[PATCH 0/3] edac: xgene: Add missing SoC register bus error handling and fix CE promotes to UE

[PATCH 0/3] edac: xgene: Add missing SoC register bus error handling and fix CE promotes to UE

[Loc Ho]

Add missing register bus error handling for SoC and fix an checking
condition for CE error promotes to UE. DT binding and documentation are
also updated.

---
Loc Ho (3):
  Documentation: Update the APM X-Gene SoC EDAC DTS binding with fix to
    the RB
  edac: xgene: Add missing SoC register bus error handling and fix CE
    promotes to UE
  arm64: Update the APM X-Gene EDAC node to the RB register resource

 .../devicetree/bindings/edac/apm-xgene-edac.txt    |    7 ++
 arch/arm64/boot/dts/apm/apm-storm.dtsi             |    6 ++
 drivers/edac/xgene_edac.c                          |   62 +++++++++++++++++++-
 3 files changed, 74 insertions(+), 1 deletions(-)

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 1/3] Documentation: Update the APM X-Gene SoC EDAC DTS binding for missing RB

[Loc Ho]

This patch updates the APM X-Gene SoC EDAC documentation binding to
reference the missing register bus for the the SoC EDAC.

Signed-off-by: Loc Ho <lho-qTEPVZfXA3Y@public.gmane.org>
---
 .../devicetree/bindings/edac/apm-xgene-edac.txt    |    7 +++++++
 1 files changed, 7 insertions(+), 0 deletions(-)

diff --git a/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt b/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt
index 78e2a31..6e66f99 100644
--- a/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt
+++ b/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt
@@ -16,6 +16,7 @@ Required properties:
 - regmap-mcba		: Regmap of the MCB-A (memory bridge) resource.
 - regmap-mcbb		: Regmap of the MCB-B (memory bridge) resource.
 - regmap-efuse		: Regmap of the PMD efuse resource.
+- regmap-rb		: Regmap of the register bus resource.
 - reg			: First resource shall be the CPU bus (PCP) resource.
 - interrupts            : Interrupt-specifier for MCU, PMD, L3, or SoC error
 			  IRQ(s).
@@ -64,6 +65,11 @@ Example:
 		reg = <0x0 0x1054a000 0x0 0x20>;
 	};
 
+	rb: rb@7e000000 {
+		compatible = "apm,xgene-rb", "syscon";
+		reg = <0x0 0x7e000000 0x0 0x10>;
+	};
+
 	edac@78800000 {
 		compatible = "apm,xgene-edac";
 		#address-cells = <2>;
@@ -73,6 +79,7 @@ Example:
 		regmap-mcba = <&mcba>;
 		regmap-mcbb = <&mcbb>;
 		regmap-efuse = <&efuse>;
+		regmap-rb = <&rb>;
 		reg = <0x0 0x78800000 0x0 0x100>;
 		interrupts = <0x0 0x20 0x4>,
 			     <0x0 0x21 0x4>,
-- 
1.7.1

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 2/3] edac: xgene: Add missing SoC register bus error handling and fix CE promotes to UE

[Loc Ho]

Add missing register bus error handling for APM X-Gene EDAC SoC and 
fix an checking condition for CE error promotes to UE.

Signed-off-by: Loc Ho <lho-qTEPVZfXA3Y@public.gmane.org>
---
 drivers/edac/xgene_edac.c |   62 ++++++++++++++++++++++++++++++++++++++++++++-
 1 files changed, 61 insertions(+), 1 deletions(-)

diff --git a/drivers/edac/xgene_edac.c b/drivers/edac/xgene_edac.c
index 41f8764..417b749 100644
--- a/drivers/edac/xgene_edac.c
+++ b/drivers/edac/xgene_edac.c
@@ -61,6 +61,7 @@ struct xgene_edac {
 	struct regmap		*mcba_map;
 	struct regmap		*mcbb_map;
 	struct regmap		*efuse_map;
+	struct regmap		*rb_map;
 	void __iomem		*pcp_csr;
 	spinlock_t		lock;
 	struct dentry           *dfs;
@@ -1057,7 +1058,7 @@ static bool xgene_edac_l3_promote_to_uc_err(u32 l3cesr, u32 l3celr)
 		case 0x041:
 			return true;
 		}
-	} else if (L3C_ELR_ERRSYN(l3celr) == 9)
+	} else if (L3C_ELR_ERRWAY(l3celr) == 9)
 		return true;
 
 	return false;
@@ -1353,6 +1354,17 @@ static int xgene_edac_l3_remove(struct xgene_edac_dev_ctx *l3)
 #define GLBL_MDED_ERRH			0x0848
 #define GLBL_MDED_ERRHMASK		0x084c
 
+/* IO Bus Registers */
+#define RBCSR				0x0000
+#define STICKYERR_MASK			BIT(0)
+#define RBEIR				0x0008
+#define AGENT_OFFLINE_ERR_MASK		BIT(30)
+#define UNIMPL_RBPAGE_ERR_MASK		BIT(29)
+#define WORD_ALIGNED_ERR_MASK		BIT(28)
+#define PAGE_ACCESS_ERR_MASK		BIT(27)
+#define WRITE_ACCESS_MASK		BIT(26)
+#define RBERRADDR_RD(src)		((src) & 0x03FFFFFF)
+
 static const char * const soc_mem_err_v1[] = {
 	"10GbE0",
 	"10GbE1",
@@ -1470,6 +1482,46 @@ static void xgene_edac_rb_report(struct edac_device_ctl_info *edac_dev)
 	u32 err_addr_hi;
 	u32 reg;
 
+	/*
+	 * Check RB acess errors
+	 * 1. Out of range
+	 * 2. Un-implemented page
+	 * 3. Un-aligned access
+	 * 4. Offline slave IP
+	 */
+	if (regmap_read(ctx->edac->rb_map, RBCSR, &reg))
+		return;
+	if (reg & STICKYERR_MASK) {
+		bool write;
+		u32 address;
+
+		dev_err(edac_dev->dev, "IOB bus access error(s)\n");
+		if (regmap_read(ctx->edac->rb_map, RBEIR, &reg))
+			return;
+		write = reg & WRITE_ACCESS_MASK ? 1 : 0;
+		address = RBERRADDR_RD(reg);
+		if (reg & AGENT_OFFLINE_ERR_MASK)
+			dev_err(edac_dev->dev,
+				"IOB bus %s access to offline agent error\n",
+				write ? "write" : "read");
+		if (reg & UNIMPL_RBPAGE_ERR_MASK)
+			dev_err(edac_dev->dev,
+				"IOB bus %s access to unimplemented page error\n",
+				write ? "write" : "read");
+		if (reg & WORD_ALIGNED_ERR_MASK)
+			dev_err(edac_dev->dev,
+				"IOB bus %s word aligned access error\n",
+				write ? "write" : "read");
+		if (reg & PAGE_ACCESS_ERR_MASK)
+			dev_err(edac_dev->dev,
+				"IOB bus %s to page out of range access error\n",
+				write ? "write" : "read");
+		if (regmap_write(ctx->edac->rb_map, RBEIR, 0))
+			return;
+		if (regmap_write(ctx->edac->rb_map, RBCSR, 0))
+			return;
+	}
+
 	/* IOB Bridge agent transaction error interrupt */
 	reg = readl(ctx->dev_csr + IOBBATRANSERRINTSTS);
 	if (!reg)
@@ -1852,6 +1904,14 @@ static int xgene_edac_probe(struct platform_device *pdev)
 		goto out_err;
 	}
 
+	edac->rb_map = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
+						       "regmap-rb");
+	if (IS_ERR(edac->rb_map)) {
+		dev_err(edac->dev, "unable to get syscon regmap rb\n");
+		rc = PTR_ERR(edac->rb_map);
+		goto out_err;
+	}
+
 	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	edac->pcp_csr = devm_ioremap_resource(&pdev->dev, res);
 	if (IS_ERR(edac->pcp_csr)) {
-- 
1.7.1

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 3/3] arm64: Update the APM X-Gene EDAC node with the RB register resource

[Loc Ho]

Add the RB register resource as requires with the updated driver.

Signed-off-by: Loc Ho <lho-qTEPVZfXA3Y@public.gmane.org>
---
 arch/arm64/boot/dts/apm/apm-storm.dtsi |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/arch/arm64/boot/dts/apm/apm-storm.dtsi b/arch/arm64/boot/dts/apm/apm-storm.dtsi
index 6c5ed11..6665906 100644
--- a/arch/arm64/boot/dts/apm/apm-storm.dtsi
+++ b/arch/arm64/boot/dts/apm/apm-storm.dtsi
@@ -444,6 +444,11 @@
 			reg = <0x0 0x1054a000 0x0 0x20>;
 		};
 
+                rb: rb@7e000000 {
+                        compatible = "apm,xgene-rb", "syscon";
+                        reg = <0x0 0x7e000000 0x0 0x10>;
+                };
+
 		edac@78800000 {
 			compatible = "apm,xgene-edac";
 			#address-cells = <2>;
@@ -453,6 +458,7 @@
 			regmap-mcba = <&mcba>;
 			regmap-mcbb = <&mcbb>;
 			regmap-efuse = <&efuse>;
+			regmap-rb = <&rb>;
 			reg = <0x0 0x78800000 0x0 0x100>;
 			interrupts = <0x0 0x20 0x4>,
 				     <0x0 0x21 0x4>,
-- 
1.7.1

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 1/3] Documentation: Update the APM X-Gene SoC EDAC DTS binding for missing RB

[Arnd Bergmann]

On Tuesday 12 January 2016 16:04:58 Loc Ho wrote:
> --- a/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt
> +++ b/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt
> @@ -16,6 +16,7 @@ Required properties:
>  - regmap-mcba          : Regmap of the MCB-A (memory bridge) resource.
>  - regmap-mcbb          : Regmap of the MCB-B (memory bridge) resource.
>  - regmap-efuse         : Regmap of the PMD efuse resource.
> +- regmap-rb            : Regmap of the register bus resource.
>  - reg                  : First resource shall be the CPU bus (PCP) resource.
>  - interrupts            : Interrupt-specifier for MCU, PMD, L3, or SoC error
>                           IRQ(s).
> @@ -64,6 +65,11 @@ Example:
>                 reg = <0x0 0x1054a000 0x0 0x20>;
>         };
>  
> +       rb: rb@7e000000 {
> +               compatible = "apm,xgene-rb", "syscon";
> +               reg = <0x0 0x7e000000 0x0 0x10>;
> +       };
> +
>         edac@78800000 {
>                 compatible = "apm,xgene-edac";
> 

When you add a property, please list that as "optional", not "required",
to ensure that the old dtb files still conform to the binding.

In the driver, make sure it is not a fatal error when the property is
missing and the driver can keep working without it.

	Arnd
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 1/3] Documentation: Update the APM X-Gene SoC EDAC DTS binding for missing RB

[Loc Ho]

Hi Arnd,

>> --- a/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt
>> +++ b/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt
>> @@ -16,6 +16,7 @@ Required properties:
>>  - regmap-mcba          : Regmap of the MCB-A (memory bridge) resource.
>>  - regmap-mcbb          : Regmap of the MCB-B (memory bridge) resource.
>>  - regmap-efuse         : Regmap of the PMD efuse resource.
>> +- regmap-rb            : Regmap of the register bus resource.
>>  - reg                  : First resource shall be the CPU bus (PCP) resource.
>>  - interrupts            : Interrupt-specifier for MCU, PMD, L3, or SoC error
>>                           IRQ(s).
>> @@ -64,6 +65,11 @@ Example:
>>                 reg = <0x0 0x1054a000 0x0 0x20>;
>>         };
>>
>> +       rb: rb@7e000000 {
>> +               compatible = "apm,xgene-rb", "syscon";
>> +               reg = <0x0 0x7e000000 0x0 0x10>;
>> +       };
>> +
>>         edac@78800000 {
>>                 compatible = "apm,xgene-edac";
>>
>
> When you add a property, please list that as "optional", not "required",
> to ensure that the old dtb files still conform to the binding.
>
> In the driver, make sure it is not a fatal error when the property is
> missing and the driver can keep working without it.

>From my view, this is required. Otherwise, the error condition will
NOT be cleared correctly and will get constant interrupt. Though, for
compatibility, we can make it optional but system will be unusable
should such error occurred. Though, it would be rare or with bad FW. I
will fix in the next version by making it optional such as it will
still probe correctly with older DT binding.

-Loc
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html