
 	drm_mm_print(&mgr->mm, printer);
 	spin_unlock(&mgr->lock);
 
-	drm_printf(printer, "man size:%llu pages, gtt available:%llu pages, usage:%lluMB\n",
+	drm_printf(printer, "man size:%llu pages, gtt available:%lld pages, usage:%lluMB\n",
 		   man->size, (u64)atomic64_read(&mgr->available),
 		   amdgpu_gtt_mgr_usage(man) >> 20);
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c

@@ -169,6 +169,12 @@ restart_ih:
 	while (adev->irq.ih.rptr != wptr) {
 		u32 ring_index = adev->irq.ih.rptr >> 2;
 
+		/* Prescreening of high-frequency interrupts */
+		if (!amdgpu_ih_prescreen_iv(adev)) {
+			adev->irq.ih.rptr &= adev->irq.ih.ptr_mask;
+			continue;
+		}
+
 		/* Before dispatching irq to IP blocks, send it to amdkfd */
 		amdgpu_amdkfd_interrupt(adev,
 				(const void *) &adev->irq.ih.ring[ring_index]);
@@ -190,3 +196,79 @@ restart_ih:
 
 	return IRQ_HANDLED;
 }
+
+/**
+ * amdgpu_ih_add_fault - Add a page fault record
+ *
+ * @adev: amdgpu device pointer
+ * @key: 64-bit encoding of PASID and address
+ *
+ * This should be called when a retry page fault interrupt is
+ * received. If this is a new page fault, it will be added to a hash
+ * table. The return value indicates whether this is a new fault, or
+ * a fault that was already known and is already being handled.
+ *
+ * If there are too many pending page faults, this will fail. Retry
+ * interrupts should be ignored in this case until there is enough
+ * free space.
+ *
+ * Returns 0 if the fault was added, 1 if the fault was already known,
+ * -ENOSPC if there are too many pending faults.
+ */
+int amdgpu_ih_add_fault(struct amdgpu_device *adev, u64 key)
+{
+	unsigned long flags;
+	int r = -ENOSPC;
+
+	if (WARN_ON_ONCE(!adev->irq.ih.faults))
+		/* Should be allocated in <IP>_ih_sw_init on GPUs that
+		 * support retry faults and require retry filtering.
+		 */
+		return r;
+
+	spin_lock_irqsave(&adev->irq.ih.faults->lock, flags);
+
+	/* Only let the hash table fill up to 50% for best performance */
+	if (adev->irq.ih.faults->count >= (1 << (AMDGPU_PAGEFAULT_HASH_BITS-1)))
+		goto unlock_out;
+
+	r = chash_table_copy_in(&adev->irq.ih.faults->hash, key, NULL);
+	if (!r)
+		adev->irq.ih.faults->count++;
+
+	/* chash_table_copy_in should never fail unless we're losing count */
+	WARN_ON_ONCE(r < 0);
+
+unlock_out:
+	spin_unlock_irqrestore(&adev->irq.ih.faults->lock, flags);
+	return r;
+}
+
+/**
+ * amdgpu_ih_clear_fault - Remove a page fault record
+ *
+ * @adev: amdgpu device pointer
+ * @key: 64-bit encoding of PASID and address
+ *
+ * This should be called when a page fault has been handled. Any
+ * future interrupt with this key will be processed as a new
+ * page fault.
+ */
+void amdgpu_ih_clear_fault(struct amdgpu_device *adev, u64 key)
+{
+	unsigned long flags;
+	int r;
+
+	if (!adev->irq.ih.faults)
+		return;
+
+	spin_lock_irqsave(&adev->irq.ih.faults->lock, flags);
+
+	r = chash_table_remove(&adev->irq.ih.faults->hash, key, NULL);
+	if (!WARN_ON_ONCE(r < 0)) {
+		adev->irq.ih.faults->count--;
+		WARN_ON_ONCE(adev->irq.ih.faults->count < 0);
+	}
+
+	spin_unlock_irqrestore(&adev->irq.ih.faults->lock, flags);
+}

drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h

@@ -24,6 +24,8 @@
 #ifndef __AMDGPU_IH_H__
 #define __AMDGPU_IH_H__
 
+#include <linux/chash.h>
+
 struct amdgpu_device;
  /*
   * vega10+ IH clients
@@ -69,6 +71,13 @@ enum amdgpu_ih_clientid
 
 #define AMDGPU_IH_CLIENTID_LEGACY 0
 
+#define AMDGPU_PAGEFAULT_HASH_BITS 8
+struct amdgpu_retryfault_hashtable {
+	DECLARE_CHASH_TABLE(hash, AMDGPU_PAGEFAULT_HASH_BITS, 8, 0);
+	spinlock_t	lock;
+	int		count;
+};
+
 /*
  * R6xx+ IH ring
  */
@@ -87,6 +96,7 @@ struct amdgpu_ih_ring {
 	bool			use_doorbell;
 	bool			use_bus_addr;
 	dma_addr_t		rb_dma_addr; /* only used when use_bus_addr = true */
+	struct amdgpu_retryfault_hashtable *faults;
 };
 
 #define AMDGPU_IH_SRC_DATA_MAX_SIZE_DW 4
@@ -109,5 +119,7 @@ int amdgpu_ih_ring_init(struct amdgpu_device *adev, unsigned ring_size,
 			bool use_bus_addr);
 void amdgpu_ih_ring_fini(struct amdgpu_device *adev);
 int amdgpu_ih_process(struct amdgpu_device *adev);
+int amdgpu_ih_add_fault(struct amdgpu_device *adev, u64 key);
+void amdgpu_ih_clear_fault(struct amdgpu_device *adev, u64 key);
 
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_job.c

@@ -65,6 +65,7 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs,
 	amdgpu_sync_create(&(*job)->sync);
 	amdgpu_sync_create(&(*job)->dep_sync);
 	amdgpu_sync_create(&(*job)->sched_sync);
+	(*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter);
 
 	return 0;
 }
@@ -103,6 +104,7 @@ static void amdgpu_job_free_cb(struct amd_sched_job *s_job)
 {
 	struct amdgpu_job *job = container_of(s_job, struct amdgpu_job, base);
 
+	amdgpu_ring_priority_put(job->ring, amd_sched_get_job_priority(s_job));
 	dma_fence_put(job->fence);
 	amdgpu_sync_free(&job->sync);
 	amdgpu_sync_free(&job->dep_sync);
@@ -139,6 +141,8 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring,
 	job->fence_ctx = entity->fence_context;
 	*f = dma_fence_get(&job->base.s_fence->finished);
 	amdgpu_job_free_resources(job);
+	amdgpu_ring_priority_get(job->ring,
+				 amd_sched_get_job_priority(&job->base));
 	amd_sched_entity_push_job(&job->base);
 
 	return 0;
@@ -177,8 +181,8 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job)
 static struct dma_fence *amdgpu_job_run(struct amd_sched_job *sched_job)
 {
 	struct dma_fence *fence = NULL;
+	struct amdgpu_device *adev;
 	struct amdgpu_job *job;
-	struct amdgpu_fpriv *fpriv = NULL;
 	int r;
 
 	if (!sched_job) {
@@ -186,23 +190,25 @@ static struct dma_fence *amdgpu_job_run(struct amd_sched_job *sched_job)
 		return NULL;
 	}
 	job = to_amdgpu_job(sched_job);
+	adev = job->adev;
 
 	BUG_ON(amdgpu_sync_peek_fence(&job->sync, NULL));
 
 	trace_amdgpu_sched_run_job(job);
-	if (job->vm)
-		fpriv = container_of(job->vm, struct amdgpu_fpriv, vm);
 	/* skip ib schedule when vram is lost */
-	if (fpriv && amdgpu_kms_vram_lost(job->adev, fpriv))
+	if (job->vram_lost_counter != atomic_read(&adev->vram_lost_counter)) {
+		dma_fence_set_error(&job->base.s_fence->finished, -ECANCELED);
 		DRM_ERROR("Skip scheduling IBs!\n");
-	else {
-		r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job, &fence);
+	} else {
+		r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job,
+				       &fence);
 		if (r)
 			DRM_ERROR("Error scheduling IBs (%d)\n", r);
 	}
 	/* if gpu reset, hw fence will be replaced here */
 	dma_fence_put(job->fence);
 	job->fence = dma_fence_get(fence);
+
 	amdgpu_job_free_resources(job);
 	return fence;
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c

@@ -28,6 +28,7 @@
 #include <drm/drmP.h>
 #include "amdgpu.h"
 #include <drm/amdgpu_drm.h>
+#include "amdgpu_sched.h"
 #include "amdgpu_uvd.h"
 #include "amdgpu_vce.h"
 
@@ -269,7 +270,6 @@ static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info,
 static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 {
 	struct amdgpu_device *adev = dev->dev_private;
-	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	struct drm_amdgpu_info *info = data;
 	struct amdgpu_mode_info *minfo = &adev->mode_info;
 	void __user *out = (void __user *)(uintptr_t)info->return_pointer;
@@ -282,8 +282,6 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
 
 	if (!info->return_size || !info->return_pointer)
 		return -EINVAL;
-	if (amdgpu_kms_vram_lost(adev, fpriv))
-		return -ENODEV;
 
 	switch (info->query) {
 	case AMDGPU_INFO_ACCEL_WORKING:
@@ -765,6 +763,9 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
 		}
 		return copy_to_user(out, &ui32, min(size, 4u)) ? -EFAULT : 0;
 	}
+	case AMDGPU_INFO_VRAM_LOST_COUNTER:
+		ui32 = atomic_read(&adev->vram_lost_counter);
+		return copy_to_user(out, &ui32, min(size, 4u)) ? -EFAULT : 0;
 	default:
 		DRM_DEBUG_KMS("Invalid request %d\n", info->query);
 		return -EINVAL;
@@ -791,12 +792,6 @@ void amdgpu_driver_lastclose_kms(struct drm_device *dev)
 	vga_switcheroo_process_delayed_switch();
 }
 
-bool amdgpu_kms_vram_lost(struct amdgpu_device *adev,
-			  struct amdgpu_fpriv *fpriv)
-{
-	return fpriv->vram_lost_counter != atomic_read(&adev->vram_lost_counter);
-}
-
 /**
  * amdgpu_driver_open_kms - drm callback for open
  *
@@ -825,7 +820,7 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 	}
 
 	r = amdgpu_vm_init(adev, &fpriv->vm,
-			   AMDGPU_VM_CONTEXT_GFX);
+			   AMDGPU_VM_CONTEXT_GFX, 0);
 	if (r) {
 		kfree(fpriv);
 		goto out_suspend;
@@ -841,8 +836,11 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 
 	if (amdgpu_sriov_vf(adev)) {
 		r = amdgpu_map_static_csa(adev, &fpriv->vm, &fpriv->csa_va);
-		if (r)
+		if (r) {
+			amdgpu_vm_fini(adev, &fpriv->vm);
+			kfree(fpriv);
 			goto out_suspend;
+		}
 	}
 
 	mutex_init(&fpriv->bo_list_lock);
@@ -850,7 +848,6 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 
 	amdgpu_ctx_mgr_init(&fpriv->ctx_mgr);
 
-	fpriv->vram_lost_counter = atomic_read(&adev->vram_lost_counter);
 	file_priv->driver_priv = fpriv;
 
 out_suspend:
@@ -1020,7 +1017,9 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_CREATE, amdgpu_gem_create_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_CTX, amdgpu_ctx_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_VM, amdgpu_vm_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_SCHED, amdgpu_sched_ioctl, DRM_MASTER),
 	DRM_IOCTL_DEF_DRV(AMDGPU_BO_LIST, amdgpu_bo_list_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_FENCE_TO_HANDLE, amdgpu_cs_fence_to_handle_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	/* KMS */
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_MMAP, amdgpu_gem_mmap_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_WAIT_IDLE, amdgpu_gem_wait_idle_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),

drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c

@@ -50,8 +50,10 @@ struct amdgpu_mn {
 	struct hlist_node	node;
 
 	/* objects protected by lock */
-	struct mutex		lock;
+	struct rw_semaphore	lock;
 	struct rb_root_cached	objects;
+	struct mutex		read_lock;
+	atomic_t		recursion;
 };
 
 struct amdgpu_mn_node {
@@ -74,7 +76,7 @@ static void amdgpu_mn_destroy(struct work_struct *work)
 	struct amdgpu_bo *bo, *next_bo;
 
 	mutex_lock(&adev->mn_lock);
-	mutex_lock(&rmn->lock);
+	down_write(&rmn->lock);
 	hash_del(&rmn->node);
 	rbtree_postorder_for_each_entry_safe(node, next_node,
 					     &rmn->objects.rb_root, it.rb) {
@@ -84,7 +86,7 @@ static void amdgpu_mn_destroy(struct work_struct *work)
 		}
 		kfree(node);
 	}
-	mutex_unlock(&rmn->lock);
+	up_write(&rmn->lock);
 	mutex_unlock(&adev->mn_lock);
 	mmu_notifier_unregister_no_release(&rmn->mn, rmn->mm);
 	kfree(rmn);
@@ -106,6 +108,53 @@ static void amdgpu_mn_release(struct mmu_notifier *mn,
 	schedule_work(&rmn->work);
 }
 
+
+/**
+ * amdgpu_mn_lock - take the write side lock for this mn
+ */
+void amdgpu_mn_lock(struct amdgpu_mn *mn)
+{
+	if (mn)
+		down_write(&mn->lock);
+}
+
+/**
+ * amdgpu_mn_unlock - drop the write side lock for this mn
+ */
+void amdgpu_mn_unlock(struct amdgpu_mn *mn)
+{
+	if (mn)
+		up_write(&mn->lock);
+}
+
+/**
+ * amdgpu_mn_read_lock - take the rmn read lock
+ *
+ * @rmn: our notifier
+ *
+ * Take the rmn read side lock.
+ */
+static void amdgpu_mn_read_lock(struct amdgpu_mn *rmn)
+{
+	mutex_lock(&rmn->read_lock);
+	if (atomic_inc_return(&rmn->recursion) == 1)
+		down_read_non_owner(&rmn->lock);
+	mutex_unlock(&rmn->read_lock);
+}
+
+/**
+ * amdgpu_mn_read_unlock - drop the rmn read lock
+ *
+ * @rmn: our notifier
+ *
+ * Drop the rmn read side lock.
+ */
+static void amdgpu_mn_read_unlock(struct amdgpu_mn *rmn)
+{
+	if (atomic_dec_return(&rmn->recursion) == 0)
+		up_read_non_owner(&rmn->lock);
+}
+
 /**
  * amdgpu_mn_invalidate_node - unmap all BOs of a node
  *
@@ -126,23 +175,12 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node,
 		if (!amdgpu_ttm_tt_affect_userptr(bo->tbo.ttm, start, end))
 			continue;
 
-		r = amdgpu_bo_reserve(bo, true);
-		if (r) {
-			DRM_ERROR("(%ld) failed to reserve user bo\n", r);
-			continue;
-		}
-
 		r = reservation_object_wait_timeout_rcu(bo->tbo.resv,
 			true, false, MAX_SCHEDULE_TIMEOUT);
 		if (r <= 0)
 			DRM_ERROR("(%ld) failed to wait for user bo\n", r);
 
-		amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
-		r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false);
-		if (r)
-			DRM_ERROR("(%ld) failed to validate user bo\n", r);
-
-		amdgpu_bo_unreserve(bo);
+		amdgpu_ttm_tt_mark_user_pages(bo->tbo.ttm);
 	}
 }
 
@@ -168,7 +206,7 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn,
 	/* notification is exclusive, but interval is inclusive */
 	end -= 1;
 
-	mutex_lock(&rmn->lock);
+	amdgpu_mn_read_lock(rmn);
 
 	it = interval_tree_iter_first(&rmn->objects, start, end);
 	while (it) {
@@ -179,13 +217,32 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn,
 
 		amdgpu_mn_invalidate_node(node, start, end);
 	}
+}
 
-	mutex_unlock(&rmn->lock);
+/**
+ * amdgpu_mn_invalidate_range_end - callback to notify about mm change
+ *
+ * @mn: our notifier
+ * @mn: the mm this callback is about
+ * @start: start of updated range
+ * @end: end of updated range
+ *
+ * Release the lock again to allow new command submissions.
+ */
+static void amdgpu_mn_invalidate_range_end(struct mmu_notifier *mn,
+					   struct mm_struct *mm,
+					   unsigned long start,
+					   unsigned long end)
+{
+	struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn);
+
+	amdgpu_mn_read_unlock(rmn);
 }
 
 static const struct mmu_notifier_ops amdgpu_mn_ops = {
 	.release = amdgpu_mn_release,
 	.invalidate_range_start = amdgpu_mn_invalidate_range_start,
+	.invalidate_range_end = amdgpu_mn_invalidate_range_end,
 };
 
 /**
@@ -195,7 +252,7 @@ static const struct mmu_notifier_ops amdgpu_mn_ops = {
  *
  * Creates a notifier context for current->mm.
  */
-static struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
+struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
 {
 	struct mm_struct *mm = current->mm;
 	struct amdgpu_mn *rmn;
@@ -220,8 +277,10 @@ static struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
 	rmn->adev = adev;
 	rmn->mm = mm;
 	rmn->mn.ops = &amdgpu_mn_ops;
-	mutex_init(&rmn->lock);
+	init_rwsem(&rmn->lock);
 	rmn->objects = RB_ROOT_CACHED;
+	mutex_init(&rmn->read_lock);
+	atomic_set(&rmn->recursion, 0);
 
 	r = __mmu_notifier_register(&rmn->mn, mm);
 	if (r)
@@ -267,7 +326,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 
 	INIT_LIST_HEAD(&bos);
 
-	mutex_lock(&rmn->lock);
+	down_write(&rmn->lock);
 
 	while ((it = interval_tree_iter_first(&rmn->objects, addr, end))) {
 		kfree(node);
@@ -281,7 +340,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 	if (!node) {
 		node = kmalloc(sizeof(struct amdgpu_mn_node), GFP_KERNEL);
 		if (!node) {
-			mutex_unlock(&rmn->lock);
+			up_write(&rmn->lock);
 			return -ENOMEM;
 		}
 	}
@@ -296,7 +355,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 
 	interval_tree_insert(&node->it, &rmn->objects);
 
-	mutex_unlock(&rmn->lock);
+	up_write(&rmn->lock);
 
 	return 0;
 }
@@ -322,7 +381,7 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo)
 		return;
 	}
 
-	mutex_lock(&rmn->lock);
+	down_write(&rmn->lock);
 
 	/* save the next list entry for later */
 	head = bo->mn_list.next;
@@ -337,6 +396,7 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo)
 		kfree(node);
 	}
 
-	mutex_unlock(&rmn->lock);
+	up_write(&rmn->lock);
 	mutex_unlock(&adev->mn_lock);
 }
+

drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h

@@ -0,0 +1,52 @@
+/*
+ * Copyright 2017 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Christian König
+ */
+#ifndef __AMDGPU_MN_H__
+#define __AMDGPU_MN_H__
+
+/*
+ * MMU Notifier
+ */
+struct amdgpu_mn;
+
+#if defined(CONFIG_MMU_NOTIFIER)
+void amdgpu_mn_lock(struct amdgpu_mn *mn);
+void amdgpu_mn_unlock(struct amdgpu_mn *mn);
+struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev);
+int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr);
+void amdgpu_mn_unregister(struct amdgpu_bo *bo);
+#else
+static inline void amdgpu_mn_lock(struct amdgpu_mn *mn) {}
+static inline void amdgpu_mn_unlock(struct amdgpu_mn *mn) {}
+static inline struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
+{
+	return NULL;
+}
+static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
+{
+	return -ENODEV;
+}
+static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {}
+#endif
+
+#endif

drivers/gpu/drm/amd/amdgpu/amdgpu_object.c

@@ -40,9 +40,7 @@
 static void amdgpu_ttm_bo_destroy(struct ttm_buffer_object *tbo)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev);
-	struct amdgpu_bo *bo;
-
-	bo = container_of(tbo, struct amdgpu_bo, tbo);
+	struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo);
 
 	amdgpu_bo_kunmap(bo);
 
@@ -64,11 +62,12 @@ bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo)
 	return false;
 }
 
-static void amdgpu_ttm_placement_init(struct amdgpu_device *adev,
-				      struct ttm_placement *placement,
-				      struct ttm_place *places,
-				      u32 domain, u64 flags)
+void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 {
+	struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev);
+	struct ttm_placement *placement = &abo->placement;
+	struct ttm_place *places = abo->placements;
+	u64 flags = abo->flags;
 	u32 c = 0;
 
 	if (domain & AMDGPU_GEM_DOMAIN_VRAM) {
@@ -151,27 +150,6 @@ static void amdgpu_ttm_placement_init(struct amdgpu_device *adev,
 	placement->busy_placement = places;
 }
 
-void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
-{
-	struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev);
-
-	amdgpu_ttm_placement_init(adev, &abo->placement, abo->placements,
-				  domain, abo->flags);
-}
-
-static void amdgpu_fill_placement_to_bo(struct amdgpu_bo *bo,
-					struct ttm_placement *placement)
-{
-	BUG_ON(placement->num_placement > (AMDGPU_GEM_DOMAIN_MAX + 1));
-
-	memcpy(bo->placements, placement->placement,
-	       placement->num_placement * sizeof(struct ttm_place));
-	bo->placement.num_placement = placement->num_placement;
-	bo->placement.num_busy_placement = placement->num_busy_placement;
-	bo->placement.placement = bo->placements;
-	bo->placement.busy_placement = bo->placements;
-}
-
 /**
  * amdgpu_bo_create_reserved - create reserved BO for kernel use
  *
@@ -303,14 +281,13 @@ void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr,
 		*cpu_addr = NULL;
 }
 
-int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
-				unsigned long size, int byte_align,
-				bool kernel, u32 domain, u64 flags,
-				struct sg_table *sg,
-				struct ttm_placement *placement,
-				struct reservation_object *resv,
-				uint64_t init_value,
-				struct amdgpu_bo **bo_ptr)
+static int amdgpu_bo_do_create(struct amdgpu_device *adev,
+			       unsigned long size, int byte_align,
+			       bool kernel, u32 domain, u64 flags,
+			       struct sg_table *sg,
+			       struct reservation_object *resv,
+			       uint64_t init_value,
+			       struct amdgpu_bo **bo_ptr)
 {
 	struct amdgpu_bo *bo;
 	enum ttm_bo_type type;
@@ -384,13 +361,17 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
 		bo->flags &= ~AMDGPU_GEM_CREATE_CPU_GTT_USWC;
 #endif
 
-	amdgpu_fill_placement_to_bo(bo, placement);
-	/* Kernel allocation are uninterruptible */
+	bo->tbo.bdev = &adev->mman.bdev;
+	amdgpu_ttm_placement_from_domain(bo, domain);
 
 	initial_bytes_moved = atomic64_read(&adev->num_bytes_moved);
+	/* Kernel allocation are uninterruptible */
 	r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type,
 				 &bo->placement, page_align, !kernel, NULL,
 				 acc_size, sg, resv, &amdgpu_ttm_bo_destroy);
+	if (unlikely(r != 0))
+		return r;
+
 	bytes_moved = atomic64_read(&adev->num_bytes_moved) -
 		      initial_bytes_moved;
 	if (adev->mc.visible_vram_size < adev->mc.real_vram_size &&
@@ -400,9 +381,6 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
 	else
 		amdgpu_cs_report_moved_bytes(adev, bytes_moved, 0);
 
-	if (unlikely(r != 0))
-		return r;
-
 	if (kernel)
 		bo->tbo.priority = 1;
 
@@ -442,27 +420,17 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev,
 				   unsigned long size, int byte_align,
 				   struct amdgpu_bo *bo)
 {
-	struct ttm_placement placement = {0};
-	struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1];
 	int r;
 
 	if (bo->shadow)
 		return 0;
 
-	memset(&placements, 0, sizeof(placements));
-	amdgpu_ttm_placement_init(adev, &placement, placements,
-				  AMDGPU_GEM_DOMAIN_GTT,
-				  AMDGPU_GEM_CREATE_CPU_GTT_USWC |
-				  AMDGPU_GEM_CREATE_SHADOW);
-
-	r = amdgpu_bo_create_restricted(adev, size, byte_align, true,
-					AMDGPU_GEM_DOMAIN_GTT,
-					AMDGPU_GEM_CREATE_CPU_GTT_USWC |
-					AMDGPU_GEM_CREATE_SHADOW,
-					NULL, &placement,
-					bo->tbo.resv,
-					0,
-					&bo->shadow);
+	r = amdgpu_bo_do_create(adev, size, byte_align, true,
+				AMDGPU_GEM_DOMAIN_GTT,
+				AMDGPU_GEM_CREATE_CPU_GTT_USWC |
+				AMDGPU_GEM_CREATE_SHADOW,
+				NULL, bo->tbo.resv, 0,
+				&bo->shadow);
 	if (!r) {
 		bo->shadow->parent = amdgpu_bo_ref(bo);
 		mutex_lock(&adev->shadow_list_lock);
@@ -484,18 +452,11 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 		     uint64_t init_value,
 		     struct amdgpu_bo **bo_ptr)
 {
-	struct ttm_placement placement = {0};
-	struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1];
 	uint64_t parent_flags = flags & ~AMDGPU_GEM_CREATE_SHADOW;
 	int r;
 
-	memset(&placements, 0, sizeof(placements));
-	amdgpu_ttm_placement_init(adev, &placement, placements,
-				  domain, parent_flags);
-
-	r = amdgpu_bo_create_restricted(adev, size, byte_align, kernel, domain,
-					parent_flags, sg, &placement, resv,
-					init_value, bo_ptr);
+	r = amdgpu_bo_do_create(adev, size, byte_align, kernel, domain,
+				parent_flags, sg, resv, init_value, bo_ptr);
 	if (r)
 		return r;
 
@@ -672,7 +633,6 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	int r, i;
-	unsigned fpfn, lpfn;
 
 	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm))
 		return -EPERM;
@@ -704,22 +664,16 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 	}
 
 	bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+	/* force to pin into visible video ram */
+	if (!(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS))
+		bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 	amdgpu_ttm_placement_from_domain(bo, domain);
 	for (i = 0; i < bo->placement.num_placement; i++) {
-		/* force to pin into visible video ram */
-		if ((bo->placements[i].flags & TTM_PL_FLAG_VRAM) &&
-		    !(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) &&
-		    (!max_offset || max_offset >
-		     adev->mc.visible_vram_size)) {
-			if (WARN_ON_ONCE(min_offset >
-					 adev->mc.visible_vram_size))
-				return -EINVAL;
-			fpfn = min_offset >> PAGE_SHIFT;
-			lpfn = adev->mc.visible_vram_size >> PAGE_SHIFT;
-		} else {
-			fpfn = min_offset >> PAGE_SHIFT;
-			lpfn = max_offset >> PAGE_SHIFT;
-		}
+		unsigned fpfn, lpfn;
+
+		fpfn = min_offset >> PAGE_SHIFT;
+		lpfn = max_offset >> PAGE_SHIFT;
+
 		if (fpfn > bo->placements[i].fpfn)
 			bo->placements[i].fpfn = fpfn;
 		if (!bo->placements[i].lpfn ||
@@ -928,8 +882,8 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,
 	if (!amdgpu_ttm_bo_is_amdgpu_bo(bo))
 		return;
 
-	abo = container_of(bo, struct amdgpu_bo, tbo);
-	amdgpu_vm_bo_invalidate(adev, abo);
+	abo = ttm_to_amdgpu_bo(bo);
+	amdgpu_vm_bo_invalidate(adev, abo, evict);
 
 	amdgpu_bo_kunmap(abo);
 
@@ -955,7 +909,7 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
 	if (!amdgpu_ttm_bo_is_amdgpu_bo(bo))
 		return 0;
 
-	abo = container_of(bo, struct amdgpu_bo, tbo);
+	abo = ttm_to_amdgpu_bo(bo);
 
 	/* Remember that this BO was accessed by the CPU */
 	abo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;

drivers/gpu/drm/amd/amdgpu/amdgpu_object.h

@@ -35,6 +35,7 @@
 
 /* bo virtual addresses in a vm */
 struct amdgpu_bo_va_mapping {
+	struct amdgpu_bo_va		*bo_va;
 	struct list_head		list;
 	struct rb_node			rb;
 	uint64_t			start;
@@ -49,12 +50,17 @@ struct amdgpu_bo_va {
 	struct amdgpu_vm_bo_base	base;
 
 	/* protected by bo being reserved */
-	struct dma_fence	        *last_pt_update;
 	unsigned			ref_count;
 
+	/* all other members protected by the VM PD being reserved */
+	struct dma_fence	        *last_pt_update;
+
 	/* mappings for this bo_va */
 	struct list_head		invalids;
 	struct list_head		valids;
+
+	/* If the mappings are cleared or filled */
+	bool				cleared;
 };
 
 struct amdgpu_bo {
@@ -88,6 +94,11 @@ struct amdgpu_bo {
 	};
 };
 
+static inline struct amdgpu_bo *ttm_to_amdgpu_bo(struct ttm_buffer_object *tbo)
+{
+	return container_of(tbo, struct amdgpu_bo, tbo);
+}
+
 /**
  * amdgpu_mem_type_to_domain - return domain corresponding to mem_type
  * @mem_type:	ttm memory type
@@ -182,6 +193,14 @@ static inline bool amdgpu_bo_gpu_accessible(struct amdgpu_bo *bo)
 	}
 }
 
+/**
+ * amdgpu_bo_explicit_sync - return whether the bo is explicitly synced
+ */
+static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo)
+{
+	return bo->flags & AMDGPU_GEM_CREATE_EXPLICIT_SYNC;
+}
+
 int amdgpu_bo_create(struct amdgpu_device *adev,
 			    unsigned long size, int byte_align,
 			    bool kernel, u32 domain, u64 flags,
@@ -189,14 +208,6 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 			    struct reservation_object *resv,
 			    uint64_t init_value,
 			    struct amdgpu_bo **bo_ptr);
-int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
-				unsigned long size, int byte_align,
-				bool kernel, u32 domain, u64 flags,
-				struct sg_table *sg,
-				struct ttm_placement *placement,
-			        struct reservation_object *resv,
-				uint64_t init_value,
-				struct amdgpu_bo **bo_ptr);
 int amdgpu_bo_create_reserved(struct amdgpu_device *adev,
 			      unsigned long size, int align,
 			      u32 domain, struct amdgpu_bo **bo_ptr,

drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c

@@ -64,17 +64,13 @@ static const struct cg_flag_name clocks[] = {
 
 void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev)
 {
-	if (adev->pp_enabled)
-		/* TODO */
-		return;
-
 	if (adev->pm.dpm_enabled) {
 		mutex_lock(&adev->pm.mutex);
 		if (power_supply_is_system_supplied() > 0)
 			adev->pm.dpm.ac_power = true;
 		else
 			adev->pm.dpm.ac_power = false;
-		if (adev->pm.funcs->enable_bapm)
+		if (adev->powerplay.pp_funcs->enable_bapm)
 			amdgpu_dpm_enable_bapm(adev, adev->pm.dpm.ac_power);
 		mutex_unlock(&adev->pm.mutex);
 	}
@@ -88,9 +84,9 @@ static ssize_t amdgpu_get_dpm_state(struct device *dev,
 	struct amdgpu_device *adev = ddev->dev_private;
 	enum amd_pm_state_type pm;
 
-	if (adev->pp_enabled) {
+	if (adev->powerplay.pp_funcs->get_current_power_state)
 		pm = amdgpu_dpm_get_current_power_state(adev);
-	} else
+	else
 		pm = adev->pm.dpm.user_state;
 
 	return snprintf(buf, PAGE_SIZE, "%s\n",
@@ -118,8 +114,8 @@ static ssize_t amdgpu_set_dpm_state(struct device *dev,
 		goto fail;
 	}
 
-	if (adev->pp_enabled) {
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_ENABLE_USER_STATE, &state, NULL);
+	if (adev->powerplay.pp_funcs->dispatch_tasks) {
+		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_ENABLE_USER_STATE, &state, NULL);
 	} else {
 		mutex_lock(&adev->pm.mutex);
 		adev->pm.dpm.user_state = state;
@@ -140,13 +136,17 @@ static ssize_t amdgpu_get_dpm_forced_performance_level(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	enum amd_dpm_forced_level level;
+	enum amd_dpm_forced_level level = 0xff;
 
 	if  ((adev->flags & AMD_IS_PX) &&
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return snprintf(buf, PAGE_SIZE, "off\n");
 
-	level = amdgpu_dpm_get_performance_level(adev);
+	if (adev->powerplay.pp_funcs->get_performance_level)
+		level = amdgpu_dpm_get_performance_level(adev);
+	else
+		level = adev->pm.dpm.forced_level;
+
 	return snprintf(buf, PAGE_SIZE, "%s\n",
 			(level == AMD_DPM_FORCED_LEVEL_AUTO) ? "auto" :
 			(level == AMD_DPM_FORCED_LEVEL_LOW) ? "low" :
@@ -167,7 +167,7 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
 	enum amd_dpm_forced_level level;
-	enum amd_dpm_forced_level current_level;
+	enum amd_dpm_forced_level current_level = 0xff;
 	int ret = 0;
 
 	/* Can't force performance level when the card is off */
@@ -175,7 +175,8 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	current_level = amdgpu_dpm_get_performance_level(adev);
+	if (adev->powerplay.pp_funcs->get_performance_level)
+		current_level = amdgpu_dpm_get_performance_level(adev);
 
 	if (strncmp("low", buf, strlen("low")) == 0) {
 		level = AMD_DPM_FORCED_LEVEL_LOW;
@@ -203,9 +204,7 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
 	if (current_level == level)
 		return count;
 
-	if (adev->pp_enabled)
-		amdgpu_dpm_force_performance_level(adev, level);
-	else {
+	if (adev->powerplay.pp_funcs->force_performance_level) {
 		mutex_lock(&adev->pm.mutex);
 		if (adev->pm.dpm.thermal_active) {
 			count = -EINVAL;
@@ -233,7 +232,7 @@ static ssize_t amdgpu_get_pp_num_states(struct device *dev,
 	struct pp_states_info data;
 	int i, buf_len;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->get_pp_num_states)
 		amdgpu_dpm_get_pp_num_states(adev, &data);
 
 	buf_len = snprintf(buf, PAGE_SIZE, "states: %d\n", data.nums);
@@ -257,8 +256,8 @@ static ssize_t amdgpu_get_pp_cur_state(struct device *dev,
 	enum amd_pm_state_type pm = 0;
 	int i = 0;
 
-	if (adev->pp_enabled) {
-
+	if (adev->powerplay.pp_funcs->get_current_power_state
+		 && adev->powerplay.pp_funcs->get_pp_num_states) {
 		pm = amdgpu_dpm_get_current_power_state(adev);
 		amdgpu_dpm_get_pp_num_states(adev, &data);
 
@@ -280,25 +279,10 @@ static ssize_t amdgpu_get_pp_force_state(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	struct pp_states_info data;
-	enum amd_pm_state_type pm = 0;
-	int i;
 
-	if (adev->pp_force_state_enabled && adev->pp_enabled) {
-		pm = amdgpu_dpm_get_current_power_state(adev);
-		amdgpu_dpm_get_pp_num_states(adev, &data);
-
-		for (i = 0; i < data.nums; i++) {
-			if (pm == data.states[i])
-				break;
-		}
-
-		if (i == data.nums)
-			i = -EINVAL;
-
-		return snprintf(buf, PAGE_SIZE, "%d\n", i);
-
-	} else
+	if (adev->pp_force_state_enabled)
+		return amdgpu_get_pp_cur_state(dev, attr, buf);
+	else
 		return snprintf(buf, PAGE_SIZE, "\n");
 }
 
@@ -315,7 +299,8 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev,
 
 	if (strlen(buf) == 1)
 		adev->pp_force_state_enabled = false;
-	else if (adev->pp_enabled) {
+	else if (adev->powerplay.pp_funcs->dispatch_tasks &&
+			adev->powerplay.pp_funcs->get_pp_num_states) {
 		struct pp_states_info data;
 
 		ret = kstrtoul(buf, 0, &idx);
@@ -330,7 +315,7 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev,
 		if (state != POWER_STATE_TYPE_INTERNAL_BOOT &&
 		    state != POWER_STATE_TYPE_DEFAULT) {
 			amdgpu_dpm_dispatch_task(adev,
-					AMD_PP_EVENT_ENABLE_USER_STATE, &state, NULL);
+					AMD_PP_TASK_ENABLE_USER_STATE, &state, NULL);
 			adev->pp_force_state_enabled = true;
 		}
 	}
@@ -347,7 +332,7 @@ static ssize_t amdgpu_get_pp_table(struct device *dev,
 	char *table = NULL;
 	int size;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->get_pp_table)
 		size = amdgpu_dpm_get_pp_table(adev, &table);
 	else
 		return 0;
@@ -368,7 +353,7 @@ static ssize_t amdgpu_set_pp_table(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->set_pp_table)
 		amdgpu_dpm_set_pp_table(adev, buf, count);
 
 	return count;
@@ -380,14 +365,11 @@ static ssize_t amdgpu_get_pp_dpm_sclk(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	ssize_t size = 0;
-
-	if (adev->pp_enabled)
-		size = amdgpu_dpm_print_clock_levels(adev, PP_SCLK, buf);
-	else if (adev->pm.funcs->print_clock_levels)
-		size = adev->pm.funcs->print_clock_levels(adev, PP_SCLK, buf);
 
-	return size;
+	if (adev->powerplay.pp_funcs->print_clock_levels)
+		return amdgpu_dpm_print_clock_levels(adev, PP_SCLK, buf);
+	else
+		return snprintf(buf, PAGE_SIZE, "\n");
 }
 
 static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev,
@@ -416,10 +398,9 @@ static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev,
 		mask |= 1 << level;
 	}
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->force_clock_level)
 		amdgpu_dpm_force_clock_level(adev, PP_SCLK, mask);
-	else if (adev->pm.funcs->force_clock_level)
-		adev->pm.funcs->force_clock_level(adev, PP_SCLK, mask);
+
 fail:
 	return count;
 }
@@ -430,14 +411,11 @@ static ssize_t amdgpu_get_pp_dpm_mclk(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	ssize_t size = 0;
 
-	if (adev->pp_enabled)
-		size = amdgpu_dpm_print_clock_levels(adev, PP_MCLK, buf);
-	else if (adev->pm.funcs->print_clock_levels)
-		size = adev->pm.funcs->print_clock_levels(adev, PP_MCLK, buf);
-
-	return size;
+	if (adev->powerplay.pp_funcs->print_clock_levels)
+		return amdgpu_dpm_print_clock_levels(adev, PP_MCLK, buf);
+	else
+		return snprintf(buf, PAGE_SIZE, "\n");
 }
 
 static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev,
@@ -465,11 +443,9 @@ static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev,
 		}
 		mask |= 1 << level;
 	}
-
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->force_clock_level)
 		amdgpu_dpm_force_clock_level(adev, PP_MCLK, mask);
-	else if (adev->pm.funcs->force_clock_level)
-		adev->pm.funcs->force_clock_level(adev, PP_MCLK, mask);
+
 fail:
 	return count;
 }
@@ -480,14 +456,11 @@ static ssize_t amdgpu_get_pp_dpm_pcie(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	ssize_t size = 0;
 
-	if (adev->pp_enabled)
-		size = amdgpu_dpm_print_clock_levels(adev, PP_PCIE, buf);
-	else if (adev->pm.funcs->print_clock_levels)
-		size = adev->pm.funcs->print_clock_levels(adev, PP_PCIE, buf);
-
-	return size;
+	if (adev->powerplay.pp_funcs->print_clock_levels)
+		return amdgpu_dpm_print_clock_levels(adev, PP_PCIE, buf);
+	else
+		return snprintf(buf, PAGE_SIZE, "\n");
 }
 
 static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev,
@@ -515,11 +488,9 @@ static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev,
 		}
 		mask |= 1 << level;
 	}
-
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->force_clock_level)
 		amdgpu_dpm_force_clock_level(adev, PP_PCIE, mask);
-	else if (adev->pm.funcs->force_clock_level)
-		adev->pm.funcs->force_clock_level(adev, PP_PCIE, mask);
+
 fail:
 	return count;
 }
@@ -532,10 +503,8 @@ static ssize_t amdgpu_get_pp_sclk_od(struct device *dev,
 	struct amdgpu_device *adev = ddev->dev_private;
 	uint32_t value = 0;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->get_sclk_od)
 		value = amdgpu_dpm_get_sclk_od(adev);
-	else if (adev->pm.funcs->get_sclk_od)
-		value = adev->pm.funcs->get_sclk_od(adev);
 
 	return snprintf(buf, PAGE_SIZE, "%d\n", value);
 }
@@ -556,12 +525,12 @@ static ssize_t amdgpu_set_pp_sclk_od(struct device *dev,
 		count = -EINVAL;
 		goto fail;
 	}
-
-	if (adev->pp_enabled) {
+	if (adev->powerplay.pp_funcs->set_sclk_od)
 		amdgpu_dpm_set_sclk_od(adev, (uint32_t)value);
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_READJUST_POWER_STATE, NULL, NULL);
-	} else if (adev->pm.funcs->set_sclk_od) {
-		adev->pm.funcs->set_sclk_od(adev, (uint32_t)value);
+
+	if (adev->powerplay.pp_funcs->dispatch_tasks) {
+		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL);
+	} else {
 		adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps;
 		amdgpu_pm_compute_clocks(adev);
 	}
@@ -578,10 +547,8 @@ static ssize_t amdgpu_get_pp_mclk_od(struct device *dev,
 	struct amdgpu_device *adev = ddev->dev_private;
 	uint32_t value = 0;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->get_mclk_od)
 		value = amdgpu_dpm_get_mclk_od(adev);
-	else if (adev->pm.funcs->get_mclk_od)
-		value = adev->pm.funcs->get_mclk_od(adev);
 
 	return snprintf(buf, PAGE_SIZE, "%d\n", value);
 }
@@ -602,12 +569,12 @@ static ssize_t amdgpu_set_pp_mclk_od(struct device *dev,
 		count = -EINVAL;
 		goto fail;
 	}
-
-	if (adev->pp_enabled) {
+	if (adev->powerplay.pp_funcs->set_mclk_od)
 		amdgpu_dpm_set_mclk_od(adev, (uint32_t)value);
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_READJUST_POWER_STATE, NULL, NULL);
-	} else if (adev->pm.funcs->set_mclk_od) {
-		adev->pm.funcs->set_mclk_od(adev, (uint32_t)value);
+
+	if (adev->powerplay.pp_funcs->dispatch_tasks) {
+		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL);
+	} else {
 		adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps;
 		amdgpu_pm_compute_clocks(adev);
 	}
@@ -621,14 +588,11 @@ static ssize_t amdgpu_get_pp_power_profile(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = ddev->dev_private;
-	int ret = 0;
+	int ret = 0xff;
 
-	if (adev->pp_enabled)
+	if (adev->powerplay.pp_funcs->get_power_profile_state)
 		ret = amdgpu_dpm_get_power_profile_state(
 				adev, query);
-	else if (adev->pm.funcs->get_power_profile_state)
-		ret = adev->pm.funcs->get_power_profile_state(
-				adev, query);
 
 	if (ret)
 		return ret;
@@ -675,15 +639,12 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev,
 	char *sub_str, buf_cpy[128], *tmp_str;
 	const char delimiter[3] = {' ', '\n', '\0'};
 	long int value;
-	int ret = 0;
+	int ret = 0xff;
 
 	if (strncmp("reset", buf, strlen("reset")) == 0) {
-		if (adev->pp_enabled)
+		if (adev->powerplay.pp_funcs->reset_power_profile_state)
 			ret = amdgpu_dpm_reset_power_profile_state(
 					adev, request);
-		else if (adev->pm.funcs->reset_power_profile_state)
-			ret = adev->pm.funcs->reset_power_profile_state(
-					adev, request);
 		if (ret) {
 			count = -EINVAL;
 			goto fail;
@@ -692,12 +653,10 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev,
 	}
 
 	if (strncmp("set", buf, strlen("set")) == 0) {
-		if (adev->pp_enabled)
+		if (adev->powerplay.pp_funcs->set_power_profile_state)
 			ret = amdgpu_dpm_set_power_profile_state(
 					adev, request);
-		else if (adev->pm.funcs->set_power_profile_state)
-			ret = adev->pm.funcs->set_power_profile_state(
-					adev, request);
+
 		if (ret) {
 			count = -EINVAL;
 			goto fail;
@@ -745,13 +704,8 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev,
 
 		loop++;
 	}
-
-	if (adev->pp_enabled)
-		ret = amdgpu_dpm_set_power_profile_state(
-				adev, request);
-	else if (adev->pm.funcs->set_power_profile_state)
-		ret = adev->pm.funcs->set_power_profile_state(
-				adev, request);
+	if (adev->powerplay.pp_funcs->set_power_profile_state)
+		ret = amdgpu_dpm_set_power_profile_state(adev, request);
 
 	if (ret)
 		count = -EINVAL;
@@ -831,7 +785,7 @@ static ssize_t amdgpu_hwmon_show_temp(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	if (!adev->pp_enabled && !adev->pm.funcs->get_temperature)
+	if (!adev->powerplay.pp_funcs->get_temperature)
 		temp = 0;
 	else
 		temp = amdgpu_dpm_get_temperature(adev);
@@ -862,7 +816,7 @@ static ssize_t amdgpu_hwmon_get_pwm1_enable(struct device *dev,
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	u32 pwm_mode = 0;
 
-	if (!adev->pp_enabled && !adev->pm.funcs->get_fan_control_mode)
+	if (!adev->powerplay.pp_funcs->get_fan_control_mode)
 		return -EINVAL;
 
 	pwm_mode = amdgpu_dpm_get_fan_control_mode(adev);
@@ -879,7 +833,7 @@ static ssize_t amdgpu_hwmon_set_pwm1_enable(struct device *dev,
 	int err;
 	int value;
 
-	if (!adev->pp_enabled && !adev->pm.funcs->set_fan_control_mode)
+	if (!adev->powerplay.pp_funcs->set_fan_control_mode)
 		return -EINVAL;
 
 	err = kstrtoint(buf, 10, &value);
@@ -919,9 +873,11 @@ static ssize_t amdgpu_hwmon_set_pwm1(struct device *dev,
 
 	value = (value * 100) / 255;
 
-	err = amdgpu_dpm_set_fan_speed_percent(adev, value);
-	if (err)
-		return err;
+	if (adev->powerplay.pp_funcs->set_fan_speed_percent) {
+		err = amdgpu_dpm_set_fan_speed_percent(adev, value);
+		if (err)
+			return err;
+	}
 
 	return count;
 }
@@ -932,11 +888,13 @@ static ssize_t amdgpu_hwmon_get_pwm1(struct device *dev,
 {
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	int err;
-	u32 speed;
+	u32 speed = 0;
 
-	err = amdgpu_dpm_get_fan_speed_percent(adev, &speed);
-	if (err)
-		return err;
+	if (adev->powerplay.pp_funcs->get_fan_speed_percent) {
+		err = amdgpu_dpm_get_fan_speed_percent(adev, &speed);
+		if (err)
+			return err;
+	}
 
 	speed = (speed * 255) / 100;
 
@@ -949,11 +907,13 @@ static ssize_t amdgpu_hwmon_get_fan1_input(struct device *dev,
 {
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	int err;
-	u32 speed;
+	u32 speed = 0;
 
-	err = amdgpu_dpm_get_fan_speed_rpm(adev, &speed);
-	if (err)
-		return err;
+	if (adev->powerplay.pp_funcs->get_fan_speed_rpm) {
+		err = amdgpu_dpm_get_fan_speed_rpm(adev, &speed);
+		if (err)
+			return err;
+	}
 
 	return sprintf(buf, "%i\n", speed);
 }
@@ -996,9 +956,6 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj,
 	     attr == &sensor_dev_attr_pwm1_min.dev_attr.attr))
 		return 0;
 
-	if (adev->pp_enabled)
-		return effective_mode;
-
 	/* Skip fan attributes if fan is not present */
 	if (adev->pm.no_fan &&
 	    (attr == &sensor_dev_attr_pwm1.dev_attr.attr ||
@@ -1008,21 +965,21 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj,
 		return 0;
 
 	/* mask fan attributes if we have no bindings for this asic to expose */
-	if ((!adev->pm.funcs->get_fan_speed_percent &&
+	if ((!adev->powerplay.pp_funcs->get_fan_speed_percent &&
 	     attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't query fan */
-	    (!adev->pm.funcs->get_fan_control_mode &&
+	    (!adev->powerplay.pp_funcs->get_fan_control_mode &&
 	     attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr)) /* can't query state */
 		effective_mode &= ~S_IRUGO;
 
-	if ((!adev->pm.funcs->set_fan_speed_percent &&
+	if ((!adev->powerplay.pp_funcs->set_fan_speed_percent &&
 	     attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't manage fan */
-	    (!adev->pm.funcs->set_fan_control_mode &&
+	    (!adev->powerplay.pp_funcs->set_fan_control_mode &&
 	     attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr)) /* can't manage state */
 		effective_mode &= ~S_IWUSR;
 
 	/* hide max/min values if we can't both query and manage the fan */
-	if ((!adev->pm.funcs->set_fan_speed_percent &&
-	     !adev->pm.funcs->get_fan_speed_percent) &&
+	if ((!adev->powerplay.pp_funcs->set_fan_speed_percent &&
+	     !adev->powerplay.pp_funcs->get_fan_speed_percent) &&
 	    (attr == &sensor_dev_attr_pwm1_max.dev_attr.attr ||
 	     attr == &sensor_dev_attr_pwm1_min.dev_attr.attr))
 		return 0;
@@ -1055,7 +1012,7 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work)
 	if (!adev->pm.dpm_enabled)
 		return;
 
-	if (adev->pm.funcs->get_temperature) {
+	if (adev->powerplay.pp_funcs->get_temperature) {
 		int temp = amdgpu_dpm_get_temperature(adev);
 
 		if (temp < adev->pm.dpm.thermal.min_temp)
@@ -1087,7 +1044,7 @@ static struct amdgpu_ps *amdgpu_dpm_pick_power_state(struct amdgpu_device *adev,
 		true : false;
 
 	/* check if the vblank period is too short to adjust the mclk */
-	if (single_display && adev->pm.funcs->vblank_too_short) {
+	if (single_display && adev->powerplay.pp_funcs->vblank_too_short) {
 		if (amdgpu_dpm_vblank_too_short(adev))
 			single_display = false;
 	}
@@ -1216,7 +1173,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev)
 	struct amdgpu_ps *ps;
 	enum amd_pm_state_type dpm_state;
 	int ret;
-	bool equal;
+	bool equal = false;
 
 	/* if dpm init failed */
 	if (!adev->pm.dpm_enabled)
@@ -1236,7 +1193,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev)
 	else
 		return;
 
-	if (amdgpu_dpm == 1) {
+	if (amdgpu_dpm == 1 && adev->powerplay.pp_funcs->print_power_state) {
 		printk("switching from power state:\n");
 		amdgpu_dpm_print_power_state(adev, adev->pm.dpm.current_ps);
 		printk("switching to power state:\n");
@@ -1245,15 +1202,17 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev)
 
 	/* update whether vce is active */
 	ps->vce_active = adev->pm.dpm.vce_active;
-
-	amdgpu_dpm_display_configuration_changed(adev);
+	if (adev->powerplay.pp_funcs->display_configuration_changed)
+		amdgpu_dpm_display_configuration_changed(adev);
 
 	ret = amdgpu_dpm_pre_set_power_state(adev);
 	if (ret)
 		return;
 
-	if ((0 != amgdpu_dpm_check_state_equal(adev, adev->pm.dpm.current_ps, adev->pm.dpm.requested_ps, &equal)))
-		equal = false;
+	if (adev->powerplay.pp_funcs->check_state_equal) {
+		if (0 != amdgpu_dpm_check_state_equal(adev, adev->pm.dpm.current_ps, adev->pm.dpm.requested_ps, &equal))
+			equal = false;
+	}
 
 	if (equal)
 		return;
@@ -1264,7 +1223,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev)
 	adev->pm.dpm.current_active_crtcs = adev->pm.dpm.new_active_crtcs;
 	adev->pm.dpm.current_active_crtc_count = adev->pm.dpm.new_active_crtc_count;
 
-	if (adev->pm.funcs->force_performance_level) {
+	if (adev->powerplay.pp_funcs->force_performance_level) {
 		if (adev->pm.dpm.thermal_active) {
 			enum amd_dpm_forced_level level = adev->pm.dpm.forced_level;
 			/* force low perf level for thermal */
@@ -1280,7 +1239,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev)
 
 void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable)
 {
-	if (adev->pp_enabled || adev->pm.funcs->powergate_uvd) {
+	if (adev->powerplay.pp_funcs->powergate_uvd) {
 		/* enable/disable UVD */
 		mutex_lock(&adev->pm.mutex);
 		amdgpu_dpm_powergate_uvd(adev, !enable);
@@ -1302,7 +1261,7 @@ void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable)
 
 void amdgpu_dpm_enable_vce(struct amdgpu_device *adev, bool enable)
 {
-	if (adev->pp_enabled || adev->pm.funcs->powergate_vce) {
+	if (adev->powerplay.pp_funcs->powergate_vce) {
 		/* enable/disable VCE */
 		mutex_lock(&adev->pm.mutex);
 		amdgpu_dpm_powergate_vce(adev, !enable);
@@ -1337,8 +1296,7 @@ void amdgpu_pm_print_power_states(struct amdgpu_device *adev)
 {
 	int i;
 
-	if (adev->pp_enabled)
-		/* TO DO */
+	if (adev->powerplay.pp_funcs->print_power_state == NULL)
 		return;
 
 	for (i = 0; i < adev->pm.dpm.num_ps; i++)
@@ -1353,10 +1311,11 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
 	if (adev->pm.sysfs_initialized)
 		return 0;
 
-	if (!adev->pp_enabled) {
-		if (adev->pm.funcs->get_temperature == NULL)
-			return 0;
-	}
+	if (adev->pm.dpm_enabled == 0)
+		return 0;
+
+	if (adev->powerplay.pp_funcs->get_temperature == NULL)
+		return 0;
 
 	adev->pm.int_hwmon_dev = hwmon_device_register_with_groups(adev->dev,
 								   DRIVER_NAME, adev,
@@ -1379,27 +1338,26 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
 		return ret;
 	}
 
-	if (adev->pp_enabled) {
-		ret = device_create_file(adev->dev, &dev_attr_pp_num_states);
-		if (ret) {
-			DRM_ERROR("failed to create device file pp_num_states\n");
-			return ret;
-		}
-		ret = device_create_file(adev->dev, &dev_attr_pp_cur_state);
-		if (ret) {
-			DRM_ERROR("failed to create device file pp_cur_state\n");
-			return ret;
-		}
-		ret = device_create_file(adev->dev, &dev_attr_pp_force_state);
-		if (ret) {
-			DRM_ERROR("failed to create device file pp_force_state\n");
-			return ret;
-		}
-		ret = device_create_file(adev->dev, &dev_attr_pp_table);
-		if (ret) {
-			DRM_ERROR("failed to create device file pp_table\n");
-			return ret;
-		}
+
+	ret = device_create_file(adev->dev, &dev_attr_pp_num_states);
+	if (ret) {
+		DRM_ERROR("failed to create device file pp_num_states\n");
+		return ret;
+	}
+	ret = device_create_file(adev->dev, &dev_attr_pp_cur_state);
+	if (ret) {
+		DRM_ERROR("failed to create device file pp_cur_state\n");
+		return ret;
+	}
+	ret = device_create_file(adev->dev, &dev_attr_pp_force_state);
+	if (ret) {
+		DRM_ERROR("failed to create device file pp_force_state\n");
+		return ret;
+	}
+	ret = device_create_file(adev->dev, &dev_attr_pp_table);
+	if (ret) {
+		DRM_ERROR("failed to create device file pp_table\n");
+		return ret;
 	}
 
 	ret = device_create_file(adev->dev, &dev_attr_pp_dpm_sclk);
@@ -1455,16 +1413,19 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
 
 void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev)
 {
+	if (adev->pm.dpm_enabled == 0)
+		return;
+
 	if (adev->pm.int_hwmon_dev)
 		hwmon_device_unregister(adev->pm.int_hwmon_dev);
 	device_remove_file(adev->dev, &dev_attr_power_dpm_state);
 	device_remove_file(adev->dev, &dev_attr_power_dpm_force_performance_level);
-	if (adev->pp_enabled) {
-		device_remove_file(adev->dev, &dev_attr_pp_num_states);
-		device_remove_file(adev->dev, &dev_attr_pp_cur_state);
-		device_remove_file(adev->dev, &dev_attr_pp_force_state);
-		device_remove_file(adev->dev, &dev_attr_pp_table);
-	}
+
+	device_remove_file(adev->dev, &dev_attr_pp_num_states);
+	device_remove_file(adev->dev, &dev_attr_pp_cur_state);
+	device_remove_file(adev->dev, &dev_attr_pp_force_state);
+	device_remove_file(adev->dev, &dev_attr_pp_table);
+
 	device_remove_file(adev->dev, &dev_attr_pp_dpm_sclk);
 	device_remove_file(adev->dev, &dev_attr_pp_dpm_mclk);
 	device_remove_file(adev->dev, &dev_attr_pp_dpm_pcie);
@@ -1495,8 +1456,8 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev)
 			amdgpu_fence_wait_empty(ring);
 	}
 
-	if (adev->pp_enabled) {
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_DISPLAY_CONFIG_CHANGE, NULL, NULL);
+	if (adev->powerplay.pp_funcs->dispatch_tasks) {
+		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_DISPLAY_CONFIG_CHANGE, NULL, NULL);
 	} else {
 		mutex_lock(&adev->pm.mutex);
 		adev->pm.dpm.new_active_crtcs = 0;
@@ -1630,15 +1591,15 @@ static int amdgpu_debugfs_pm_info(struct seq_file *m, void *data)
 	if  ((adev->flags & AMD_IS_PX) &&
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) {
 		seq_printf(m, "PX asic powered off\n");
-	} else if (adev->pp_enabled) {
-		return amdgpu_debugfs_pm_info_pp(m, adev);
-	} else {
+	} else if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level) {
 		mutex_lock(&adev->pm.mutex);
-		if (adev->pm.funcs->debugfs_print_current_performance_level)
-			adev->pm.funcs->debugfs_print_current_performance_level(adev, m);
+		if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level)
+			adev->powerplay.pp_funcs->debugfs_print_current_performance_level(adev, m);
 		else
 			seq_printf(m, "Debugfs support not implemented for this asic\n");
 		mutex_unlock(&adev->pm.mutex);
+	} else {
+		return amdgpu_debugfs_pm_info_pp(m, adev);
 	}
 
 	return 0;

drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c

@@ -34,24 +34,6 @@
 #include "cik_dpm.h"
 #include "vi_dpm.h"
 
-static int amdgpu_create_pp_handle(struct amdgpu_device *adev)
-{
-	struct amd_pp_init pp_init;
-	struct amd_powerplay *amd_pp;
-	int ret;
-
-	amd_pp = &(adev->powerplay);
-	pp_init.chip_family = adev->family;
-	pp_init.chip_id = adev->asic_type;
-	pp_init.pm_en = (amdgpu_dpm != 0 && !amdgpu_sriov_vf(adev)) ? true : false;
-	pp_init.feature_mask = amdgpu_pp_feature_mask;
-	pp_init.device = amdgpu_cgs_create_device(adev);
-	ret = amd_powerplay_create(&pp_init, &(amd_pp->pp_handle));
-	if (ret)
-		return -EINVAL;
-	return 0;
-}
-
 static int amdgpu_pp_early_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -59,7 +41,6 @@ static int amdgpu_pp_early_init(void *handle)
 	int ret = 0;
 
 	amd_pp = &(adev->powerplay);
-	adev->pp_enabled = false;
 	amd_pp->pp_handle = (void *)adev;
 
 	switch (adev->asic_type) {
@@ -73,9 +54,7 @@ static int amdgpu_pp_early_init(void *handle)
 	case CHIP_STONEY:
 	case CHIP_VEGA10:
 	case CHIP_RAVEN:
-		adev->pp_enabled = true;
-		if (amdgpu_create_pp_handle(adev))
-			return -EINVAL;
+		amd_pp->cgs_device = amdgpu_cgs_create_device(adev);
 		amd_pp->ip_funcs = &pp_ip_funcs;
 		amd_pp->pp_funcs = &pp_dpm_funcs;
 		break;
@@ -87,17 +66,26 @@ static int amdgpu_pp_early_init(void *handle)
 	case CHIP_OLAND:
 	case CHIP_HAINAN:
 		amd_pp->ip_funcs = &si_dpm_ip_funcs;
+		amd_pp->pp_funcs = &si_dpm_funcs;
 	break;
 #endif
 #ifdef CONFIG_DRM_AMDGPU_CIK
 	case CHIP_BONAIRE:
 	case CHIP_HAWAII:
-		amd_pp->ip_funcs = &ci_dpm_ip_funcs;
+		if (amdgpu_dpm == -1) {
+			amd_pp->ip_funcs = &ci_dpm_ip_funcs;
+			amd_pp->pp_funcs = &ci_dpm_funcs;
+		} else {
+			amd_pp->cgs_device = amdgpu_cgs_create_device(adev);
+			amd_pp->ip_funcs = &pp_ip_funcs;
+			amd_pp->pp_funcs = &pp_dpm_funcs;
+		}
 		break;
 	case CHIP_KABINI:
 	case CHIP_MULLINS:
 	case CHIP_KAVERI:
 		amd_pp->ip_funcs = &kv_dpm_ip_funcs;
+		amd_pp->pp_funcs = &kv_dpm_funcs;
 		break;
 #endif
 	default:
@@ -107,12 +95,9 @@ static int amdgpu_pp_early_init(void *handle)
 
 	if (adev->powerplay.ip_funcs->early_init)
 		ret = adev->powerplay.ip_funcs->early_init(
-					adev->powerplay.pp_handle);
+					amd_pp->cgs_device ? amd_pp->cgs_device :
+					amd_pp->pp_handle);
 
-	if (ret == PP_DPM_DISABLED) {
-		adev->pm.dpm_enabled = false;
-		return 0;
-	}
 	return ret;
 }
 
@@ -126,11 +111,6 @@ static int amdgpu_pp_late_init(void *handle)
 		ret = adev->powerplay.ip_funcs->late_init(
 					adev->powerplay.pp_handle);
 
-	if (adev->pp_enabled && adev->pm.dpm_enabled) {
-		amdgpu_pm_sysfs_init(adev);
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_COMPLETE_INIT, NULL, NULL);
-	}
-
 	return ret;
 }
 
@@ -165,21 +145,13 @@ static int amdgpu_pp_hw_init(void *handle)
 	int ret = 0;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	if (adev->pp_enabled && adev->firmware.load_type == AMDGPU_FW_LOAD_SMU)
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_SMU)
 		amdgpu_ucode_init_bo(adev);
 
 	if (adev->powerplay.ip_funcs->hw_init)
 		ret = adev->powerplay.ip_funcs->hw_init(
 					adev->powerplay.pp_handle);
 
-	if (ret == PP_DPM_DISABLED) {
-		adev->pm.dpm_enabled = false;
-		return 0;
-	}
-
-	if ((amdgpu_dpm != 0) && !amdgpu_sriov_vf(adev))
-		adev->pm.dpm_enabled = true;
-
 	return ret;
 }
 
@@ -188,14 +160,11 @@ static int amdgpu_pp_hw_fini(void *handle)
 	int ret = 0;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	if (adev->pp_enabled && adev->pm.dpm_enabled)
-		amdgpu_pm_sysfs_fini(adev);
-
 	if (adev->powerplay.ip_funcs->hw_fini)
 		ret = adev->powerplay.ip_funcs->hw_fini(
 					adev->powerplay.pp_handle);
 
-	if (adev->pp_enabled && adev->firmware.load_type == AMDGPU_FW_LOAD_SMU)
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_SMU)
 		amdgpu_ucode_fini_bo(adev);
 
 	return ret;
@@ -209,9 +178,8 @@ static void amdgpu_pp_late_fini(void *handle)
 		adev->powerplay.ip_funcs->late_fini(
 			  adev->powerplay.pp_handle);
 
-
-	if (adev->pp_enabled)
-		amd_powerplay_destroy(adev->powerplay.pp_handle);
+	if (adev->powerplay.cgs_device)
+		amdgpu_cgs_destroy_device(adev->powerplay.cgs_device);
 }
 
 static int amdgpu_pp_suspend(void *handle)

drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c

@@ -57,6 +57,40 @@ void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr)
 	ttm_bo_kunmap(&bo->dma_buf_vmap);
 }
 
+int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
+{
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
+	unsigned asize = amdgpu_bo_size(bo);
+	int ret;
+
+	if (!vma->vm_file)
+		return -ENODEV;
+
+	if (adev == NULL)
+		return -ENODEV;
+
+	/* Check for valid size. */
+	if (asize < vma->vm_end - vma->vm_start)
+		return -EINVAL;
+
+	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) ||
+	    (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) {
+		return -EPERM;
+	}
+	vma->vm_pgoff += amdgpu_bo_mmap_offset(bo) >> PAGE_SHIFT;
+
+	/* prime mmap does not need to check access, so allow here */
+	ret = drm_vma_node_allow(&obj->vma_node, vma->vm_file->private_data);
+	if (ret)
+		return ret;
+
+	ret = ttm_bo_mmap(vma->vm_file, vma, &adev->mman.bdev);
+	drm_vma_node_revoke(&obj->vma_node, vma->vm_file->private_data);
+
+	return ret;
+}
+
 struct drm_gem_object *
 amdgpu_gem_prime_import_sg_table(struct drm_device *dev,
 				 struct dma_buf_attachment *attach,
@@ -136,7 +170,8 @@ struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev,
 {
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
 
-	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm))
+	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) ||
+	    bo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
 		return ERR_PTR(-EPERM);
 
 	return drm_gem_prime_export(dev, gobj, flags);

drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c

@@ -57,21 +57,23 @@ static int psp_sw_init(void *handle)
 		psp->prep_cmd_buf = psp_v3_1_prep_cmd_buf;
 		psp->ring_init = psp_v3_1_ring_init;
 		psp->ring_create = psp_v3_1_ring_create;
+		psp->ring_stop = psp_v3_1_ring_stop;
 		psp->ring_destroy = psp_v3_1_ring_destroy;
 		psp->cmd_submit = psp_v3_1_cmd_submit;
 		psp->compare_sram_data = psp_v3_1_compare_sram_data;
 		psp->smu_reload_quirk = psp_v3_1_smu_reload_quirk;
+		psp->mode1_reset = psp_v3_1_mode1_reset;
 		break;
 	case CHIP_RAVEN:
-#if 0
 		psp->init_microcode = psp_v10_0_init_microcode;
-#endif
 		psp->prep_cmd_buf = psp_v10_0_prep_cmd_buf;
 		psp->ring_init = psp_v10_0_ring_init;
 		psp->ring_create = psp_v10_0_ring_create;
+		psp->ring_stop = psp_v10_0_ring_stop;
 		psp->ring_destroy = psp_v10_0_ring_destroy;
 		psp->cmd_submit = psp_v10_0_cmd_submit;
 		psp->compare_sram_data = psp_v10_0_compare_sram_data;
+		psp->mode1_reset = psp_v10_0_mode1_reset;
 		break;
 	default:
 		return -EINVAL;
@@ -90,6 +92,12 @@ static int psp_sw_init(void *handle)
 
 static int psp_sw_fini(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	release_firmware(adev->psp.sos_fw);
+	adev->psp.sos_fw = NULL;
+	release_firmware(adev->psp.asd_fw);
+	adev->psp.asd_fw = NULL;
 	return 0;
 }
 
@@ -253,15 +261,18 @@ static int psp_asd_load(struct psp_context *psp)
 
 static int psp_hw_start(struct psp_context *psp)
 {
+	struct amdgpu_device *adev = psp->adev;
 	int ret;
 
-	ret = psp_bootloader_load_sysdrv(psp);
-	if (ret)
-		return ret;
+	if (!amdgpu_sriov_vf(adev) || !adev->in_sriov_reset) {
+		ret = psp_bootloader_load_sysdrv(psp);
+		if (ret)
+			return ret;
 
-	ret = psp_bootloader_load_sos(psp);
-	if (ret)
-		return ret;
+		ret = psp_bootloader_load_sos(psp);
+		if (ret)
+			return ret;
+	}
 
 	ret = psp_ring_create(psp, PSP_RING_TYPE__KM);
 	if (ret)
@@ -453,6 +464,16 @@ static int psp_hw_fini(void *handle)
 
 static int psp_suspend(void *handle)
 {
+	int ret;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct psp_context *psp = &adev->psp;
+
+	ret = psp_ring_stop(psp, PSP_RING_TYPE__KM);
+	if (ret) {
+		DRM_ERROR("PSP ring stop failed\n");
+		return ret;
+	}
+
 	return 0;
 }
 
@@ -487,6 +508,22 @@ failed:
 	return ret;
 }
 
+static bool psp_check_reset(void* handle)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	if (adev->flags & AMD_IS_APU)
+		return true;
+
+	return false;
+}
+
+static int psp_reset(void* handle)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	return psp_mode1_reset(&adev->psp);
+}
+
 static bool psp_check_fw_loading_status(struct amdgpu_device *adev,
 					enum AMDGPU_UCODE_ID ucode_type)
 {
@@ -530,8 +567,9 @@ const struct amd_ip_funcs psp_ip_funcs = {
 	.suspend = psp_suspend,
 	.resume = psp_resume,
 	.is_idle = NULL,
+	.check_soft_reset = psp_check_reset,
 	.wait_for_idle = NULL,
-	.soft_reset = NULL,
+	.soft_reset = psp_reset,
 	.set_clockgating_state = psp_set_clockgating_state,
 	.set_powergating_state = psp_set_powergating_state,
 };

drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h

@@ -66,6 +66,8 @@ struct psp_context
 			    struct psp_gfx_cmd_resp *cmd);
 	int (*ring_init)(struct psp_context *psp, enum psp_ring_type ring_type);
 	int (*ring_create)(struct psp_context *psp, enum psp_ring_type ring_type);
+	int (*ring_stop)(struct psp_context *psp,
+			    enum psp_ring_type ring_type);
 	int (*ring_destroy)(struct psp_context *psp,
 			    enum psp_ring_type ring_type);
 	int (*cmd_submit)(struct psp_context *psp, struct amdgpu_firmware_info *ucode,
@@ -74,6 +76,7 @@ struct psp_context
 				  struct amdgpu_firmware_info *ucode,
 				  enum AMDGPU_UCODE_ID ucode_type);
 	bool (*smu_reload_quirk)(struct psp_context *psp);
+	int (*mode1_reset)(struct psp_context *psp);
 
 	/* fence buffer */
 	struct amdgpu_bo 		*fw_pri_bo;
@@ -123,6 +126,7 @@ struct amdgpu_psp_funcs {
 #define psp_prep_cmd_buf(ucode, type) (psp)->prep_cmd_buf((ucode), (type))
 #define psp_ring_init(psp, type) (psp)->ring_init((psp), (type))
 #define psp_ring_create(psp, type) (psp)->ring_create((psp), (type))
+#define psp_ring_stop(psp, type) (psp)->ring_stop((psp), (type))
 #define psp_ring_destroy(psp, type) ((psp)->ring_destroy((psp), (type)))
 #define psp_cmd_submit(psp, ucode, cmd_mc, fence_mc, index) \
 		(psp)->cmd_submit((psp), (ucode), (cmd_mc), (fence_mc), (index))
@@ -136,6 +140,8 @@ struct amdgpu_psp_funcs {
 		((psp)->bootloader_load_sos ? (psp)->bootloader_load_sos((psp)) : 0)
 #define psp_smu_reload_quirk(psp) \
 		((psp)->smu_reload_quirk ? (psp)->smu_reload_quirk((psp)) : false)
+#define psp_mode1_reset(psp) \
+		((psp)->mode1_reset ? (psp)->mode1_reset((psp)) : false)
 
 extern const struct amd_ip_funcs psp_ip_funcs;
 

drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c

@@ -121,7 +121,7 @@ static enum amdgpu_ring_type amdgpu_hw_ip_to_ring_type(int hw_ip)
 
 static int amdgpu_lru_map(struct amdgpu_device *adev,
 			  struct amdgpu_queue_mapper *mapper,
-			  int user_ring,
+			  int user_ring, bool lru_pipe_order,
 			  struct amdgpu_ring **out_ring)
 {
 	int r, i, j;
@@ -139,7 +139,7 @@ static int amdgpu_lru_map(struct amdgpu_device *adev,
 	}
 
 	r = amdgpu_ring_lru_get(adev, ring_type, ring_blacklist,
-				j, out_ring);
+				j, lru_pipe_order, out_ring);
 	if (r)
 		return r;
 
@@ -284,8 +284,10 @@ int amdgpu_queue_mgr_map(struct amdgpu_device *adev,
 		r = amdgpu_identity_map(adev, mapper, ring, out_ring);
 		break;
 	case AMDGPU_HW_IP_DMA:
+		r = amdgpu_lru_map(adev, mapper, ring, false, out_ring);
+		break;
 	case AMDGPU_HW_IP_COMPUTE:
-		r = amdgpu_lru_map(adev, mapper, ring, out_ring);
+		r = amdgpu_lru_map(adev, mapper, ring, true, out_ring);
 		break;
 	default:
 		*out_ring = NULL;

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c

@@ -136,7 +136,8 @@ void amdgpu_ring_commit(struct amdgpu_ring *ring)
 	if (ring->funcs->end_use)
 		ring->funcs->end_use(ring);
 
-	amdgpu_ring_lru_touch(ring->adev, ring);
+	if (ring->funcs->type != AMDGPU_RING_TYPE_KIQ)
+		amdgpu_ring_lru_touch(ring->adev, ring);
 }
 
 /**
@@ -154,6 +155,75 @@ void amdgpu_ring_undo(struct amdgpu_ring *ring)
 		ring->funcs->end_use(ring);
 }
 
+/**
+ * amdgpu_ring_priority_put - restore a ring's priority
+ *
+ * @ring: amdgpu_ring structure holding the information
+ * @priority: target priority
+ *
+ * Release a request for executing at @priority
+ */
+void amdgpu_ring_priority_put(struct amdgpu_ring *ring,
+			      enum amd_sched_priority priority)
+{
+	int i;
+
+	if (!ring->funcs->set_priority)
+		return;
+
+	if (atomic_dec_return(&ring->num_jobs[priority]) > 0)
+		return;
+
+	/* no need to restore if the job is already at the lowest priority */
+	if (priority == AMD_SCHED_PRIORITY_NORMAL)
+		return;
+
+	mutex_lock(&ring->priority_mutex);
+	/* something higher prio is executing, no need to decay */
+	if (ring->priority > priority)
+		goto out_unlock;
+
+	/* decay priority to the next level with a job available */
+	for (i = priority; i >= AMD_SCHED_PRIORITY_MIN; i--) {
+		if (i == AMD_SCHED_PRIORITY_NORMAL
+				|| atomic_read(&ring->num_jobs[i])) {
+			ring->priority = i;
+			ring->funcs->set_priority(ring, i);
+			break;
+		}
+	}
+
+out_unlock:
+	mutex_unlock(&ring->priority_mutex);
+}
+
+/**
+ * amdgpu_ring_priority_get - change the ring's priority
+ *
+ * @ring: amdgpu_ring structure holding the information
+ * @priority: target priority
+ *
+ * Request a ring's priority to be raised to @priority (refcounted).
+ */
+void amdgpu_ring_priority_get(struct amdgpu_ring *ring,
+			      enum amd_sched_priority priority)
+{
+	if (!ring->funcs->set_priority)
+		return;
+
+	atomic_inc(&ring->num_jobs[priority]);
+
+	mutex_lock(&ring->priority_mutex);
+	if (priority <= ring->priority)
+		goto out_unlock;
+
+	ring->priority = priority;
+	ring->funcs->set_priority(ring, priority);
+
+out_unlock:
+	mutex_unlock(&ring->priority_mutex);
+}
+
 /**
  * amdgpu_ring_init - init driver ring struct.
  *
@@ -169,7 +239,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 		     unsigned max_dw, struct amdgpu_irq_src *irq_src,
 		     unsigned irq_type)
 {
-	int r;
+	int r, i;
 	int sched_hw_submission = amdgpu_sched_hw_submission;
 
 	/* Set the hw submission limit higher for KIQ because
@@ -247,9 +317,14 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	}
 
 	ring->max_dw = max_dw;
+	ring->priority = AMD_SCHED_PRIORITY_NORMAL;
+	mutex_init(&ring->priority_mutex);
 	INIT_LIST_HEAD(&ring->lru_list);
 	amdgpu_ring_lru_touch(adev, ring);
 
+	for (i = 0; i < AMD_SCHED_PRIORITY_MAX; ++i)
+		atomic_set(&ring->num_jobs[i], 0);
+
 	if (amdgpu_debugfs_ring_init(adev, ring)) {
 		DRM_ERROR("Failed to register debugfs file for rings !\n");
 	}
@@ -315,14 +390,16 @@ static bool amdgpu_ring_is_blacklisted(struct amdgpu_ring *ring,
  * @type: amdgpu_ring_type enum
  * @blacklist: blacklisted ring ids array
  * @num_blacklist: number of entries in @blacklist
+ * @lru_pipe_order: find a ring from the least recently used pipe
  * @ring: output ring
  *
  * Retrieve the amdgpu_ring structure for the least recently used ring of
  * a specific IP block (all asics).
  * Returns 0 on success, error on failure.
  */
-int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist,
-			int num_blacklist, struct amdgpu_ring **ring)
+int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type,
+			int *blacklist,	int num_blacklist,
+			bool lru_pipe_order, struct amdgpu_ring **ring)
 {
 	struct amdgpu_ring *entry;
 
@@ -337,10 +414,23 @@ int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist,
 		if (amdgpu_ring_is_blacklisted(entry, blacklist, num_blacklist))
 			continue;
 
-		*ring = entry;
-		amdgpu_ring_lru_touch_locked(adev, *ring);
-		break;
+		if (!*ring) {
+			*ring = entry;
+
+			/* We are done for ring LRU */
+			if (!lru_pipe_order)
+				break;
+		}
+
+		/* Move all rings on the same pipe to the end of the list */
+		if (entry->pipe == (*ring)->pipe)
+			amdgpu_ring_lru_touch_locked(adev, entry);
 	}
+
+	/* Move the ring we found to the end of the list */
+	if (*ring)
+		amdgpu_ring_lru_touch_locked(adev, *ring);
+
 	spin_unlock(&adev->ring_lru_list_lock);
 
 	if (!*ring) {

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h

@@ -24,6 +24,7 @@
 #ifndef __AMDGPU_RING_H__
 #define __AMDGPU_RING_H__
 
+#include <drm/amdgpu_drm.h>
 #include "gpu_scheduler.h"
 
 /* max number of rings */
@@ -56,6 +57,7 @@ struct amdgpu_device;
 struct amdgpu_ring;
 struct amdgpu_ib;
 struct amdgpu_cs_parser;
+struct amdgpu_job;
 
 /*
  * Fences.
@@ -88,8 +90,12 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
 void amdgpu_fence_driver_suspend(struct amdgpu_device *adev);
 void amdgpu_fence_driver_resume(struct amdgpu_device *adev);
 int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **fence);
+int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s);
 void amdgpu_fence_process(struct amdgpu_ring *ring);
 int amdgpu_fence_wait_empty(struct amdgpu_ring *ring);
+signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring,
+				      uint32_t wait_seq,
+				      signed long timeout);
 unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring);
 
 /*
@@ -147,6 +153,9 @@ struct amdgpu_ring_funcs {
 	void (*emit_rreg)(struct amdgpu_ring *ring, uint32_t reg);
 	void (*emit_wreg)(struct amdgpu_ring *ring, uint32_t reg, uint32_t val);
 	void (*emit_tmz)(struct amdgpu_ring *ring, bool start);
+	/* priority functions */
+	void (*set_priority) (struct amdgpu_ring *ring,
+			      enum amd_sched_priority priority);
 };
 
 struct amdgpu_ring {
@@ -187,6 +196,12 @@ struct amdgpu_ring {
 	volatile u32		*cond_exe_cpu_addr;
 	unsigned		vm_inv_eng;
 	bool			has_compute_vm_bug;
+
+	atomic_t		num_jobs[AMD_SCHED_PRIORITY_MAX];
+	struct mutex		priority_mutex;
+	/* protected by priority_mutex */
+	int			priority;
+
 #if defined(CONFIG_DEBUG_FS)
 	struct dentry *ent;
 #endif
@@ -197,12 +212,17 @@ void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count);
 void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
 void amdgpu_ring_commit(struct amdgpu_ring *ring);
 void amdgpu_ring_undo(struct amdgpu_ring *ring);
+void amdgpu_ring_priority_get(struct amdgpu_ring *ring,
+			      enum amd_sched_priority priority);
+void amdgpu_ring_priority_put(struct amdgpu_ring *ring,
+			      enum amd_sched_priority priority);
 int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 		     unsigned ring_size, struct amdgpu_irq_src *irq_src,
 		     unsigned irq_type);
 void amdgpu_ring_fini(struct amdgpu_ring *ring);
-int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist,
-			int num_blacklist, struct amdgpu_ring **ring);
+int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type,
+			int *blacklist, int num_blacklist,
+			bool lru_pipe_order, struct amdgpu_ring **ring);
 void amdgpu_ring_lru_touch(struct amdgpu_device *adev, struct amdgpu_ring *ring);
 static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring)
 {

drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c

@@ -0,0 +1,109 @@
+/*
+ * Copyright 2017 Valve Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Andres Rodriguez <andresx7@gmail.com>
+ */
+
+#include <linux/fdtable.h>
+#include <linux/pid.h>
+#include <drm/amdgpu_drm.h>
+#include "amdgpu.h"
+
+#include "amdgpu_vm.h"
+
+enum amd_sched_priority amdgpu_to_sched_priority(int amdgpu_priority)
+{
+	switch (amdgpu_priority) {
+	case AMDGPU_CTX_PRIORITY_VERY_HIGH:
+		return AMD_SCHED_PRIORITY_HIGH_HW;
+	case AMDGPU_CTX_PRIORITY_HIGH:
+		return AMD_SCHED_PRIORITY_HIGH_SW;
+	case AMDGPU_CTX_PRIORITY_NORMAL:
+		return AMD_SCHED_PRIORITY_NORMAL;
+	case AMDGPU_CTX_PRIORITY_LOW:
+	case AMDGPU_CTX_PRIORITY_VERY_LOW:
+		return AMD_SCHED_PRIORITY_LOW;
+	case AMDGPU_CTX_PRIORITY_UNSET:
+		return AMD_SCHED_PRIORITY_UNSET;
+	default:
+		WARN(1, "Invalid context priority %d\n", amdgpu_priority);
+		return AMD_SCHED_PRIORITY_INVALID;
+	}
+}
+
+static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev,
+						  int fd,
+						  enum amd_sched_priority priority)
+{
+	struct file *filp = fcheck(fd);
+	struct drm_file *file;
+	struct pid *pid;
+	struct amdgpu_fpriv *fpriv;
+	struct amdgpu_ctx *ctx;
+	uint32_t id;
+
+	if (!filp)
+		return -EINVAL;
+
+	pid = get_pid(((struct drm_file *)filp->private_data)->pid);
+
+	mutex_lock(&adev->ddev->filelist_mutex);
+	list_for_each_entry(file, &adev->ddev->filelist, lhead) {
+		if (file->pid != pid)
+			continue;
+
+		fpriv = file->driver_priv;
+		idr_for_each_entry(&fpriv->ctx_mgr.ctx_handles, ctx, id)
+				amdgpu_ctx_priority_override(ctx, priority);
+	}
+	mutex_unlock(&adev->ddev->filelist_mutex);
+
+	put_pid(pid);
+
+	return 0;
+}
+
+int amdgpu_sched_ioctl(struct drm_device *dev, void *data,
+		       struct drm_file *filp)
+{
+	union drm_amdgpu_sched *args = data;
+	struct amdgpu_device *adev = dev->dev_private;
+	enum amd_sched_priority priority;
+	int r;
+
+	priority = amdgpu_to_sched_priority(args->in.priority);
+	if (args->in.flags || priority == AMD_SCHED_PRIORITY_INVALID)
+		return -EINVAL;
+
+	switch (args->in.op) {
+	case AMDGPU_SCHED_OP_PROCESS_PRIORITY_OVERRIDE:
+		r = amdgpu_sched_process_priority_override(adev,
+							   args->in.fd,
+							   priority);
+		break;
+	default:
+		DRM_ERROR("Invalid sched op specified: %d\n", args->in.op);
+		r = -EINVAL;
+		break;
+	}
+
+	return r;
+}

drivers/gpu/drm/amd/powerplay/eventmgr/eventinit.h → drivers/gpu/drm/amd/amdgpu/amdgpu_sched.h

@@ -1,5 +1,5 @@
 /*
- * Copyright 2015 Advanced Micro Devices, Inc.
+ * Copyright 2017 Valve Corporation
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -19,16 +19,16 @@
  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
  * OTHER DEALINGS IN THE SOFTWARE.
  *
+ * Authors: Andres Rodriguez <andresx7@gmail.com>
  */
 
-#ifndef _EVENTINIT_H_
-#define _EVENTINIT_H_
+#ifndef __AMDGPU_SCHED_H__
+#define __AMDGPU_SCHED_H__
 
-#define PEM_CURRENT_POWERPLAY_FEATURE_VERSION 4
+#include <drm/drmP.h>
 
-void pem_init_feature_info(struct pp_eventmgr *eventmgr);
-void pem_uninit_featureInfo(struct pp_eventmgr *eventmgr);
-int pem_register_interrupts(struct pp_eventmgr *eventmgr);
-int pem_unregister_interrupts(struct pp_eventmgr *eventmgr);
+enum amd_sched_priority amdgpu_to_sched_priority(int amdgpu_priority);
+int amdgpu_sched_ioctl(struct drm_device *dev, void *data,
+		       struct drm_file *filp);
 
-#endif /* _EVENTINIT_H_ */
+#endif // __AMDGPU_SCHED_H__

drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c

@@ -169,14 +169,14 @@ int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync,
  *
  * @sync: sync object to add fences from reservation object to
  * @resv: reservation object with embedded fence
- * @shared: true if we should only sync to the exclusive fence
+ * @explicit_sync: true if we should only sync to the exclusive fence
  *
  * Sync to the fence
  */
 int amdgpu_sync_resv(struct amdgpu_device *adev,
 		     struct amdgpu_sync *sync,
 		     struct reservation_object *resv,
-		     void *owner)
+		     void *owner, bool explicit_sync)
 {
 	struct reservation_object_list *flist;
 	struct dma_fence *f;
@@ -191,6 +191,9 @@ int amdgpu_sync_resv(struct amdgpu_device *adev,
 	f = reservation_object_get_excl(resv);
 	r = amdgpu_sync_fence(adev, sync, f);
 
+	if (explicit_sync)
+		return r;
+
 	flist = reservation_object_get_list(resv);
 	if (!flist || r)
 		return r;

drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h

@@ -45,7 +45,8 @@ int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync,
 int amdgpu_sync_resv(struct amdgpu_device *adev,
 		     struct amdgpu_sync *sync,
 		     struct reservation_object *resv,
-		     void *owner);
+		     void *owner,
+		     bool explicit_sync);
 struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync,
 				     struct amdgpu_ring *ring);
 struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync);

drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h

@@ -15,62 +15,6 @@
 #define AMDGPU_JOB_GET_TIMELINE_NAME(job) \
 	 job->base.s_fence->finished.ops->get_timeline_name(&job->base.s_fence->finished)
 
-TRACE_EVENT(amdgpu_ttm_tt_populate,
-	    TP_PROTO(struct amdgpu_device *adev, uint64_t dma_address, uint64_t phys_address),
-	    TP_ARGS(adev, dma_address, phys_address),
-	    TP_STRUCT__entry(
-				__field(uint16_t, domain)
-				__field(uint8_t, bus)
-				__field(uint8_t, slot)
-				__field(uint8_t, func)
-				__field(uint64_t, dma)
-				__field(uint64_t, phys)
-			    ),
-	    TP_fast_assign(
-			   __entry->domain = pci_domain_nr(adev->pdev->bus);
-			   __entry->bus = adev->pdev->bus->number;
-			   __entry->slot = PCI_SLOT(adev->pdev->devfn);
-			   __entry->func = PCI_FUNC(adev->pdev->devfn);
-			   __entry->dma = dma_address;
-			   __entry->phys = phys_address;
-			   ),
-	    TP_printk("%04x:%02x:%02x.%x: 0x%llx => 0x%llx",
-		      (unsigned)__entry->domain,
-		      (unsigned)__entry->bus,
-		      (unsigned)__entry->slot,
-		      (unsigned)__entry->func,
-		      (unsigned long long)__entry->dma,
-		      (unsigned long long)__entry->phys)
-);
-
-TRACE_EVENT(amdgpu_ttm_tt_unpopulate,
-	    TP_PROTO(struct amdgpu_device *adev, uint64_t dma_address, uint64_t phys_address),
-	    TP_ARGS(adev, dma_address, phys_address),
-	    TP_STRUCT__entry(
-				__field(uint16_t, domain)
-				__field(uint8_t, bus)
-				__field(uint8_t, slot)
-				__field(uint8_t, func)
-				__field(uint64_t, dma)
-				__field(uint64_t, phys)
-			    ),
-	    TP_fast_assign(
-			   __entry->domain = pci_domain_nr(adev->pdev->bus);
-			   __entry->bus = adev->pdev->bus->number;
-			   __entry->slot = PCI_SLOT(adev->pdev->devfn);
-			   __entry->func = PCI_FUNC(adev->pdev->devfn);
-			   __entry->dma = dma_address;
-			   __entry->phys = phys_address;
-			   ),
-	    TP_printk("%04x:%02x:%02x.%x: 0x%llx => 0x%llx",
-		      (unsigned)__entry->domain,
-		      (unsigned)__entry->bus,
-		      (unsigned)__entry->slot,
-		      (unsigned)__entry->func,
-		      (unsigned long long)__entry->dma,
-		      (unsigned long long)__entry->phys)
-);
-
 TRACE_EVENT(amdgpu_mm_rreg,
 	    TP_PROTO(unsigned did, uint32_t reg, uint32_t value),
 	    TP_ARGS(did, reg, value),
@@ -474,5 +418,5 @@ TRACE_EVENT(amdgpu_ttm_bo_move,
 
 /* This part must be outside protection */
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH .
+#define TRACE_INCLUDE_PATH ../../drivers/gpu/drm/amd/amdgpu
 #include <trace/define_trace.h>

drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c

@@ -1,5 +1,24 @@
 // SPDX-License-Identifier: GPL-2.0
 /* Copyright Red Hat Inc 2010.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
  * Author : Dave Airlie <airlied@redhat.com>
  */
 #include <drm/drmP.h>

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

@@ -42,7 +42,9 @@
 #include <linux/swap.h>
 #include <linux/pagemap.h>
 #include <linux/debugfs.h>
+#include <linux/iommu.h>
 #include "amdgpu.h"
+#include "amdgpu_object.h"
 #include "amdgpu_trace.h"
 #include "bif/bif_4_1_d.h"
 
@@ -208,7 +210,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
 		placement->num_busy_placement = 1;
 		return;
 	}
-	abo = container_of(bo, struct amdgpu_bo, tbo);
+	abo = ttm_to_amdgpu_bo(bo);
 	switch (bo->mem.mem_type) {
 	case TTM_PL_VRAM:
 		if (adev->mman.buffer_funcs &&
@@ -256,7 +258,7 @@ gtt:
 
 static int amdgpu_verify_access(struct ttm_buffer_object *bo, struct file *filp)
 {
-	struct amdgpu_bo *abo = container_of(bo, struct amdgpu_bo, tbo);
+	struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo);
 
 	if (amdgpu_ttm_tt_get_usermm(bo->ttm))
 		return -EPERM;
@@ -288,97 +290,177 @@ static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo,
 	return addr;
 }
 
-static int amdgpu_move_blit(struct ttm_buffer_object *bo,
-			    bool evict, bool no_wait_gpu,
-			    struct ttm_mem_reg *new_mem,
-			    struct ttm_mem_reg *old_mem)
+/**
+ * amdgpu_find_mm_node - Helper function finds the drm_mm_node
+ *  corresponding to @offset. It also modifies the offset to be
+ *  within the drm_mm_node returned
+ */
+static struct drm_mm_node *amdgpu_find_mm_node(struct ttm_mem_reg *mem,
+					       unsigned long *offset)
 {
-	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
-	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
+	struct drm_mm_node *mm_node = mem->mm_node;
 
-	struct drm_mm_node *old_mm, *new_mm;
-	uint64_t old_start, old_size, new_start, new_size;
-	unsigned long num_pages;
-	struct dma_fence *fence = NULL;
-	int r;
+	while (*offset >= (mm_node->size << PAGE_SHIFT)) {
+		*offset -= (mm_node->size << PAGE_SHIFT);
+		++mm_node;
+	}
+	return mm_node;
+}
 
-	BUILD_BUG_ON((PAGE_SIZE % AMDGPU_GPU_PAGE_SIZE) != 0);
+/**
+ * amdgpu_copy_ttm_mem_to_mem - Helper function for copy
+ *
+ * The function copies @size bytes from {src->mem + src->offset} to
+ * {dst->mem + dst->offset}. src->bo and dst->bo could be same BO for a
+ * move and different for a BO to BO copy.
+ *
+ * @f: Returns the last fence if multiple jobs are submitted.
+ */
+int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
+			       struct amdgpu_copy_mem *src,
+			       struct amdgpu_copy_mem *dst,
+			       uint64_t size,
+			       struct reservation_object *resv,
+			       struct dma_fence **f)
+{
+	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
+	struct drm_mm_node *src_mm, *dst_mm;
+	uint64_t src_node_start, dst_node_start, src_node_size,
+		 dst_node_size, src_page_offset, dst_page_offset;
+	struct dma_fence *fence = NULL;
+	int r = 0;
+	const uint64_t GTT_MAX_BYTES = (AMDGPU_GTT_MAX_TRANSFER_SIZE *
+					AMDGPU_GPU_PAGE_SIZE);
 
 	if (!ring->ready) {
 		DRM_ERROR("Trying to move memory with ring turned off.\n");
 		return -EINVAL;
 	}
 
-	old_mm = old_mem->mm_node;
-	old_size = old_mm->size;
-	old_start = amdgpu_mm_node_addr(bo, old_mm, old_mem);
+	src_mm = amdgpu_find_mm_node(src->mem, &src->offset);
+	src_node_start = amdgpu_mm_node_addr(src->bo, src_mm, src->mem) +
+					     src->offset;
+	src_node_size = (src_mm->size << PAGE_SHIFT) - src->offset;
+	src_page_offset = src_node_start & (PAGE_SIZE - 1);
 
-	new_mm = new_mem->mm_node;
-	new_size = new_mm->size;
-	new_start = amdgpu_mm_node_addr(bo, new_mm, new_mem);
+	dst_mm = amdgpu_find_mm_node(dst->mem, &dst->offset);
+	dst_node_start = amdgpu_mm_node_addr(dst->bo, dst_mm, dst->mem) +
+					     dst->offset;
+	dst_node_size = (dst_mm->size << PAGE_SHIFT) - dst->offset;
+	dst_page_offset = dst_node_start & (PAGE_SIZE - 1);
 
-	num_pages = new_mem->num_pages;
 	mutex_lock(&adev->mman.gtt_window_lock);
-	while (num_pages) {
-		unsigned long cur_pages = min(min(old_size, new_size),
-					      (u64)AMDGPU_GTT_MAX_TRANSFER_SIZE);
-		uint64_t from = old_start, to = new_start;
+
+	while (size) {
+		unsigned long cur_size;
+		uint64_t from = src_node_start, to = dst_node_start;
 		struct dma_fence *next;
 
-		if (old_mem->mem_type == TTM_PL_TT &&
-		    !amdgpu_gtt_mgr_is_allocated(old_mem)) {
-			r = amdgpu_map_buffer(bo, old_mem, cur_pages,
-					      old_start, 0, ring, &from);
+		/* Copy size cannot exceed GTT_MAX_BYTES. So if src or dst
+		 * begins at an offset, then adjust the size accordingly
+		 */
+		cur_size = min3(min(src_node_size, dst_node_size), size,
+				GTT_MAX_BYTES);
+		if (cur_size + src_page_offset > GTT_MAX_BYTES ||
+		    cur_size + dst_page_offset > GTT_MAX_BYTES)
+			cur_size -= max(src_page_offset, dst_page_offset);
+
+		/* Map only what needs to be accessed. Map src to window 0 and
+		 * dst to window 1
+		 */
+		if (src->mem->mem_type == TTM_PL_TT &&
+		    !amdgpu_gtt_mgr_is_allocated(src->mem)) {
+			r = amdgpu_map_buffer(src->bo, src->mem,
+					PFN_UP(cur_size + src_page_offset),
+					src_node_start, 0, ring,
+					&from);
 			if (r)
 				goto error;
+			/* Adjust the offset because amdgpu_map_buffer returns
+			 * start of mapped page
+			 */
+			from += src_page_offset;
 		}
 
-		if (new_mem->mem_type == TTM_PL_TT &&
-		    !amdgpu_gtt_mgr_is_allocated(new_mem)) {
-			r = amdgpu_map_buffer(bo, new_mem, cur_pages,
-					      new_start, 1, ring, &to);
+		if (dst->mem->mem_type == TTM_PL_TT &&
+		    !amdgpu_gtt_mgr_is_allocated(dst->mem)) {
+			r = amdgpu_map_buffer(dst->bo, dst->mem,
+					PFN_UP(cur_size + dst_page_offset),
+					dst_node_start, 1, ring,
+					&to);
 			if (r)
 				goto error;
+			to += dst_page_offset;
 		}
 
-		r = amdgpu_copy_buffer(ring, from, to,
-				       cur_pages * PAGE_SIZE,
-				       bo->resv, &next, false, true);
+		r = amdgpu_copy_buffer(ring, from, to, cur_size,
+				       resv, &next, false, true);
 		if (r)
 			goto error;
 
 		dma_fence_put(fence);
 		fence = next;
 
-		num_pages -= cur_pages;
-		if (!num_pages)
+		size -= cur_size;
+		if (!size)
 			break;
 
-		old_size -= cur_pages;
-		if (!old_size) {
-			old_start = amdgpu_mm_node_addr(bo, ++old_mm, old_mem);
-			old_size = old_mm->size;
+		src_node_size -= cur_size;
+		if (!src_node_size) {
+			src_node_start = amdgpu_mm_node_addr(src->bo, ++src_mm,
+							     src->mem);
+			src_node_size = (src_mm->size << PAGE_SHIFT);
 		} else {
-			old_start += cur_pages * PAGE_SIZE;
+			src_node_start += cur_size;
+			src_page_offset = src_node_start & (PAGE_SIZE - 1);
 		}
-
-		new_size -= cur_pages;
-		if (!new_size) {
-			new_start = amdgpu_mm_node_addr(bo, ++new_mm, new_mem);
-			new_size = new_mm->size;
+		dst_node_size -= cur_size;
+		if (!dst_node_size) {
+			dst_node_start = amdgpu_mm_node_addr(dst->bo, ++dst_mm,
+							     dst->mem);
+			dst_node_size = (dst_mm->size << PAGE_SHIFT);
 		} else {
-			new_start += cur_pages * PAGE_SIZE;
+			dst_node_start += cur_size;
+			dst_page_offset = dst_node_start & (PAGE_SIZE - 1);
 		}
 	}
+error:
 	mutex_unlock(&adev->mman.gtt_window_lock);
+	if (f)
+		*f = dma_fence_get(fence);
+	dma_fence_put(fence);
+	return r;
+}
+
+
+static int amdgpu_move_blit(struct ttm_buffer_object *bo,
+			    bool evict, bool no_wait_gpu,
+			    struct ttm_mem_reg *new_mem,
+			    struct ttm_mem_reg *old_mem)
+{
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
+	struct amdgpu_copy_mem src, dst;
+	struct dma_fence *fence = NULL;
+	int r;
+
+	src.bo = bo;
+	dst.bo = bo;
+	src.mem = old_mem;
+	dst.mem = new_mem;
+	src.offset = 0;
+	dst.offset = 0;
+
+	r = amdgpu_ttm_copy_mem_to_mem(adev, &src, &dst,
+				       new_mem->num_pages << PAGE_SHIFT,
+				       bo->resv, &fence);
+	if (r)
+		goto error;
 
 	r = ttm_bo_pipeline_move(bo, fence, evict, new_mem);
 	dma_fence_put(fence);
 	return r;
 
 error:
-	mutex_unlock(&adev->mman.gtt_window_lock);
-
 	if (fence)
 		dma_fence_wait(fence, false);
 	dma_fence_put(fence);
@@ -483,7 +565,7 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo,
 	int r;
 
 	/* Can't move a pinned BO */
-	abo = container_of(bo, struct amdgpu_bo, tbo);
+	abo = ttm_to_amdgpu_bo(bo);
 	if (WARN_ON_ONCE(abo->pin_count > 0))
 		return -EINVAL;
 
@@ -581,13 +663,12 @@ static void amdgpu_ttm_io_mem_free(struct ttm_bo_device *bdev, struct ttm_mem_re
 static unsigned long amdgpu_ttm_io_mem_pfn(struct ttm_buffer_object *bo,
 					   unsigned long page_offset)
 {
-	struct drm_mm_node *mm = bo->mem.mm_node;
-	uint64_t size = mm->size;
-	uint64_t offset = page_offset;
+	struct drm_mm_node *mm;
+	unsigned long offset = (page_offset << PAGE_SHIFT);
 
-	page_offset = do_div(offset, size);
-	mm += offset;
-	return (bo->mem.bus.base >> PAGE_SHIFT) + mm->start + page_offset;
+	mm = amdgpu_find_mm_node(&bo->mem, &offset);
+	return (bo->mem.bus.base >> PAGE_SHIFT) + mm->start +
+		(offset >> PAGE_SHIFT);
 }
 
 /*
@@ -608,6 +689,7 @@ struct amdgpu_ttm_tt {
 	spinlock_t              guptasklock;
 	struct list_head        guptasks;
 	atomic_t		mmu_invalidations;
+	uint32_t		last_set_pages;
 	struct list_head        list;
 };
 
@@ -621,6 +703,8 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 	if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY))
 		flags |= FOLL_WRITE;
 
+	down_read(&current->mm->mmap_sem);
+
 	if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) {
 		/* check that we only use anonymous memory
 		   to prevent problems with writeback */
@@ -628,8 +712,10 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 		struct vm_area_struct *vma;
 
 		vma = find_vma(gtt->usermm, gtt->userptr);
-		if (!vma || vma->vm_file || vma->vm_end < end)
+		if (!vma || vma->vm_file || vma->vm_end < end) {
+			up_read(&current->mm->mmap_sem);
 			return -EPERM;
+		}
 	}
 
 	do {
@@ -656,42 +742,44 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 
 	} while (pinned < ttm->num_pages);
 
+	up_read(&current->mm->mmap_sem);
 	return 0;
 
 release_pages:
 	release_pages(pages, pinned);
+	up_read(&current->mm->mmap_sem);
 	return r;
 }
 
-static void amdgpu_trace_dma_map(struct ttm_tt *ttm)
+void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages)
 {
-	struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
 	unsigned i;
 
-	if (unlikely(trace_amdgpu_ttm_tt_populate_enabled())) {
-		for (i = 0; i < ttm->num_pages; i++) {
-			trace_amdgpu_ttm_tt_populate(
-				adev,
-				gtt->ttm.dma_address[i],
-				page_to_phys(ttm->pages[i]));
-		}
+	gtt->last_set_pages = atomic_read(&gtt->mmu_invalidations);
+	for (i = 0; i < ttm->num_pages; ++i) {
+		if (ttm->pages[i])
+			put_page(ttm->pages[i]);
+
+		ttm->pages[i] = pages ? pages[i] : NULL;
 	}
 }
 
-static void amdgpu_trace_dma_unmap(struct ttm_tt *ttm)
+void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm)
 {
-	struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
 	unsigned i;
 
-	if (unlikely(trace_amdgpu_ttm_tt_unpopulate_enabled())) {
-		for (i = 0; i < ttm->num_pages; i++) {
-			trace_amdgpu_ttm_tt_unpopulate(
-				adev,
-				gtt->ttm.dma_address[i],
-				page_to_phys(ttm->pages[i]));
-		}
+	for (i = 0; i < ttm->num_pages; ++i) {
+		struct page *page = ttm->pages[i];
+
+		if (!page)
+			continue;
+
+		if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY))
+			set_page_dirty(page);
+
+		mark_page_accessed(page);
 	}
 }
 
@@ -721,8 +809,6 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm)
 	drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages,
 					 gtt->ttm.dma_address, ttm->num_pages);
 
-	amdgpu_trace_dma_map(ttm);
-
 	return 0;
 
 release_sg:
@@ -734,7 +820,6 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-	struct sg_page_iter sg_iter;
 
 	int write = !(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY);
 	enum dma_data_direction direction = write ?
@@ -747,16 +832,7 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm)
 	/* free the sg table and pages again */
 	dma_unmap_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction);
 
-	for_each_sg_page(ttm->sg->sgl, &sg_iter, ttm->sg->nents, 0) {
-		struct page *page = sg_page_iter_page(&sg_iter);
-		if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY))
-			set_page_dirty(page);
-
-		mark_page_accessed(page);
-		put_page(page);
-	}
-
-	amdgpu_trace_dma_unmap(ttm);
+	amdgpu_ttm_tt_mark_user_pages(ttm);
 
 	sg_free_table(ttm->sg);
 }
@@ -818,7 +894,6 @@ int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem)
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
 	struct ttm_tt *ttm = bo->ttm;
 	struct ttm_mem_reg tmp;
-
 	struct ttm_placement placement;
 	struct ttm_place placements;
 	int r;
@@ -834,7 +909,8 @@ int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem)
 	placement.busy_placement = &placements;
 	placements.fpfn = 0;
 	placements.lpfn = adev->mc.gart_size >> PAGE_SHIFT;
-	placements.flags = bo->mem.placement | TTM_PL_FLAG_TT;
+	placements.flags = (bo->mem.placement & ~TTM_PL_MASK_MEM) |
+		TTM_PL_FLAG_TT;
 
 	r = ttm_bo_mem_space(bo, &placement, &tmp, true, false);
 	if (unlikely(r))
@@ -941,8 +1017,6 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-	unsigned i;
-	int r;
 	bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
 
 	if (ttm->state != tt_unpopulated)
@@ -962,52 +1036,26 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm)
 		drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages,
 						 gtt->ttm.dma_address, ttm->num_pages);
 		ttm->state = tt_unbound;
-		r = 0;
-		goto trace_mappings;
+		return 0;
 	}
 
 #ifdef CONFIG_SWIOTLB
 	if (swiotlb_nr_tbl()) {
-		r = ttm_dma_populate(&gtt->ttm, adev->dev);
-		goto trace_mappings;
+		return ttm_dma_populate(&gtt->ttm, adev->dev);
 	}
 #endif
 
-	r = ttm_pool_populate(ttm);
-	if (r) {
-		return r;
-	}
-
-	for (i = 0; i < ttm->num_pages; i++) {
-		gtt->ttm.dma_address[i] = pci_map_page(adev->pdev, ttm->pages[i],
-						       0, PAGE_SIZE,
-						       PCI_DMA_BIDIRECTIONAL);
-		if (pci_dma_mapping_error(adev->pdev, gtt->ttm.dma_address[i])) {
-			while (i--) {
-				pci_unmap_page(adev->pdev, gtt->ttm.dma_address[i],
-					       PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
-				gtt->ttm.dma_address[i] = 0;
-			}
-			ttm_pool_unpopulate(ttm);
-			return -EFAULT;
-		}
-	}
-
-	r = 0;
-trace_mappings:
-	if (likely(!r))
-		amdgpu_trace_dma_map(ttm);
-	return r;
+	return ttm_populate_and_map_pages(adev->dev, &gtt->ttm);
 }
 
 static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm)
 {
 	struct amdgpu_device *adev;
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-	unsigned i;
 	bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
 
 	if (gtt && gtt->userptr) {
+		amdgpu_ttm_tt_set_user_pages(ttm, NULL);
 		kfree(ttm->sg);
 		ttm->page_flags &= ~TTM_PAGE_FLAG_SG;
 		return;
@@ -1018,8 +1066,6 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm)
 
 	adev = amdgpu_ttm_adev(ttm->bdev);
 
-	amdgpu_trace_dma_unmap(ttm);
-
 #ifdef CONFIG_SWIOTLB
 	if (swiotlb_nr_tbl()) {
 		ttm_dma_unpopulate(&gtt->ttm, adev->dev);
@@ -1027,14 +1073,7 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm)
 	}
 #endif
 
-	for (i = 0; i < ttm->num_pages; i++) {
-		if (gtt->ttm.dma_address[i]) {
-			pci_unmap_page(adev->pdev, gtt->ttm.dma_address[i],
-				       PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
-		}
-	}
-
-	ttm_pool_unpopulate(ttm);
+	ttm_unmap_and_unpopulate_pages(adev->dev, &gtt->ttm);
 }
 
 int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
@@ -1051,6 +1090,7 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
 	spin_lock_init(&gtt->guptasklock);
 	INIT_LIST_HEAD(&gtt->guptasks);
 	atomic_set(&gtt->mmu_invalidations, 0);
+	gtt->last_set_pages = 0;
 
 	return 0;
 }
@@ -1103,6 +1143,16 @@ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
 	return prev_invalidated != *last_invalidated;
 }
 
+bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm)
+{
+	struct amdgpu_ttm_tt *gtt = (void *)ttm;
+
+	if (gtt == NULL || !gtt->userptr)
+		return false;
+
+	return atomic_read(&gtt->mmu_invalidations) != gtt->last_set_pages;
+}
+
 bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm)
 {
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
@@ -1143,9 +1193,6 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo,
 	unsigned long num_pages = bo->mem.num_pages;
 	struct drm_mm_node *node = bo->mem.mm_node;
 
-	if (bo->mem.start != AMDGPU_BO_INVALID_OFFSET)
-		return ttm_bo_eviction_valuable(bo, place);
-
 	switch (bo->mem.mem_type) {
 	case TTM_PL_TT:
 		return true;
@@ -1160,7 +1207,7 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo,
 			num_pages -= node->size;
 			++node;
 		}
-		break;
+		return false;
 
 	default:
 		break;
@@ -1173,9 +1220,9 @@ static int amdgpu_ttm_access_memory(struct ttm_buffer_object *bo,
 				    unsigned long offset,
 				    void *buf, int len, int write)
 {
-	struct amdgpu_bo *abo = container_of(bo, struct amdgpu_bo, tbo);
+	struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo);
 	struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev);
-	struct drm_mm_node *nodes = abo->tbo.mem.mm_node;
+	struct drm_mm_node *nodes;
 	uint32_t value = 0;
 	int ret = 0;
 	uint64_t pos;
@@ -1184,10 +1231,7 @@ static int amdgpu_ttm_access_memory(struct ttm_buffer_object *bo,
 	if (bo->mem.mem_type != TTM_PL_VRAM)
 		return -EIO;
 
-	while (offset >= (nodes->size << PAGE_SHIFT)) {
-		offset -= nodes->size << PAGE_SHIFT;
-		++nodes;
-	}
+	nodes = amdgpu_find_mm_node(&abo->tbo.mem, &offset);
 	pos = (nodes->start << PAGE_SHIFT) + offset;
 
 	while (len && pos < adev->mc.mc_vram_size) {
@@ -1202,14 +1246,14 @@ static int amdgpu_ttm_access_memory(struct ttm_buffer_object *bo,
 		}
 
 		spin_lock_irqsave(&adev->mmio_idx_lock, flags);
-		WREG32(mmMM_INDEX, ((uint32_t)aligned_pos) | 0x80000000);
-		WREG32(mmMM_INDEX_HI, aligned_pos >> 31);
+		WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)aligned_pos) | 0x80000000);
+		WREG32_NO_KIQ(mmMM_INDEX_HI, aligned_pos >> 31);
 		if (!write || mask != 0xffffffff)
-			value = RREG32(mmMM_DATA);
+			value = RREG32_NO_KIQ(mmMM_DATA);
 		if (write) {
 			value &= ~mask;
 			value |= (*(uint32_t *)buf << shift) & mask;
-			WREG32(mmMM_DATA, value);
+			WREG32_NO_KIQ(mmMM_DATA, value);
 		}
 		spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
 		if (!write) {
@@ -1286,6 +1330,15 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 	/* Change the size here instead of the init above so only lpfn is affected */
 	amdgpu_ttm_set_active_vram_size(adev, adev->mc.visible_vram_size);
 
+	/*
+	 *The reserved vram for firmware must be pinned to the specified
+	 *place on the VRAM, so reserve it early.
+	 */
+	r = amdgpu_fw_reserve_vram_init(adev);
+	if (r) {
+		return r;
+	}
+
 	r = amdgpu_bo_create_kernel(adev, adev->mc.stolen_size, PAGE_SIZE,
 				    AMDGPU_GEM_DOMAIN_VRAM,
 				    &adev->stolen_vga_memory,
@@ -1510,7 +1563,8 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 	job->vm_needs_flush = vm_needs_flush;
 	if (resv) {
 		r = amdgpu_sync_resv(adev, &job->sync, resv,
-				     AMDGPU_FENCE_OWNER_UNDEFINED);
+				     AMDGPU_FENCE_OWNER_UNDEFINED,
+				     false);
 		if (r) {
 			DRM_ERROR("sync failed (%d).\n", r);
 			goto error_free;
@@ -1557,8 +1611,8 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 		       struct dma_fence **fence)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	/* max_bytes applies to SDMA_OP_PTEPDE as well as SDMA_OP_CONST_FILL*/
-	uint32_t max_bytes = adev->mman.buffer_funcs->fill_max_bytes;
+	uint32_t max_bytes = 8 *
+			adev->vm_manager.vm_pte_funcs->set_max_nums_pte_pde;
 	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
 
 	struct drm_mm_node *mm_node;
@@ -1590,8 +1644,8 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 		++mm_node;
 	}
 
-	/* 10 double words for each SDMA_OP_PTEPDE cmd */
-	num_dw = num_loops * 10;
+	/* num of dwords for each SDMA_OP_PTEPDE cmd */
+	num_dw = num_loops * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw;
 
 	/* for IB padding */
 	num_dw += 64;
@@ -1602,7 +1656,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 
 	if (resv) {
 		r = amdgpu_sync_resv(adev, &job->sync, resv,
-				     AMDGPU_FENCE_OWNER_UNDEFINED);
+				     AMDGPU_FENCE_OWNER_UNDEFINED, false);
 		if (r) {
 			DRM_ERROR("sync failed (%d).\n", r);
 			goto error_free;
@@ -1697,9 +1751,9 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf,
 			return result;
 
 		spin_lock_irqsave(&adev->mmio_idx_lock, flags);
-		WREG32(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000);
-		WREG32(mmMM_INDEX_HI, *pos >> 31);
-		value = RREG32(mmMM_DATA);
+		WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000);
+		WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31);
+		value = RREG32_NO_KIQ(mmMM_DATA);
 		spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
 
 		r = put_user(value, (uint32_t *)buf);
@@ -1715,10 +1769,50 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf,
 	return result;
 }
 
+static ssize_t amdgpu_ttm_vram_write(struct file *f, const char __user *buf,
+				    size_t size, loff_t *pos)
+{
+	struct amdgpu_device *adev = file_inode(f)->i_private;
+	ssize_t result = 0;
+	int r;
+
+	if (size & 0x3 || *pos & 0x3)
+		return -EINVAL;
+
+	if (*pos >= adev->mc.mc_vram_size)
+		return -ENXIO;
+
+	while (size) {
+		unsigned long flags;
+		uint32_t value;
+
+		if (*pos >= adev->mc.mc_vram_size)
+			return result;
+
+		r = get_user(value, (uint32_t *)buf);
+		if (r)
+			return r;
+
+		spin_lock_irqsave(&adev->mmio_idx_lock, flags);
+		WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000);
+		WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31);
+		WREG32_NO_KIQ(mmMM_DATA, value);
+		spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
+
+		result += 4;
+		buf += 4;
+		*pos += 4;
+		size -= 4;
+	}
+
+	return result;
+}
+
 static const struct file_operations amdgpu_ttm_vram_fops = {
 	.owner = THIS_MODULE,
 	.read = amdgpu_ttm_vram_read,
-	.llseek = default_llseek
+	.write = amdgpu_ttm_vram_write,
+	.llseek = default_llseek,
 };
 
 #ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
@@ -1770,6 +1864,53 @@ static const struct file_operations amdgpu_ttm_gtt_fops = {
 
 #endif
 
+static ssize_t amdgpu_iova_to_phys_read(struct file *f, char __user *buf,
+				   size_t size, loff_t *pos)
+{
+	struct amdgpu_device *adev = file_inode(f)->i_private;
+	int r;
+	uint64_t phys;
+	struct iommu_domain *dom;
+
+	// always return 8 bytes
+	if (size != 8)
+		return -EINVAL;
+
+	// only accept page addresses
+	if (*pos & 0xFFF)
+		return -EINVAL;
+
+	dom = iommu_get_domain_for_dev(adev->dev);
+	if (dom)
+		phys = iommu_iova_to_phys(dom, *pos);
+	else
+		phys = *pos;
+
+	r = copy_to_user(buf, &phys, 8);
+	if (r)
+		return -EFAULT;
+
+	return 8;
+}
+
+static const struct file_operations amdgpu_ttm_iova_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_iova_to_phys_read,
+	.llseek = default_llseek
+};
+
+static const struct {
+	char *name;
+	const struct file_operations *fops;
+	int domain;
+} ttm_debugfs_entries[] = {
+	{ "amdgpu_vram", &amdgpu_ttm_vram_fops, TTM_PL_VRAM },
+#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
+	{ "amdgpu_gtt", &amdgpu_ttm_gtt_fops, TTM_PL_TT },
+#endif
+	{ "amdgpu_iova", &amdgpu_ttm_iova_fops, TTM_PL_SYSTEM },
+};
+
 #endif
 
 static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
@@ -1780,22 +1921,21 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
 	struct drm_minor *minor = adev->ddev->primary;
 	struct dentry *ent, *root = minor->debugfs_root;
 
-	ent = debugfs_create_file("amdgpu_vram", S_IFREG | S_IRUGO, root,
-				  adev, &amdgpu_ttm_vram_fops);
-	if (IS_ERR(ent))
-		return PTR_ERR(ent);
-	i_size_write(ent->d_inode, adev->mc.mc_vram_size);
-	adev->mman.vram = ent;
-
-#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
-	ent = debugfs_create_file("amdgpu_gtt", S_IFREG | S_IRUGO, root,
-				  adev, &amdgpu_ttm_gtt_fops);
-	if (IS_ERR(ent))
-		return PTR_ERR(ent);
-	i_size_write(ent->d_inode, adev->mc.gart_size);
-	adev->mman.gtt = ent;
+	for (count = 0; count < ARRAY_SIZE(ttm_debugfs_entries); count++) {
+		ent = debugfs_create_file(
+				ttm_debugfs_entries[count].name,
+				S_IFREG | S_IRUGO, root,
+				adev,
+				ttm_debugfs_entries[count].fops);
+		if (IS_ERR(ent))
+			return PTR_ERR(ent);
+		if (ttm_debugfs_entries[count].domain == TTM_PL_VRAM)
+			i_size_write(ent->d_inode, adev->mc.mc_vram_size);
+		else if (ttm_debugfs_entries[count].domain == TTM_PL_TT)
+			i_size_write(ent->d_inode, adev->mc.gart_size);
+		adev->mman.debugfs_entries[count] = ent;
+	}
 
-#endif
 	count = ARRAY_SIZE(amdgpu_ttm_debugfs_list);
 
 #ifdef CONFIG_SWIOTLB
@@ -1805,7 +1945,6 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
 
 	return amdgpu_debugfs_add_files(adev, amdgpu_ttm_debugfs_list, count);
 #else
-
 	return 0;
 #endif
 }
@@ -1813,14 +1952,9 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev)
 static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev)
 {
 #if defined(CONFIG_DEBUG_FS)
+	unsigned i;
 
-	debugfs_remove(adev->mman.vram);
-	adev->mman.vram = NULL;
-
-#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
-	debugfs_remove(adev->mman.gtt);
-	adev->mman.gtt = NULL;
-#endif
-
+	for (i = 0; i < ARRAY_SIZE(ttm_debugfs_entries); i++)
+		debugfs_remove(adev->mman.debugfs_entries[i]);
 #endif
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h

@@ -24,6 +24,7 @@
 #ifndef __AMDGPU_TTM_H__
 #define __AMDGPU_TTM_H__
 
+#include "amdgpu.h"
 #include "gpu_scheduler.h"
 
 #define AMDGPU_PL_GDS		(TTM_PL_PRIV + 0)
@@ -45,8 +46,7 @@ struct amdgpu_mman {
 	bool				initialized;
 
 #if defined(CONFIG_DEBUG_FS)
-	struct dentry			*vram;
-	struct dentry			*gtt;
+	struct dentry			*debugfs_entries[8];
 #endif
 
 	/* buffer handling */
@@ -58,6 +58,12 @@ struct amdgpu_mman {
 	struct amd_sched_entity			entity;
 };
 
+struct amdgpu_copy_mem {
+	struct ttm_buffer_object	*bo;
+	struct ttm_mem_reg		*mem;
+	unsigned long			offset;
+};
+
 extern const struct ttm_mem_type_manager_func amdgpu_gtt_mgr_func;
 extern const struct ttm_mem_type_manager_func amdgpu_vram_mgr_func;
 
@@ -72,6 +78,12 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 		       struct reservation_object *resv,
 		       struct dma_fence **fence, bool direct_submit,
 		       bool vm_needs_flush);
+int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
+			       struct amdgpu_copy_mem *src,
+			       struct amdgpu_copy_mem *dst,
+			       uint64_t size,
+			       struct reservation_object *resv,
+			       struct dma_fence **f);
 int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 			uint64_t src_data,
 			struct reservation_object *resv,
@@ -82,4 +94,20 @@ bool amdgpu_ttm_is_bound(struct ttm_tt *ttm);
 int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem);
 int amdgpu_ttm_recover_gart(struct amdgpu_device *adev);
 
+int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages);
+void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages);
+void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm);
+int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
+				     uint32_t flags);
+bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm);
+struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm);
+bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
+				  unsigned long end);
+bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
+				       int *last_invalidated);
+bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm);
+bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm);
+uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm,
+				 struct ttm_mem_reg *mem);
+
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c

@@ -270,12 +270,8 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type)
 		else
 			return AMDGPU_FW_LOAD_SMU;
 	case CHIP_VEGA10:
-		if (!load_type)
-			return AMDGPU_FW_LOAD_DIRECT;
-		else
-			return AMDGPU_FW_LOAD_PSP;
 	case CHIP_RAVEN:
-		if (load_type != 2)
+		if (!load_type)
 			return AMDGPU_FW_LOAD_DIRECT;
 		else
 			return AMDGPU_FW_LOAD_PSP;
@@ -364,8 +360,6 @@ static int amdgpu_ucode_patch_jt(struct amdgpu_firmware_info *ucode,
 int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 {
 	struct amdgpu_bo **bo = &adev->firmware.fw_buf;
-	uint64_t fw_mc_addr;
-	void *fw_buf_ptr = NULL;
 	uint64_t fw_offset = 0;
 	int i, err;
 	struct amdgpu_firmware_info *ucode = NULL;
@@ -376,37 +370,39 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 		return 0;
 	}
 
-	err = amdgpu_bo_create(adev, adev->firmware.fw_size, PAGE_SIZE, true,
-				amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
-				AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
-				NULL, NULL, 0, bo);
-	if (err) {
-		dev_err(adev->dev, "(%d) Firmware buffer allocate failed\n", err);
-		goto failed;
-	}
+	if (!amdgpu_sriov_vf(adev) || !adev->in_sriov_reset) {
+		err = amdgpu_bo_create(adev, adev->firmware.fw_size, PAGE_SIZE, true,
+					amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
+					AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
+					NULL, NULL, 0, bo);
+		if (err) {
+			dev_err(adev->dev, "(%d) Firmware buffer allocate failed\n", err);
+			goto failed;
+		}
 
-	err = amdgpu_bo_reserve(*bo, false);
-	if (err) {
-		dev_err(adev->dev, "(%d) Firmware buffer reserve failed\n", err);
-		goto failed_reserve;
-	}
+		err = amdgpu_bo_reserve(*bo, false);
+		if (err) {
+			dev_err(adev->dev, "(%d) Firmware buffer reserve failed\n", err);
+			goto failed_reserve;
+		}
 
-	err = amdgpu_bo_pin(*bo, amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
-				&fw_mc_addr);
-	if (err) {
-		dev_err(adev->dev, "(%d) Firmware buffer pin failed\n", err);
-		goto failed_pin;
-	}
+		err = amdgpu_bo_pin(*bo, amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
+					&adev->firmware.fw_buf_mc);
+		if (err) {
+			dev_err(adev->dev, "(%d) Firmware buffer pin failed\n", err);
+			goto failed_pin;
+		}
 
-	err = amdgpu_bo_kmap(*bo, &fw_buf_ptr);
-	if (err) {
-		dev_err(adev->dev, "(%d) Firmware buffer kmap failed\n", err);
-		goto failed_kmap;
-	}
+		err = amdgpu_bo_kmap(*bo, &adev->firmware.fw_buf_ptr);
+		if (err) {
+			dev_err(adev->dev, "(%d) Firmware buffer kmap failed\n", err);
+			goto failed_kmap;
+		}
 
-	amdgpu_bo_unreserve(*bo);
+		amdgpu_bo_unreserve(*bo);
+	}
 
-	memset(fw_buf_ptr, 0, adev->firmware.fw_size);
+	memset(adev->firmware.fw_buf_ptr, 0, adev->firmware.fw_size);
 
 	/*
 	 * if SMU loaded firmware, it needn't add SMC, UVD, and VCE
@@ -425,14 +421,14 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 		ucode = &adev->firmware.ucode[i];
 		if (ucode->fw) {
 			header = (const struct common_firmware_header *)ucode->fw->data;
-			amdgpu_ucode_init_single_fw(adev, ucode, fw_mc_addr + fw_offset,
-						    (void *)((uint8_t *)fw_buf_ptr + fw_offset));
+			amdgpu_ucode_init_single_fw(adev, ucode, adev->firmware.fw_buf_mc + fw_offset,
+						    adev->firmware.fw_buf_ptr + fw_offset);
 			if (i == AMDGPU_UCODE_ID_CP_MEC1 &&
 			    adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
 				const struct gfx_firmware_header_v1_0 *cp_hdr;
 				cp_hdr = (const struct gfx_firmware_header_v1_0 *)ucode->fw->data;
-				amdgpu_ucode_patch_jt(ucode, fw_mc_addr + fw_offset,
-						    fw_buf_ptr + fw_offset);
+				amdgpu_ucode_patch_jt(ucode,  adev->firmware.fw_buf_mc + fw_offset,
+						    adev->firmware.fw_buf_ptr + fw_offset);
 				fw_offset += ALIGN(le32_to_cpu(cp_hdr->jt_size) << 2, PAGE_SIZE);
 			}
 			fw_offset += ALIGN(ucode->ucode_size, PAGE_SIZE);

drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c

@@ -269,6 +269,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 
 int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
 {
+	int i;
 	kfree(adev->uvd.saved_bo);
 
 	amd_sched_entity_fini(&adev->uvd.ring.sched, &adev->uvd.entity);
@@ -279,6 +280,9 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
 
 	amdgpu_ring_fini(&adev->uvd.ring);
 
+	for (i = 0; i < AMDGPU_MAX_UVD_ENC_RINGS; ++i)
+		amdgpu_ring_fini(&adev->uvd.ring_enc[i]);
+
 	release_firmware(adev->uvd.fw);
 
 	return 0;
@@ -410,10 +414,10 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx)
 	uint64_t addr = amdgpu_uvd_get_addr_from_ctx(ctx);
 	int r = 0;
 
-	mapping = amdgpu_cs_find_mapping(ctx->parser, addr, &bo);
-	if (mapping == NULL) {
+	r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping);
+	if (r) {
 		DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr);
-		return -EINVAL;
+		return r;
 	}
 
 	if (!ctx->parser->adev->uvd.address_64_bit) {
@@ -737,10 +741,10 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
 	uint64_t addr = amdgpu_uvd_get_addr_from_ctx(ctx);
 	int r;
 
-	mapping = amdgpu_cs_find_mapping(ctx->parser, addr, &bo);
-	if (mapping == NULL) {
+	r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping);
+	if (r) {
 		DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr);
-		return -EINVAL;
+		return r;
 	}
 
 	start = amdgpu_bo_gpu_offset(bo);
@@ -917,10 +921,6 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx)
 		return -EINVAL;
 	}
 
-	r = amdgpu_cs_sysvm_access_required(parser);
-	if (r)
-		return r;
-
 	ctx.parser = parser;
 	ctx.buf_sizes = buf_sizes;
 	ctx.ib_idx = ib_idx;

drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c

@@ -559,6 +559,7 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx,
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_bo *bo;
 	uint64_t addr;
+	int r;
 
 	if (index == 0xffffffff)
 		index = 0;
@@ -567,11 +568,11 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx,
 	       ((uint64_t)amdgpu_get_ib_value(p, ib_idx, hi)) << 32;
 	addr += ((uint64_t)size) * ((uint64_t)index);
 
-	mapping = amdgpu_cs_find_mapping(p, addr, &bo);
-	if (mapping == NULL) {
+	r = amdgpu_cs_find_mapping(p, addr, &bo, &mapping);
+	if (r) {
 		DRM_ERROR("Can't find BO for addr 0x%010Lx %d %d %d %d\n",
 			  addr, lo, hi, size, index);
-		return -EINVAL;
+		return r;
 	}
 
 	if ((addr + (uint64_t)size) >
@@ -647,15 +648,11 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx)
 	uint32_t allocated = 0;
 	uint32_t tmp, handle = 0;
 	uint32_t *size = &tmp;
-	int i, r, idx = 0;
+	int i, r = 0, idx = 0;
 
 	p->job->vm = NULL;
 	ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo);
 
-	r = amdgpu_cs_sysvm_access_required(p);
-	if (r)
-		return r;
-
 	while (idx < ib->length_dw) {
 		uint32_t len = amdgpu_get_ib_value(p, ib_idx, idx);
 		uint32_t cmd = amdgpu_get_ib_value(p, ib_idx, idx + 1);

drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c

@@ -25,30 +25,26 @@
 #include "amdgpu_vf_error.h"
 #include "mxgpu_ai.h"
 
-#define AMDGPU_VF_ERROR_ENTRY_SIZE    16 
-
-/* struct error_entry - amdgpu VF error information. */
-struct amdgpu_vf_error_buffer {
-	int read_count;
-	int write_count;
-	uint16_t code[AMDGPU_VF_ERROR_ENTRY_SIZE];
-	uint16_t flags[AMDGPU_VF_ERROR_ENTRY_SIZE];
-	uint64_t data[AMDGPU_VF_ERROR_ENTRY_SIZE];
-};
-
-struct amdgpu_vf_error_buffer admgpu_vf_errors;
-
-
-void amdgpu_vf_error_put(uint16_t sub_error_code, uint16_t error_flags, uint64_t error_data)
+void amdgpu_vf_error_put(struct amdgpu_device *adev,
+			 uint16_t sub_error_code,
+			 uint16_t error_flags,
+			 uint64_t error_data)
 {
 	int index;
-	uint16_t error_code = AMDGIM_ERROR_CODE(AMDGIM_ERROR_CATEGORY_VF, sub_error_code);
+	uint16_t error_code;
 
-	index = admgpu_vf_errors.write_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
-	admgpu_vf_errors.code [index] = error_code;
-	admgpu_vf_errors.flags [index] = error_flags;
-	admgpu_vf_errors.data [index] = error_data;
-	admgpu_vf_errors.write_count ++;
+	if (!amdgpu_sriov_vf(adev))
+		return;
+
+	error_code = AMDGIM_ERROR_CODE(AMDGIM_ERROR_CATEGORY_VF, sub_error_code);
+
+	mutex_lock(&adev->virt.vf_errors.lock);
+	index = adev->virt.vf_errors.write_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
+	adev->virt.vf_errors.code [index] = error_code;
+	adev->virt.vf_errors.flags [index] = error_flags;
+	adev->virt.vf_errors.data [index] = error_data;
+	adev->virt.vf_errors.write_count ++;
+	mutex_unlock(&adev->virt.vf_errors.lock);
 }
 
 
@@ -58,7 +54,8 @@ void amdgpu_vf_error_trans_all(struct amdgpu_device *adev)
 	u32 data1, data2, data3;
 	int index;
 
-	if ((NULL == adev) || (!amdgpu_sriov_vf(adev)) || (!adev->virt.ops) || (!adev->virt.ops->trans_msg)) {
+	if ((NULL == adev) || (!amdgpu_sriov_vf(adev)) ||
+	    (!adev->virt.ops) || (!adev->virt.ops->trans_msg)) {
 		return;
 	}
 /*
@@ -68,18 +65,22 @@ void amdgpu_vf_error_trans_all(struct amdgpu_device *adev)
 		return;
 	}
 */
+
+	mutex_lock(&adev->virt.vf_errors.lock);
 	/* The errors are overlay of array, correct read_count as full. */
-	if (admgpu_vf_errors.write_count - admgpu_vf_errors.read_count > AMDGPU_VF_ERROR_ENTRY_SIZE) {
-		admgpu_vf_errors.read_count = admgpu_vf_errors.write_count - AMDGPU_VF_ERROR_ENTRY_SIZE;
+	if (adev->virt.vf_errors.write_count - adev->virt.vf_errors.read_count > AMDGPU_VF_ERROR_ENTRY_SIZE) {
+		adev->virt.vf_errors.read_count = adev->virt.vf_errors.write_count - AMDGPU_VF_ERROR_ENTRY_SIZE;
 	}
 
-	while (admgpu_vf_errors.read_count < admgpu_vf_errors.write_count) {
-		index =admgpu_vf_errors.read_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
-		data1 = AMDGIM_ERROR_CODE_FLAGS_TO_MAILBOX (admgpu_vf_errors.code[index], admgpu_vf_errors.flags[index]);
-		data2 = admgpu_vf_errors.data[index] & 0xFFFFFFFF;
-		data3 = (admgpu_vf_errors.data[index] >> 32) & 0xFFFFFFFF;
+	while (adev->virt.vf_errors.read_count < adev->virt.vf_errors.write_count) {
+		index =adev->virt.vf_errors.read_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
+		data1 = AMDGIM_ERROR_CODE_FLAGS_TO_MAILBOX(adev->virt.vf_errors.code[index],
+							   adev->virt.vf_errors.flags[index]);
+		data2 = adev->virt.vf_errors.data[index] & 0xFFFFFFFF;
+		data3 = (adev->virt.vf_errors.data[index] >> 32) & 0xFFFFFFFF;
 
 		adev->virt.ops->trans_msg(adev, IDH_LOG_VF_ERROR, data1, data2, data3);
-		admgpu_vf_errors.read_count ++;
+		adev->virt.vf_errors.read_count ++;
 	}
+	mutex_unlock(&adev->virt.vf_errors.lock);
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h

@@ -56,7 +56,10 @@ enum AMDGIM_ERROR_CATEGORY {
 	AMDGIM_ERROR_CATEGORY_MAX
 };
 
-void amdgpu_vf_error_put(uint16_t sub_error_code, uint16_t error_flags, uint64_t error_data);
+void amdgpu_vf_error_put(struct amdgpu_device *adev,
+			 uint16_t sub_error_code,
+			 uint16_t error_flags,
+			 uint64_t error_data);
 void amdgpu_vf_error_trans_all (struct amdgpu_device *adev);
 
 #endif /* __VF_ERROR_H__ */

drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c

@@ -22,7 +22,7 @@
  */
 
 #include "amdgpu.h"
-#define MAX_KIQ_REG_WAIT	100000
+#define MAX_KIQ_REG_WAIT	100000000 /* in usecs */
 
 int amdgpu_allocate_static_csa(struct amdgpu_device *adev)
 {
@@ -114,27 +114,25 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev)
 uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
 {
 	signed long r;
-	uint32_t val;
-	struct dma_fence *f;
+	unsigned long flags;
+	uint32_t val, seq;
 	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
 	struct amdgpu_ring *ring = &kiq->ring;
 
 	BUG_ON(!ring->funcs->emit_rreg);
 
-	mutex_lock(&kiq->ring_mutex);
+	spin_lock_irqsave(&kiq->ring_lock, flags);
 	amdgpu_ring_alloc(ring, 32);
 	amdgpu_ring_emit_rreg(ring, reg);
-	amdgpu_fence_emit(ring, &f);
+	amdgpu_fence_emit_polling(ring, &seq);
 	amdgpu_ring_commit(ring);
-	mutex_unlock(&kiq->ring_mutex);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
 
-	r = dma_fence_wait_timeout(f, false, msecs_to_jiffies(MAX_KIQ_REG_WAIT));
-	dma_fence_put(f);
+	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
 	if (r < 1) {
-		DRM_ERROR("wait for kiq fence error: %ld.\n", r);
+		DRM_ERROR("wait for kiq fence error: %ld\n", r);
 		return ~0;
 	}
-
 	val = adev->wb.wb[adev->virt.reg_val_offs];
 
 	return val;
@@ -143,23 +141,23 @@ uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
 void amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
 {
 	signed long r;
-	struct dma_fence *f;
+	unsigned long flags;
+	uint32_t seq;
 	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
 	struct amdgpu_ring *ring = &kiq->ring;
 
 	BUG_ON(!ring->funcs->emit_wreg);
 
-	mutex_lock(&kiq->ring_mutex);
+	spin_lock_irqsave(&kiq->ring_lock, flags);
 	amdgpu_ring_alloc(ring, 32);
 	amdgpu_ring_emit_wreg(ring, reg, v);
-	amdgpu_fence_emit(ring, &f);
+	amdgpu_fence_emit_polling(ring, &seq);
 	amdgpu_ring_commit(ring);
-	mutex_unlock(&kiq->ring_mutex);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
 
-	r = dma_fence_wait_timeout(f, false, msecs_to_jiffies(MAX_KIQ_REG_WAIT));
+	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
 	if (r < 1)
-		DRM_ERROR("wait for kiq fence error: %ld.\n", r);
-	dma_fence_put(f);
+		DRM_ERROR("wait for kiq fence error: %ld\n", r);
 }
 
 /**
@@ -274,3 +272,80 @@ void amdgpu_virt_free_mm_table(struct amdgpu_device *adev)
 			      (void *)&adev->virt.mm_table.cpu_addr);
 	adev->virt.mm_table.gpu_addr = 0;
 }
+
+
+int amdgpu_virt_fw_reserve_get_checksum(void *obj,
+					unsigned long obj_size,
+					unsigned int key,
+					unsigned int chksum)
+{
+	unsigned int ret = key;
+	unsigned long i = 0;
+	unsigned char *pos;
+
+	pos = (char *)obj;
+	/* calculate checksum */
+	for (i = 0; i < obj_size; ++i)
+		ret += *(pos + i);
+	/* minus the chksum itself */
+	pos = (char *)&chksum;
+	for (i = 0; i < sizeof(chksum); ++i)
+		ret -= *(pos + i);
+	return ret;
+}
+
+void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev)
+{
+	uint32_t pf2vf_ver = 0;
+	uint32_t pf2vf_size = 0;
+	uint32_t checksum = 0;
+	uint32_t checkval;
+	char *str;
+
+	adev->virt.fw_reserve.p_pf2vf = NULL;
+	adev->virt.fw_reserve.p_vf2pf = NULL;
+
+	if (adev->fw_vram_usage.va != NULL) {
+		adev->virt.fw_reserve.p_pf2vf =
+			(struct amdgim_pf2vf_info_header *)(
+			adev->fw_vram_usage.va + AMDGIM_DATAEXCHANGE_OFFSET);
+		pf2vf_ver = adev->virt.fw_reserve.p_pf2vf->version;
+		AMDGPU_FW_VRAM_PF2VF_READ(adev, header.size, &pf2vf_size);
+		AMDGPU_FW_VRAM_PF2VF_READ(adev, checksum, &checksum);
+
+		/* pf2vf message must be in 4K */
+		if (pf2vf_size > 0 && pf2vf_size < 4096) {
+			checkval = amdgpu_virt_fw_reserve_get_checksum(
+				adev->virt.fw_reserve.p_pf2vf, pf2vf_size,
+				adev->virt.fw_reserve.checksum_key, checksum);
+			if (checkval == checksum) {
+				adev->virt.fw_reserve.p_vf2pf =
+					((void *)adev->virt.fw_reserve.p_pf2vf +
+					pf2vf_size);
+				memset((void *)adev->virt.fw_reserve.p_vf2pf, 0,
+					sizeof(amdgim_vf2pf_info));
+				AMDGPU_FW_VRAM_VF2PF_WRITE(adev, header.version,
+					AMDGPU_FW_VRAM_VF2PF_VER);
+				AMDGPU_FW_VRAM_VF2PF_WRITE(adev, header.size,
+					sizeof(amdgim_vf2pf_info));
+				AMDGPU_FW_VRAM_VF2PF_READ(adev, driver_version,
+					&str);
+#ifdef MODULE
+				if (THIS_MODULE->version != NULL)
+					strcpy(str, THIS_MODULE->version);
+				else
+#endif
+					strcpy(str, "N/A");
+				AMDGPU_FW_VRAM_VF2PF_WRITE(adev, driver_cert,
+					0);
+				AMDGPU_FW_VRAM_VF2PF_WRITE(adev, checksum,
+					amdgpu_virt_fw_reserve_get_checksum(
+					adev->virt.fw_reserve.p_vf2pf,
+					pf2vf_size,
+					adev->virt.fw_reserve.checksum_key, 0));
+			}
+		}
+	}
+}
+
+

drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h

@@ -36,6 +36,18 @@ struct amdgpu_mm_table {
 	uint64_t		gpu_addr;
 };
 
+#define AMDGPU_VF_ERROR_ENTRY_SIZE    16
+
+/* struct error_entry - amdgpu VF error information. */
+struct amdgpu_vf_error_buffer {
+	struct mutex lock;
+	int read_count;
+	int write_count;
+	uint16_t code[AMDGPU_VF_ERROR_ENTRY_SIZE];
+	uint16_t flags[AMDGPU_VF_ERROR_ENTRY_SIZE];
+	uint64_t data[AMDGPU_VF_ERROR_ENTRY_SIZE];
+};
+
 /**
  * struct amdgpu_virt_ops - amdgpu device virt operations
  */
@@ -46,6 +58,179 @@ struct amdgpu_virt_ops {
 	void (*trans_msg)(struct amdgpu_device *adev, u32 req, u32 data1, u32 data2, u32 data3);
 };
 
+/*
+ * Firmware Reserve Frame buffer
+ */
+struct amdgpu_virt_fw_reserve {
+	struct amdgim_pf2vf_info_header *p_pf2vf;
+	struct amdgim_vf2pf_info_header *p_vf2pf;
+	unsigned int checksum_key;
+};
+/*
+ * Defination between PF and VF
+ * Structures forcibly aligned to 4 to keep the same style as PF.
+ */
+#define AMDGIM_DATAEXCHANGE_OFFSET		(64 * 1024)
+
+#define AMDGIM_GET_STRUCTURE_RESERVED_SIZE(total, u8, u16, u32, u64) \
+		(total - (((u8)+3) / 4 + ((u16)+1) / 2 + (u32) + (u64)*2))
+
+enum AMDGIM_FEATURE_FLAG {
+	/* GIM supports feature of Error log collecting */
+	AMDGIM_FEATURE_ERROR_LOG_COLLECT = 0x1,
+	/* GIM supports feature of loading uCodes */
+	AMDGIM_FEATURE_GIM_LOAD_UCODES   = 0x2,
+};
+
+struct amdgim_pf2vf_info_header {
+	/* the total structure size in byte. */
+	uint32_t size;
+	/* version of this structure, written by the GIM */
+	uint32_t version;
+} __aligned(4);
+struct  amdgim_pf2vf_info_v1 {
+	/* header contains size and version */
+	struct amdgim_pf2vf_info_header header;
+	/* max_width * max_height */
+	unsigned int uvd_enc_max_pixels_count;
+	/* 16x16 pixels/sec, codec independent */
+	unsigned int uvd_enc_max_bandwidth;
+	/* max_width * max_height */
+	unsigned int vce_enc_max_pixels_count;
+	/* 16x16 pixels/sec, codec independent */
+	unsigned int vce_enc_max_bandwidth;
+	/* MEC FW position in kb from the start of visible frame buffer */
+	unsigned int mecfw_kboffset;
+	/* The features flags of the GIM driver supports. */
+	unsigned int feature_flags;
+	/* use private key from mailbox 2 to create chueksum */
+	unsigned int checksum;
+} __aligned(4);
+
+struct  amdgim_pf2vf_info_v2 {
+	/* header contains size and version */
+	struct amdgim_pf2vf_info_header header;
+	/* use private key from mailbox 2 to create chueksum */
+	uint32_t checksum;
+	/* The features flags of the GIM driver supports. */
+	uint32_t feature_flags;
+	/* max_width * max_height */
+	uint32_t uvd_enc_max_pixels_count;
+	/* 16x16 pixels/sec, codec independent */
+	uint32_t uvd_enc_max_bandwidth;
+	/* max_width * max_height */
+	uint32_t vce_enc_max_pixels_count;
+	/* 16x16 pixels/sec, codec independent */
+	uint32_t vce_enc_max_bandwidth;
+	/* MEC FW position in kb from the start of VF visible frame buffer */
+	uint64_t mecfw_kboffset;
+	/* MEC FW size in KB */
+	uint32_t mecfw_ksize;
+	/* UVD FW position in kb from the start of VF visible frame buffer */
+	uint64_t uvdfw_kboffset;
+	/* UVD FW size in KB */
+	uint32_t uvdfw_ksize;
+	/* VCE FW position in kb from the start of VF visible frame buffer */
+	uint64_t vcefw_kboffset;
+	/* VCE FW size in KB */
+	uint32_t vcefw_ksize;
+	uint32_t reserved[AMDGIM_GET_STRUCTURE_RESERVED_SIZE(256, 0, 0, (9 + sizeof(struct amdgim_pf2vf_info_header)/sizeof(uint32_t)), 3)];
+} __aligned(4);
+
+
+struct amdgim_vf2pf_info_header {
+	/* the total structure size in byte. */
+	uint32_t size;
+	/*version of this structure, written by the guest */
+	uint32_t version;
+} __aligned(4);
+
+struct amdgim_vf2pf_info_v1 {
+	/* header contains size and version */
+	struct amdgim_vf2pf_info_header header;
+	/* driver version */
+	char driver_version[64];
+	/* driver certification, 1=WHQL, 0=None */
+	unsigned int driver_cert;
+	/* guest OS type and version: need a define */
+	unsigned int os_info;
+	/* in the unit of 1M */
+	unsigned int fb_usage;
+	/* guest gfx engine usage percentage */
+	unsigned int gfx_usage;
+	/* guest gfx engine health percentage */
+	unsigned int gfx_health;
+	/* guest compute engine usage percentage */
+	unsigned int compute_usage;
+	/* guest compute engine health percentage */
+	unsigned int compute_health;
+	/* guest vce engine usage percentage. 0xffff means N/A. */
+	unsigned int vce_enc_usage;
+	/* guest vce engine health percentage. 0xffff means N/A. */
+	unsigned int vce_enc_health;
+	/* guest uvd engine usage percentage. 0xffff means N/A. */
+	unsigned int uvd_enc_usage;
+	/* guest uvd engine usage percentage. 0xffff means N/A. */
+	unsigned int uvd_enc_health;
+	unsigned int checksum;
+} __aligned(4);
+
+struct amdgim_vf2pf_info_v2 {
+	/* header contains size and version */
+	struct amdgim_vf2pf_info_header header;
+	uint32_t checksum;
+	/* driver version */
+	uint8_t driver_version[64];
+	/* driver certification, 1=WHQL, 0=None */
+	uint32_t driver_cert;
+	/* guest OS type and version: need a define */
+	uint32_t os_info;
+	/* in the unit of 1M */
+	uint32_t fb_usage;
+	/* guest gfx engine usage percentage */
+	uint32_t gfx_usage;
+	/* guest gfx engine health percentage */
+	uint32_t gfx_health;
+	/* guest compute engine usage percentage */
+	uint32_t compute_usage;
+	/* guest compute engine health percentage */
+	uint32_t compute_health;
+	/* guest vce engine usage percentage. 0xffff means N/A. */
+	uint32_t vce_enc_usage;
+	/* guest vce engine health percentage. 0xffff means N/A. */
+	uint32_t vce_enc_health;
+	/* guest uvd engine usage percentage. 0xffff means N/A. */
+	uint32_t uvd_enc_usage;
+	/* guest uvd engine usage percentage. 0xffff means N/A. */
+	uint32_t uvd_enc_health;
+	uint32_t reserved[AMDGIM_GET_STRUCTURE_RESERVED_SIZE(256, 64, 0, (12 + sizeof(struct amdgim_vf2pf_info_header)/sizeof(uint32_t)), 0)];
+} __aligned(4);
+
+#define AMDGPU_FW_VRAM_VF2PF_VER 2
+typedef struct amdgim_vf2pf_info_v2 amdgim_vf2pf_info ;
+
+#define AMDGPU_FW_VRAM_VF2PF_WRITE(adev, field, val) \
+	do { \
+		((amdgim_vf2pf_info *)adev->virt.fw_reserve.p_vf2pf)->field = (val); \
+	} while (0)
+
+#define AMDGPU_FW_VRAM_VF2PF_READ(adev, field, val) \
+	do { \
+		(*val) = ((amdgim_vf2pf_info *)adev->virt.fw_reserve.p_vf2pf)->field; \
+	} while (0)
+
+#define AMDGPU_FW_VRAM_PF2VF_READ(adev, field, val) \
+	do { \
+		if (!adev->virt.fw_reserve.p_pf2vf) \
+			*(val) = 0; \
+		else { \
+			if (adev->virt.fw_reserve.p_pf2vf->version == 1) \
+				*(val) = ((struct amdgim_pf2vf_info_v1 *)adev->virt.fw_reserve.p_pf2vf)->field; \
+			if (adev->virt.fw_reserve.p_pf2vf->version == 2) \
+				*(val) = ((struct amdgim_pf2vf_info_v2 *)adev->virt.fw_reserve.p_pf2vf)->field; \
+		} \
+	} while (0)
+
 /* GPU virtualization */
 struct amdgpu_virt {
 	uint32_t			caps;
@@ -59,6 +244,8 @@ struct amdgpu_virt {
 	struct work_struct		flr_work;
 	struct amdgpu_mm_table		mm_table;
 	const struct amdgpu_virt_ops	*ops;
+	struct amdgpu_vf_error_buffer   vf_errors;
+	struct amdgpu_virt_fw_reserve	fw_reserve;
 };
 
 #define AMDGPU_CSA_SIZE    (8 * 1024)
@@ -101,5 +288,9 @@ int amdgpu_virt_reset_gpu(struct amdgpu_device *adev);
 int amdgpu_sriov_gpu_reset(struct amdgpu_device *adev, struct amdgpu_job *job);
 int amdgpu_virt_alloc_mm_table(struct amdgpu_device *adev);
 void amdgpu_virt_free_mm_table(struct amdgpu_device *adev);
+int amdgpu_virt_fw_reserve_get_checksum(void *obj, unsigned long obj_size,
+					unsigned int key,
+					unsigned int chksum);
+void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev);
 
 #endif

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

@@ -27,11 +27,58 @@
  */
 #include <linux/dma-fence-array.h>
 #include <linux/interval_tree_generic.h>
+#include <linux/idr.h>
 #include <drm/drmP.h>
 #include <drm/amdgpu_drm.h>
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 
+/*
+ * PASID manager
+ *
+ * PASIDs are global address space identifiers that can be shared
+ * between the GPU, an IOMMU and the driver. VMs on different devices
+ * may use the same PASID if they share the same address
+ * space. Therefore PASIDs are allocated using a global IDA. VMs are
+ * looked up from the PASID per amdgpu_device.
+ */
+static DEFINE_IDA(amdgpu_vm_pasid_ida);
+
+/**
+ * amdgpu_vm_alloc_pasid - Allocate a PASID
+ * @bits: Maximum width of the PASID in bits, must be at least 1
+ *
+ * Allocates a PASID of the given width while keeping smaller PASIDs
+ * available if possible.
+ *
+ * Returns a positive integer on success. Returns %-EINVAL if bits==0.
+ * Returns %-ENOSPC if no PASID was available. Returns %-ENOMEM on
+ * memory allocation failure.
+ */
+int amdgpu_vm_alloc_pasid(unsigned int bits)
+{
+	int pasid = -EINVAL;
+
+	for (bits = min(bits, 31U); bits > 0; bits--) {
+		pasid = ida_simple_get(&amdgpu_vm_pasid_ida,
+				       1U << (bits - 1), 1U << bits,
+				       GFP_KERNEL);
+		if (pasid != -ENOSPC)
+			break;
+	}
+
+	return pasid;
+}
+
+/**
+ * amdgpu_vm_free_pasid - Free a PASID
+ * @pasid: PASID to free
+ */
+void amdgpu_vm_free_pasid(unsigned int pasid)
+{
+	ida_simple_remove(&amdgpu_vm_pasid_ida, pasid);
+}
+
 /*
  * GPUVM
  * GPUVM is similar to the legacy gart on older asics, however
@@ -140,7 +187,7 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
 			 struct list_head *validated,
 			 struct amdgpu_bo_list_entry *entry)
 {
-	entry->robj = vm->root.bo;
+	entry->robj = vm->root.base.bo;
 	entry->priority = 0;
 	entry->tv.bo = &entry->robj->tbo;
 	entry->tv.shared = true;
@@ -149,86 +196,80 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
 }
 
 /**
- * amdgpu_vm_validate_layer - validate a single page table level
+ * amdgpu_vm_validate_pt_bos - validate the page table BOs
  *
- * @parent: parent page table level
+ * @adev: amdgpu device pointer
+ * @vm: vm providing the BOs
  * @validate: callback to do the validation
  * @param: parameter for the validation callback
  *
  * Validate the page table BOs on command submission if neccessary.
  */
-static int amdgpu_vm_validate_level(struct amdgpu_vm_pt *parent,
-				    int (*validate)(void *, struct amdgpu_bo *),
-				    void *param, bool use_cpu_for_update,
-				    struct ttm_bo_global *glob)
+int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+			      int (*validate)(void *p, struct amdgpu_bo *bo),
+			      void *param)
 {
-	unsigned i;
+	struct ttm_bo_global *glob = adev->mman.bdev.glob;
 	int r;
 
-	if (use_cpu_for_update) {
-		r = amdgpu_bo_kmap(parent->bo, NULL);
-		if (r)
-			return r;
-	}
-
-	if (!parent->entries)
-		return 0;
+	spin_lock(&vm->status_lock);
+	while (!list_empty(&vm->evicted)) {
+		struct amdgpu_vm_bo_base *bo_base;
+		struct amdgpu_bo *bo;
 
-	for (i = 0; i <= parent->last_entry_used; ++i) {
-		struct amdgpu_vm_pt *entry = &parent->entries[i];
+		bo_base = list_first_entry(&vm->evicted,
+					   struct amdgpu_vm_bo_base,
+					   vm_status);
+		spin_unlock(&vm->status_lock);
 
-		if (!entry->bo)
-			continue;
+		bo = bo_base->bo;
+		BUG_ON(!bo);
+		if (bo->parent) {
+			r = validate(param, bo);
+			if (r)
+				return r;
 
-		r = validate(param, entry->bo);
-		if (r)
-			return r;
+			spin_lock(&glob->lru_lock);
+			ttm_bo_move_to_lru_tail(&bo->tbo);
+			if (bo->shadow)
+				ttm_bo_move_to_lru_tail(&bo->shadow->tbo);
+			spin_unlock(&glob->lru_lock);
+		}
 
-		spin_lock(&glob->lru_lock);
-		ttm_bo_move_to_lru_tail(&entry->bo->tbo);
-		if (entry->bo->shadow)
-			ttm_bo_move_to_lru_tail(&entry->bo->shadow->tbo);
-		spin_unlock(&glob->lru_lock);
+		if (bo->tbo.type == ttm_bo_type_kernel &&
+		    vm->use_cpu_for_update) {
+			r = amdgpu_bo_kmap(bo, NULL);
+			if (r)
+				return r;
+		}
 
-		/*
-		 * Recurse into the sub directory. This is harmless because we
-		 * have only a maximum of 5 layers.
-		 */
-		r = amdgpu_vm_validate_level(entry, validate, param,
-					     use_cpu_for_update, glob);
-		if (r)
-			return r;
+		spin_lock(&vm->status_lock);
+		if (bo->tbo.type != ttm_bo_type_kernel)
+			list_move(&bo_base->vm_status, &vm->moved);
+		else
+			list_move(&bo_base->vm_status, &vm->relocated);
 	}
+	spin_unlock(&vm->status_lock);
 
-	return r;
+	return 0;
 }
 
 /**
- * amdgpu_vm_validate_pt_bos - validate the page table BOs
+ * amdgpu_vm_ready - check VM is ready for updates
  *
- * @adev: amdgpu device pointer
- * @vm: vm providing the BOs
- * @validate: callback to do the validation
- * @param: parameter for the validation callback
+ * @vm: VM to check
  *
- * Validate the page table BOs on command submission if neccessary.
+ * Check if all VM PDs/PTs are ready for updates
  */
-int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			      int (*validate)(void *p, struct amdgpu_bo *bo),
-			      void *param)
+bool amdgpu_vm_ready(struct amdgpu_vm *vm)
 {
-	uint64_t num_evictions;
+	bool ready;
 
-	/* We only need to validate the page tables
-	 * if they aren't already valid.
-	 */
-	num_evictions = atomic64_read(&adev->num_evictions);
-	if (num_evictions == vm->last_eviction_counter)
-		return 0;
+	spin_lock(&vm->status_lock);
+	ready = list_empty(&vm->evicted);
+	spin_unlock(&vm->status_lock);
 
-	return amdgpu_vm_validate_level(&vm->root, validate, param,
-					vm->use_cpu_for_update,
-					adev->mman.bdev.glob);
+	return ready;
 }
 
 /**
@@ -287,18 +328,19 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev,
 				AMDGPU_GEM_CREATE_SHADOW);
 
 	if (vm->pte_support_ats) {
-		init_value = AMDGPU_PTE_SYSTEM;
+		init_value = AMDGPU_PTE_DEFAULT_ATC;
 		if (level != adev->vm_manager.num_level - 1)
 			init_value |= AMDGPU_PDE_PTE;
+
 	}
 
 	/* walk over the address space and allocate the page tables */
 	for (pt_idx = from; pt_idx <= to; ++pt_idx) {
-		struct reservation_object *resv = vm->root.bo->tbo.resv;
+		struct reservation_object *resv = vm->root.base.bo->tbo.resv;
 		struct amdgpu_vm_pt *entry = &parent->entries[pt_idx];
 		struct amdgpu_bo *pt;
 
-		if (!entry->bo) {
+		if (!entry->base.bo) {
 			r = amdgpu_bo_create(adev,
 					     amdgpu_vm_bo_size(adev, level),
 					     AMDGPU_GPU_PAGE_SIZE, true,
@@ -319,9 +361,14 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev,
 			/* Keep a reference to the root directory to avoid
 			* freeing them up in the wrong order.
 			*/
-			pt->parent = amdgpu_bo_ref(vm->root.bo);
-
-			entry->bo = pt;
+			pt->parent = amdgpu_bo_ref(parent->base.bo);
+
+			entry->base.vm = vm;
+			entry->base.bo = pt;
+			list_add_tail(&entry->base.bo_list, &pt->va);
+			spin_lock(&vm->status_lock);
+			list_add(&entry->base.vm_status, &vm->relocated);
+			spin_unlock(&vm->status_lock);
 			entry->addr = 0;
 		}
 
@@ -988,7 +1035,7 @@ static int amdgpu_vm_wait_pd(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	int r;
 
 	amdgpu_sync_create(&sync);
-	amdgpu_sync_resv(adev, &sync, vm->root.bo->tbo.resv, owner);
+	amdgpu_sync_resv(adev, &sync, vm->root.base.bo->tbo.resv, owner, false);
 	r = amdgpu_sync_wait(&sync, true);
 	amdgpu_sync_free(&sync);
 
@@ -1007,18 +1054,17 @@ static int amdgpu_vm_wait_pd(struct amdgpu_device *adev, struct amdgpu_vm *vm,
  */
 static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 				  struct amdgpu_vm *vm,
-				  struct amdgpu_vm_pt *parent,
-				  unsigned level)
+				  struct amdgpu_vm_pt *parent)
 {
 	struct amdgpu_bo *shadow;
 	struct amdgpu_ring *ring = NULL;
 	uint64_t pd_addr, shadow_addr = 0;
-	uint32_t incr = amdgpu_vm_bo_size(adev, level + 1);
 	uint64_t last_pde = ~0, last_pt = ~0, last_shadow = ~0;
 	unsigned count = 0, pt_idx, ndw = 0;
 	struct amdgpu_job *job;
 	struct amdgpu_pte_update_params params;
 	struct dma_fence *fence = NULL;
+	uint32_t incr;
 
 	int r;
 
@@ -1027,10 +1073,10 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 
 	memset(&params, 0, sizeof(params));
 	params.adev = adev;
-	shadow = parent->bo->shadow;
+	shadow = parent->base.bo->shadow;
 
 	if (vm->use_cpu_for_update) {
-		pd_addr = (unsigned long)amdgpu_bo_kptr(parent->bo);
+		pd_addr = (unsigned long)amdgpu_bo_kptr(parent->base.bo);
 		r = amdgpu_vm_wait_pd(adev, vm, AMDGPU_FENCE_OWNER_VM);
 		if (unlikely(r))
 			return r;
@@ -1046,7 +1092,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 		/* assume the worst case */
 		ndw += parent->last_entry_used * 6;
 
-		pd_addr = amdgpu_bo_gpu_offset(parent->bo);
+		pd_addr = amdgpu_bo_gpu_offset(parent->base.bo);
 
 		if (shadow) {
 			shadow_addr = amdgpu_bo_gpu_offset(shadow);
@@ -1066,12 +1112,17 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 
 	/* walk over the address space and update the directory */
 	for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) {
-		struct amdgpu_bo *bo = parent->entries[pt_idx].bo;
+		struct amdgpu_vm_pt *entry = &parent->entries[pt_idx];
+		struct amdgpu_bo *bo = entry->base.bo;
 		uint64_t pde, pt;
 
 		if (bo == NULL)
 			continue;
 
+		spin_lock(&vm->status_lock);
+		list_del_init(&entry->base.vm_status);
+		spin_unlock(&vm->status_lock);
+
 		pt = amdgpu_bo_gpu_offset(bo);
 		pt = amdgpu_gart_get_vm_pde(adev, pt);
 		/* Don't update huge pages here */
@@ -1082,6 +1133,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 		parent->entries[pt_idx].addr = pt | AMDGPU_PTE_VALID;
 
 		pde = pd_addr + pt_idx * 8;
+		incr = amdgpu_bo_size(bo);
 		if (((last_pde + 8 * count) != pde) ||
 		    ((last_pt + incr * count) != pt) ||
 		    (count == AMDGPU_VM_MAX_UPDATE_SIZE)) {
@@ -1109,7 +1161,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 	}
 
 	if (count) {
-		if (vm->root.bo->shadow)
+		if (vm->root.base.bo->shadow)
 			params.func(&params, last_shadow, last_pt,
 				    count, incr, AMDGPU_PTE_VALID);
 
@@ -1122,12 +1174,13 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 			amdgpu_job_free(job);
 		} else {
 			amdgpu_ring_pad_ib(ring, params.ib);
-			amdgpu_sync_resv(adev, &job->sync, parent->bo->tbo.resv,
-					 AMDGPU_FENCE_OWNER_VM);
+			amdgpu_sync_resv(adev, &job->sync,
+					 parent->base.bo->tbo.resv,
+					 AMDGPU_FENCE_OWNER_VM, false);
 			if (shadow)
 				amdgpu_sync_resv(adev, &job->sync,
 						 shadow->tbo.resv,
-						 AMDGPU_FENCE_OWNER_VM);
+						 AMDGPU_FENCE_OWNER_VM, false);
 
 			WARN_ON(params.ib->length_dw > ndw);
 			r = amdgpu_job_submit(job, ring, &vm->entity,
@@ -1135,26 +1188,11 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev,
 			if (r)
 				goto error_free;
 
-			amdgpu_bo_fence(parent->bo, fence, true);
-			dma_fence_put(vm->last_dir_update);
-			vm->last_dir_update = dma_fence_get(fence);
-			dma_fence_put(fence);
+			amdgpu_bo_fence(parent->base.bo, fence, true);
+			dma_fence_put(vm->last_update);
+			vm->last_update = fence;
 		}
 	}
-	/*
-	 * Recurse into the subdirectories. This recursion is harmless because
-	 * we only have a maximum of 5 layers.
-	 */
-	for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) {
-		struct amdgpu_vm_pt *entry = &parent->entries[pt_idx];
-
-		if (!entry->bo)
-			continue;
-
-		r = amdgpu_vm_update_level(adev, vm, entry, level + 1);
-		if (r)
-			return r;
-	}
 
 	return 0;
 
@@ -1170,7 +1208,8 @@ error_free:
  *
  * Mark all PD level as invalid after an error.
  */
-static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent)
+static void amdgpu_vm_invalidate_level(struct amdgpu_vm *vm,
+				       struct amdgpu_vm_pt *parent)
 {
 	unsigned pt_idx;
 
@@ -1181,11 +1220,15 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent)
 	for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) {
 		struct amdgpu_vm_pt *entry = &parent->entries[pt_idx];
 
-		if (!entry->bo)
+		if (!entry->base.bo)
 			continue;
 
 		entry->addr = ~0ULL;
-		amdgpu_vm_invalidate_level(entry);
+		spin_lock(&vm->status_lock);
+		if (list_empty(&entry->base.vm_status))
+			list_add(&entry->base.vm_status, &vm->relocated);
+		spin_unlock(&vm->status_lock);
+		amdgpu_vm_invalidate_level(vm, entry);
 	}
 }
 
@@ -1201,11 +1244,40 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent)
 int amdgpu_vm_update_directories(struct amdgpu_device *adev,
 				 struct amdgpu_vm *vm)
 {
-	int r;
+	int r = 0;
 
-	r = amdgpu_vm_update_level(adev, vm, &vm->root, 0);
-	if (r)
-		amdgpu_vm_invalidate_level(&vm->root);
+	spin_lock(&vm->status_lock);
+	while (!list_empty(&vm->relocated)) {
+		struct amdgpu_vm_bo_base *bo_base;
+		struct amdgpu_bo *bo;
+
+		bo_base = list_first_entry(&vm->relocated,
+					   struct amdgpu_vm_bo_base,
+					   vm_status);
+		spin_unlock(&vm->status_lock);
+
+		bo = bo_base->bo->parent;
+		if (bo) {
+			struct amdgpu_vm_bo_base *parent;
+			struct amdgpu_vm_pt *pt;
+
+			parent = list_first_entry(&bo->va,
+						  struct amdgpu_vm_bo_base,
+						  bo_list);
+			pt = container_of(parent, struct amdgpu_vm_pt, base);
+
+			r = amdgpu_vm_update_level(adev, vm, pt);
+			if (r) {
+				amdgpu_vm_invalidate_level(vm, &vm->root);
+				return r;
+			}
+			spin_lock(&vm->status_lock);
+		} else {
+			spin_lock(&vm->status_lock);
+			list_del_init(&bo_base->vm_status);
+		}
+	}
+	spin_unlock(&vm->status_lock);
 
 	if (vm->use_cpu_for_update) {
 		/* Flush HDP */
@@ -1236,7 +1308,7 @@ void amdgpu_vm_get_entry(struct amdgpu_pte_update_params *p, uint64_t addr,
 	*entry = &p->vm->root;
 	while ((*entry)->entries) {
 		idx = addr >> (p->adev->vm_manager.block_size * level--);
-		idx %= amdgpu_bo_size((*entry)->bo) / 8;
+		idx %= amdgpu_bo_size((*entry)->base.bo) / 8;
 		*parent = *entry;
 		*entry = &(*entry)->entries[idx];
 	}
@@ -1272,7 +1344,7 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p,
 	    p->src ||
 	    !(flags & AMDGPU_PTE_VALID)) {
 
-		dst = amdgpu_bo_gpu_offset(entry->bo);
+		dst = amdgpu_bo_gpu_offset(entry->base.bo);
 		dst = amdgpu_gart_get_vm_pde(p->adev, dst);
 		flags = AMDGPU_PTE_VALID;
 	} else {
@@ -1298,18 +1370,18 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p,
 		tmp = p->pages_addr;
 		p->pages_addr = NULL;
 
-		pd_addr = (unsigned long)amdgpu_bo_kptr(parent->bo);
+		pd_addr = (unsigned long)amdgpu_bo_kptr(parent->base.bo);
 		pde = pd_addr + (entry - parent->entries) * 8;
 		amdgpu_vm_cpu_set_ptes(p, pde, dst, 1, 0, flags);
 
 		p->pages_addr = tmp;
 	} else {
-		if (parent->bo->shadow) {
-			pd_addr = amdgpu_bo_gpu_offset(parent->bo->shadow);
+		if (parent->base.bo->shadow) {
+			pd_addr = amdgpu_bo_gpu_offset(parent->base.bo->shadow);
 			pde = pd_addr + (entry - parent->entries) * 8;
 			amdgpu_vm_do_set_ptes(p, pde, dst, 1, 0, flags);
 		}
-		pd_addr = amdgpu_bo_gpu_offset(parent->bo);
+		pd_addr = amdgpu_bo_gpu_offset(parent->base.bo);
 		pde = pd_addr + (entry - parent->entries) * 8;
 		amdgpu_vm_do_set_ptes(p, pde, dst, 1, 0, flags);
 	}
@@ -1360,7 +1432,7 @@ static int amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params,
 		if (entry->addr & AMDGPU_PDE_PTE)
 			continue;
 
-		pt = entry->bo;
+		pt = entry->base.bo;
 		if (use_cpu_update) {
 			pe_start = (unsigned long)amdgpu_bo_kptr(pt);
 		} else {
@@ -1396,8 +1468,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params	*params,
 				uint64_t start, uint64_t end,
 				uint64_t dst, uint64_t flags)
 {
-	int r;
-
 	/**
 	 * The MC L1 TLB supports variable sized pages, based on a fragment
 	 * field in the PTE. When this field is set to a non-zero value, page
@@ -1416,39 +1486,38 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params	*params,
 	 * Userspace can support this by aligning virtual base address and
 	 * allocation size to the fragment size.
 	 */
-	unsigned pages_per_frag = params->adev->vm_manager.fragment_size;
-	uint64_t frag_flags = AMDGPU_PTE_FRAG(pages_per_frag);
-	uint64_t frag_align = 1 << pages_per_frag;
-
-	uint64_t frag_start = ALIGN(start, frag_align);
-	uint64_t frag_end = end & ~(frag_align - 1);
+	unsigned max_frag = params->adev->vm_manager.fragment_size;
+	int r;
 
 	/* system pages are non continuously */
-	if (params->src || !(flags & AMDGPU_PTE_VALID) ||
-	    (frag_start >= frag_end))
+	if (params->src || !(flags & AMDGPU_PTE_VALID))
 		return amdgpu_vm_update_ptes(params, start, end, dst, flags);
 
-	/* handle the 4K area at the beginning */
-	if (start != frag_start) {
-		r = amdgpu_vm_update_ptes(params, start, frag_start,
-					  dst, flags);
+	while (start != end) {
+		uint64_t frag_flags, frag_end;
+		unsigned frag;
+
+		/* This intentionally wraps around if no bit is set */
+		frag = min((unsigned)ffs(start) - 1,
+			   (unsigned)fls64(end - start) - 1);
+		if (frag >= max_frag) {
+			frag_flags = AMDGPU_PTE_FRAG(max_frag);
+			frag_end = end & ~((1ULL << max_frag) - 1);
+		} else {
+			frag_flags = AMDGPU_PTE_FRAG(frag);
+			frag_end = start + (1 << frag);
+		}
+
+		r = amdgpu_vm_update_ptes(params, start, frag_end, dst,
+					  flags | frag_flags);
 		if (r)
 			return r;
-		dst += (frag_start - start) * AMDGPU_GPU_PAGE_SIZE;
-	}
 
-	/* handle the area in the middle */
-	r = amdgpu_vm_update_ptes(params, frag_start, frag_end, dst,
-				  flags | frag_flags);
-	if (r)
-		return r;
-
-	/* handle the 4K area at the end */
-	if (frag_end != end) {
-		dst += (frag_end - frag_start) * AMDGPU_GPU_PAGE_SIZE;
-		r = amdgpu_vm_update_ptes(params, frag_end, end, dst, flags);
+		dst += (frag_end - start) * AMDGPU_GPU_PAGE_SIZE;
+		start = frag_end;
 	}
-	return r;
+
+	return 0;
 }
 
 /**
@@ -1456,7 +1525,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params	*params,
  *
  * @adev: amdgpu_device pointer
  * @exclusive: fence we need to sync to
- * @src: address where to copy page table entries from
  * @pages_addr: DMA addresses to use for mapping
  * @vm: requested vm
  * @start: start of mapped range
@@ -1470,7 +1538,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params	*params,
  */
 static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 				       struct dma_fence *exclusive,
-				       uint64_t src,
 				       dma_addr_t *pages_addr,
 				       struct amdgpu_vm *vm,
 				       uint64_t start, uint64_t last,
@@ -1488,7 +1555,6 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	memset(&params, 0, sizeof(params));
 	params.adev = adev;
 	params.vm = vm;
-	params.src = src;
 
 	/* sync to everything on unmapping */
 	if (!(flags & AMDGPU_PTE_VALID))
@@ -1517,10 +1583,12 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	nptes = last - start + 1;
 
 	/*
-	 * reserve space for one command every (1 << BLOCK_SIZE)
+	 * reserve space for two commands every (1 << BLOCK_SIZE)
 	 *  entries or 2k dwords (whatever is smaller)
+         *
+         * The second command is for the shadow pagetables.
 	 */
-	ncmds = (nptes >> min(adev->vm_manager.block_size, 11u)) + 1;
+	ncmds = ((nptes >> min(adev->vm_manager.block_size, 11u)) + 1) * 2;
 
 	/* padding, etc. */
 	ndw = 64;
@@ -1528,15 +1596,9 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	/* one PDE write for each huge page */
 	ndw += ((nptes >> adev->vm_manager.block_size) + 1) * 6;
 
-	if (src) {
-		/* only copy commands needed */
-		ndw += ncmds * 7;
-
-		params.func = amdgpu_vm_do_copy_ptes;
-
-	} else if (pages_addr) {
+	if (pages_addr) {
 		/* copy commands needed */
-		ndw += ncmds * 7;
+		ndw += ncmds * adev->vm_manager.vm_pte_funcs->copy_pte_num_dw;
 
 		/* and also PTEs */
 		ndw += nptes * 2;
@@ -1545,10 +1607,11 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 
 	} else {
 		/* set page commands needed */
-		ndw += ncmds * 10;
+		ndw += ncmds * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw;
 
-		/* two extra commands for begin/end of fragment */
-		ndw += 2 * 10;
+		/* extra commands for begin/end fragments */
+		ndw += 2 * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw
+				* adev->vm_manager.fragment_size;
 
 		params.func = amdgpu_vm_do_set_ptes;
 	}
@@ -1559,7 +1622,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 
 	params.ib = &job->ibs[0];
 
-	if (!src && pages_addr) {
+	if (pages_addr) {
 		uint64_t *pte;
 		unsigned i;
 
@@ -1580,12 +1643,12 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	if (r)
 		goto error_free;
 
-	r = amdgpu_sync_resv(adev, &job->sync, vm->root.bo->tbo.resv,
-			     owner);
+	r = amdgpu_sync_resv(adev, &job->sync, vm->root.base.bo->tbo.resv,
+			     owner, false);
 	if (r)
 		goto error_free;
 
-	r = reservation_object_reserve_shared(vm->root.bo->tbo.resv);
+	r = reservation_object_reserve_shared(vm->root.base.bo->tbo.resv);
 	if (r)
 		goto error_free;
 
@@ -1600,14 +1663,14 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	if (r)
 		goto error_free;
 
-	amdgpu_bo_fence(vm->root.bo, f, true);
+	amdgpu_bo_fence(vm->root.base.bo, f, true);
 	dma_fence_put(*fence);
 	*fence = f;
 	return 0;
 
 error_free:
 	amdgpu_job_free(job);
-	amdgpu_vm_invalidate_level(&vm->root);
+	amdgpu_vm_invalidate_level(vm, &vm->root);
 	return r;
 }
 
@@ -1636,7 +1699,8 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
 				      struct drm_mm_node *nodes,
 				      struct dma_fence **fence)
 {
-	uint64_t pfn, src = 0, start = mapping->start;
+	unsigned min_linear_pages = 1 << adev->vm_manager.fragment_size;
+	uint64_t pfn, start = mapping->start;
 	int r;
 
 	/* normally,bo_va->flags only contians READABLE and WIRTEABLE bit go here
@@ -1670,6 +1734,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
 	}
 
 	do {
+		dma_addr_t *dma_addr = NULL;
 		uint64_t max_entries;
 		uint64_t addr, last;
 
@@ -1683,16 +1748,32 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
 		}
 
 		if (pages_addr) {
+			uint64_t count;
+
 			max_entries = min(max_entries, 16ull * 1024ull);
-			addr = 0;
+			for (count = 1; count < max_entries; ++count) {
+				uint64_t idx = pfn + count;
+
+				if (pages_addr[idx] !=
+				    (pages_addr[idx - 1] + PAGE_SIZE))
+					break;
+			}
+
+			if (count < min_linear_pages) {
+				addr = pfn << PAGE_SHIFT;
+				dma_addr = pages_addr;
+			} else {
+				addr = pages_addr[pfn];
+				max_entries = count;
+			}
+
 		} else if (flags & AMDGPU_PTE_VALID) {
 			addr += adev->vm_manager.vram_base_offset;
+			addr += pfn << PAGE_SHIFT;
 		}
-		addr += pfn << PAGE_SHIFT;
 
 		last = min((uint64_t)mapping->last, start + max_entries - 1);
-		r = amdgpu_vm_bo_update_mapping(adev, exclusive,
-						src, pages_addr, vm,
+		r = amdgpu_vm_bo_update_mapping(adev, exclusive, dma_addr, vm,
 						start, last, flags, addr,
 						fence);
 		if (r)
@@ -1730,7 +1811,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
 	dma_addr_t *pages_addr = NULL;
 	struct ttm_mem_reg *mem;
 	struct drm_mm_node *nodes;
-	struct dma_fence *exclusive;
+	struct dma_fence *exclusive, **last_update;
 	uint64_t flags;
 	int r;
 
@@ -1756,38 +1837,43 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
 	else
 		flags = 0x0;
 
-	spin_lock(&vm->status_lock);
-	if (!list_empty(&bo_va->base.vm_status))
+	if (clear || (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv))
+		last_update = &vm->last_update;
+	else
+		last_update = &bo_va->last_pt_update;
+
+	if (!clear && bo_va->base.moved) {
+		bo_va->base.moved = false;
 		list_splice_init(&bo_va->valids, &bo_va->invalids);
-	spin_unlock(&vm->status_lock);
+
+	} else if (bo_va->cleared != clear) {
+		list_splice_init(&bo_va->valids, &bo_va->invalids);
+	}
 
 	list_for_each_entry(mapping, &bo_va->invalids, list) {
 		r = amdgpu_vm_bo_split_mapping(adev, exclusive, pages_addr, vm,
 					       mapping, flags, nodes,
-					       &bo_va->last_pt_update);
+					       last_update);
 		if (r)
 			return r;
 	}
 
-	if (trace_amdgpu_vm_bo_mapping_enabled()) {
-		list_for_each_entry(mapping, &bo_va->valids, list)
-			trace_amdgpu_vm_bo_mapping(mapping);
-
-		list_for_each_entry(mapping, &bo_va->invalids, list)
-			trace_amdgpu_vm_bo_mapping(mapping);
+	if (vm->use_cpu_for_update) {
+		/* Flush HDP */
+		mb();
+		amdgpu_gart_flush_gpu_tlb(adev, 0);
 	}
 
 	spin_lock(&vm->status_lock);
-	list_splice_init(&bo_va->invalids, &bo_va->valids);
 	list_del_init(&bo_va->base.vm_status);
-	if (clear)
-		list_add(&bo_va->base.vm_status, &vm->cleared);
 	spin_unlock(&vm->status_lock);
 
-	if (vm->use_cpu_for_update) {
-		/* Flush HDP */
-		mb();
-		amdgpu_gart_flush_gpu_tlb(adev, 0);
+	list_splice_init(&bo_va->invalids, &bo_va->valids);
+	bo_va->cleared = clear;
+
+	if (trace_amdgpu_vm_bo_mapping_enabled()) {
+		list_for_each_entry(mapping, &bo_va->valids, list)
+			trace_amdgpu_vm_bo_mapping(mapping);
 	}
 
 	return 0;
@@ -1895,7 +1981,7 @@ static void amdgpu_vm_free_mapping(struct amdgpu_device *adev,
  */
 static void amdgpu_vm_prt_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
-	struct reservation_object *resv = vm->root.bo->tbo.resv;
+	struct reservation_object *resv = vm->root.base.bo->tbo.resv;
 	struct dma_fence *excl, **shared;
 	unsigned i, shared_count;
 	int r;
@@ -1951,9 +2037,9 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 		list_del(&mapping->list);
 
 		if (vm->pte_support_ats)
-			init_pte_value = AMDGPU_PTE_SYSTEM;
+			init_pte_value = AMDGPU_PTE_DEFAULT_ATC;
 
-		r = amdgpu_vm_bo_update_mapping(adev, NULL, 0, NULL, vm,
+		r = amdgpu_vm_bo_update_mapping(adev, NULL, NULL, vm,
 						mapping->start, mapping->last,
 						init_pte_value, 0, &f);
 		amdgpu_vm_free_mapping(adev, vm, mapping, f);
@@ -1975,29 +2061,35 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 }
 
 /**
- * amdgpu_vm_clear_moved - clear moved BOs in the PT
+ * amdgpu_vm_handle_moved - handle moved BOs in the PT
  *
  * @adev: amdgpu_device pointer
  * @vm: requested vm
+ * @sync: sync object to add fences to
  *
- * Make sure all moved BOs are cleared in the PT.
+ * Make sure all BOs which are moved are updated in the PTs.
  * Returns 0 for success.
  *
- * PTs have to be reserved and mutex must be locked!
+ * PTs have to be reserved!
  */
-int amdgpu_vm_clear_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			    struct amdgpu_sync *sync)
+int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
+			   struct amdgpu_vm *vm)
 {
-	struct amdgpu_bo_va *bo_va = NULL;
+	bool clear;
 	int r = 0;
 
 	spin_lock(&vm->status_lock);
 	while (!list_empty(&vm->moved)) {
+		struct amdgpu_bo_va *bo_va;
+
 		bo_va = list_first_entry(&vm->moved,
 			struct amdgpu_bo_va, base.vm_status);
 		spin_unlock(&vm->status_lock);
 
-		r = amdgpu_vm_bo_update(adev, bo_va, true);
+		/* Per VM BOs never need to bo cleared in the page tables */
+		clear = bo_va->base.bo->tbo.resv != vm->root.base.bo->tbo.resv;
+
+		r = amdgpu_vm_bo_update(adev, bo_va, clear);
 		if (r)
 			return r;
 
@@ -2005,9 +2097,6 @@ int amdgpu_vm_clear_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	}
 	spin_unlock(&vm->status_lock);
 
-	if (bo_va)
-		r = amdgpu_sync_fence(adev, sync, bo_va->last_pt_update);
-
 	return r;
 }
 
@@ -2049,6 +2138,39 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
 	return bo_va;
 }
 
+
+/**
+ * amdgpu_vm_bo_insert_mapping - insert a new mapping
+ *
+ * @adev: amdgpu_device pointer
+ * @bo_va: bo_va to store the address
+ * @mapping: the mapping to insert
+ *
+ * Insert a new mapping into all structures.
+ */
+static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev,
+				    struct amdgpu_bo_va *bo_va,
+				    struct amdgpu_bo_va_mapping *mapping)
+{
+	struct amdgpu_vm *vm = bo_va->base.vm;
+	struct amdgpu_bo *bo = bo_va->base.bo;
+
+	mapping->bo_va = bo_va;
+	list_add(&mapping->list, &bo_va->invalids);
+	amdgpu_vm_it_insert(mapping, &vm->va);
+
+	if (mapping->flags & AMDGPU_PTE_PRT)
+		amdgpu_vm_prt_get(adev);
+
+	if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv) {
+		spin_lock(&vm->status_lock);
+		if (list_empty(&bo_va->base.vm_status))
+			list_add(&bo_va->base.vm_status, &vm->moved);
+		spin_unlock(&vm->status_lock);
+	}
+	trace_amdgpu_vm_bo_map(bo_va, mapping);
+}
+
 /**
  * amdgpu_vm_bo_map - map bo inside a vm
  *
@@ -2100,17 +2222,12 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 	if (!mapping)
 		return -ENOMEM;
 
-	INIT_LIST_HEAD(&mapping->list);
 	mapping->start = saddr;
 	mapping->last = eaddr;
 	mapping->offset = offset;
 	mapping->flags = flags;
 
-	list_add(&mapping->list, &bo_va->invalids);
-	amdgpu_vm_it_insert(mapping, &vm->va);
-
-	if (flags & AMDGPU_PTE_PRT)
-		amdgpu_vm_prt_get(adev);
+	amdgpu_vm_bo_insert_map(adev, bo_va, mapping);
 
 	return 0;
 }
@@ -2137,7 +2254,6 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
 {
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_bo *bo = bo_va->base.bo;
-	struct amdgpu_vm *vm = bo_va->base.vm;
 	uint64_t eaddr;
 	int r;
 
@@ -2171,11 +2287,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
 	mapping->offset = offset;
 	mapping->flags = flags;
 
-	list_add(&mapping->list, &bo_va->invalids);
-	amdgpu_vm_it_insert(mapping, &vm->va);
-
-	if (flags & AMDGPU_PTE_PRT)
-		amdgpu_vm_prt_get(adev);
+	amdgpu_vm_bo_insert_map(adev, bo_va, mapping);
 
 	return 0;
 }
@@ -2221,6 +2333,7 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 
 	list_del(&mapping->list);
 	amdgpu_vm_it_remove(mapping, &vm->va);
+	mapping->bo_va = NULL;
 	trace_amdgpu_vm_bo_unmap(bo_va, mapping);
 
 	if (valid)
@@ -2306,6 +2419,7 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 		if (tmp->last > eaddr)
 		    tmp->last = eaddr;
 
+		tmp->bo_va = NULL;
 		list_add(&tmp->list, &vm->freed);
 		trace_amdgpu_vm_bo_unmap(NULL, tmp);
 	}
@@ -2331,6 +2445,19 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 	return 0;
 }
 
+/**
+ * amdgpu_vm_bo_lookup_mapping - find mapping by address
+ *
+ * @vm: the requested VM
+ *
+ * Find a mapping by it's address.
+ */
+struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm,
+							 uint64_t addr)
+{
+	return amdgpu_vm_it_iter_first(&vm->va, addr, addr);
+}
+
 /**
  * amdgpu_vm_bo_rmv - remove a bo to a specific vm
  *
@@ -2356,6 +2483,7 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
 	list_for_each_entry_safe(mapping, next, &bo_va->valids, list) {
 		list_del(&mapping->list);
 		amdgpu_vm_it_remove(mapping, &vm->va);
+		mapping->bo_va = NULL;
 		trace_amdgpu_vm_bo_unmap(bo_va, mapping);
 		list_add(&mapping->list, &vm->freed);
 	}
@@ -2380,15 +2508,36 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
  * Mark @bo as invalid.
  */
 void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
-			     struct amdgpu_bo *bo)
+			     struct amdgpu_bo *bo, bool evicted)
 {
 	struct amdgpu_vm_bo_base *bo_base;
 
 	list_for_each_entry(bo_base, &bo->va, bo_list) {
+		struct amdgpu_vm *vm = bo_base->vm;
+
+		bo_base->moved = true;
+		if (evicted && bo->tbo.resv == vm->root.base.bo->tbo.resv) {
+			spin_lock(&bo_base->vm->status_lock);
+			if (bo->tbo.type == ttm_bo_type_kernel)
+				list_move(&bo_base->vm_status, &vm->evicted);
+			else
+				list_move_tail(&bo_base->vm_status,
+					       &vm->evicted);
+			spin_unlock(&bo_base->vm->status_lock);
+			continue;
+		}
+
+		if (bo->tbo.type == ttm_bo_type_kernel) {
+			spin_lock(&bo_base->vm->status_lock);
+			if (list_empty(&bo_base->vm_status))
+				list_add(&bo_base->vm_status, &vm->relocated);
+			spin_unlock(&bo_base->vm->status_lock);
+			continue;
+		}
+
 		spin_lock(&bo_base->vm->status_lock);
 		if (list_empty(&bo_base->vm_status))
-			list_add(&bo_base->vm_status,
-				 &bo_base->vm->moved);
+			list_add(&bo_base->vm_status, &vm->moved);
 		spin_unlock(&bo_base->vm->status_lock);
 	}
 }
@@ -2412,7 +2561,8 @@ static uint32_t amdgpu_vm_get_block_size(uint64_t vm_size)
  * @adev: amdgpu_device pointer
  * @fragment_size_default: the default fragment size if it's set auto
  */
-void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev, uint32_t fragment_size_default)
+void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev,
+				 uint32_t fragment_size_default)
 {
 	if (amdgpu_vm_fragment_size == -1)
 		adev->vm_manager.fragment_size = fragment_size_default;
@@ -2426,7 +2576,8 @@ void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev, uint32_t fragment_s
  * @adev: amdgpu_device pointer
  * @vm_size: the default vm size if it's set auto
  */
-void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size, uint32_t fragment_size_default)
+void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size,
+			   uint32_t fragment_size_default)
 {
 	/* adjust vm size firstly */
 	if (amdgpu_vm_size == -1)
@@ -2458,7 +2609,7 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size, uint32_
  * Init @vm fields.
  */
 int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-		   int vm_context)
+		   int vm_context, unsigned int pasid)
 {
 	const unsigned align = min(AMDGPU_VM_PTB_ALIGN_SIZE,
 		AMDGPU_VM_PTE_COUNT(adev) * 8);
@@ -2474,8 +2625,9 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
 		vm->reserved_vmid[i] = NULL;
 	spin_lock_init(&vm->status_lock);
+	INIT_LIST_HEAD(&vm->evicted);
+	INIT_LIST_HEAD(&vm->relocated);
 	INIT_LIST_HEAD(&vm->moved);
-	INIT_LIST_HEAD(&vm->cleared);
 	INIT_LIST_HEAD(&vm->freed);
 
 	/* create scheduler entity for page table updates */
@@ -2497,7 +2649,9 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 
 		if (adev->asic_type == CHIP_RAVEN) {
 			vm->pte_support_ats = true;
-			init_pde_value = AMDGPU_PTE_SYSTEM | AMDGPU_PDE_PTE;
+			init_pde_value = AMDGPU_PTE_DEFAULT_ATC
+					| AMDGPU_PDE_PTE;
+
 		}
 	} else
 		vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
@@ -2506,7 +2660,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			 vm->use_cpu_for_update ? "CPU" : "SDMA");
 	WARN_ONCE((vm->use_cpu_for_update & !amdgpu_vm_is_large_bar(adev)),
 		  "CPU update of VM recommended only for large BAR system\n");
-	vm->last_dir_update = NULL;
+	vm->last_update = NULL;
 
 	flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
 			AMDGPU_GEM_CREATE_VRAM_CLEARED;
@@ -2519,30 +2673,47 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	r = amdgpu_bo_create(adev, amdgpu_vm_bo_size(adev, 0), align, true,
 			     AMDGPU_GEM_DOMAIN_VRAM,
 			     flags,
-			     NULL, NULL, init_pde_value, &vm->root.bo);
+			     NULL, NULL, init_pde_value, &vm->root.base.bo);
 	if (r)
 		goto error_free_sched_entity;
 
-	r = amdgpu_bo_reserve(vm->root.bo, false);
-	if (r)
-		goto error_free_root;
-
-	vm->last_eviction_counter = atomic64_read(&adev->num_evictions);
+	vm->root.base.vm = vm;
+	list_add_tail(&vm->root.base.bo_list, &vm->root.base.bo->va);
+	INIT_LIST_HEAD(&vm->root.base.vm_status);
 
 	if (vm->use_cpu_for_update) {
-		r = amdgpu_bo_kmap(vm->root.bo, NULL);
+		r = amdgpu_bo_reserve(vm->root.base.bo, false);
+		if (r)
+			goto error_free_root;
+
+		r = amdgpu_bo_kmap(vm->root.base.bo, NULL);
+		amdgpu_bo_unreserve(vm->root.base.bo);
 		if (r)
 			goto error_free_root;
 	}
 
-	amdgpu_bo_unreserve(vm->root.bo);
+	if (pasid) {
+		unsigned long flags;
+
+		spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags);
+		r = idr_alloc(&adev->vm_manager.pasid_idr, vm, pasid, pasid + 1,
+			      GFP_ATOMIC);
+		spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags);
+		if (r < 0)
+			goto error_free_root;
+
+		vm->pasid = pasid;
+	}
+
+	INIT_KFIFO(vm->faults);
+	vm->fault_credit = 16;
 
 	return 0;
 
 error_free_root:
-	amdgpu_bo_unref(&vm->root.bo->shadow);
-	amdgpu_bo_unref(&vm->root.bo);
-	vm->root.bo = NULL;
+	amdgpu_bo_unref(&vm->root.base.bo->shadow);
+	amdgpu_bo_unref(&vm->root.base.bo);
+	vm->root.base.bo = NULL;
 
 error_free_sched_entity:
 	amd_sched_entity_fini(&ring->sched, &vm->entity);
@@ -2561,9 +2732,11 @@ static void amdgpu_vm_free_levels(struct amdgpu_vm_pt *level)
 {
 	unsigned i;
 
-	if (level->bo) {
-		amdgpu_bo_unref(&level->bo->shadow);
-		amdgpu_bo_unref(&level->bo);
+	if (level->base.bo) {
+		list_del(&level->base.bo_list);
+		list_del(&level->base.vm_status);
+		amdgpu_bo_unref(&level->base.bo->shadow);
+		amdgpu_bo_unref(&level->base.bo);
 	}
 
 	if (level->entries)
@@ -2586,7 +2759,21 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
 	struct amdgpu_bo_va_mapping *mapping, *tmp;
 	bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt;
-	int i;
+	struct amdgpu_bo *root;
+	u64 fault;
+	int i, r;
+
+	/* Clear pending page faults from IH when the VM is destroyed */
+	while (kfifo_get(&vm->faults, &fault))
+		amdgpu_ih_clear_fault(adev, fault);
+
+	if (vm->pasid) {
+		unsigned long flags;
+
+		spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags);
+		idr_remove(&adev->vm_manager.pasid_idr, vm->pasid);
+		spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags);
+	}
 
 	amd_sched_entity_fini(vm->entity.sched, &vm->entity);
 
@@ -2609,12 +2796,50 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 		amdgpu_vm_free_mapping(adev, vm, mapping, NULL);
 	}
 
-	amdgpu_vm_free_levels(&vm->root);
-	dma_fence_put(vm->last_dir_update);
+	root = amdgpu_bo_ref(vm->root.base.bo);
+	r = amdgpu_bo_reserve(root, true);
+	if (r) {
+		dev_err(adev->dev, "Leaking page tables because BO reservation failed\n");
+	} else {
+		amdgpu_vm_free_levels(&vm->root);
+		amdgpu_bo_unreserve(root);
+	}
+	amdgpu_bo_unref(&root);
+	dma_fence_put(vm->last_update);
 	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
 		amdgpu_vm_free_reserved_vmid(adev, vm, i);
 }
 
+/**
+ * amdgpu_vm_pasid_fault_credit - Check fault credit for given PASID
+ *
+ * @adev: amdgpu_device pointer
+ * @pasid: PASID do identify the VM
+ *
+ * This function is expected to be called in interrupt context. Returns
+ * true if there was fault credit, false otherwise
+ */
+bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev,
+				  unsigned int pasid)
+{
+	struct amdgpu_vm *vm;
+
+	spin_lock(&adev->vm_manager.pasid_lock);
+	vm = idr_find(&adev->vm_manager.pasid_idr, pasid);
+	spin_unlock(&adev->vm_manager.pasid_lock);
+	if (!vm)
+		/* VM not found, can't track fault credit */
+		return true;
+
+	/* No lock needed. only accessed by IRQ handler */
+	if (!vm->fault_credit)
+		/* Too many faults in this VM */
+		return false;
+
+	vm->fault_credit--;
+	return true;
+}
+
 /**
  * amdgpu_vm_manager_init - init the VM manager
  *
@@ -2668,6 +2893,8 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 	adev->vm_manager.vm_update_mode = 0;
 #endif
 
+	idr_init(&adev->vm_manager.pasid_idr);
+	spin_lock_init(&adev->vm_manager.pasid_lock);
 }
 
 /**
@@ -2681,6 +2908,9 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev)
 {
 	unsigned i, j;
 
+	WARN_ON(!idr_is_empty(&adev->vm_manager.pasid_idr));
+	idr_destroy(&adev->vm_manager.pasid_idr);
+
 	for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
 		struct amdgpu_vm_id_manager *id_mgr =
 			&adev->vm_manager.id_mgr[i];

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h

@@ -25,6 +25,7 @@
 #define __AMDGPU_VM_H__
 
 #include <linux/rbtree.h>
+#include <linux/idr.h>
 
 #include "gpu_scheduler.h"
 #include "amdgpu_sync.h"
@@ -72,6 +73,16 @@ struct amdgpu_bo_list_entry;
 #define AMDGPU_PTE_MTYPE(a)    ((uint64_t)a << 57)
 #define AMDGPU_PTE_MTYPE_MASK	AMDGPU_PTE_MTYPE(3ULL)
 
+/* For Raven */
+#define AMDGPU_MTYPE_CC 2
+
+#define AMDGPU_PTE_DEFAULT_ATC  (AMDGPU_PTE_SYSTEM      \
+                                | AMDGPU_PTE_SNOOPED    \
+                                | AMDGPU_PTE_EXECUTABLE \
+                                | AMDGPU_PTE_READABLE   \
+                                | AMDGPU_PTE_WRITEABLE  \
+                                | AMDGPU_PTE_MTYPE(AMDGPU_MTYPE_CC))
+
 /* How to programm VM fault handling */
 #define AMDGPU_VM_FAULT_STOP_NEVER	0
 #define AMDGPU_VM_FAULT_STOP_FIRST	1
@@ -105,17 +116,24 @@ struct amdgpu_vm_bo_base {
 
 	/* protected by spinlock */
 	struct list_head		vm_status;
+
+	/* protected by the BO being reserved */
+	bool				moved;
 };
 
 struct amdgpu_vm_pt {
-	struct amdgpu_bo	*bo;
-	uint64_t		addr;
+	struct amdgpu_vm_bo_base	base;
+	uint64_t			addr;
 
 	/* array of page tables, one for each directory entry */
-	struct amdgpu_vm_pt	*entries;
-	unsigned		last_entry_used;
+	struct amdgpu_vm_pt		*entries;
+	unsigned			last_entry_used;
 };
 
+#define AMDGPU_VM_FAULT(pasid, addr) (((u64)(pasid) << 48) | (addr))
+#define AMDGPU_VM_FAULT_PASID(fault) ((u64)(fault) >> 48)
+#define AMDGPU_VM_FAULT_ADDR(fault)  ((u64)(fault) & 0xfffffffff000ULL)
+
 struct amdgpu_vm {
 	/* tree of virtual addresses mapped */
 	struct rb_root_cached	va;
@@ -123,19 +141,21 @@ struct amdgpu_vm {
 	/* protecting invalidated */
 	spinlock_t		status_lock;
 
+	/* BOs who needs a validation */
+	struct list_head	evicted;
+
+	/* PT BOs which relocated and their parent need an update */
+	struct list_head	relocated;
+
 	/* BOs moved, but not yet updated in the PT */
 	struct list_head	moved;
 
-	/* BOs cleared in the PT because of a move */
-	struct list_head	cleared;
-
 	/* BO mappings freed, but not yet updated in the PT */
 	struct list_head	freed;
 
 	/* contains the page directory */
 	struct amdgpu_vm_pt     root;
-	struct dma_fence	*last_dir_update;
-	uint64_t		last_eviction_counter;
+	struct dma_fence	*last_update;
 
 	/* protecting freed */
 	spinlock_t		freed_lock;
@@ -143,8 +163,9 @@ struct amdgpu_vm {
 	/* Scheduler entity for page table updates */
 	struct amd_sched_entity	entity;
 
-	/* client id */
+	/* client id and PASID (TODO: replace client_id with PASID) */
 	u64                     client_id;
+	unsigned int		pasid;
 	/* dedicated to vm */
 	struct amdgpu_vm_id	*reserved_vmid[AMDGPU_MAX_VMHUBS];
 
@@ -153,6 +174,12 @@ struct amdgpu_vm {
 
 	/* Flag to indicate ATS support from PTE for GFX9 */
 	bool			pte_support_ats;
+
+	/* Up to 128 pending retry page faults */
+	DECLARE_KFIFO(faults, u64, 128);
+
+	/* Limit non-retry fault storms */
+	unsigned int		fault_credit;
 };
 
 struct amdgpu_vm_id {
@@ -215,16 +242,27 @@ struct amdgpu_vm_manager {
 	 * BIT1[= 0] Compute updated by SDMA [= 1] by CPU
 	 */
 	int					vm_update_mode;
+
+	/* PASID to VM mapping, will be used in interrupt context to
+	 * look up VM of a page fault
+	 */
+	struct idr				pasid_idr;
+	spinlock_t				pasid_lock;
 };
 
+int amdgpu_vm_alloc_pasid(unsigned int bits);
+void amdgpu_vm_free_pasid(unsigned int pasid);
 void amdgpu_vm_manager_init(struct amdgpu_device *adev);
 void amdgpu_vm_manager_fini(struct amdgpu_device *adev);
 int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-		   int vm_context);
+		   int vm_context, unsigned int pasid);
 void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm);
+bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev,
+				  unsigned int pasid);
 void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
 			 struct list_head *validated,
 			 struct amdgpu_bo_list_entry *entry);
+bool amdgpu_vm_ready(struct amdgpu_vm *vm);
 int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			      int (*callback)(void *p, struct amdgpu_bo *bo),
 			      void *param);
@@ -243,13 +281,13 @@ int amdgpu_vm_update_directories(struct amdgpu_device *adev,
 int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 			  struct amdgpu_vm *vm,
 			  struct dma_fence **fence);
-int amdgpu_vm_clear_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			  struct amdgpu_sync *sync);
+int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
+			   struct amdgpu_vm *vm);
 int amdgpu_vm_bo_update(struct amdgpu_device *adev,
 			struct amdgpu_bo_va *bo_va,
 			bool clear);
 void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
-			     struct amdgpu_bo *bo);
+			     struct amdgpu_bo *bo, bool evicted);
 struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm,
 				       struct amdgpu_bo *bo);
 struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
@@ -269,6 +307,8 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 				struct amdgpu_vm *vm,
 				uint64_t saddr, uint64_t size);
+struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm,
+							 uint64_t addr);
 void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
 		      struct amdgpu_bo_va *bo_va);
 void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev,

drivers/gpu/drm/amd/amdgpu/atom.c

@@ -1343,8 +1343,11 @@ struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios)
 		idx = 0x80;
 
 	str = CSTR(idx);
-	if (*str != '\0')
+	if (*str != '\0') {
 		pr_info("ATOM BIOS: %s\n", str);
+		strlcpy(ctx->vbios_version, str, sizeof(ctx->vbios_version));
+	}
+
 
 	return ctx;
 }

drivers/gpu/drm/amd/amdgpu/atom.h

@@ -140,6 +140,7 @@ struct atom_context {
 	int io_mode;
 	uint32_t *scratch;
 	int scratch_size_bytes;
+	char vbios_version[20];
 };
 
 extern int amdgpu_atom_debug;

drivers/gpu/drm/amd/amdgpu/ci_dpm.c

@@ -307,7 +307,6 @@ static int ci_set_power_limit(struct amdgpu_device *adev, u32 n);
 static int ci_set_overdrive_target_tdp(struct amdgpu_device *adev,
 				       u32 target_tdp);
 static int ci_update_uvd_dpm(struct amdgpu_device *adev, bool gate);
-static void ci_dpm_set_dpm_funcs(struct amdgpu_device *adev);
 static void ci_dpm_set_irq_funcs(struct amdgpu_device *adev);
 
 static PPSMC_Result amdgpu_ci_send_msg_to_smc_with_parameter(struct amdgpu_device *adev,
@@ -883,8 +882,9 @@ static int ci_power_control_set_level(struct amdgpu_device *adev)
 	return ret;
 }
 
-static void ci_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate)
+static void ci_dpm_powergate_uvd(void *handle, bool gate)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	pi->uvd_power_gated = gate;
@@ -901,8 +901,9 @@ static void ci_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate)
 	}
 }
 
-static bool ci_dpm_vblank_too_short(struct amdgpu_device *adev)
+static bool ci_dpm_vblank_too_short(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	u32 vblank_time = amdgpu_dpm_get_vblank_time(adev);
 	u32 switch_limit = adev->mc.vram_type == AMDGPU_VRAM_TYPE_GDDR5 ? 450 : 300;
 
@@ -1210,11 +1211,12 @@ static int ci_fan_ctrl_stop_smc_fan_control(struct amdgpu_device *adev)
 	}
 }
 
-static int ci_dpm_get_fan_speed_percent(struct amdgpu_device *adev,
+static int ci_dpm_get_fan_speed_percent(void *handle,
 					u32 *speed)
 {
 	u32 duty, duty100;
 	u64 tmp64;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	if (adev->pm.no_fan)
 		return -ENOENT;
@@ -1237,12 +1239,13 @@ static int ci_dpm_get_fan_speed_percent(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int ci_dpm_set_fan_speed_percent(struct amdgpu_device *adev,
+static int ci_dpm_set_fan_speed_percent(void *handle,
 					u32 speed)
 {
 	u32 tmp;
 	u32 duty, duty100;
 	u64 tmp64;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	if (adev->pm.no_fan)
@@ -1271,8 +1274,10 @@ static int ci_dpm_set_fan_speed_percent(struct amdgpu_device *adev,
 	return 0;
 }
 
-static void ci_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode)
+static void ci_dpm_set_fan_control_mode(void *handle, u32 mode)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
 	switch (mode) {
 	case AMD_FAN_CTRL_NONE:
 		if (adev->pm.dpm.fan.ucode_fan_control)
@@ -1292,8 +1297,9 @@ static void ci_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode)
 	}
 }
 
-static u32 ci_dpm_get_fan_control_mode(struct amdgpu_device *adev)
+static u32 ci_dpm_get_fan_control_mode(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	if (pi->fan_is_controlled_by_smc)
@@ -4378,9 +4384,10 @@ static u32 ci_get_lowest_enabled_level(struct amdgpu_device *adev,
 }
 
 
-static int ci_dpm_force_performance_level(struct amdgpu_device *adev,
+static int ci_dpm_force_performance_level(void *handle,
 					  enum amd_dpm_forced_level level)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	u32 tmp, levels, i;
 	int ret;
@@ -5291,8 +5298,9 @@ static void ci_update_requested_ps(struct amdgpu_device *adev,
 	adev->pm.dpm.requested_ps = &pi->requested_rps;
 }
 
-static int ci_dpm_pre_set_power_state(struct amdgpu_device *adev)
+static int ci_dpm_pre_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps;
 	struct amdgpu_ps *new_ps = &requested_ps;
@@ -5304,8 +5312,9 @@ static int ci_dpm_pre_set_power_state(struct amdgpu_device *adev)
 	return 0;
 }
 
-static void ci_dpm_post_set_power_state(struct amdgpu_device *adev)
+static void ci_dpm_post_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct amdgpu_ps *new_ps = &pi->requested_rps;
 
@@ -5479,8 +5488,9 @@ static void ci_dpm_disable(struct amdgpu_device *adev)
 	ci_update_current_ps(adev, boot_ps);
 }
 
-static int ci_dpm_set_power_state(struct amdgpu_device *adev)
+static int ci_dpm_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct amdgpu_ps *new_ps = &pi->requested_rps;
 	struct amdgpu_ps *old_ps = &pi->current_rps;
@@ -5551,8 +5561,10 @@ static void ci_dpm_reset_asic(struct amdgpu_device *adev)
 }
 #endif
 
-static void ci_dpm_display_configuration_changed(struct amdgpu_device *adev)
+static void ci_dpm_display_configuration_changed(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
 	ci_program_display_gap(adev);
 }
 
@@ -6105,9 +6117,10 @@ static int ci_dpm_init(struct amdgpu_device *adev)
 }
 
 static void
-ci_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
+ci_dpm_debugfs_print_current_performance_level(void *handle,
 					       struct seq_file *m)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct amdgpu_ps *rps = &pi->current_rps;
 	u32 sclk = ci_get_average_sclk_freq(adev);
@@ -6131,12 +6144,13 @@ ci_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
 	seq_printf(m, "GPU load: %u %%\n", activity_percent);
 }
 
-static void ci_dpm_print_power_state(struct amdgpu_device *adev,
-				     struct amdgpu_ps *rps)
+static void ci_dpm_print_power_state(void *handle, void *current_ps)
 {
+	struct amdgpu_ps *rps = (struct amdgpu_ps *)current_ps;
 	struct ci_ps *ps = ci_get_ps(rps);
 	struct ci_pl *pl;
 	int i;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	amdgpu_dpm_print_class_info(rps->class, rps->class2);
 	amdgpu_dpm_print_cap_info(rps->caps);
@@ -6158,20 +6172,23 @@ static inline bool ci_are_power_levels_equal(const struct ci_pl *ci_cpl1,
 		  (ci_cpl1->pcie_lane == ci_cpl2->pcie_lane));
 }
 
-static int ci_check_state_equal(struct amdgpu_device *adev,
-				struct amdgpu_ps *cps,
-				struct amdgpu_ps *rps,
+static int ci_check_state_equal(void *handle,
+				void *current_ps,
+				void *request_ps,
 				bool *equal)
 {
 	struct ci_ps *ci_cps;
 	struct ci_ps *ci_rps;
 	int i;
+	struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps;
+	struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	if (adev == NULL || cps == NULL || rps == NULL || equal == NULL)
 		return -EINVAL;
 
-	ci_cps = ci_get_ps(cps);
-	ci_rps = ci_get_ps(rps);
+	ci_cps = ci_get_ps((struct amdgpu_ps *)cps);
+	ci_rps = ci_get_ps((struct amdgpu_ps *)rps);
 
 	if (ci_cps == NULL) {
 		*equal = false;
@@ -6199,8 +6216,9 @@ static int ci_check_state_equal(struct amdgpu_device *adev,
 	return 0;
 }
 
-static u32 ci_dpm_get_sclk(struct amdgpu_device *adev, bool low)
+static u32 ci_dpm_get_sclk(void *handle, bool low)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_ps *requested_state = ci_get_ps(&pi->requested_rps);
 
@@ -6210,8 +6228,9 @@ static u32 ci_dpm_get_sclk(struct amdgpu_device *adev, bool low)
 		return requested_state->performance_levels[requested_state->performance_level_count - 1].sclk;
 }
 
-static u32 ci_dpm_get_mclk(struct amdgpu_device *adev, bool low)
+static u32 ci_dpm_get_mclk(void *handle, bool low)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_ps *requested_state = ci_get_ps(&pi->requested_rps);
 
@@ -6222,10 +6241,11 @@ static u32 ci_dpm_get_mclk(struct amdgpu_device *adev, bool low)
 }
 
 /* get temperature in millidegrees */
-static int ci_dpm_get_temp(struct amdgpu_device *adev)
+static int ci_dpm_get_temp(void *handle)
 {
 	u32 temp;
 	int actual_temp = 0;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	temp = (RREG32_SMC(ixCG_MULT_THERMAL_STATUS) & CG_MULT_THERMAL_STATUS__CTF_TEMP_MASK) >>
 		CG_MULT_THERMAL_STATUS__CTF_TEMP__SHIFT;
@@ -6261,7 +6281,6 @@ static int ci_dpm_early_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	ci_dpm_set_dpm_funcs(adev);
 	ci_dpm_set_irq_funcs(adev);
 
 	return 0;
@@ -6346,7 +6365,6 @@ static int ci_dpm_sw_fini(void *handle)
 	flush_work(&adev->pm.dpm.thermal.work);
 
 	mutex_lock(&adev->pm.mutex);
-	amdgpu_pm_sysfs_fini(adev);
 	ci_dpm_fini(adev);
 	mutex_unlock(&adev->pm.mutex);
 
@@ -6551,9 +6569,10 @@ static int ci_dpm_set_powergating_state(void *handle,
 	return 0;
 }
 
-static int ci_dpm_print_clock_levels(struct amdgpu_device *adev,
+static int ci_dpm_print_clock_levels(void *handle,
 		enum pp_clock_type type, char *buf)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_single_dpm_table *sclk_table = &pi->dpm_table.sclk_table;
 	struct ci_single_dpm_table *mclk_table = &pi->dpm_table.mclk_table;
@@ -6618,9 +6637,10 @@ static int ci_dpm_print_clock_levels(struct amdgpu_device *adev,
 	return size;
 }
 
-static int ci_dpm_force_clock_level(struct amdgpu_device *adev,
+static int ci_dpm_force_clock_level(void *handle,
 		enum pp_clock_type type, uint32_t mask)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	if (adev->pm.dpm.forced_level & (AMD_DPM_FORCED_LEVEL_AUTO |
@@ -6664,8 +6684,9 @@ static int ci_dpm_force_clock_level(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int ci_dpm_get_sclk_od(struct amdgpu_device *adev)
+static int ci_dpm_get_sclk_od(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_single_dpm_table *sclk_table = &(pi->dpm_table.sclk_table);
 	struct ci_single_dpm_table *golden_sclk_table =
@@ -6680,8 +6701,9 @@ static int ci_dpm_get_sclk_od(struct amdgpu_device *adev)
 	return value;
 }
 
-static int ci_dpm_set_sclk_od(struct amdgpu_device *adev, uint32_t value)
+static int ci_dpm_set_sclk_od(void *handle, uint32_t value)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_ps *ps = ci_get_ps(adev->pm.dpm.requested_ps);
 	struct ci_single_dpm_table *golden_sclk_table =
@@ -6698,8 +6720,9 @@ static int ci_dpm_set_sclk_od(struct amdgpu_device *adev, uint32_t value)
 	return 0;
 }
 
-static int ci_dpm_get_mclk_od(struct amdgpu_device *adev)
+static int ci_dpm_get_mclk_od(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_single_dpm_table *mclk_table = &(pi->dpm_table.mclk_table);
 	struct ci_single_dpm_table *golden_mclk_table =
@@ -6714,8 +6737,9 @@ static int ci_dpm_get_mclk_od(struct amdgpu_device *adev)
 	return value;
 }
 
-static int ci_dpm_set_mclk_od(struct amdgpu_device *adev, uint32_t value)
+static int ci_dpm_set_mclk_od(void *handle, uint32_t value)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct ci_ps *ps = ci_get_ps(adev->pm.dpm.requested_ps);
 	struct ci_single_dpm_table *golden_mclk_table =
@@ -6732,9 +6756,10 @@ static int ci_dpm_set_mclk_od(struct amdgpu_device *adev, uint32_t value)
 	return 0;
 }
 
-static int ci_dpm_get_power_profile_state(struct amdgpu_device *adev,
+static int ci_dpm_get_power_profile_state(void *handle,
 		struct amd_pp_profile *query)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	if (!pi || !query)
@@ -6851,9 +6876,10 @@ static int ci_set_power_profile_state(struct amdgpu_device *adev,
 	return result;
 }
 
-static int ci_dpm_set_power_profile_state(struct amdgpu_device *adev,
+static int ci_dpm_set_power_profile_state(void *handle,
 		struct amd_pp_profile *request)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	int ret = -1;
 
@@ -6906,9 +6932,10 @@ static int ci_dpm_set_power_profile_state(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int ci_dpm_reset_power_profile_state(struct amdgpu_device *adev,
+static int ci_dpm_reset_power_profile_state(void *handle,
 		struct amd_pp_profile *request)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 
 	if (!pi || !request)
@@ -6927,9 +6954,10 @@ static int ci_dpm_reset_power_profile_state(struct amdgpu_device *adev,
 		return -EINVAL;
 }
 
-static int ci_dpm_switch_power_profile(struct amdgpu_device *adev,
+static int ci_dpm_switch_power_profile(void *handle,
 		enum amd_pp_profile_type type)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct ci_power_info *pi = ci_get_pi(adev);
 	struct amd_pp_profile request = {0};
 
@@ -6944,11 +6972,12 @@ static int ci_dpm_switch_power_profile(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int ci_dpm_read_sensor(struct amdgpu_device *adev, int idx,
+static int ci_dpm_read_sensor(void *handle, int idx,
 			      void *value, int *size)
 {
 	u32 activity_percent = 50;
 	int ret;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	/* size must be at least 4 bytes for all sensors */
 	if (*size < 4)
@@ -7003,7 +7032,7 @@ const struct amd_ip_funcs ci_dpm_ip_funcs = {
 	.set_powergating_state = ci_dpm_set_powergating_state,
 };
 
-static const struct amdgpu_dpm_funcs ci_dpm_funcs = {
+const struct amd_pm_funcs ci_dpm_funcs = {
 	.get_temperature = &ci_dpm_get_temp,
 	.pre_set_power_state = &ci_dpm_pre_set_power_state,
 	.set_power_state = &ci_dpm_set_power_state,
@@ -7035,12 +7064,6 @@ static const struct amdgpu_dpm_funcs ci_dpm_funcs = {
 	.read_sensor = ci_dpm_read_sensor,
 };
 
-static void ci_dpm_set_dpm_funcs(struct amdgpu_device *adev)
-{
-	if (adev->pm.funcs == NULL)
-		adev->pm.funcs = &ci_dpm_funcs;
-}
-
 static const struct amdgpu_irq_src_funcs ci_dpm_irq_funcs = {
 	.set = ci_dpm_set_interrupt_state,
 	.process = ci_dpm_process_interrupt,

drivers/gpu/drm/amd/amdgpu/cik_dpm.h

@@ -26,5 +26,6 @@
 
 extern const struct amd_ip_funcs ci_dpm_ip_funcs;
 extern const struct amd_ip_funcs kv_dpm_ip_funcs;
-
+extern const struct amd_pm_funcs ci_dpm_funcs;
+extern const struct amd_pm_funcs kv_dpm_funcs;
 #endif

drivers/gpu/drm/amd/amdgpu/cik_ih.c

@@ -228,6 +228,34 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev)
  * [127:96] - reserved
  */
 
+/**
+ * cik_ih_prescreen_iv - prescreen an interrupt vector
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns true if the interrupt vector should be further processed.
+ */
+static bool cik_ih_prescreen_iv(struct amdgpu_device *adev)
+{
+	u32 ring_index = adev->irq.ih.rptr >> 2;
+	u16 pasid;
+
+	switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) {
+	case 146:
+	case 147:
+		pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16;
+		if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid))
+			return true;
+		break;
+	default:
+		/* Not a VM fault */
+		return true;
+	}
+
+	adev->irq.ih.rptr += 16;
+	return false;
+}
+
  /**
  * cik_ih_decode_iv - decode an interrupt vector
  *
@@ -433,6 +461,7 @@ static const struct amd_ip_funcs cik_ih_ip_funcs = {
 
 static const struct amdgpu_ih_funcs cik_ih_funcs = {
 	.get_wptr = cik_ih_get_wptr,
+	.prescreen_iv = cik_ih_prescreen_iv,
 	.decode_iv = cik_ih_decode_iv,
 	.set_rptr = cik_ih_set_rptr
 };

drivers/gpu/drm/amd/amdgpu/cik_sdma.c

@@ -1387,8 +1387,13 @@ static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev)
 }
 
 static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = {
+	.copy_pte_num_dw = 7,
 	.copy_pte = cik_sdma_vm_copy_pte,
+
 	.write_pte = cik_sdma_vm_write_pte,
+
+	.set_max_nums_pte_pde = 0x1fffff >> 3,
+	.set_pte_pde_num_dw = 10,
 	.set_pte_pde = cik_sdma_vm_set_pte_pde,
 };
 

drivers/gpu/drm/amd/amdgpu/cz_ih.c

@@ -207,6 +207,34 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev)
 	return (wptr & adev->irq.ih.ptr_mask);
 }
 
+/**
+ * cz_ih_prescreen_iv - prescreen an interrupt vector
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns true if the interrupt vector should be further processed.
+ */
+static bool cz_ih_prescreen_iv(struct amdgpu_device *adev)
+{
+	u32 ring_index = adev->irq.ih.rptr >> 2;
+	u16 pasid;
+
+	switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) {
+	case 146:
+	case 147:
+		pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16;
+		if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid))
+			return true;
+		break;
+	default:
+		/* Not a VM fault */
+		return true;
+	}
+
+	adev->irq.ih.rptr += 16;
+	return false;
+}
+
 /**
  * cz_ih_decode_iv - decode an interrupt vector
  *
@@ -414,6 +442,7 @@ static const struct amd_ip_funcs cz_ih_ip_funcs = {
 
 static const struct amdgpu_ih_funcs cz_ih_funcs = {
 	.get_wptr = cz_ih_get_wptr,
+	.prescreen_iv = cz_ih_prescreen_iv,
 	.decode_iv = cz_ih_decode_iv,
 	.set_rptr = cz_ih_set_rptr
 };

drivers/gpu/drm/amd/amdgpu/dce_virtual.c

@@ -288,7 +288,7 @@ dce_virtual_encoder(struct drm_connector *connector)
 		if (connector->encoder_ids[i] == 0)
 			break;
 
-		encoder = drm_encoder_find(connector->dev, connector->encoder_ids[i]);
+		encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]);
 		if (!encoder)
 			continue;
 
@@ -298,7 +298,7 @@ dce_virtual_encoder(struct drm_connector *connector)
 
 	/* pick the first one */
 	if (enc_id)
-		return drm_encoder_find(connector->dev, enc_id);
+		return drm_encoder_find(connector->dev, NULL, enc_id);
 	return NULL;
 }
 

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

@@ -20,6 +20,7 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  *
  */
+#include <linux/kernel.h>
 #include <linux/firmware.h>
 #include <drm/drmP.h>
 #include "amdgpu.h"
@@ -125,24 +126,39 @@ MODULE_FIRMWARE("amdgpu/fiji_mec2.bin");
 MODULE_FIRMWARE("amdgpu/fiji_rlc.bin");
 
 MODULE_FIRMWARE("amdgpu/polaris11_ce.bin");
+MODULE_FIRMWARE("amdgpu/polaris11_ce_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_pfp.bin");
+MODULE_FIRMWARE("amdgpu/polaris11_pfp_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_me.bin");
+MODULE_FIRMWARE("amdgpu/polaris11_me_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_mec.bin");
+MODULE_FIRMWARE("amdgpu/polaris11_mec_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_mec2.bin");
+MODULE_FIRMWARE("amdgpu/polaris11_mec2_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_rlc.bin");
 
 MODULE_FIRMWARE("amdgpu/polaris10_ce.bin");
+MODULE_FIRMWARE("amdgpu/polaris10_ce_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_pfp.bin");
+MODULE_FIRMWARE("amdgpu/polaris10_pfp_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_me.bin");
+MODULE_FIRMWARE("amdgpu/polaris10_me_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_mec.bin");
+MODULE_FIRMWARE("amdgpu/polaris10_mec_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_mec2.bin");
+MODULE_FIRMWARE("amdgpu/polaris10_mec2_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_rlc.bin");
 
 MODULE_FIRMWARE("amdgpu/polaris12_ce.bin");
+MODULE_FIRMWARE("amdgpu/polaris12_ce_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_pfp.bin");
+MODULE_FIRMWARE("amdgpu/polaris12_pfp_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_me.bin");
+MODULE_FIRMWARE("amdgpu/polaris12_me_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_mec.bin");
+MODULE_FIRMWARE("amdgpu/polaris12_mec_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_mec2.bin");
+MODULE_FIRMWARE("amdgpu/polaris12_mec2_2.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_rlc.bin");
 
 static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] =
@@ -918,8 +934,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 		BUG();
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-	err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev);
+	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp_2.bin", chip_name);
+		err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev);
+		if (err == -ENOENT) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
+			err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev);
+		}
+	} else {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
+		err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev);
+	}
 	if (err)
 		goto out;
 	err = amdgpu_ucode_validate(adev->gfx.pfp_fw);
@@ -929,8 +954,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
 	adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-	err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev);
+	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me_2.bin", chip_name);
+		err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev);
+		if (err == -ENOENT) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
+			err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev);
+		}
+	} else {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
+		err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev);
+	}
 	if (err)
 		goto out;
 	err = amdgpu_ucode_validate(adev->gfx.me_fw);
@@ -941,8 +975,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 
 	adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-	err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev);
+	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce_2.bin", chip_name);
+		err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev);
+		if (err == -ENOENT) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
+			err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev);
+		}
+	} else {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
+		err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev);
+	}
 	if (err)
 		goto out;
 	err = amdgpu_ucode_validate(adev->gfx.ce_fw);
@@ -1012,8 +1055,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++)
 		adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
-	err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
+	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec_2.bin", chip_name);
+		err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
+		if (err == -ENOENT) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
+			err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
+		}
+	} else {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
+		err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
+	}
 	if (err)
 		goto out;
 	err = amdgpu_ucode_validate(adev->gfx.mec_fw);
@@ -1025,8 +1077,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 
 	if ((adev->asic_type != CHIP_STONEY) &&
 	    (adev->asic_type != CHIP_TOPAZ)) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
-		err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev);
+		if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2_2.bin", chip_name);
+			err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev);
+			if (err == -ENOENT) {
+				snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
+				err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev);
+			}
+		} else {
+			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
+			err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev);
+		}
 		if (!err) {
 			err = amdgpu_ucode_validate(adev->gfx.mec2_fw);
 			if (err)
@@ -2053,6 +2114,7 @@ static int gfx_v8_0_sw_fini(void *handle)
 	amdgpu_gfx_compute_mqd_sw_fini(adev);
 	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring, &adev->gfx.kiq.irq);
 	amdgpu_gfx_kiq_fini(adev);
+	amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL);
 
 	gfx_v8_0_mec_fini(adev);
 	gfx_v8_0_rlc_fini(adev);
@@ -3891,10 +3953,10 @@ static int gfx_v8_0_init_save_restore_list(struct amdgpu_device *adev)
 				adev->gfx.rlc.reg_list_format_size_bytes >> 2,
 				unique_indices,
 				&indices_count,
-				sizeof(unique_indices) / sizeof(int),
+				ARRAY_SIZE(unique_indices),
 				indirect_start_offsets,
 				&offset_count,
-				sizeof(indirect_start_offsets)/sizeof(int));
+				ARRAY_SIZE(indirect_start_offsets));
 
 	/* save and restore list */
 	WREG32_FIELD(RLC_SRM_CNTL, AUTO_INCR_ADDR, 1);
@@ -3916,14 +3978,14 @@ static int gfx_v8_0_init_save_restore_list(struct amdgpu_device *adev)
 	/* starting offsets starts */
 	WREG32(mmRLC_GPM_SCRATCH_ADDR,
 		adev->gfx.rlc.starting_offsets_start);
-	for (i = 0; i < sizeof(indirect_start_offsets)/sizeof(int); i++)
+	for (i = 0; i < ARRAY_SIZE(indirect_start_offsets); i++)
 		WREG32(mmRLC_GPM_SCRATCH_DATA,
 				indirect_start_offsets[i]);
 
 	/* unique indices */
 	temp = mmRLC_SRM_INDEX_CNTL_ADDR_0;
 	data = mmRLC_SRM_INDEX_CNTL_DATA_0;
-	for (i = 0; i < sizeof(unique_indices) / sizeof(int); i++) {
+	for (i = 0; i < ARRAY_SIZE(unique_indices); i++) {
 		if (unique_indices[i] != 0) {
 			WREG32(temp + i, unique_indices[i] & 0x3FFFF);
 			WREG32(data + i, unique_indices[i] >> 20);
@@ -4071,18 +4133,12 @@ static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev)
 	gfx_v8_0_rlc_reset(adev);
 	gfx_v8_0_init_pg(adev);
 
-	if (!adev->pp_enabled) {
-		if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) {
-			/* legacy rlc firmware loading */
-			r = gfx_v8_0_rlc_load_microcode(adev);
-			if (r)
-				return r;
-		} else {
-			r = adev->smu.smumgr_funcs->check_fw_load_finish(adev,
-							AMDGPU_UCODE_ID_RLC_G);
-			if (r)
-				return -EINVAL;
-		}
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+		/* legacy rlc firmware loading */
+		r = gfx_v8_0_rlc_load_microcode(adev);
+		if (r)
+			return r;
 	}
 
 	gfx_v8_0_rlc_start(adev);
@@ -4577,12 +4633,10 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring)
 	mqd->compute_static_thread_mgmt_se2 = 0xffffffff;
 	mqd->compute_static_thread_mgmt_se3 = 0xffffffff;
 	mqd->compute_misc_reserved = 0x00000003;
-	if (!(adev->flags & AMD_IS_APU)) {
-		mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr
-					     + offsetof(struct vi_mqd_allocation, dynamic_cu_mask));
-		mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr
-					     + offsetof(struct vi_mqd_allocation, dynamic_cu_mask));
-	}
+	mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr
+						     + offsetof(struct vi_mqd_allocation, dynamic_cu_mask));
+	mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr
+						     + offsetof(struct vi_mqd_allocation, dynamic_cu_mask));
 	eop_base_addr = ring->eop_gpu_addr >> 8;
 	mqd->cp_hqd_eop_base_addr_lo = eop_base_addr;
 	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
@@ -4753,7 +4807,7 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 
 	gfx_v8_0_kiq_setting(ring);
 
-	if (adev->gfx.in_reset) { /* for GPU_RESET case */
+	if (adev->in_sriov_reset) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation));
@@ -4790,7 +4844,7 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring)
 	struct vi_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.compute_ring[0];
 
-	if (!adev->gfx.in_reset && !adev->gfx.in_suspend) {
+	if (!adev->in_sriov_reset && !adev->gfx.in_suspend) {
 		memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation));
 		((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
 		((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
@@ -4802,7 +4856,7 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring)
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
-	} else if (adev->gfx.in_reset) { /* for GPU_RESET case */
+	} else if (adev->in_sriov_reset) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation));
@@ -4900,43 +4954,15 @@ static int gfx_v8_0_cp_resume(struct amdgpu_device *adev)
 	if (!(adev->flags & AMD_IS_APU))
 		gfx_v8_0_enable_gui_idle_interrupt(adev, false);
 
-	if (!adev->pp_enabled) {
-		if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) {
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
 			/* legacy firmware loading */
-			r = gfx_v8_0_cp_gfx_load_microcode(adev);
-			if (r)
-				return r;
+		r = gfx_v8_0_cp_gfx_load_microcode(adev);
+		if (r)
+			return r;
 
-			r = gfx_v8_0_cp_compute_load_microcode(adev);
-			if (r)
-				return r;
-		} else {
-			r = adev->smu.smumgr_funcs->check_fw_load_finish(adev,
-							AMDGPU_UCODE_ID_CP_CE);
-			if (r)
-				return -EINVAL;
-
-			r = adev->smu.smumgr_funcs->check_fw_load_finish(adev,
-							AMDGPU_UCODE_ID_CP_PFP);
-			if (r)
-				return -EINVAL;
-
-			r = adev->smu.smumgr_funcs->check_fw_load_finish(adev,
-							AMDGPU_UCODE_ID_CP_ME);
-			if (r)
-				return -EINVAL;
-
-			if (adev->asic_type == CHIP_TOPAZ) {
-				r = gfx_v8_0_cp_compute_load_microcode(adev);
-				if (r)
-					return r;
-			} else {
-				r = adev->smu.smumgr_funcs->check_fw_load_finish(adev,
-										 AMDGPU_UCODE_ID_CP_MEC1);
-				if (r)
-					return -EINVAL;
-			}
-		}
+		r = gfx_v8_0_cp_compute_load_microcode(adev);
+		if (r)
+			return r;
 	}
 
 	r = gfx_v8_0_cp_gfx_resume(adev);
@@ -4975,12 +5001,69 @@ static int gfx_v8_0_hw_init(void *handle)
 	return r;
 }
 
+static int gfx_v8_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	uint32_t scratch, tmp = 0;
+	int r, i;
+
+	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	if (r) {
+		DRM_ERROR("Failed to get scratch reg (%d).\n", r);
+		return r;
+	}
+	WREG32(scratch, 0xCAFEDEAD);
+
+	r = amdgpu_ring_alloc(kiq_ring, 10);
+	if (r) {
+		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
+		amdgpu_gfx_scratch_free(adev, scratch);
+		return r;
+	}
+
+	/* unmap queues */
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+						PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */
+						PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) |
+						PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) |
+						PACKET3_UNMAP_QUEUES_NUM_QUEUES(1));
+	amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index));
+	amdgpu_ring_write(kiq_ring, 0);
+	amdgpu_ring_write(kiq_ring, 0);
+	amdgpu_ring_write(kiq_ring, 0);
+	/* write to scratch for completion */
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
+	amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
+	amdgpu_ring_write(kiq_ring, 0xDEADBEEF);
+	amdgpu_ring_commit(kiq_ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(scratch);
+		if (tmp == 0xDEADBEEF)
+			break;
+		DRM_UDELAY(1);
+	}
+	if (i >= adev->usec_timeout) {
+		DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp);
+		r = -EINVAL;
+	}
+	amdgpu_gfx_scratch_free(adev, scratch);
+	return r;
+}
+
 static int gfx_v8_0_hw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int i;
 
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+
+	/* disable KCQ to avoid CPC touch memory not valid anymore */
+	for (i = 0; i < adev->gfx.num_compute_rings; i++)
+		gfx_v8_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]);
+
 	if (amdgpu_sriov_vf(adev)) {
 		pr_debug("For SRIOV client, shouldn't do anything.\n");
 		return 0;
@@ -5902,7 +5985,6 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev,
 {
 	uint32_t msg_id, pp_state = 0;
 	uint32_t pp_support_state = 0;
-	void *pp_handle = adev->powerplay.pp_handle;
 
 	if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_CGLS)) {
 		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) {
@@ -5920,7 +6002,8 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev,
 				PP_BLOCK_GFX_CG,
 				pp_support_state,
 				pp_state);
-		amd_set_clockgating_by_smu(pp_handle, msg_id);
+		if (adev->powerplay.pp_funcs->set_clockgating_by_smu)
+			amdgpu_dpm_set_clockgating_by_smu(adev, msg_id);
 	}
 
 	if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS)) {
@@ -5941,7 +6024,8 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev,
 				PP_BLOCK_GFX_MG,
 				pp_support_state,
 				pp_state);
-		amd_set_clockgating_by_smu(pp_handle, msg_id);
+		if (adev->powerplay.pp_funcs->set_clockgating_by_smu)
+			amdgpu_dpm_set_clockgating_by_smu(adev, msg_id);
 	}
 
 	return 0;
@@ -5953,7 +6037,6 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev,
 
 	uint32_t msg_id, pp_state = 0;
 	uint32_t pp_support_state = 0;
-	void *pp_handle = adev->powerplay.pp_handle;
 
 	if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_CGLS)) {
 		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) {
@@ -5971,7 +6054,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev,
 				PP_BLOCK_GFX_CG,
 				pp_support_state,
 				pp_state);
-		amd_set_clockgating_by_smu(pp_handle, msg_id);
+		if (adev->powerplay.pp_funcs->set_clockgating_by_smu)
+			amdgpu_dpm_set_clockgating_by_smu(adev, msg_id);
 	}
 
 	if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_GFX_3D_CGLS)) {
@@ -5990,7 +6074,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev,
 				PP_BLOCK_GFX_3D,
 				pp_support_state,
 				pp_state);
-		amd_set_clockgating_by_smu(pp_handle, msg_id);
+		if (adev->powerplay.pp_funcs->set_clockgating_by_smu)
+			amdgpu_dpm_set_clockgating_by_smu(adev, msg_id);
 	}
 
 	if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS)) {
@@ -6011,7 +6096,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev,
 				PP_BLOCK_GFX_MG,
 				pp_support_state,
 				pp_state);
-		amd_set_clockgating_by_smu(pp_handle, msg_id);
+		if (adev->powerplay.pp_funcs->set_clockgating_by_smu)
+			amdgpu_dpm_set_clockgating_by_smu(adev, msg_id);
 	}
 
 	if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) {
@@ -6026,7 +6112,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev,
 				PP_BLOCK_GFX_RLC,
 				pp_support_state,
 				pp_state);
-		amd_set_clockgating_by_smu(pp_handle, msg_id);
+		if (adev->powerplay.pp_funcs->set_clockgating_by_smu)
+			amdgpu_dpm_set_clockgating_by_smu(adev, msg_id);
 	}
 
 	if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) {
@@ -6040,7 +6127,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev,
 			PP_BLOCK_GFX_CP,
 			pp_support_state,
 			pp_state);
-		amd_set_clockgating_by_smu(pp_handle, msg_id);
+		if (adev->powerplay.pp_funcs->set_clockgating_by_smu)
+			amdgpu_dpm_set_clockgating_by_smu(adev, msg_id);
 	}
 
 	return 0;
@@ -6307,6 +6395,104 @@ static void gfx_v8_0_ring_set_wptr_compute(struct amdgpu_ring *ring)
 	WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr));
 }
 
+static void gfx_v8_0_ring_set_pipe_percent(struct amdgpu_ring *ring,
+					   bool acquire)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int pipe_num, tmp, reg;
+	int pipe_percent = acquire ? SPI_WCL_PIPE_PERCENT_GFX__VALUE_MASK : 0x1;
+
+	pipe_num = ring->me * adev->gfx.mec.num_pipe_per_mec + ring->pipe;
+
+	/* first me only has 2 entries, GFX and HP3D */
+	if (ring->me > 0)
+		pipe_num -= 2;
+
+	reg = mmSPI_WCL_PIPE_PERCENT_GFX + pipe_num;
+	tmp = RREG32(reg);
+	tmp = REG_SET_FIELD(tmp, SPI_WCL_PIPE_PERCENT_GFX, VALUE, pipe_percent);
+	WREG32(reg, tmp);
+}
+
+static void gfx_v8_0_pipe_reserve_resources(struct amdgpu_device *adev,
+					    struct amdgpu_ring *ring,
+					    bool acquire)
+{
+	int i, pipe;
+	bool reserve;
+	struct amdgpu_ring *iring;
+
+	mutex_lock(&adev->gfx.pipe_reserve_mutex);
+	pipe = amdgpu_gfx_queue_to_bit(adev, ring->me, ring->pipe, 0);
+	if (acquire)
+		set_bit(pipe, adev->gfx.pipe_reserve_bitmap);
+	else
+		clear_bit(pipe, adev->gfx.pipe_reserve_bitmap);
+
+	if (!bitmap_weight(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES)) {
+		/* Clear all reservations - everyone reacquires all resources */
+		for (i = 0; i < adev->gfx.num_gfx_rings; ++i)
+			gfx_v8_0_ring_set_pipe_percent(&adev->gfx.gfx_ring[i],
+						       true);
+
+		for (i = 0; i < adev->gfx.num_compute_rings; ++i)
+			gfx_v8_0_ring_set_pipe_percent(&adev->gfx.compute_ring[i],
+						       true);
+	} else {
+		/* Lower all pipes without a current reservation */
+		for (i = 0; i < adev->gfx.num_gfx_rings; ++i) {
+			iring = &adev->gfx.gfx_ring[i];
+			pipe = amdgpu_gfx_queue_to_bit(adev,
+						       iring->me,
+						       iring->pipe,
+						       0);
+			reserve = test_bit(pipe, adev->gfx.pipe_reserve_bitmap);
+			gfx_v8_0_ring_set_pipe_percent(iring, reserve);
+		}
+
+		for (i = 0; i < adev->gfx.num_compute_rings; ++i) {
+			iring = &adev->gfx.compute_ring[i];
+			pipe = amdgpu_gfx_queue_to_bit(adev,
+						       iring->me,
+						       iring->pipe,
+						       0);
+			reserve = test_bit(pipe, adev->gfx.pipe_reserve_bitmap);
+			gfx_v8_0_ring_set_pipe_percent(iring, reserve);
+		}
+	}
+
+	mutex_unlock(&adev->gfx.pipe_reserve_mutex);
+}
+
+static void gfx_v8_0_hqd_set_priority(struct amdgpu_device *adev,
+				      struct amdgpu_ring *ring,
+				      bool acquire)
+{
+	uint32_t pipe_priority = acquire ? 0x2 : 0x0;
+	uint32_t queue_priority = acquire ? 0xf : 0x0;
+
+	mutex_lock(&adev->srbm_mutex);
+	vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+
+	WREG32(mmCP_HQD_PIPE_PRIORITY, pipe_priority);
+	WREG32(mmCP_HQD_QUEUE_PRIORITY, queue_priority);
+
+	vi_srbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+static void gfx_v8_0_ring_set_priority_compute(struct amdgpu_ring *ring,
+					       enum amd_sched_priority priority)
+{
+	struct amdgpu_device *adev = ring->adev;
+	bool acquire = priority == AMD_SCHED_PRIORITY_HIGH_HW;
+
+	if (ring->funcs->type != AMDGPU_RING_TYPE_COMPUTE)
+		return;
+
+	gfx_v8_0_hqd_set_priority(adev, ring, acquire);
+	gfx_v8_0_pipe_reserve_resources(adev, ring, acquire);
+}
+
 static void gfx_v8_0_ring_emit_fence_compute(struct amdgpu_ring *ring,
 					     u64 addr, u64 seq,
 					     unsigned flags)
@@ -6752,6 +6938,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = {
 	.test_ib = gfx_v8_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.set_priority = gfx_v8_0_ring_set_priority_compute,
 };
 
 static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_kiq = {
@@ -6960,7 +7147,7 @@ static void gfx_v8_0_ring_emit_ce_meta(struct amdgpu_ring *ring)
 {
 	uint64_t ce_payload_addr;
 	int cnt_ce;
-	static union {
+	union {
 		struct vi_ce_ib_state regular;
 		struct vi_ce_ib_state_chained_ib chained;
 	} ce_payload = {};
@@ -6989,7 +7176,7 @@ static void gfx_v8_0_ring_emit_de_meta(struct amdgpu_ring *ring)
 {
 	uint64_t de_payload_addr, gds_addr, csa_addr;
 	int cnt_de;
-	static union {
+	union {
 		struct vi_de_ib_state regular;
 		struct vi_de_ib_state_chained_ib chained;
 	} de_payload = {};

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

@@ -20,6 +20,7 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  *
  */
+#include <linux/kernel.h>
 #include <linux/firmware.h>
 #include <drm/drmP.h>
 #include "amdgpu.h"
@@ -66,38 +67,70 @@ MODULE_FIRMWARE("amdgpu/raven_rlc.bin");
 
 static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] =
 {
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE),
-	       	SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14)},
-	{SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE),
-		SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15)}
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)},
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14) },
+	{ SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15),
+	  SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15) }
 };
 
 static const u32 golden_settings_gc_9_0[] =
@@ -352,6 +385,25 @@ err1:
         return r;
 }
 
+
+static void gfx_v9_0_free_microcode(struct amdgpu_device *adev)
+{
+	release_firmware(adev->gfx.pfp_fw);
+	adev->gfx.pfp_fw = NULL;
+	release_firmware(adev->gfx.me_fw);
+	adev->gfx.me_fw = NULL;
+	release_firmware(adev->gfx.ce_fw);
+	adev->gfx.ce_fw = NULL;
+	release_firmware(adev->gfx.rlc_fw);
+	adev->gfx.rlc_fw = NULL;
+	release_firmware(adev->gfx.mec_fw);
+	adev->gfx.mec_fw = NULL;
+	release_firmware(adev->gfx.mec2_fw);
+	adev->gfx.mec2_fw = NULL;
+
+	kfree(adev->gfx.rlc.register_list_format);
+}
+
 static int gfx_v9_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
@@ -1120,30 +1172,22 @@ static int gfx_v9_0_ngg_en(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring = &adev->gfx.gfx_ring[0];
 	int r;
-	u32 data;
-	u32 size;
-	u32 base;
+	u32 data, base;
 
 	if (!amdgpu_ngg)
 		return 0;
 
 	/* Program buffer size */
-	data = 0;
-	size = adev->gfx.ngg.buf[NGG_PRIM].size / 256;
-	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE, size);
-
-	size = adev->gfx.ngg.buf[NGG_POS].size / 256;
-	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE, size);
-
+	data = REG_SET_FIELD(0, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE,
+			     adev->gfx.ngg.buf[NGG_PRIM].size >> 8);
+	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE,
+			     adev->gfx.ngg.buf[NGG_POS].size >> 8);
 	WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_1, data);
 
-	data = 0;
-	size = adev->gfx.ngg.buf[NGG_CNTL].size / 256;
-	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE, size);
-
-	size = adev->gfx.ngg.buf[NGG_PARAM].size / 1024;
-	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE, size);
-
+	data = REG_SET_FIELD(0, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE,
+			     adev->gfx.ngg.buf[NGG_CNTL].size >> 8);
+	data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE,
+			     adev->gfx.ngg.buf[NGG_PARAM].size >> 10);
 	WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_2, data);
 
 	/* Program buffer base address */
@@ -1306,7 +1350,10 @@ static int gfx_v9_0_sw_init(void *handle)
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
 		ring = &adev->gfx.gfx_ring[i];
 		ring->ring_obj = NULL;
-		sprintf(ring->name, "gfx");
+		if (!i)
+			sprintf(ring->name, "gfx");
+		else
+			sprintf(ring->name, "gfx_%d", i);
 		ring->use_doorbell = true;
 		ring->doorbell_index = AMDGPU_DOORBELL64_GFX_RING0 << 1;
 		r = amdgpu_ring_init(adev, ring, 1024,
@@ -1346,7 +1393,7 @@ static int gfx_v9_0_sw_init(void *handle)
 		return r;
 
 	/* create MQD for all compute queues as wel as KIQ for SRIOV case */
-	r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct v9_mqd));
+	r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation));
 	if (r)
 		return r;
 
@@ -1398,9 +1445,11 @@ static int gfx_v9_0_sw_fini(void *handle)
 	amdgpu_gfx_compute_mqd_sw_fini(adev);
 	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring, &adev->gfx.kiq.irq);
 	amdgpu_gfx_kiq_fini(adev);
+	amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL);
 
 	gfx_v9_0_mec_fini(adev);
 	gfx_v9_0_ngg_fini(adev);
+	gfx_v9_0_free_microcode(adev);
 
 	return 0;
 }
@@ -1682,10 +1731,10 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev)
 				adev->gfx.rlc.reg_list_format_size_bytes >> 2,
 				unique_indirect_regs,
 				&unique_indirect_reg_count,
-				sizeof(unique_indirect_regs)/sizeof(int),
+				ARRAY_SIZE(unique_indirect_regs),
 				indirect_start_offsets,
 				&indirect_start_offsets_count,
-				sizeof(indirect_start_offsets)/sizeof(int));
+				ARRAY_SIZE(indirect_start_offsets));
 
 	/* enable auto inc in case it is disabled */
 	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL));
@@ -1722,12 +1771,12 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev)
 	/* write the starting offsets to RLC scratch ram */
 	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_ADDR),
 		adev->gfx.rlc.starting_offsets_start);
-	for (i = 0; i < sizeof(indirect_start_offsets)/sizeof(int); i++)
+	for (i = 0; i < ARRAY_SIZE(indirect_start_offsets); i++)
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_DATA),
 			indirect_start_offsets[i]);
 
 	/* load unique indirect regs*/
-	for (i = 0; i < sizeof(unique_indirect_regs)/sizeof(int); i++) {
+	for (i = 0; i < ARRAY_SIZE(unique_indirect_regs); i++) {
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_ADDR_0) + i,
 			unique_indirect_regs[i] & 0x3FFFF);
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_DATA_0) + i,
@@ -1740,11 +1789,7 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev)
 
 static void gfx_v9_0_enable_save_restore_machine(struct amdgpu_device *adev)
 {
-	u32 tmp = 0;
-
-	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL));
-	tmp |= RLC_SRM_CNTL__SRM_ENABLE_MASK;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL), tmp);
+	WREG32_FIELD15(GC, 0, RLC_SRM_CNTL, SRM_ENABLE, 1);
 }
 
 static void pwr_10_0_gfxip_control_over_cgpg(struct amdgpu_device *adev,
@@ -1822,16 +1867,11 @@ static void gfx_v9_0_enable_sck_slow_down_on_power_up(struct amdgpu_device *adev
 	uint32_t default_data = 0;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-
-	if (enable == true) {
-		data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK;
-		if (default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	} else {
-		data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK;
-		if(default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	}
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     SMU_CLK_SLOWDOWN_ON_PU_ENABLE,
+			     enable ? 1 : 0);
+	if (default_data != data)
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
 
 static void gfx_v9_0_enable_sck_slow_down_on_power_down(struct amdgpu_device *adev,
@@ -1841,16 +1881,11 @@ static void gfx_v9_0_enable_sck_slow_down_on_power_down(struct amdgpu_device *ad
 	uint32_t default_data = 0;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-
-	if (enable == true) {
-		data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK;
-		if(default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	} else {
-		data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK;
-		if(default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	}
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     SMU_CLK_SLOWDOWN_ON_PD_ENABLE,
+			     enable ? 1 : 0);
+	if(default_data != data)
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
 
 static void gfx_v9_0_enable_cp_power_gating(struct amdgpu_device *adev,
@@ -1860,16 +1895,11 @@ static void gfx_v9_0_enable_cp_power_gating(struct amdgpu_device *adev,
 	uint32_t default_data = 0;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-
-	if (enable == true) {
-		data &= ~RLC_PG_CNTL__CP_PG_DISABLE_MASK;
-		if(default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	} else {
-		data |= RLC_PG_CNTL__CP_PG_DISABLE_MASK;
-		if(default_data != data)
-			WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
-	}
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     CP_PG_DISABLE,
+			     enable ? 0 : 1);
+	if(default_data != data)
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
 
 static void gfx_v9_0_enable_gfx_cg_power_gating(struct amdgpu_device *adev,
@@ -1878,10 +1908,9 @@ static void gfx_v9_0_enable_gfx_cg_power_gating(struct amdgpu_device *adev,
 	uint32_t data, default_data;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-	if (enable == true)
-		data |= RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK;
-	else
-		data &= ~RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK;
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     GFX_POWER_GATING_ENABLE,
+			     enable ? 1 : 0);
 	if(default_data != data)
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
@@ -1892,10 +1921,9 @@ static void gfx_v9_0_enable_gfx_pipeline_powergating(struct amdgpu_device *adev,
 	uint32_t data, default_data;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-	if (enable == true)
-		data |= RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK;
-	else
-		data &= ~RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK;
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     GFX_PIPELINE_PG_ENABLE,
+			     enable ? 1 : 0);
 	if(default_data != data)
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 
@@ -1910,10 +1938,9 @@ static void gfx_v9_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *ade
 	uint32_t data, default_data;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-	if (enable == true)
-		data |= RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK;
-	else
-		data &= ~RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK;
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     STATIC_PER_CU_PG_ENABLE,
+			     enable ? 1 : 0);
 	if(default_data != data)
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
@@ -1924,10 +1951,9 @@ static void gfx_v9_0_enable_gfx_dynamic_mg_power_gating(struct amdgpu_device *ad
 	uint32_t data, default_data;
 
 	default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL));
-	if (enable == true)
-		data |= RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK;
-	else
-		data &= ~RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK;
+	data = REG_SET_FIELD(data, RLC_PG_CNTL,
+			     DYN_PER_CU_PG_ENABLE,
+			     enable ? 1 : 0);
 	if(default_data != data)
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data);
 }
@@ -1967,13 +1993,8 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev)
 
 void gfx_v9_0_rlc_stop(struct amdgpu_device *adev)
 {
-	u32 tmp = RREG32_SOC15(GC, 0, mmRLC_CNTL);
-
-	tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0);
-	WREG32_SOC15(GC, 0, mmRLC_CNTL, tmp);
-
+	WREG32_FIELD15(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 0);
 	gfx_v9_0_enable_gui_idle_interrupt(adev, false);
-
 	gfx_v9_0_wait_for_rlc_serdes(adev);
 }
 
@@ -2045,8 +2066,10 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev)
 {
 	int r;
 
-	if (amdgpu_sriov_vf(adev))
+	if (amdgpu_sriov_vf(adev)) {
+		gfx_v9_0_init_csb(adev);
 		return 0;
+	}
 
 	gfx_v9_0_rlc_stop(adev);
 
@@ -2463,6 +2486,13 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring)
 	mqd->compute_static_thread_mgmt_se3 = 0xffffffff;
 	mqd->compute_misc_reserved = 0x00000003;
 
+	mqd->dynamic_cu_mask_addr_lo =
+		lower_32_bits(ring->mqd_gpu_addr
+			      + offsetof(struct v9_mqd_allocation, dynamic_cu_mask));
+	mqd->dynamic_cu_mask_addr_hi =
+		upper_32_bits(ring->mqd_gpu_addr
+			      + offsetof(struct v9_mqd_allocation, dynamic_cu_mask));
+
 	eop_base_addr = ring->eop_gpu_addr >> 8;
 	mqd->cp_hqd_eop_base_addr_lo = eop_base_addr;
 	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
@@ -2486,10 +2516,10 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring)
 				    DOORBELL_SOURCE, 0);
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
 				    DOORBELL_HIT, 0);
-	}
-	else
+	} else {
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
 					 DOORBELL_EN, 0);
+	}
 
 	mqd->cp_hqd_pq_doorbell_control = tmp;
 
@@ -2692,10 +2722,10 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 
 	gfx_v9_0_kiq_setting(ring);
 
-	if (adev->gfx.in_reset) { /* for GPU_RESET case */
+	if (adev->in_sriov_reset) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
@@ -2707,7 +2737,9 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 		soc15_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	} else {
-		memset((void *)mqd, 0, sizeof(*mqd));
+		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
+		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
+		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
 		mutex_lock(&adev->srbm_mutex);
 		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
 		gfx_v9_0_mqd_init(ring);
@@ -2716,7 +2748,7 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
 	}
 
 	return 0;
@@ -2728,8 +2760,10 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
 	struct v9_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.compute_ring[0];
 
-	if (!adev->gfx.in_reset && !adev->gfx.in_suspend) {
-		memset((void *)mqd, 0, sizeof(*mqd));
+	if (!adev->in_sriov_reset && !adev->gfx.in_suspend) {
+		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
+		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
+		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
 		mutex_lock(&adev->srbm_mutex);
 		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
 		gfx_v9_0_mqd_init(ring);
@@ -2737,11 +2771,11 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
-	} else if (adev->gfx.in_reset) { /* for GPU_RESET case */
+			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
+	} else if (adev->in_sriov_reset) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
@@ -2882,12 +2916,70 @@ static int gfx_v9_0_hw_init(void *handle)
 	return r;
 }
 
+static int gfx_v9_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	uint32_t scratch, tmp = 0;
+	int r, i;
+
+	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	if (r) {
+		DRM_ERROR("Failed to get scratch reg (%d).\n", r);
+		return r;
+	}
+	WREG32(scratch, 0xCAFEDEAD);
+
+	r = amdgpu_ring_alloc(kiq_ring, 10);
+	if (r) {
+		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
+		amdgpu_gfx_scratch_free(adev, scratch);
+		return r;
+	}
+
+	/* unmap queues */
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+						PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */
+						PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) |
+						PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) |
+						PACKET3_UNMAP_QUEUES_NUM_QUEUES(1));
+	amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index));
+	amdgpu_ring_write(kiq_ring, 0);
+	amdgpu_ring_write(kiq_ring, 0);
+	amdgpu_ring_write(kiq_ring, 0);
+	/* write to scratch for completion */
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
+	amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
+	amdgpu_ring_write(kiq_ring, 0xDEADBEEF);
+	amdgpu_ring_commit(kiq_ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(scratch);
+		if (tmp == 0xDEADBEEF)
+			break;
+		DRM_UDELAY(1);
+	}
+	if (i >= adev->usec_timeout) {
+		DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp);
+		r = -EINVAL;
+	}
+	amdgpu_gfx_scratch_free(adev, scratch);
+	return r;
+}
+
+
 static int gfx_v9_0_hw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int i;
 
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+
+	/* disable KCQ to avoid CPC touch memory not valid anymore */
+	for (i = 0; i < adev->gfx.num_compute_rings; i++)
+		gfx_v9_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]);
+
 	if (amdgpu_sriov_vf(adev)) {
 		pr_debug("For SRIOV client, shouldn't do anything.\n");
 		return 0;
@@ -2930,15 +3022,10 @@ static bool gfx_v9_0_is_idle(void *handle)
 static int gfx_v9_0_wait_for_idle(void *handle)
 {
 	unsigned i;
-	u32 tmp;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		/* read MC_STATUS */
-		tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS) &
-			GRBM_STATUS__GUI_ACTIVE_MASK;
-
-		if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE))
+		if (gfx_v9_0_is_idle(handle))
 			return 0;
 		udelay(1);
 	}
@@ -3497,9 +3584,11 @@ static void gfx_v9_0_ring_set_wptr_gfx(struct amdgpu_ring *ring)
 static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 {
 	u32 ref_and_mask, reg_mem_engine;
-	struct nbio_hdp_flush_reg *nbio_hf_reg;
+	const struct nbio_hdp_flush_reg *nbio_hf_reg;
 
-	if (ring->adev->asic_type == CHIP_VEGA10)
+	if (ring->adev->flags & AMD_IS_APU)
+		nbio_hf_reg = &nbio_v7_0_hdp_flush_reg;
+	else
 		nbio_hf_reg = &nbio_v6_1_hdp_flush_reg;
 
 	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
@@ -3528,7 +3617,7 @@ static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 static void gfx_v9_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring)
 {
 	gfx_v9_0_write_data_to_reg(ring, 0, true,
-				   SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 1);
+				   SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1);
 }
 
 static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
@@ -3718,7 +3807,7 @@ static void gfx_v9_ring_emit_sb(struct amdgpu_ring *ring)
 
 static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring)
 {
-	static struct v9_ce_ib_state ce_payload = {0};
+	struct v9_ce_ib_state ce_payload = {0};
 	uint64_t csa_addr;
 	int cnt;
 
@@ -3737,7 +3826,7 @@ static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring)
 
 static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring)
 {
-	static struct v9_de_ib_state de_payload = {0};
+	struct v9_de_ib_state de_payload = {0};
 	uint64_t csa_addr, gds_addr;
 	int cnt;
 
@@ -3757,6 +3846,12 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring)
 	amdgpu_ring_write_multiple(ring, (void *)&de_payload, sizeof(de_payload) >> 2);
 }
 
+static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start)
+{
+	amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0));
+	amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */
+}
+
 static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
 {
 	uint32_t dw2 = 0;
@@ -3764,6 +3859,8 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
 	if (amdgpu_sriov_vf(ring->adev))
 		gfx_v9_0_ring_emit_ce_meta(ring);
 
+	gfx_v9_0_ring_emit_tmz(ring, true);
+
 	dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */
 	if (flags & AMDGPU_HAVE_CTX_SWITCH) {
 		/* set load_global_config & load_global_uconfig */
@@ -3814,12 +3911,6 @@ static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne
 		ring->ring[offset] = (ring->ring_size>>2) - offset + cur;
 }
 
-static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start)
-{
-	amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0));
-	amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */
-}
-
 static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)
 {
 	struct amdgpu_device *adev = ring->adev;

drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c

@@ -319,6 +319,12 @@ void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev,
 			WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
 	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
 			EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	if (!value) {
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_NO_RETRY_FAULT, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_RETRY_FAULT, 1);
+    }
 	WREG32_SOC15(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL, tmp);
 }
 

drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c

@@ -831,7 +831,7 @@ static int gmc_v6_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	amdgpu_vm_adjust_size(adev, 64, 4);
+	amdgpu_vm_adjust_size(adev, 64, 9);
 	adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18;
 
 	adev->mc.mc_mask = 0xffffffffffULL;
@@ -901,6 +901,8 @@ static int gmc_v6_0_sw_fini(void *handle)
 	gmc_v6_0_gart_fini(adev);
 	amdgpu_gem_force_release(adev);
 	amdgpu_bo_fini(adev);
+	release_firmware(adev->mc.fw);
+	adev->mc.fw = NULL;
 
 	return 0;
 }

drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c

@@ -970,7 +970,7 @@ static int gmc_v7_0_sw_init(void *handle)
 	 * Currently set to 4GB ((1 << 20) 4k pages).
 	 * Max GPUVM size for cayman and SI is 40 bits.
 	 */
-	amdgpu_vm_adjust_size(adev, 64, 4);
+	amdgpu_vm_adjust_size(adev, 64, 9);
 	adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18;
 
 	/* Set the internal MC address mask
@@ -1050,6 +1050,8 @@ static int gmc_v7_0_sw_fini(void *handle)
 	gmc_v7_0_gart_fini(adev);
 	amdgpu_gem_force_release(adev);
 	amdgpu_bo_fini(adev);
+	release_firmware(adev->mc.fw);
+	adev->mc.fw = NULL;
 
 	return 0;
 }

drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c

@@ -1067,7 +1067,7 @@ static int gmc_v8_0_sw_init(void *handle)
 	 * Currently set to 4GB ((1 << 20) 4k pages).
 	 * Max GPUVM size for cayman and SI is 40 bits.
 	 */
-	amdgpu_vm_adjust_size(adev, 64, 4);
+	amdgpu_vm_adjust_size(adev, 64, 9);
 	adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18;
 
 	/* Set the internal MC address mask
@@ -1147,6 +1147,8 @@ static int gmc_v8_0_sw_fini(void *handle)
 	gmc_v8_0_gart_fini(adev);
 	amdgpu_gem_force_release(adev);
 	amdgpu_bo_fini(adev);
+	release_firmware(adev->mc.fw);
+	adev->mc.fw = NULL;
 
 	return 0;
 }

drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c

@@ -32,6 +32,8 @@
 #include "vega10/DC/dce_12_0_offset.h"
 #include "vega10/DC/dce_12_0_sh_mask.h"
 #include "vega10/vega10_enum.h"
+#include "vega10/MMHUB/mmhub_1_0_offset.h"
+#include "vega10/ATHUB/athub_1_0_offset.h"
 
 #include "soc15_common.h"
 
@@ -71,13 +73,25 @@ static const u32 golden_settings_vega10_hdp[] =
 	0xf6e, 0x0fffffff, 0x00000000,
 };
 
+static const u32 golden_settings_mmhub_1_0_0[] =
+{
+	SOC15_REG_OFFSET(MMHUB, 0, mmDAGB1_WRCLI2), 0x00000007, 0xfe5fe0fa,
+	SOC15_REG_OFFSET(MMHUB, 0, mmMMEA1_DRAM_WR_CLI2GRP_MAP0), 0x00000030, 0x55555565
+};
+
+static const u32 golden_settings_athub_1_0_0[] =
+{
+	SOC15_REG_OFFSET(ATHUB, 0, mmRPB_ARB_CNTL), 0x0000ff00, 0x00000800,
+	SOC15_REG_OFFSET(ATHUB, 0, mmRPB_ARB_CNTL2), 0x00ff00ff, 0x00080008
+};
+
 static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *src,
 					unsigned type,
 					enum amdgpu_interrupt_state state)
 {
 	struct amdgpu_vmhub *hub;
-	u32 tmp, reg, bits, i;
+	u32 tmp, reg, bits, i, j;
 
 	bits = VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
 		VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
@@ -89,43 +103,26 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
-		/* MM HUB */
-		hub = &adev->vmhub[AMDGPU_MMHUB];
-		for (i = 0; i< 16; i++) {
-			reg = hub->vm_context0_cntl + i;
-			tmp = RREG32(reg);
-			tmp &= ~bits;
-			WREG32(reg, tmp);
-		}
-
-		/* GFX HUB */
-		hub = &adev->vmhub[AMDGPU_GFXHUB];
-		for (i = 0; i < 16; i++) {
-			reg = hub->vm_context0_cntl + i;
-			tmp = RREG32(reg);
-			tmp &= ~bits;
-			WREG32(reg, tmp);
+		for (j = 0; j < AMDGPU_MAX_VMHUBS; j++) {
+			hub = &adev->vmhub[j];
+			for (i = 0; i < 16; i++) {
+				reg = hub->vm_context0_cntl + i;
+				tmp = RREG32(reg);
+				tmp &= ~bits;
+				WREG32(reg, tmp);
+			}
 		}
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
-		/* MM HUB */
-		hub = &adev->vmhub[AMDGPU_MMHUB];
-		for (i = 0; i< 16; i++) {
-			reg = hub->vm_context0_cntl + i;
-			tmp = RREG32(reg);
-			tmp |= bits;
-			WREG32(reg, tmp);
-		}
-
-		/* GFX HUB */
-		hub = &adev->vmhub[AMDGPU_GFXHUB];
-		for (i = 0; i < 16; i++) {
-			reg = hub->vm_context0_cntl + i;
-			tmp = RREG32(reg);
-			tmp |= bits;
-			WREG32(reg, tmp);
+		for (j = 0; j < AMDGPU_MAX_VMHUBS; j++) {
+			hub = &adev->vmhub[j];
+			for (i = 0; i < 16; i++) {
+				reg = hub->vm_context0_cntl + i;
+				tmp = RREG32(reg);
+				tmp |= bits;
+				WREG32(reg, tmp);
+			}
 		}
-		break;
 	default:
 		break;
 	}
@@ -682,8 +679,17 @@ static void gmc_v9_0_init_golden_registers(struct amdgpu_device *adev)
 {
 	switch (adev->asic_type) {
 	case CHIP_VEGA10:
+		amdgpu_program_register_sequence(adev,
+						golden_settings_mmhub_1_0_0,
+						(const u32)ARRAY_SIZE(golden_settings_mmhub_1_0_0));
+		amdgpu_program_register_sequence(adev,
+						golden_settings_athub_1_0_0,
+						(const u32)ARRAY_SIZE(golden_settings_athub_1_0_0));
 		break;
 	case CHIP_RAVEN:
+		amdgpu_program_register_sequence(adev,
+						golden_settings_athub_1_0_0,
+						(const u32)ARRAY_SIZE(golden_settings_athub_1_0_0));
 		break;
 	default:
 		break;
@@ -713,12 +719,6 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
-	/* After HDP is initialized, flush HDP.*/
-	if (adev->flags & AMD_IS_APU)
-		nbio_v7_0_hdp_flush(adev);
-	else
-		nbio_v6_1_hdp_flush(adev);
-
 	switch (adev->asic_type) {
 	case CHIP_RAVEN:
 		mmhub_v1_0_initialize_power_gating(adev);
@@ -736,13 +736,16 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
-	tmp = RREG32_SOC15(HDP, 0, mmHDP_MISC_CNTL);
-	tmp |= HDP_MISC_CNTL__FLUSH_INVALIDATE_CACHE_MASK;
-	WREG32_SOC15(HDP, 0, mmHDP_MISC_CNTL, tmp);
+	WREG32_FIELD15(HDP, 0, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 1);
 
 	tmp = RREG32_SOC15(HDP, 0, mmHDP_HOST_PATH_CNTL);
 	WREG32_SOC15(HDP, 0, mmHDP_HOST_PATH_CNTL, tmp);
 
+	/* After HDP is initialized, flush HDP.*/
+	if (adev->flags & AMD_IS_APU)
+		nbio_v7_0_hdp_flush(adev);
+	else
+		nbio_v6_1_hdp_flush(adev);
 
 	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS)
 		value = false;
@@ -751,7 +754,6 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 
 	gfxhub_v1_0_set_fault_enable_default(adev, value);
 	mmhub_v1_0_set_fault_enable_default(adev, value);
-
 	gmc_v9_0_gart_flush_gpu_tlb(adev, 0);
 
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
@@ -770,17 +772,11 @@ static int gmc_v9_0_hw_init(void *handle)
 	gmc_v9_0_init_golden_registers(adev);
 
 	if (adev->mode_info.num_crtc) {
-		u32 tmp;
-
 		/* Lockout access through VGA aperture*/
-		tmp = RREG32_SOC15(DCE, 0, mmVGA_HDP_CONTROL);
-		tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1);
-		WREG32_SOC15(DCE, 0, mmVGA_HDP_CONTROL, tmp);
+		WREG32_FIELD15(DCE, 0, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1);
 
 		/* disable VGA render */
-		tmp = RREG32_SOC15(DCE, 0, mmVGA_RENDER_CONTROL);
-		tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0);
-		WREG32_SOC15(DCE, 0, mmVGA_RENDER_CONTROL, tmp);
+		WREG32_FIELD15(DCE, 0, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0);
 	}
 
 	r = gmc_v9_0_gart_enable(adev);
@@ -822,9 +818,7 @@ static int gmc_v9_0_suspend(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	gmc_v9_0_hw_fini(adev);
-
-	return 0;
+	return gmc_v9_0_hw_fini(adev);
 }
 
 static int gmc_v9_0_resume(void *handle)

drivers/gpu/drm/amd/amdgpu/iceland_ih.c

@@ -207,6 +207,34 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev)
 	return (wptr & adev->irq.ih.ptr_mask);
 }
 
+/**
+ * iceland_ih_prescreen_iv - prescreen an interrupt vector
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns true if the interrupt vector should be further processed.
+ */
+static bool iceland_ih_prescreen_iv(struct amdgpu_device *adev)
+{
+	u32 ring_index = adev->irq.ih.rptr >> 2;
+	u16 pasid;
+
+	switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) {
+	case 146:
+	case 147:
+		pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16;
+		if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid))
+			return true;
+		break;
+	default:
+		/* Not a VM fault */
+		return true;
+	}
+
+	adev->irq.ih.rptr += 16;
+	return false;
+}
+
 /**
  * iceland_ih_decode_iv - decode an interrupt vector
  *
@@ -412,6 +440,7 @@ static const struct amd_ip_funcs iceland_ih_ip_funcs = {
 
 static const struct amdgpu_ih_funcs iceland_ih_funcs = {
 	.get_wptr = iceland_ih_get_wptr,
+	.prescreen_iv = iceland_ih_prescreen_iv,
 	.decode_iv = iceland_ih_decode_iv,
 	.set_rptr = iceland_ih_set_rptr
 };

drivers/gpu/drm/amd/amdgpu/kv_dpm.c

@@ -42,7 +42,6 @@
 #define KV_MINIMUM_ENGINE_CLOCK         800
 #define SMC_RAM_END                     0x40000
 
-static void kv_dpm_set_dpm_funcs(struct amdgpu_device *adev);
 static void kv_dpm_set_irq_funcs(struct amdgpu_device *adev);
 static int kv_enable_nb_dpm(struct amdgpu_device *adev,
 			    bool enable);
@@ -64,7 +63,7 @@ static int kv_set_thermal_temperature_range(struct amdgpu_device *adev,
 					    int min_temp, int max_temp);
 static int kv_init_fps_limits(struct amdgpu_device *adev);
 
-static void kv_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate);
+static void kv_dpm_powergate_uvd(void *handle, bool gate);
 static void kv_dpm_powergate_vce(struct amdgpu_device *adev, bool gate);
 static void kv_dpm_powergate_samu(struct amdgpu_device *adev, bool gate);
 static void kv_dpm_powergate_acp(struct amdgpu_device *adev, bool gate);
@@ -1245,8 +1244,9 @@ static void kv_update_requested_ps(struct amdgpu_device *adev,
 	adev->pm.dpm.requested_ps = &pi->requested_rps;
 }
 
-static void kv_dpm_enable_bapm(struct amdgpu_device *adev, bool enable)
+static void kv_dpm_enable_bapm(void *handle, bool enable)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	int ret;
 
@@ -1672,8 +1672,9 @@ static int kv_update_acp_dpm(struct amdgpu_device *adev, bool gate)
 	return kv_enable_acp_dpm(adev, !gate);
 }
 
-static void kv_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate)
+static void kv_dpm_powergate_uvd(void *handle, bool gate)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	int ret;
 
@@ -1868,10 +1869,11 @@ static int kv_enable_nb_dpm(struct amdgpu_device *adev,
 	return ret;
 }
 
-static int kv_dpm_force_performance_level(struct amdgpu_device *adev,
+static int kv_dpm_force_performance_level(void *handle,
 					  enum amd_dpm_forced_level level)
 {
 	int ret;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	if (level == AMD_DPM_FORCED_LEVEL_HIGH) {
 		ret = kv_force_dpm_highest(adev);
@@ -1892,8 +1894,9 @@ static int kv_dpm_force_performance_level(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int kv_dpm_pre_set_power_state(struct amdgpu_device *adev)
+static int kv_dpm_pre_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps;
 	struct amdgpu_ps *new_ps = &requested_ps;
@@ -1907,8 +1910,9 @@ static int kv_dpm_pre_set_power_state(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int kv_dpm_set_power_state(struct amdgpu_device *adev)
+static int kv_dpm_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	struct amdgpu_ps *new_ps = &pi->requested_rps;
 	struct amdgpu_ps *old_ps = &pi->current_rps;
@@ -1981,8 +1985,9 @@ static int kv_dpm_set_power_state(struct amdgpu_device *adev)
 	return 0;
 }
 
-static void kv_dpm_post_set_power_state(struct amdgpu_device *adev)
+static void kv_dpm_post_set_power_state(void *handle)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	struct amdgpu_ps *new_ps = &pi->requested_rps;
 
@@ -2848,9 +2853,10 @@ static int kv_dpm_init(struct amdgpu_device *adev)
 }
 
 static void
-kv_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
+kv_dpm_debugfs_print_current_performance_level(void *handle,
 					       struct seq_file *m)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	u32 current_index =
 		(RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX) &
@@ -2875,11 +2881,12 @@ kv_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
 }
 
 static void
-kv_dpm_print_power_state(struct amdgpu_device *adev,
-			 struct amdgpu_ps *rps)
+kv_dpm_print_power_state(void *handle, void *request_ps)
 {
 	int i;
+	struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps;
 	struct kv_ps *ps = kv_get_ps(rps);
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	amdgpu_dpm_print_class_info(rps->class, rps->class2);
 	amdgpu_dpm_print_cap_info(rps->caps);
@@ -2905,13 +2912,14 @@ static void kv_dpm_fini(struct amdgpu_device *adev)
 	amdgpu_free_extended_power_table(adev);
 }
 
-static void kv_dpm_display_configuration_changed(struct amdgpu_device *adev)
+static void kv_dpm_display_configuration_changed(void *handle)
 {
 
 }
 
-static u32 kv_dpm_get_sclk(struct amdgpu_device *adev, bool low)
+static u32 kv_dpm_get_sclk(void *handle, bool low)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	struct kv_ps *requested_state = kv_get_ps(&pi->requested_rps);
 
@@ -2921,18 +2929,20 @@ static u32 kv_dpm_get_sclk(struct amdgpu_device *adev, bool low)
 		return requested_state->levels[requested_state->num_levels - 1].sclk;
 }
 
-static u32 kv_dpm_get_mclk(struct amdgpu_device *adev, bool low)
+static u32 kv_dpm_get_mclk(void *handle, bool low)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 
 	return pi->sys_info.bootup_uma_clk;
 }
 
 /* get temperature in millidegrees */
-static int kv_dpm_get_temp(struct amdgpu_device *adev)
+static int kv_dpm_get_temp(void *handle)
 {
 	u32 temp;
 	int actual_temp = 0;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	temp = RREG32_SMC(0xC0300E0C);
 
@@ -2950,7 +2960,6 @@ static int kv_dpm_early_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	kv_dpm_set_dpm_funcs(adev);
 	kv_dpm_set_irq_funcs(adev);
 
 	return 0;
@@ -2960,16 +2969,10 @@ static int kv_dpm_late_init(void *handle)
 {
 	/* powerdown unused blocks for now */
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int ret;
 
 	if (!amdgpu_dpm)
 		return 0;
 
-	/* init the sysfs and debugfs files late */
-	ret = amdgpu_pm_sysfs_init(adev);
-	if (ret)
-		return ret;
-
 	kv_dpm_powergate_acp(adev, true);
 	kv_dpm_powergate_samu(adev, true);
 
@@ -3031,7 +3034,6 @@ static int kv_dpm_sw_fini(void *handle)
 	flush_work(&adev->pm.dpm.thermal.work);
 
 	mutex_lock(&adev->pm.mutex);
-	amdgpu_pm_sysfs_fini(adev);
 	kv_dpm_fini(adev);
 	mutex_unlock(&adev->pm.mutex);
 
@@ -3222,14 +3224,17 @@ static inline bool kv_are_power_levels_equal(const struct kv_pl *kv_cpl1,
 		  (kv_cpl1->force_nbp_state == kv_cpl2->force_nbp_state));
 }
 
-static int kv_check_state_equal(struct amdgpu_device *adev,
-				struct amdgpu_ps *cps,
-				struct amdgpu_ps *rps,
+static int kv_check_state_equal(void *handle,
+				void *current_ps,
+				void *request_ps,
 				bool *equal)
 {
 	struct kv_ps *kv_cps;
 	struct kv_ps *kv_rps;
 	int i;
+	struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps;
+	struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps;
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
 	if (adev == NULL || cps == NULL || rps == NULL || equal == NULL)
 		return -EINVAL;
@@ -3262,9 +3267,10 @@ static int kv_check_state_equal(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int kv_dpm_read_sensor(struct amdgpu_device *adev, int idx,
+static int kv_dpm_read_sensor(void *handle, int idx,
 			      void *value, int *size)
 {
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	struct kv_power_info *pi = kv_get_pi(adev);
 	uint32_t sclk;
 	u32 pl_index =
@@ -3312,7 +3318,7 @@ const struct amd_ip_funcs kv_dpm_ip_funcs = {
 	.set_powergating_state = kv_dpm_set_powergating_state,
 };
 
-static const struct amdgpu_dpm_funcs kv_dpm_funcs = {
+const struct amd_pm_funcs kv_dpm_funcs = {
 	.get_temperature = &kv_dpm_get_temp,
 	.pre_set_power_state = &kv_dpm_pre_set_power_state,
 	.set_power_state = &kv_dpm_set_power_state,
@@ -3330,12 +3336,6 @@ static const struct amdgpu_dpm_funcs kv_dpm_funcs = {
 	.read_sensor = &kv_dpm_read_sensor,
 };
 
-static void kv_dpm_set_dpm_funcs(struct amdgpu_device *adev)
-{
-	if (adev->pm.funcs == NULL)
-		adev->pm.funcs = &kv_dpm_funcs;
-}
-
 static const struct amdgpu_irq_src_funcs kv_dpm_irq_funcs = {
 	.set = kv_dpm_set_interrupt_state,
 	.process = kv_dpm_process_interrupt,

drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c

@@ -273,7 +273,7 @@ static const struct pctl_data pctl0_data[] = {
 	{0x135, 0x12a810},
 	{0x149, 0x7a82c}
 };
-#define PCTL0_DATA_LEN (sizeof(pctl0_data)/sizeof(pctl0_data[0]))
+#define PCTL0_DATA_LEN (ARRAY_SIZE(pctl0_data))
 
 #define PCTL0_RENG_EXEC_END_PTR 0x151
 #define PCTL0_STCTRL_REG_SAVE_RANGE0_BASE  0xa640
@@ -309,7 +309,7 @@ static const struct pctl_data pctl1_data[] = {
 	{0x1f0, 0x5000a7f6},
 	{0x1f1, 0x5000a7e4}
 };
-#define PCTL1_DATA_LEN (sizeof(pctl1_data)/sizeof(pctl1_data[0]))
+#define PCTL1_DATA_LEN (ARRAY_SIZE(pctl1_data))
 
 #define PCTL1_RENG_EXEC_END_PTR 0x1f1
 #define PCTL1_STCTRL_REG_SAVE_RANGE0_BASE  0xa000
@@ -561,6 +561,13 @@ void mmhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, bool value)
 			WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
 	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
 			EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	if (!value) {
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_NO_RETRY_FAULT, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_RETRY_FAULT, 1);
+    }
+
 	WREG32_SOC15(MMHUB, 0, mmVM_L2_PROTECTION_FAULT_CNTL, tmp);
 }