diff --git a/Documentation/devicetree/bindings/pci/layerscape-pci.txt b/Documentation/devicetree/bindings/pci/layerscape-pci.txt
index ef683b2fd23a..41e9f55a1467 100644
--- a/Documentation/devicetree/bindings/pci/layerscape-pci.txt
+++ b/Documentation/devicetree/bindings/pci/layerscape-pci.txt
@@ -24,6 +24,9 @@ Required properties:
   The first entry must be a link to the SCFG device node
   The second entry must be '0' or '1' based on physical PCIe controller index.
   This is used to get SCFG PEXN registers
+- dma-coherent: Indicates that the hardware IP block can ensure the coherency
+  of the data transferred from/to the IP block. This can avoid the software
+  cache flush/invalid actions, and improve the performance significantly.
 
 Example:
 
@@ -38,6 +41,7 @@ Example:
 		#address-cells = <3>;
 		#size-cells = <2>;
 		device_type = "pci";
+		dma-coherent;
 		num-lanes = <4>;
 		bus-range = <0x0 0xff>;
 		ranges = <0x81000000 0x0 0x00000000 0x40 0x00010000 0x0 0x00010000   /* downstream I/O */
diff --git a/arch/arm64/boot/dts/freescale/fsl-ls1043a.dtsi b/arch/arm64/boot/dts/freescale/fsl-ls1043a.dtsi
index de0323b48b1e..19572d85c80d 100644
--- a/arch/arm64/boot/dts/freescale/fsl-ls1043a.dtsi
+++ b/arch/arm64/boot/dts/freescale/fsl-ls1043a.dtsi
@@ -51,7 +51,7 @@
 	#size-cells = <2>;
 
 	cpus {
-		#address-cells = <2>;
+		#address-cells = <1>;
 		#size-cells = <0>;
 
 		/*
@@ -63,29 +63,37 @@
 		cpu0: cpu@0 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a53";
-			reg = <0x0 0x0>;
+			reg = <0x0>;
 			clocks = <&clockgen 1 0>;
+			next-level-cache = <&l2>;
 		};
 
 		cpu1: cpu@1 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a53";
-			reg = <0x0 0x1>;
+			reg = <0x1>;
 			clocks = <&clockgen 1 0>;
+			next-level-cache = <&l2>;
 		};
 
 		cpu2: cpu@2 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a53";
-			reg = <0x0 0x2>;
+			reg = <0x2>;
 			clocks = <&clockgen 1 0>;
+			next-level-cache = <&l2>;
 		};
 
 		cpu3: cpu@3 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a53";
-			reg = <0x0 0x3>;
+			reg = <0x3>;
 			clocks = <&clockgen 1 0>;
+			next-level-cache = <&l2>;
+		};
+
+		l2: l2-cache {
+			compatible = "cache";
 		};
 	};
 
@@ -422,6 +430,7 @@
 			interrupts = <0 60 0x4>;
 			dr_mode = "host";
 			snps,quirk-frame-length-adjustment = <0x20>;
+			snps,dis_rxdet_inp3_quirk;
 		};
 
 		usb1: usb3@3000000 {
@@ -430,6 +439,7 @@
 			interrupts = <0 61 0x4>;
 			dr_mode = "host";
 			snps,quirk-frame-length-adjustment = <0x20>;
+			snps,dis_rxdet_inp3_quirk;
 		};
 
 		usb2: usb3@3100000 {
@@ -438,6 +448,7 @@
 			interrupts = <0 63 0x4>;
 			dr_mode = "host";
 			snps,quirk-frame-length-adjustment = <0x20>;
+			snps,dis_rxdet_inp3_quirk;
 		};
 
 		sata: sata@3200000 {
@@ -479,6 +490,7 @@
 			#address-cells = <3>;
 			#size-cells = <2>;
 			device_type = "pci";
+			dma-coherent;
 			num-lanes = <4>;
 			bus-range = <0x0 0xff>;
 			ranges = <0x81000000 0x0 0x00000000 0x40 0x00010000 0x0 0x00010000   /* downstream I/O */
@@ -503,6 +515,7 @@
 			#address-cells = <3>;
 			#size-cells = <2>;
 			device_type = "pci";
+			dma-coherent;
 			num-lanes = <2>;
 			bus-range = <0x0 0xff>;
 			ranges = <0x81000000 0x0 0x00000000 0x48 0x00010000 0x0 0x00010000   /* downstream I/O */
@@ -527,6 +540,7 @@
 			#address-cells = <3>;
 			#size-cells = <2>;
 			device_type = "pci";
+			dma-coherent;
 			num-lanes = <2>;
 			bus-range = <0x0 0xff>;
 			ranges = <0x81000000 0x0 0x00000000 0x50 0x00010000 0x0 0x00010000   /* downstream I/O */
diff --git a/arch/arm64/boot/dts/freescale/fsl-ls2080a.dtsi b/arch/arm64/boot/dts/freescale/fsl-ls2080a.dtsi
index 3187c822afa3..21023a388c29 100644
--- a/arch/arm64/boot/dts/freescale/fsl-ls2080a.dtsi
+++ b/arch/arm64/boot/dts/freescale/fsl-ls2080a.dtsi
@@ -51,7 +51,7 @@
 	#size-cells = <2>;
 
 	cpus {
-		#address-cells = <2>;
+		#address-cells = <1>;
 		#size-cells = <0>;
 
 		/*
@@ -65,57 +65,81 @@
 		cpu@0 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a57";
-			reg = <0x0 0x0>;
+			reg = <0x0>;
 			clocks = <&clockgen 1 0>;
+			next-level-cache = <&cluster0_l2>;
 		};
 
 		cpu@1 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a57";
-			reg = <0x0 0x1>;
+			reg = <0x1>;
 			clocks = <&clockgen 1 0>;
+			next-level-cache = <&cluster0_l2>;
 		};
 
 		cpu@100 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a57";
-			reg = <0x0 0x100>;
+			reg = <0x100>;
 			clocks = <&clockgen 1 1>;
+			next-level-cache = <&cluster1_l2>;
 		};
 
 		cpu@101 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a57";
-			reg = <0x0 0x101>;
+			reg = <0x101>;
 			clocks = <&clockgen 1 1>;
+			next-level-cache = <&cluster1_l2>;
 		};
 
 		cpu@200 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a57";
-			reg = <0x0 0x200>;
+			reg = <0x200>;
 			clocks = <&clockgen 1 2>;
+			next-level-cache = <&cluster2_l2>;
 		};
 
 		cpu@201 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a57";
-			reg = <0x0 0x201>;
+			reg = <0x201>;
 			clocks = <&clockgen 1 2>;
+			next-level-cache = <&cluster2_l2>;
 		};
 
 		cpu@300 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a57";
-			reg = <0x0 0x300>;
+			reg = <0x300>;
 			clocks = <&clockgen 1 3>;
+			next-level-cache = <&cluster3_l2>;
 		};
 
 		cpu@301 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a57";
-			reg = <0x0 0x301>;
+			reg = <0x301>;
 			clocks = <&clockgen 1 3>;
+			next-level-cache = <&cluster3_l2>;
+		};
+
+		cluster0_l2: l2-cache0 {
+			compatible = "cache";
+		};
+
+		cluster1_l2: l2-cache1 {
+			compatible = "cache";
+		};
+
+		cluster2_l2: l2-cache2 {
+			compatible = "cache";
+		};
+
+		cluster3_l2: l2-cache3 {
+			compatible = "cache";
 		};
 	};
 
@@ -672,6 +696,7 @@
 			interrupts = <0 80 0x4>; /* Level high type */
 			dr_mode = "host";
 			snps,quirk-frame-length-adjustment = <0x20>;
+			snps,dis_rxdet_inp3_quirk;
 		};
 
 		usb1: usb3@3110000 {
@@ -681,6 +706,7 @@
 			interrupts = <0 81 0x4>; /* Level high type */
 			dr_mode = "host";
 			snps,quirk-frame-length-adjustment = <0x20>;
+			snps,dis_rxdet_inp3_quirk;
 		};
 
 		ccn@4000000 {