[PATCH v4 0/4] edac: Add L3/SoC support to the APM X-Gene SoC EDAC driver

[PATCH v4 0/4] edac: Add L3/SoC support to the APM X-Gene SoC EDAC driver

[Loc Ho]

v4:
* Remove function for EDAC debugfs node and replace with export variable
* Switch the driver to use the exported edac_debugfs variable
* Revert the if statement logic for function as necessary
* Remove un-necessary code in debugfs creation functions

v3:
* Add an function to retrieve the EDAC debugfs node
* Move all debugfs node under EDAC code debugfs node
* Update L3 to check for v1 errata
* Rename error label for L3/SoC add routines
* Re-structure SoC EDAC functions for code readability
* Inline the function xgene_edac_soc_mem_data
* Remove un-necessary { }

v2:
* Update binding documentation accordingly
* Change all single bit defines to BIT(x)
* Add support for L3 version 1 and 2 HW's
* Change to use debug file system for error injection
* In L3/SoC instance add function, allocate EDAC context after all
  initalization successed
* Support raw or detail info for SoC EDAC error reporting

v1:
* Add L3/SoC support to the APM X-Gene SoC EDAC driver
---
Loc Ho (4):
  edac: Export edac_debugfs
  Documentation: Update the APM X-Gene SoC EDAC DTS binding for L3/SoC
    subnodes
  edac: Add L3/SoC EDAC support to the APM X-Gene SoC EDAC driver
  arm64: Add L3/SoC DT subnodes to the APM X-Gene SoC EDAC node

 .../devicetree/bindings/edac/apm-xgene-edac.txt    |   23 +
 arch/arm64/boot/dts/apm/apm-storm.dtsi             |   10 +
 drivers/edac/edac_core.h                           |    1 +
 drivers/edac/edac_mc_sysfs.c                       |    5 +-
 drivers/edac/xgene_edac.c                          | 1169 ++++++++++++++++----
 5 files changed, 1012 insertions(+), 196 deletions(-)

[PATCH v4 1/4] edac: Export edac_debugfs

[Loc Ho]

This patch exports and expose the edac_debugfs file node. This allows
EDAC driver to create debugfs node under the EDAC debugfs node.

Signed-off-by: Loc Ho <lho@apm.com>
---
 drivers/edac/edac_core.h     |    1 +
 drivers/edac/edac_mc_sysfs.c |    5 +++--
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/edac/edac_core.h b/drivers/edac/edac_core.h
index ad42587..a978a4f 100644
--- a/drivers/edac/edac_core.h
+++ b/drivers/edac/edac_core.h
@@ -511,5 +511,6 @@ extern void edac_pci_remove_sysfs(struct edac_pci_ctl_info *pci);
  * edac misc APIs
  */
 extern char *edac_op_state_to_string(int op_state);
+extern struct dentry *edac_debugfs;
 
 #endif				/* _EDAC_CORE_H_ */
diff --git a/drivers/edac/edac_mc_sysfs.c b/drivers/edac/edac_mc_sysfs.c
index 33df7d9..236e76c 100644
--- a/drivers/edac/edac_mc_sysfs.c
+++ b/drivers/edac/edac_mc_sysfs.c
@@ -896,9 +896,10 @@ static struct device_type mci_attr_type = {
 	.release	= mci_attr_release,
 };
 
-#ifdef CONFIG_EDAC_DEBUG
-static struct dentry *edac_debugfs;
+struct dentry *edac_debugfs;
+EXPORT_SYMBOL_GPL(edac_debugfs);
 
+#ifdef CONFIG_EDAC_DEBUG
 int __init edac_debugfs_init(void)
 {
 	edac_debugfs = debugfs_create_dir("edac", NULL);
-- 
1.7.1

[PATCH v4 2/4] Documentation: Update the APM X-Gene SoC EDAC DTS binding for L3/SoC subnodes

[Loc Ho]

This patch updates documentation for the APM X-Gene SoC EDAC DTS binding
for L3/SoC subnodes.

Signed-off-by: Loc Ho <lho@apm.com>
---
 .../devicetree/bindings/edac/apm-xgene-edac.txt    |   23 ++++++++++++++++++++
 1 files changed, 23 insertions(+), 0 deletions(-)

diff --git a/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt b/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt
index 78edb80..78e2a31 100644
--- a/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt
+++ b/Documentation/devicetree/bindings/edac/apm-xgene-edac.txt
@@ -5,6 +5,8 @@ The follow error types are supported:
 
   memory controller	- Memory controller
   PMD (L1/L2)		- Processor module unit (PMD) L1/L2 cache
+  L3			- L3 cache controller
+  SoC			- SoC IP's such as Ethernet, SATA, and etc
 
 The following section describes the EDAC DT node binding.
 
@@ -30,6 +32,17 @@ Required properties for PMD subnode:
 - reg			: First resource shall be the PMD resource.
 - pmd-controller	: Instance number of the PMD controller.
 
+Required properties for L3 subnode:
+- compatible		: Shall be "apm,xgene-edac-l3" or
+                          "apm,xgene-edac-l3-v2".
+- reg			: First resource shall be the L3 EDAC resource.
+
+Required properties for SoC subnode:
+- compatible		: Shall be "apm,xgene-edac-soc-v1" for revision 1 or
+                          "apm,xgene-edac-l3-soc" for general value reporting
+                          only.
+- reg			: First resource shall be the SoC EDAC resource.
+
 Example:
 	csw: csw at 7e200000 {
 		compatible = "apm,xgene-csw", "syscon";
@@ -76,4 +89,14 @@ Example:
 			reg = <0x0 0x7c000000 0x0 0x200000>;
 			pmd-controller = <0>;
 		};
+
+		edacl3 at 7e600000 {
+			compatible = "apm,xgene-edac-l3";
+			reg = <0x0 0x7e600000 0x0 0x1000>;
+		};
+
+		edacsoc at 7e930000 {
+			compatible = "apm,xgene-edac-soc-v1";
+			reg = <0x0 0x7e930000 0x0 0x1000>;
+		};
 	};
-- 
1.7.1

[PATCH v4 3/4] edac: Add L3/SoC EDAC support to the APM X-Gene SoC EDAC driver

[Loc Ho]

This patch adds EDAC support for the L3 and SoC components.

Signed-off-by: Loc Ho <lho@apm.com>
---
 drivers/edac/xgene_edac.c | 1169 +++++++++++++++++++++++++++++++++++++--------
 1 files changed, 975 insertions(+), 194 deletions(-)

diff --git a/drivers/edac/xgene_edac.c b/drivers/edac/xgene_edac.c
index ba06904..c7a5ea6 100644
--- a/drivers/edac/xgene_edac.c
+++ b/drivers/edac/xgene_edac.c
@@ -66,6 +66,8 @@ struct xgene_edac {
 
 	struct list_head	mcus;
 	struct list_head	pmds;
+	struct list_head	l3s;
+	struct list_head	socs;
 
 	struct mutex		mc_lock;
 	int			mc_active_mask;
@@ -536,140 +538,134 @@ static void xgene_edac_pmd_l1_check(struct edac_device_ctl_info *edac_dev,
 	pg_f = ctx->pmd_csr + cpu_idx * CPU_CSR_STRIDE + CPU_MEMERR_CPU_PAGE;
 
 	val = readl(pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET);
-	if (val) {
-		dev_err(edac_dev->dev,
-			"CPU%d L1 memory error ICF 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n",
-			ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
-			MEMERR_CPU_ICFESR_ERRWAY_RD(val),
-			MEMERR_CPU_ICFESR_ERRINDEX_RD(val),
-			MEMERR_CPU_ICFESR_ERRINFO_RD(val));
-		if (val & MEMERR_CPU_ICFESR_CERR_MASK)
-			dev_err(edac_dev->dev,
-				"One or more correctable error\n");
-		if (val & MEMERR_CPU_ICFESR_MULTCERR_MASK)
-			dev_err(edac_dev->dev, "Multiple correctable error\n");
-		switch (MEMERR_CPU_ICFESR_ERRTYPE_RD(val)) {
-		case 1:
-			dev_err(edac_dev->dev, "L1 TLB multiple hit\n");
-			break;
-		case 2:
-			dev_err(edac_dev->dev, "Way select multiple hit\n");
-			break;
-		case 3:
-			dev_err(edac_dev->dev, "Physical tag parity error\n");
-			break;
-		case 4:
-		case 5:
-			dev_err(edac_dev->dev, "L1 data parity error\n");
-			break;
-		case 6:
-			dev_err(edac_dev->dev, "L1 pre-decode parity error\n");
-			break;
-		}
+	if (!val)
+		goto chk_lsu;
+	dev_err(edac_dev->dev,
+		"CPU%d L1 memory error ICF 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n",
+		ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
+		MEMERR_CPU_ICFESR_ERRWAY_RD(val),
+		MEMERR_CPU_ICFESR_ERRINDEX_RD(val),
+		MEMERR_CPU_ICFESR_ERRINFO_RD(val));
+	if (val & MEMERR_CPU_ICFESR_CERR_MASK)
+		dev_err(edac_dev->dev, "One or more correctable error\n");
+	if (val & MEMERR_CPU_ICFESR_MULTCERR_MASK)
+		dev_err(edac_dev->dev, "Multiple correctable error\n");
+	switch (MEMERR_CPU_ICFESR_ERRTYPE_RD(val)) {
+	case 1:
+		dev_err(edac_dev->dev, "L1 TLB multiple hit\n");
+		break;
+	case 2:
+		dev_err(edac_dev->dev, "Way select multiple hit\n");
+		break;
+	case 3:
+		dev_err(edac_dev->dev, "Physical tag parity error\n");
+		break;
+	case 4:
+	case 5:
+		dev_err(edac_dev->dev, "L1 data parity error\n");
+		break;
+	case 6:
+		dev_err(edac_dev->dev, "L1 pre-decode parity error\n");
+		break;
+	}
 
-		/* Clear any HW errors */
-		writel(val, pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET);
+	/* Clear any HW errors */
+	writel(val, pg_f + MEMERR_CPU_ICFESR_PAGE_OFFSET);
 
-		if (val & (MEMERR_CPU_ICFESR_CERR_MASK |
-			   MEMERR_CPU_ICFESR_MULTCERR_MASK))
-			edac_device_handle_ce(edac_dev, 0, 0,
-					      edac_dev->ctl_name);
-	}
+	if (val & (MEMERR_CPU_ICFESR_CERR_MASK |
+		   MEMERR_CPU_ICFESR_MULTCERR_MASK))
+		edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
 
+chk_lsu:
 	val = readl(pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET);
-	if (val) {
+	if (!val)
+		goto chk_mmu;
+	dev_err(edac_dev->dev,
+		"CPU%d memory error LSU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n",
+		ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
+		MEMERR_CPU_LSUESR_ERRWAY_RD(val),
+		MEMERR_CPU_LSUESR_ERRINDEX_RD(val),
+		MEMERR_CPU_LSUESR_ERRINFO_RD(val));
+	if (val & MEMERR_CPU_LSUESR_CERR_MASK)
+		dev_err(edac_dev->dev, "One or more correctable error\n");
+	if (val & MEMERR_CPU_LSUESR_MULTCERR_MASK)
+		dev_err(edac_dev->dev, "Multiple correctable error\n");
+	switch (MEMERR_CPU_LSUESR_ERRTYPE_RD(val)) {
+	case 0:
+		dev_err(edac_dev->dev, "Load tag error\n");
+		break;
+	case 1:
+		dev_err(edac_dev->dev, "Load data error\n");
+		break;
+	case 2:
+		dev_err(edac_dev->dev, "WSL multihit error\n");
+		break;
+	case 3:
+		dev_err(edac_dev->dev, "Store tag error\n");
+		break;
+	case 4:
 		dev_err(edac_dev->dev,
-			"CPU%d memory error LSU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X\n",
-			ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
-			MEMERR_CPU_LSUESR_ERRWAY_RD(val),
-			MEMERR_CPU_LSUESR_ERRINDEX_RD(val),
-			MEMERR_CPU_LSUESR_ERRINFO_RD(val));
-		if (val & MEMERR_CPU_LSUESR_CERR_MASK)
-			dev_err(edac_dev->dev,
-				"One or more correctable error\n");
-		if (val & MEMERR_CPU_LSUESR_MULTCERR_MASK)
-			dev_err(edac_dev->dev, "Multiple correctable error\n");
-		switch (MEMERR_CPU_LSUESR_ERRTYPE_RD(val)) {
-		case 0:
-			dev_err(edac_dev->dev, "Load tag error\n");
-			break;
-		case 1:
-			dev_err(edac_dev->dev, "Load data error\n");
-			break;
-		case 2:
-			dev_err(edac_dev->dev, "WSL multihit error\n");
-			break;
-		case 3:
-			dev_err(edac_dev->dev, "Store tag error\n");
-			break;
-		case 4:
-			dev_err(edac_dev->dev,
-				"DTB multihit from load pipeline error\n");
-			break;
-		case 5:
-			dev_err(edac_dev->dev,
-				"DTB multihit from store pipeline error\n");
-			break;
-		}
+			"DTB multihit from load pipeline error\n");
+		break;
+	case 5:
+		dev_err(edac_dev->dev,
+			"DTB multihit from store pipeline error\n");
+		break;
+	}
 
-		/* Clear any HW errors */
-		writel(val, pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET);
+	/* Clear any HW errors */
+	writel(val, pg_f + MEMERR_CPU_LSUESR_PAGE_OFFSET);
 
-		if (val & (MEMERR_CPU_LSUESR_CERR_MASK |
-			   MEMERR_CPU_LSUESR_MULTCERR_MASK))
-			edac_device_handle_ce(edac_dev, 0, 0,
-					      edac_dev->ctl_name);
-	}
+	if (val & (MEMERR_CPU_LSUESR_CERR_MASK |
+		   MEMERR_CPU_LSUESR_MULTCERR_MASK))
+		edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
 
+chk_mmu:
 	val = readl(pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET);
-	if (val) {
-		dev_err(edac_dev->dev,
-			"CPU%d memory error MMU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X %s\n",
-			ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
-			MEMERR_CPU_MMUESR_ERRWAY_RD(val),
-			MEMERR_CPU_MMUESR_ERRINDEX_RD(val),
-			MEMERR_CPU_MMUESR_ERRINFO_RD(val),
-			val & MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK ? "LSU" :
-								     "ICF");
-		if (val & MEMERR_CPU_MMUESR_CERR_MASK)
-			dev_err(edac_dev->dev,
-				"One or more correctable error\n");
-		if (val & MEMERR_CPU_MMUESR_MULTCERR_MASK)
-			dev_err(edac_dev->dev, "Multiple correctable error\n");
-		switch (MEMERR_CPU_MMUESR_ERRTYPE_RD(val)) {
-		case 0:
-			dev_err(edac_dev->dev, "Stage 1 UTB hit error\n");
-			break;
-		case 1:
-			dev_err(edac_dev->dev, "Stage 1 UTB miss error\n");
-			break;
-		case 2:
-			dev_err(edac_dev->dev, "Stage 1 UTB allocate error\n");
-			break;
-		case 3:
-			dev_err(edac_dev->dev,
-				"TMO operation single bank error\n");
-			break;
-		case 4:
-			dev_err(edac_dev->dev, "Stage 2 UTB error\n");
-			break;
-		case 5:
-			dev_err(edac_dev->dev, "Stage 2 UTB miss error\n");
-			break;
-		case 6:
-			dev_err(edac_dev->dev, "Stage 2 UTB allocate error\n");
-			break;
-		case 7:
-			dev_err(edac_dev->dev,
-				"TMO operation multiple bank error\n");
-			break;
-		}
+	if (!val)
+		return;
+	dev_err(edac_dev->dev,
+		"CPU%d memory error MMU 0x%08X Way 0x%02X Index 0x%02X Info 0x%02X %s\n",
+		ctx->pmd * MAX_CPU_PER_PMD + cpu_idx, val,
+		MEMERR_CPU_MMUESR_ERRWAY_RD(val),
+		MEMERR_CPU_MMUESR_ERRINDEX_RD(val),
+		MEMERR_CPU_MMUESR_ERRINFO_RD(val),
+		val & MEMERR_CPU_MMUESR_ERRREQSTR_LSU_MASK ? "LSU" : "ICF");
+	if (val & MEMERR_CPU_MMUESR_CERR_MASK)
+		dev_err(edac_dev->dev, "One or more correctable error\n");
+	if (val & MEMERR_CPU_MMUESR_MULTCERR_MASK)
+		dev_err(edac_dev->dev, "Multiple correctable error\n");
+	switch (MEMERR_CPU_MMUESR_ERRTYPE_RD(val)) {
+	case 0:
+		dev_err(edac_dev->dev, "Stage 1 UTB hit error\n");
+		break;
+	case 1:
+		dev_err(edac_dev->dev, "Stage 1 UTB miss error\n");
+		break;
+	case 2:
+		dev_err(edac_dev->dev, "Stage 1 UTB allocate error\n");
+		break;
+	case 3:
+		dev_err(edac_dev->dev, "TMO operation single bank error\n");
+		break;
+	case 4:
+		dev_err(edac_dev->dev, "Stage 2 UTB error\n");
+		break;
+	case 5:
+		dev_err(edac_dev->dev, "Stage 2 UTB miss error\n");
+		break;
+	case 6:
+		dev_err(edac_dev->dev, "Stage 2 UTB allocate error\n");
+		break;
+	case 7:
+		dev_err(edac_dev->dev, "TMO operation multiple bank error\n");
+		break;
+	}
 
-		/* Clear any HW errors */
-		writel(val, pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET);
+	/* Clear any HW errors */
+	writel(val, pg_f + MEMERR_CPU_MMUESR_PAGE_OFFSET);
 
-		edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
-	}
+	edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
 }
 
 static void xgene_edac_pmd_l2_check(struct edac_device_ctl_info *edac_dev)
@@ -684,60 +680,56 @@ static void xgene_edac_pmd_l2_check(struct edac_device_ctl_info *edac_dev)
 	/* Check L2 */
 	pg_e = ctx->pmd_csr + CPU_MEMERR_L2C_PAGE;
 	val = readl(pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET);
-	if (val) {
-		val_lo = readl(pg_e + MEMERR_L2C_L2EALR_PAGE_OFFSET);
-		val_hi = readl(pg_e + MEMERR_L2C_L2EAHR_PAGE_OFFSET);
-		dev_err(edac_dev->dev,
-			"PMD%d memory error L2C L2ESR 0x%08X @ 0x%08X.%08X\n",
-			ctx->pmd, val, val_hi, val_lo);
-		dev_err(edac_dev->dev,
-			"ErrSyndrome 0x%02X ErrWay 0x%02X ErrCpu %d ErrGroup 0x%02X ErrAction 0x%02X\n",
-			MEMERR_L2C_L2ESR_ERRSYN_RD(val),
-			MEMERR_L2C_L2ESR_ERRWAY_RD(val),
-			MEMERR_L2C_L2ESR_ERRCPU_RD(val),
-			MEMERR_L2C_L2ESR_ERRGROUP_RD(val),
-			MEMERR_L2C_L2ESR_ERRACTION_RD(val));
-
-		if (val & MEMERR_L2C_L2ESR_ERR_MASK)
-			dev_err(edac_dev->dev,
-				"One or more correctable error\n");
-		if (val & MEMERR_L2C_L2ESR_MULTICERR_MASK)
-			dev_err(edac_dev->dev, "Multiple correctable error\n");
-		if (val & MEMERR_L2C_L2ESR_UCERR_MASK)
-			dev_err(edac_dev->dev,
-				"One or more uncorrectable error\n");
-		if (val & MEMERR_L2C_L2ESR_MULTUCERR_MASK)
-			dev_err(edac_dev->dev,
-				"Multiple uncorrectable error\n");
-
-		switch (MEMERR_L2C_L2ESR_ERRTYPE_RD(val)) {
-		case 0:
-			dev_err(edac_dev->dev, "Outbound SDB parity error\n");
-			break;
-		case 1:
-			dev_err(edac_dev->dev, "Inbound SDB parity error\n");
-			break;
-		case 2:
-			dev_err(edac_dev->dev, "Tag ECC error\n");
-			break;
-		case 3:
-			dev_err(edac_dev->dev, "Data ECC error\n");
-			break;
-		}
+	if (!val)
+		goto chk_l2c;
+	val_lo = readl(pg_e + MEMERR_L2C_L2EALR_PAGE_OFFSET);
+	val_hi = readl(pg_e + MEMERR_L2C_L2EAHR_PAGE_OFFSET);
+	dev_err(edac_dev->dev,
+		"PMD%d memory error L2C L2ESR 0x%08X @ 0x%08X.%08X\n",
+		ctx->pmd, val, val_hi, val_lo);
+	dev_err(edac_dev->dev,
+		"ErrSyndrome 0x%02X ErrWay 0x%02X ErrCpu %d ErrGroup 0x%02X ErrAction 0x%02X\n",
+		MEMERR_L2C_L2ESR_ERRSYN_RD(val),
+		MEMERR_L2C_L2ESR_ERRWAY_RD(val),
+		MEMERR_L2C_L2ESR_ERRCPU_RD(val),
+		MEMERR_L2C_L2ESR_ERRGROUP_RD(val),
+		MEMERR_L2C_L2ESR_ERRACTION_RD(val));
+
+	if (val & MEMERR_L2C_L2ESR_ERR_MASK)
+		dev_err(edac_dev->dev, "One or more correctable error\n");
+	if (val & MEMERR_L2C_L2ESR_MULTICERR_MASK)
+		dev_err(edac_dev->dev, "Multiple correctable error\n");
+	if (val & MEMERR_L2C_L2ESR_UCERR_MASK)
+		dev_err(edac_dev->dev, "One or more uncorrectable error\n");
+	if (val & MEMERR_L2C_L2ESR_MULTUCERR_MASK)
+		dev_err(edac_dev->dev, "Multiple uncorrectable error\n");
+
+	switch (MEMERR_L2C_L2ESR_ERRTYPE_RD(val)) {
+	case 0:
+		dev_err(edac_dev->dev, "Outbound SDB parity error\n");
+		break;
+	case 1:
+		dev_err(edac_dev->dev, "Inbound SDB parity error\n");
+		break;
+	case 2:
+		dev_err(edac_dev->dev, "Tag ECC error\n");
+		break;
+	case 3:
+		dev_err(edac_dev->dev, "Data ECC error\n");
+		break;
+	}
 
-		/* Clear any HW errors */
-		writel(val, pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET);
+	/* Clear any HW errors */
+	writel(val, pg_e + MEMERR_L2C_L2ESR_PAGE_OFFSET);
 
-		if (val & (MEMERR_L2C_L2ESR_ERR_MASK |
-			   MEMERR_L2C_L2ESR_MULTICERR_MASK))
-			edac_device_handle_ce(edac_dev, 0, 0,
-					      edac_dev->ctl_name);
-		if (val & (MEMERR_L2C_L2ESR_UCERR_MASK |
-			   MEMERR_L2C_L2ESR_MULTUCERR_MASK))
-			edac_device_handle_ue(edac_dev, 0, 0,
-					      edac_dev->ctl_name);
-	}
+	if (val & (MEMERR_L2C_L2ESR_ERR_MASK |
+		   MEMERR_L2C_L2ESR_MULTICERR_MASK))
+		edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+	if (val & (MEMERR_L2C_L2ESR_UCERR_MASK |
+		   MEMERR_L2C_L2ESR_MULTUCERR_MASK))
+		edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
 
+chk_l2c:
 	/* Check if any memory request timed out on L2 cache */
 	pg_d = ctx->pmd_csr + CPU_L2C_PAGE;
 	val = readl(pg_d + CPUX_L2C_L2RTOSR_PAGE_OFFSET);
@@ -877,26 +869,15 @@ static const struct file_operations xgene_edac_pmd_debug_inject_fops[] = {
 	{ }
 };
 
-static void xgene_edac_pmd_create_debugfs_nodes(
-	struct edac_device_ctl_info *edac_dev)
+static void
+xgene_edac_pmd_create_debugfs_nodes(struct edac_device_ctl_info *edac_dev)
 {
 	struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
 	struct dentry *edac_debugfs;
 	char name[30];
 
-	if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
+	if (!ctx->edac->dfs)
 		return;
-
-	/*
-	 * Todo: Switch to common EDAC debug file system for edac device
-	 *       when available.
-	 */
-	if (!ctx->edac->dfs) {
-		ctx->edac->dfs = debugfs_create_dir(edac_dev->dev->kobj.name,
-						    NULL);
-		if (!ctx->edac->dfs)
-			return;
-	}
 	sprintf(name, "PMD%d", ctx->pmd);
 	edac_debugfs = debugfs_create_dir(name, ctx->edac->dfs);
 	if (!edac_debugfs)
@@ -1016,10 +997,786 @@ static int xgene_edac_pmd_remove(struct xgene_edac_pmd_ctx *pmd)
 	return 0;
 }
 
+/* L3 Error device */
+#define L3C_ESR				(0x0A * 4)
+#define  L3C_ESR_DATATAG_MASK		BIT(9)
+#define  L3C_ESR_MULTIHIT_MASK		BIT(8)
+#define  L3C_ESR_UCEVICT_MASK		BIT(6)
+#define  L3C_ESR_MULTIUCERR_MASK	BIT(5)
+#define  L3C_ESR_MULTICERR_MASK		BIT(4)
+#define  L3C_ESR_UCERR_MASK		BIT(3)
+#define  L3C_ESR_CERR_MASK		BIT(2)
+#define  L3C_ESR_UCERRINTR_MASK		BIT(1)
+#define  L3C_ESR_CERRINTR_MASK		BIT(0)
+#define L3C_ECR				(0x0B * 4)
+#define  L3C_ECR_UCINTREN		BIT(3)
+#define  L3C_ECR_CINTREN		BIT(2)
+#define  L3C_UCERREN			BIT(1)
+#define  L3C_CERREN			BIT(0)
+#define L3C_ELR				(0x0C * 4)
+#define  L3C_ELR_ERRSYN(src)		((src & 0xFF800000) >> 23)
+#define  L3C_ELR_ERRWAY(src)		((src & 0x007E0000) >> 17)
+#define  L3C_ELR_AGENTID(src)		((src & 0x0001E000) >> 13)
+#define  L3C_ELR_ERRGRP(src)		((src & 0x00000F00) >> 8)
+#define  L3C_ELR_OPTYPE(src)		((src & 0x000000F0) >> 4)
+#define  L3C_ELR_PADDRHIGH(src)		(src & 0x0000000F)
+#define L3C_AELR			(0x0D * 4)
+#define L3C_BELR			(0x0E * 4)
+#define  L3C_BELR_BANK(src)		(src & 0x0000000F)
+
+struct xgene_edac_dev_ctx {
+	struct list_head	next;
+	struct device		ddev;
+	char			*name;
+	struct xgene_edac	*edac;
+	struct edac_device_ctl_info *edac_dev;
+	int			edac_idx;
+	void __iomem		*dev_csr;
+	int			version;
+};
+
+static bool xgene_edac_l3_v1_errata_chk(u32 l3cesr, u32 l3celr)
+{
+	/*
+	 * L3 version 1 has certain conditions in which correctable error
+	 * needs to be flagged as un-correctable error. This function
+	 * check for such conditions.
+	 */
+	if (l3cesr & L3C_ESR_DATATAG_MASK) {
+		switch (L3C_ELR_ERRSYN(l3celr)) {
+		case 0x13C:
+		case 0x0B4:
+		case 0x007:
+		case 0x00D:
+		case 0x00E:
+		case 0x019:
+		case 0x01A:
+		case 0x01C:
+		case 0x04E:
+		case 0x041:
+			return true;
+		}
+	} else if (L3C_ELR_ERRSYN(l3celr) == 9) {
+		return true;
+	}
+
+	return false;
+}
+
+static void xgene_edac_l3_check(struct edac_device_ctl_info *edac_dev)
+{
+	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+	u32 l3cesr;
+	u32 l3celr;
+	u32 l3caelr;
+	u32 l3cbelr;
+
+	l3cesr = readl(ctx->dev_csr + L3C_ESR);
+	if (!(l3cesr & (L3C_ESR_UCERR_MASK | L3C_ESR_CERR_MASK)))
+		return;
+
+	if (l3cesr & L3C_ESR_UCERR_MASK)
+		dev_err(edac_dev->dev, "L3C uncorrectable error\n");
+	if (l3cesr & L3C_ESR_CERR_MASK)
+		dev_warn(edac_dev->dev, "L3C correctable error\n");
+
+	l3celr = readl(ctx->dev_csr + L3C_ELR);
+	l3caelr = readl(ctx->dev_csr + L3C_AELR);
+	l3cbelr = readl(ctx->dev_csr + L3C_BELR);
+	if (l3cesr & L3C_ESR_MULTIHIT_MASK)
+		dev_err(edac_dev->dev, "L3C multiple hit error\n");
+	if (l3cesr & L3C_ESR_UCEVICT_MASK)
+		dev_err(edac_dev->dev,
+			"L3C dropped eviction of line with error\n");
+	if (l3cesr & L3C_ESR_MULTIUCERR_MASK)
+		dev_err(edac_dev->dev, "L3C multiple uncorrectable error\n");
+	if (l3cesr & L3C_ESR_DATATAG_MASK)
+		dev_err(edac_dev->dev,
+			"L3C data error syndrome 0x%X group 0x%X\n",
+			L3C_ELR_ERRSYN(l3celr), L3C_ELR_ERRGRP(l3celr));
+	else
+		dev_err(edac_dev->dev,
+			"L3C tag error syndrome 0x%X Way of Tag 0x%X Agent ID 0x%X Operation type 0x%X\n",
+			L3C_ELR_ERRSYN(l3celr), L3C_ELR_ERRWAY(l3celr),
+			L3C_ELR_AGENTID(l3celr), L3C_ELR_OPTYPE(l3celr));
+	/*
+	 * NOTE: Address [41:38] in L3C_ELR_PADDRHIGH(l3celr).
+	 *       Address [37:6] in l3caelr. Lower 6 bits are zero.
+	 */
+	dev_err(edac_dev->dev, "L3C error address 0x%08X.%08X bank %d\n",
+		L3C_ELR_PADDRHIGH(l3celr) << 6 | (l3caelr >> 26),
+		(l3caelr & 0x3FFFFFFF) << 6, L3C_BELR_BANK(l3cbelr));
+	dev_err(edac_dev->dev,
+		"L3C error status register value 0x%X\n", l3cesr);
+
+	/* Clear L3C error interrupt */
+	writel(0, ctx->dev_csr + L3C_ESR);
+
+	if (ctx->version <= 1 &&
+	    xgene_edac_l3_v1_errata_chk(l3cesr, l3celr)) {
+		/*
+		 * Version 1 of the L3C has broken single bit correctable
+		 * logic for certain condition. When such condition is
+		 * detected, map them as uncorrectable.
+		 */
+		edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+		return;
+	}
+	if (l3cesr & L3C_ESR_CERR_MASK)
+		edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+	if (l3cesr & L3C_ESR_UCERR_MASK)
+		edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+}
+
+static void xgene_edac_l3_hw_init(struct edac_device_ctl_info *edac_dev,
+				  bool enable)
+{
+	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+	u32 val;
+
+	val = readl(ctx->dev_csr + L3C_ECR);
+	val |= L3C_UCERREN | L3C_CERREN;
+	/* On disable, we just disable interrupt but keep error enabled */
+	if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
+		if (enable)
+			val |= L3C_ECR_UCINTREN | L3C_ECR_CINTREN;
+		else
+			val &= ~(L3C_ECR_UCINTREN | L3C_ECR_CINTREN);
+	}
+	writel(val, ctx->dev_csr + L3C_ECR);
+
+	if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
+		/* Enable/disable L3 error top level interrupt */
+		if (enable) {
+			xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
+					       L3C_UNCORR_ERR_MASK);
+			xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK,
+					       L3C_CORR_ERR_MASK);
+		} else {
+			xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
+					       L3C_UNCORR_ERR_MASK);
+			xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK,
+					       L3C_CORR_ERR_MASK);
+		}
+	}
+}
+
+static ssize_t xgene_edac_l3_inject_ctrl_write(struct file *file,
+					       const char __user *data,
+					       size_t count, loff_t *ppos)
+{
+	struct edac_device_ctl_info *edac_dev = file->private_data;
+	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+
+	/* Generate all errors */
+	writel(0xFFFFFFFF, ctx->dev_csr + L3C_ESR);
+	return count;
+}
+
+static const struct file_operations xgene_edac_l3_debug_inject_fops = {
+	.open = simple_open,
+	.write = xgene_edac_l3_inject_ctrl_write,
+	.llseek = generic_file_llseek
+};
+
+static void xgene_edac_l3_create_debugfs_nodes(
+	struct edac_device_ctl_info *edac_dev)
+{
+	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+	struct dentry *edac_debugfs;
+	char name[30];
+
+	if (!ctx->edac->dfs)
+		return;
+	sprintf(name, "l3c%d", ctx->edac_idx);
+	edac_debugfs = debugfs_create_dir(name, ctx->edac->dfs);
+	if (!edac_debugfs)
+		return;
+
+	debugfs_create_file("l3_inject_ctrl", S_IWUSR, edac_debugfs, edac_dev,
+			    &xgene_edac_l3_debug_inject_fops);
+}
+
+static int xgene_edac_l3_add(struct xgene_edac *edac, struct device_node *np,
+			     int version)
+{
+	struct edac_device_ctl_info *edac_dev;
+	struct xgene_edac_dev_ctx *ctx;
+	struct resource res;
+	void __iomem *dev_csr;
+	int edac_idx;
+	int rc = 0;
+
+	if (!devres_open_group(edac->dev, xgene_edac_l3_add, GFP_KERNEL))
+		return -ENOMEM;
+
+	rc = of_address_to_resource(np, 0, &res);
+	if (rc < 0) {
+		dev_err(edac->dev, "no L3 resource address\n");
+		goto err_release_group;
+	}
+	dev_csr = devm_ioremap_resource(edac->dev, &res);
+	if (IS_ERR(dev_csr)) {
+		dev_err(edac->dev,
+			"devm_ioremap_resource failed for L3 resource address\n");
+		rc = PTR_ERR(dev_csr);
+		goto err_release_group;
+	}
+
+	edac_idx = edac_device_alloc_index();
+	edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx),
+					      "l3c", 1, "l3c", 1, 0, NULL, 0,
+					      edac_idx);
+	if (!edac_dev) {
+		rc = -ENOMEM;
+		goto err_release_group;
+	}
+
+	ctx = edac_dev->pvt_info;
+	ctx->dev_csr = dev_csr;
+	ctx->name = "xgene_l3_err";
+	ctx->edac_idx = edac_idx;
+	ctx->edac = edac;
+	ctx->edac_dev = edac_dev;
+	ctx->ddev = *edac->dev;
+	ctx->version = version;
+	edac_dev->dev = &ctx->ddev;
+	edac_dev->ctl_name = ctx->name;
+	edac_dev->dev_name = ctx->name;
+	edac_dev->mod_name = EDAC_MOD_STR;
+
+	if (edac_op_state == EDAC_OPSTATE_POLL)
+		edac_dev->edac_check = xgene_edac_l3_check;
+
+	xgene_edac_l3_create_debugfs_nodes(edac_dev);
+
+	rc = edac_device_add_device(edac_dev);
+	if (rc > 0) {
+		dev_err(edac->dev, "failed edac_device_add_device()\n");
+		rc = -ENOMEM;
+		goto err_ctl_free;
+	}
+
+	if (edac_op_state == EDAC_OPSTATE_INT)
+		edac_dev->op_state = OP_RUNNING_INTERRUPT;
+
+	list_add(&ctx->next, &edac->l3s);
+
+	xgene_edac_l3_hw_init(edac_dev, 1);
+
+	devres_remove_group(edac->dev, xgene_edac_l3_add);
+
+	dev_info(edac->dev, "X-Gene EDAC L3 registered\n");
+	return 0;
+
+err_ctl_free:
+	edac_device_free_ctl_info(edac_dev);
+err_release_group:
+	devres_release_group(edac->dev, xgene_edac_l3_add);
+	return rc;
+}
+
+static int xgene_edac_l3_remove(struct xgene_edac_dev_ctx *l3)
+{
+	struct edac_device_ctl_info *edac_dev = l3->edac_dev;
+
+	xgene_edac_l3_hw_init(edac_dev, 0);
+	edac_device_del_device(l3->edac->dev);
+	edac_device_free_ctl_info(edac_dev);
+	return 0;
+}
+
+/* SoC Error device */
+#define IOBAXIS0TRANSERRINTSTS		0x0000
+#define  IOBAXIS0_M_ILLEGAL_ACCESS_MASK	BIT(1)
+#define  IOBAXIS0_ILLEGAL_ACCESS_MASK	BIT(0)
+#define IOBAXIS0TRANSERRINTMSK		0x0004
+#define IOBAXIS0TRANSERRREQINFOL	0x0008
+#define IOBAXIS0TRANSERRREQINFOH	0x000c
+#define  REQTYPE_RD(src)		(((src) & BIT(0)))
+#define  ERRADDRH_RD(src)		(((src) & 0xffc00000) >> 22)
+#define IOBAXIS1TRANSERRINTSTS		0x0010
+#define IOBAXIS1TRANSERRINTMSK		0x0014
+#define IOBAXIS1TRANSERRREQINFOL	0x0018
+#define IOBAXIS1TRANSERRREQINFOH	0x001c
+#define IOBPATRANSERRINTSTS		0x0020
+#define  IOBPA_M_REQIDRAM_CORRUPT_MASK	BIT(7)
+#define  IOBPA_REQIDRAM_CORRUPT_MASK	BIT(6)
+#define  IOBPA_M_TRANS_CORRUPT_MASK	BIT(5)
+#define  IOBPA_TRANS_CORRUPT_MASK	BIT(4)
+#define  IOBPA_M_WDATA_CORRUPT_MASK	BIT(3)
+#define  IOBPA_WDATA_CORRUPT_MASK	BIT(2)
+#define  IOBPA_M_RDATA_CORRUPT_MASK	BIT(1)
+#define  IOBPA_RDATA_CORRUPT_MASK	BIT(0)
+#define IOBBATRANSERRINTSTS		0x0030
+#define  M_ILLEGAL_ACCESS_MASK		BIT(15)
+#define  ILLEGAL_ACCESS_MASK		BIT(14)
+#define  M_WIDRAM_CORRUPT_MASK		BIT(13)
+#define  WIDRAM_CORRUPT_MASK		BIT(12)
+#define  M_RIDRAM_CORRUPT_MASK		BIT(11)
+#define  RIDRAM_CORRUPT_MASK		BIT(10)
+#define  M_TRANS_CORRUPT_MASK		BIT(9)
+#define  TRANS_CORRUPT_MASK		BIT(8)
+#define  M_WDATA_CORRUPT_MASK		BIT(7)
+#define  WDATA_CORRUPT_MASK		BIT(6)
+#define  M_RBM_POISONED_REQ_MASK	BIT(5)
+#define  RBM_POISONED_REQ_MASK		BIT(4)
+#define  M_XGIC_POISONED_REQ_MASK	BIT(3)
+#define  XGIC_POISONED_REQ_MASK		BIT(2)
+#define  M_WRERR_RESP_MASK		BIT(1)
+#define  WRERR_RESP_MASK		BIT(0)
+#define IOBBATRANSERRREQINFOL		0x0038
+#define IOBBATRANSERRREQINFOH		0x003c
+#define  REQTYPE_F2_RD(src)		((src) & BIT(0))
+#define  ERRADDRH_F2_RD(src)		(((src) & 0xffc00000) >> 22)
+#define IOBBATRANSERRCSWREQID		0x0040
+#define XGICTRANSERRINTSTS		0x0050
+#define  M_WR_ACCESS_ERR_MASK		BIT(3)
+#define  WR_ACCESS_ERR_MASK		BIT(2)
+#define  M_RD_ACCESS_ERR_MASK		BIT(1)
+#define  RD_ACCESS_ERR_MASK		BIT(0)
+#define XGICTRANSERRINTMSK		0x0054
+#define XGICTRANSERRREQINFO		0x0058
+#define  REQTYPE_MASK			BIT(26)
+#define  ERRADDR_RD(src)		((src) & 0x03ffffff)
+#define GLBL_ERR_STS			0x0800
+#define  MDED_ERR_MASK			BIT(3)
+#define  DED_ERR_MASK			BIT(2)
+#define  MSEC_ERR_MASK			BIT(1)
+#define  SEC_ERR_MASK			BIT(0)
+#define GLBL_SEC_ERRL			0x0810
+#define GLBL_SEC_ERRH			0x0818
+#define GLBL_MSEC_ERRL			0x0820
+#define GLBL_MSEC_ERRH			0x0828
+#define GLBL_DED_ERRL			0x0830
+#define GLBL_DED_ERRLMASK		0x0834
+#define GLBL_DED_ERRH			0x0838
+#define GLBL_DED_ERRHMASK		0x083c
+#define GLBL_MDED_ERRL			0x0840
+#define GLBL_MDED_ERRLMASK		0x0844
+#define GLBL_MDED_ERRH			0x0848
+#define GLBL_MDED_ERRHMASK		0x084c
+
+static const char * const soc_mem_err_v1[] = {
+	"10GbE0",
+	"10GbE1",
+	"Security",
+	"SATA45",
+	"SATA23/ETH23",
+	"SATA01/ETH01",
+	"USB1",
+	"USB0",
+	"QML",
+	"QM0",
+	"QM1 (XGbE01)",
+	"PCIE4",
+	"PCIE3",
+	"PCIE2",
+	"PCIE1",
+	"PCIE0",
+	"CTX Manager",
+	"OCM",
+	"1GbE",
+	"CLE",
+	"AHBC",
+	"PktDMA",
+	"GFC",
+	"MSLIM",
+	"10GbE2",
+	"10GbE3",
+	"QM2 (XGbE23)",
+	"IOB",
+	"unknown",
+	"unknown",
+	"unknown",
+	"unknown",
+};
+
+static void xgene_edac_iob_gic_report(struct edac_device_ctl_info *edac_dev)
+{
+	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+	u32 err_addr_lo;
+	u32 err_addr_hi;
+	u32 reg;
+	u32 info;
+
+	/* GIC transaction error interrupt */
+	reg = readl(ctx->dev_csr + XGICTRANSERRINTSTS);
+	if (!reg)
+		goto chk_iob_err;
+	dev_err(edac_dev->dev, "XGIC transaction error\n");
+	if (reg & RD_ACCESS_ERR_MASK)
+		dev_err(edac_dev->dev, "XGIC read size error\n");
+	if (reg & M_RD_ACCESS_ERR_MASK)
+		dev_err(edac_dev->dev, "Multiple XGIC read size error\n");
+	if (reg & WR_ACCESS_ERR_MASK)
+		dev_err(edac_dev->dev, "XGIC write size error\n");
+	if (reg & M_WR_ACCESS_ERR_MASK)
+		dev_err(edac_dev->dev, "Multiple XGIC write size error\n");
+	info = readl(ctx->dev_csr + XGICTRANSERRREQINFO);
+	dev_err(edac_dev->dev, "XGIC %s access @ 0x%08X (0x%08X)\n",
+		info & REQTYPE_MASK ? "read" : "write", ERRADDR_RD(info),
+		info);
+	writel(reg, ctx->dev_csr + XGICTRANSERRINTSTS);
+
+chk_iob_err:
+	/* IOB memory error */
+	reg = readl(ctx->dev_csr + GLBL_ERR_STS);
+	if (!reg)
+		return;
+	if (reg & SEC_ERR_MASK) {
+		err_addr_lo = readl(ctx->dev_csr + GLBL_SEC_ERRL);
+		err_addr_hi = readl(ctx->dev_csr + GLBL_SEC_ERRH);
+		dev_err(edac_dev->dev,
+			"IOB single-bit correctable memory at 0x%08X.%08X error\n",
+			err_addr_lo, err_addr_hi);
+		writel(err_addr_lo, ctx->dev_csr + GLBL_SEC_ERRL);
+		writel(err_addr_hi, ctx->dev_csr + GLBL_SEC_ERRH);
+	}
+	if (reg & MSEC_ERR_MASK) {
+		err_addr_lo = readl(ctx->dev_csr + GLBL_MSEC_ERRL);
+		err_addr_hi = readl(ctx->dev_csr + GLBL_MSEC_ERRH);
+		dev_err(edac_dev->dev,
+			"IOB multiple single-bit correctable memory at 0x%08X.%08X error\n",
+			err_addr_lo, err_addr_hi);
+		writel(err_addr_lo, ctx->dev_csr + GLBL_MSEC_ERRL);
+		writel(err_addr_hi, ctx->dev_csr + GLBL_MSEC_ERRH);
+	}
+	if (reg & (SEC_ERR_MASK | MSEC_ERR_MASK))
+		edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+
+	if (reg & DED_ERR_MASK) {
+		err_addr_lo = readl(ctx->dev_csr + GLBL_DED_ERRL);
+		err_addr_hi = readl(ctx->dev_csr + GLBL_DED_ERRH);
+		dev_err(edac_dev->dev,
+			"IOB double-bit uncorrectable memory at 0x%08X.%08X error\n",
+			err_addr_lo, err_addr_hi);
+		writel(err_addr_lo, ctx->dev_csr + GLBL_DED_ERRL);
+		writel(err_addr_hi, ctx->dev_csr + GLBL_DED_ERRH);
+	}
+	if (reg & MDED_ERR_MASK) {
+		err_addr_lo = readl(ctx->dev_csr + GLBL_MDED_ERRL);
+		err_addr_hi = readl(ctx->dev_csr + GLBL_MDED_ERRH);
+		dev_err(edac_dev->dev,
+			"Multiple IOB double-bit uncorrectable memory at 0x%08X.%08X error\n",
+			err_addr_lo, err_addr_hi);
+		writel(err_addr_lo, ctx->dev_csr + GLBL_MDED_ERRL);
+		writel(err_addr_hi, ctx->dev_csr + GLBL_MDED_ERRH);
+	}
+	if (reg & (DED_ERR_MASK | MDED_ERR_MASK))
+		edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+}
+
+static void xgene_edac_rb_report(struct edac_device_ctl_info *edac_dev)
+{
+	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+	u32 err_addr_lo;
+	u32 err_addr_hi;
+	u32 reg;
+
+	/* IOB Bridge agent transaction error interrupt */
+	reg = readl(ctx->dev_csr + IOBBATRANSERRINTSTS);
+	if (!reg)
+		return;
+
+	dev_err(edac_dev->dev, "IOB bridge agent (BA) transaction error\n");
+	if (reg & WRERR_RESP_MASK)
+		dev_err(edac_dev->dev, "IOB BA write response error\n");
+	if (reg & M_WRERR_RESP_MASK)
+		dev_err(edac_dev->dev,
+			"Multiple IOB BA write response error\n");
+	if (reg & XGIC_POISONED_REQ_MASK)
+		dev_err(edac_dev->dev, "IOB BA XGIC poisoned write error\n");
+	if (reg & M_XGIC_POISONED_REQ_MASK)
+		dev_err(edac_dev->dev,
+			"Multiple IOB BA XGIC poisoned write error\n");
+	if (reg & RBM_POISONED_REQ_MASK)
+		dev_err(edac_dev->dev, "IOB BA RBM poisoned write error\n");
+	if (reg & M_RBM_POISONED_REQ_MASK)
+		dev_err(edac_dev->dev,
+			"Multiple IOB BA RBM poisoned write error\n");
+	if (reg & WDATA_CORRUPT_MASK)
+		dev_err(edac_dev->dev, "IOB BA write error\n");
+	if (reg & M_WDATA_CORRUPT_MASK)
+		dev_err(edac_dev->dev, "Multiple IOB BA write error\n");
+	if (reg & TRANS_CORRUPT_MASK)
+		dev_err(edac_dev->dev, "IOB BA transaction error\n");
+	if (reg & M_TRANS_CORRUPT_MASK)
+		dev_err(edac_dev->dev, "Multiple IOB BA transaction error\n");
+	if (reg & RIDRAM_CORRUPT_MASK)
+		dev_err(edac_dev->dev,
+			"IOB BA RDIDRAM read transaction ID error\n");
+	if (reg & M_RIDRAM_CORRUPT_MASK)
+		dev_err(edac_dev->dev,
+			"Multiple IOB BA RDIDRAM read transaction ID error\n");
+	if (reg & WIDRAM_CORRUPT_MASK)
+		dev_err(edac_dev->dev,
+			"IOB BA RDIDRAM write transaction ID error\n");
+	if (reg & M_WIDRAM_CORRUPT_MASK)
+		dev_err(edac_dev->dev,
+			"Multiple IOB BA RDIDRAM write transaction ID error\n");
+	if (reg & ILLEGAL_ACCESS_MASK)
+		dev_err(edac_dev->dev,
+			"IOB BA XGIC/RB illegal access error\n");
+	if (reg & M_ILLEGAL_ACCESS_MASK)
+		dev_err(edac_dev->dev,
+			"Multiple IOB BA XGIC/RB illegal access error\n");
+
+	err_addr_lo = readl(ctx->dev_csr + IOBBATRANSERRREQINFOL);
+	err_addr_hi = readl(ctx->dev_csr + IOBBATRANSERRREQINFOH);
+	dev_err(edac_dev->dev, "IOB BA %s access at 0x%02X.%08X (0x%08X)\n",
+		REQTYPE_F2_RD(err_addr_hi) ? "read" : "write",
+		ERRADDRH_F2_RD(err_addr_hi), err_addr_lo, err_addr_hi);
+	if (reg & WRERR_RESP_MASK)
+		dev_err(edac_dev->dev, "IOB BA requestor ID 0x%08X\n",
+			readl(ctx->dev_csr + IOBBATRANSERRCSWREQID));
+	writel(reg, ctx->dev_csr + IOBBATRANSERRINTSTS);
+}
+
+static void xgene_edac_pa_report(struct edac_device_ctl_info *edac_dev)
+{
+	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+	u32 err_addr_lo;
+	u32 err_addr_hi;
+	u32 reg;
+
+	/* IOB Processing agent transaction error interrupt */
+	reg = readl(ctx->dev_csr + IOBPATRANSERRINTSTS);
+	if (!reg)
+		goto chk_iob_axi0;
+	dev_err(edac_dev->dev, "IOB procesing agent (PA) transaction error\n");
+	if (reg & IOBPA_RDATA_CORRUPT_MASK)
+		dev_err(edac_dev->dev, "IOB PA read data RAM error\n");
+	if (reg & IOBPA_M_RDATA_CORRUPT_MASK)
+		dev_err(edac_dev->dev,
+			"Mutilple IOB PA read data RAM error\n");
+	if (reg & IOBPA_WDATA_CORRUPT_MASK)
+		dev_err(edac_dev->dev, "IOB PA write data RAM error\n");
+	if (reg & IOBPA_M_WDATA_CORRUPT_MASK)
+		dev_err(edac_dev->dev,
+			"Mutilple IOB PA write data RAM error\n");
+	if (reg & IOBPA_TRANS_CORRUPT_MASK)
+		dev_err(edac_dev->dev, "IOB PA transaction error\n");
+	if (reg & IOBPA_M_TRANS_CORRUPT_MASK)
+		dev_err(edac_dev->dev, "Mutilple IOB PA transaction error\n");
+	if (reg & IOBPA_REQIDRAM_CORRUPT_MASK)
+		dev_err(edac_dev->dev, "IOB PA transaction ID RAM error\n");
+	if (reg & IOBPA_M_REQIDRAM_CORRUPT_MASK)
+		dev_err(edac_dev->dev,
+			"Multiple IOB PA transaction ID RAM error\n");
+	writel(reg, ctx->dev_csr + IOBPATRANSERRINTSTS);
+
+chk_iob_axi0:
+	/* IOB AXI0 Error */
+	reg = readl(ctx->dev_csr + IOBAXIS0TRANSERRINTSTS);
+	if (!reg)
+		goto chk_iob_axi1;
+	err_addr_lo = readl(ctx->dev_csr + IOBAXIS0TRANSERRREQINFOL);
+	err_addr_hi = readl(ctx->dev_csr + IOBAXIS0TRANSERRREQINFOH);
+	dev_err(edac_dev->dev,
+		"%sAXI slave 0 illegal %s access @ 0x%02X.%08X (0x%08X)\n",
+		reg & IOBAXIS0_M_ILLEGAL_ACCESS_MASK ? "Multiple " : "",
+		REQTYPE_RD(err_addr_hi) ? "read" : "write",
+		ERRADDRH_RD(err_addr_hi), err_addr_lo, err_addr_hi);
+	writel(reg, ctx->dev_csr + IOBAXIS0TRANSERRINTSTS);
+
+chk_iob_axi1:
+	/* IOB AXI1 Error */
+	reg = readl(ctx->dev_csr + IOBAXIS1TRANSERRINTSTS);
+	if (!reg)
+		return;
+	err_addr_lo = readl(ctx->dev_csr + IOBAXIS1TRANSERRREQINFOL);
+	err_addr_hi = readl(ctx->dev_csr + IOBAXIS1TRANSERRREQINFOH);
+	dev_err(edac_dev->dev,
+		"%sAXI slave 1 illegal %s access @ 0x%02X.%08X (0x%08X)\n",
+		reg & IOBAXIS0_M_ILLEGAL_ACCESS_MASK ? "Multiple " : "",
+		REQTYPE_RD(err_addr_hi) ? "read" : "write",
+		ERRADDRH_RD(err_addr_hi), err_addr_lo, err_addr_hi);
+	writel(reg, ctx->dev_csr + IOBAXIS1TRANSERRINTSTS);
+}
+
+static void xgene_edac_soc_check(struct edac_device_ctl_info *edac_dev)
+{
+	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+	const char * const *soc_mem_err = NULL;
+	u32 pcp_hp_stat;
+	u32 pcp_lp_stat;
+	u32 reg;
+	int i;
+
+	xgene_edac_pcp_rd(ctx->edac, PCPHPERRINTSTS, &pcp_hp_stat);
+	xgene_edac_pcp_rd(ctx->edac, PCPLPERRINTSTS, &pcp_lp_stat);
+	xgene_edac_pcp_rd(ctx->edac, MEMERRINTSTS, &reg);
+	if (!((pcp_hp_stat & (IOB_PA_ERR_MASK | IOB_BA_ERR_MASK |
+			     IOB_XGIC_ERR_MASK | IOB_RB_ERR_MASK)) ||
+	      (pcp_lp_stat & CSW_SWITCH_TRACE_ERR_MASK) || reg))
+		return;
+
+	if (pcp_hp_stat & IOB_XGIC_ERR_MASK)
+		xgene_edac_iob_gic_report(edac_dev);
+
+	if (pcp_hp_stat & (IOB_RB_ERR_MASK | IOB_BA_ERR_MASK))
+		xgene_edac_rb_report(edac_dev);
+
+	if (pcp_hp_stat & IOB_PA_ERR_MASK)
+		xgene_edac_pa_report(edac_dev);
+
+	if (pcp_lp_stat & CSW_SWITCH_TRACE_ERR_MASK) {
+		dev_info(edac_dev->dev,
+			 "CSW switch trace correctable memory parity error\n");
+		edac_device_handle_ce(edac_dev, 0, 0, edac_dev->ctl_name);
+	}
+
+	if (!reg)
+		return;
+	if (ctx->version == 1)
+		soc_mem_err = soc_mem_err_v1;
+	if (!soc_mem_err) {
+		dev_err(edac_dev->dev, "SoC memory parity error 0x%08X\n",
+			reg);
+		edac_device_handle_ue(edac_dev, 0, 0, edac_dev->ctl_name);
+		return;
+	}
+	for (i = 0; i < 31; i++) {
+		if (reg & (1 << i)) {
+			dev_err(edac_dev->dev, "%s memory parity error\n",
+				soc_mem_err[i]);
+			edac_device_handle_ue(edac_dev, 0, 0,
+					      edac_dev->ctl_name);
+		}
+	}
+}
+
+static void xgene_edac_soc_hw_init(struct edac_device_ctl_info *edac_dev,
+				   bool enable)
+{
+	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
+
+	/* Enable SoC IP error interrupt */
+	if (edac_dev->op_state == OP_RUNNING_INTERRUPT) {
+		if (enable) {
+			xgene_edac_pcp_clrbits(ctx->edac, PCPHPERRINTMSK,
+					       IOB_PA_ERR_MASK |
+					       IOB_BA_ERR_MASK |
+					       IOB_XGIC_ERR_MASK |
+					       IOB_RB_ERR_MASK);
+			xgene_edac_pcp_clrbits(ctx->edac, PCPLPERRINTMSK,
+					       CSW_SWITCH_TRACE_ERR_MASK);
+		} else {
+			xgene_edac_pcp_setbits(ctx->edac, PCPHPERRINTMSK,
+					       IOB_PA_ERR_MASK |
+					       IOB_BA_ERR_MASK |
+					       IOB_XGIC_ERR_MASK |
+					       IOB_RB_ERR_MASK);
+			xgene_edac_pcp_setbits(ctx->edac, PCPLPERRINTMSK,
+					       CSW_SWITCH_TRACE_ERR_MASK);
+		}
+
+		writel(enable ? 0x0 : 0xFFFFFFFF,
+		       ctx->dev_csr + IOBAXIS0TRANSERRINTMSK);
+		writel(enable ? 0x0 : 0xFFFFFFFF,
+		       ctx->dev_csr + IOBAXIS1TRANSERRINTMSK);
+		writel(enable ? 0x0 : 0xFFFFFFFF,
+		       ctx->dev_csr + XGICTRANSERRINTMSK);
+
+		xgene_edac_pcp_setbits(ctx->edac, MEMERRINTMSK,
+				       enable ? 0x0 : 0xFFFFFFFF);
+	}
+}
+
+static int xgene_edac_soc_add(struct xgene_edac *edac, struct device_node *np,
+			      int version)
+{
+	struct edac_device_ctl_info *edac_dev;
+	struct xgene_edac_dev_ctx *ctx;
+	void __iomem *dev_csr;
+	struct resource res;
+	int edac_idx;
+	int rc = 0;
+
+	if (!devres_open_group(edac->dev, xgene_edac_soc_add, GFP_KERNEL))
+		return -ENOMEM;
+
+	rc = of_address_to_resource(np, 0, &res);
+	if (rc < 0) {
+		dev_err(edac->dev, "no SoC resource address\n");
+		goto err_release_group;
+	}
+	dev_csr = devm_ioremap_resource(edac->dev, &res);
+	if (IS_ERR(dev_csr)) {
+		dev_err(edac->dev,
+			"devm_ioremap_resource failed for soc resource address\n");
+		rc = PTR_ERR(dev_csr);
+		goto err_release_group;
+	}
+
+	edac_idx = edac_device_alloc_index();
+	edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx),
+					      "SOC", 1, "SOC", 1, 2, NULL, 0,
+					      edac_idx);
+	if (!edac_dev) {
+		rc = -ENOMEM;
+		goto err_release_group;
+	}
+
+	ctx = edac_dev->pvt_info;
+	ctx->dev_csr = dev_csr;
+	ctx->name = "xgene_soc_err";
+	ctx->edac_idx = edac_idx;
+	ctx->edac = edac;
+	ctx->edac_dev = edac_dev;
+	ctx->ddev = *edac->dev;
+	ctx->version = version;
+	edac_dev->dev = &ctx->ddev;
+	edac_dev->ctl_name = ctx->name;
+	edac_dev->dev_name = ctx->name;
+	edac_dev->mod_name = EDAC_MOD_STR;
+
+	if (edac_op_state == EDAC_OPSTATE_POLL)
+		edac_dev->edac_check = xgene_edac_soc_check;
+
+	rc = edac_device_add_device(edac_dev);
+	if (rc > 0) {
+		dev_err(edac->dev, "failed edac_device_add_device()\n");
+		rc = -ENOMEM;
+		goto err_ctl_free;
+	}
+
+	if (edac_op_state == EDAC_OPSTATE_INT)
+		edac_dev->op_state = OP_RUNNING_INTERRUPT;
+
+	list_add(&ctx->next, &edac->socs);
+
+	xgene_edac_soc_hw_init(edac_dev, 1);
+
+	devres_remove_group(edac->dev, xgene_edac_soc_add);
+
+	dev_info(edac->dev, "X-Gene EDAC SoC registered\n");
+
+	return 0;
+
+err_ctl_free:
+	edac_device_free_ctl_info(edac_dev);
+err_release_group:
+	devres_release_group(edac->dev, xgene_edac_soc_add);
+	return rc;
+}
+
+static int xgene_edac_soc_remove(struct xgene_edac_dev_ctx *soc)
+{
+	struct edac_device_ctl_info *edac_dev = soc->edac_dev;
+
+	xgene_edac_soc_hw_init(edac_dev, 0);
+	edac_device_del_device(soc->edac->dev);
+	edac_device_free_ctl_info(edac_dev);
+	return 0;
+}
+
 static irqreturn_t xgene_edac_isr(int irq, void *dev_id)
 {
 	struct xgene_edac *ctx = dev_id;
 	struct xgene_edac_pmd_ctx *pmd;
+	struct xgene_edac_dev_ctx *node;
 	unsigned int pcp_hp_stat;
 	unsigned int pcp_lp_stat;
 
@@ -1030,9 +1787,8 @@ static irqreturn_t xgene_edac_isr(int irq, void *dev_id)
 	    (MCU_CORR_ERR_MASK & pcp_lp_stat)) {
 		struct xgene_edac_mc_ctx *mcu;
 
-		list_for_each_entry(mcu, &ctx->mcus, next) {
+		list_for_each_entry(mcu, &ctx->mcus, next)
 			xgene_edac_mc_check(mcu->mci);
-		}
 	}
 
 	list_for_each_entry(pmd, &ctx->pmds, next) {
@@ -1040,6 +1796,12 @@ static irqreturn_t xgene_edac_isr(int irq, void *dev_id)
 			xgene_edac_pmd_check(pmd->edac_dev);
 	}
 
+	list_for_each_entry(node, &ctx->l3s, next)
+		xgene_edac_l3_check(node->edac_dev);
+
+	list_for_each_entry(node, &ctx->socs, next)
+		xgene_edac_soc_check(node->edac_dev);
+
 	return IRQ_HANDLED;
 }
 
@@ -1058,6 +1820,8 @@ static int xgene_edac_probe(struct platform_device *pdev)
 	platform_set_drvdata(pdev, edac);
 	INIT_LIST_HEAD(&edac->mcus);
 	INIT_LIST_HEAD(&edac->pmds);
+	INIT_LIST_HEAD(&edac->l3s);
+	INIT_LIST_HEAD(&edac->socs);
 	spin_lock_init(&edac->lock);
 	mutex_init(&edac->mc_lock);
 
@@ -1122,6 +1886,8 @@ static int xgene_edac_probe(struct platform_device *pdev)
 		}
 	}
 
+	edac->dfs = edac_debugfs;
+
 	for_each_child_of_node(pdev->dev.of_node, child) {
 		if (!of_device_is_available(child))
 			continue;
@@ -1131,6 +1897,14 @@ static int xgene_edac_probe(struct platform_device *pdev)
 			xgene_edac_pmd_add(edac, child, 1);
 		if (of_device_is_compatible(child, "apm,xgene-edac-pmd-v2"))
 			xgene_edac_pmd_add(edac, child, 2);
+		if (of_device_is_compatible(child, "apm,xgene-edac-l3"))
+			xgene_edac_l3_add(edac, child, 1);
+		if (of_device_is_compatible(child, "apm,xgene-edac-l3-v2"))
+			xgene_edac_l3_add(edac, child, 2);
+		if (of_device_is_compatible(child, "apm,xgene-edac-soc"))
+			xgene_edac_soc_add(edac, child, 0);
+		if (of_device_is_compatible(child, "apm,xgene-edac-soc-v1"))
+			xgene_edac_soc_add(edac, child, 1);
 	}
 
 	return 0;
@@ -1146,14 +1920,21 @@ static int xgene_edac_remove(struct platform_device *pdev)
 	struct xgene_edac_mc_ctx *temp_mcu;
 	struct xgene_edac_pmd_ctx *pmd;
 	struct xgene_edac_pmd_ctx *temp_pmd;
+	struct xgene_edac_dev_ctx *node;
+	struct xgene_edac_dev_ctx *temp_node;
 
-	list_for_each_entry_safe(mcu, temp_mcu, &edac->mcus, next) {
+	list_for_each_entry_safe(mcu, temp_mcu, &edac->mcus, next)
 		xgene_edac_mc_remove(mcu);
-	}
 
-	list_for_each_entry_safe(pmd, temp_pmd, &edac->pmds, next) {
+	list_for_each_entry_safe(pmd, temp_pmd, &edac->pmds, next)
 		xgene_edac_pmd_remove(pmd);
-	}
+
+	list_for_each_entry_safe(node, temp_node, &edac->l3s, next)
+		xgene_edac_l3_remove(node);
+
+	list_for_each_entry_safe(node, temp_node, &edac->socs, next)
+		xgene_edac_soc_remove(node);
+
 	return 0;
 }
 
-- 
1.7.1

[PATCH v4 4/4] arm64: Add L3/SoC DT subnodes to the APM X-Gene SoC EDAC node

[Loc Ho]

This patch adds L3/SoC DT subnodes to the APM X-Gene SoC EDAC node.

Signed-off-by: Loc Ho <lho@apm.com>
---
 arch/arm64/boot/dts/apm/apm-storm.dtsi |   10 ++++++++++
 1 files changed, 10 insertions(+), 0 deletions(-)

diff --git a/arch/arm64/boot/dts/apm/apm-storm.dtsi b/arch/arm64/boot/dts/apm/apm-storm.dtsi
index 58093ed..8ab6b0e 100644
--- a/arch/arm64/boot/dts/apm/apm-storm.dtsi
+++ b/arch/arm64/boot/dts/apm/apm-storm.dtsi
@@ -477,6 +477,16 @@
 				reg = <0x0 0x7c600000 0x0 0x200000>;
 				pmd-controller = <3>;
 			};
+
+			edacl3 at 7e600000 {
+				compatible = "apm,xgene-edac-l3";
+				reg = <0x0 0x7e600000 0x0 0x1000>;
+			};
+
+			edacsoc at 7e930000 {
+				compatible = "apm,xgene-edac-soc-v1";
+				reg = <0x0 0x7e930000 0x0 0x1000>;
+			};
 		};
 
 		pcie0: pcie at 1f2b0000 {
-- 
1.7.1

[PATCH v4 0/4] edac: Add L3/SoC support to the APM X-Gene SoC EDAC driver

[Loc Ho]

Hi Borislav,

Are you okay with this version to pull in for 4.3 release?

-Loc

On Thu, Aug 13, 2015 at 11:46 PM, Loc Ho <lho@apm.com> wrote:
> v4:
> * Remove function for EDAC debugfs node and replace with export variable
> * Switch the driver to use the exported edac_debugfs variable
> * Revert the if statement logic for function as necessary
> * Remove un-necessary code in debugfs creation functions
>
> v3:
> * Add an function to retrieve the EDAC debugfs node
> * Move all debugfs node under EDAC code debugfs node
> * Update L3 to check for v1 errata
> * Rename error label for L3/SoC add routines
> * Re-structure SoC EDAC functions for code readability
> * Inline the function xgene_edac_soc_mem_data
> * Remove un-necessary { }
>
> v2:
> * Update binding documentation accordingly
> * Change all single bit defines to BIT(x)
> * Add support for L3 version 1 and 2 HW's
> * Change to use debug file system for error injection
> * In L3/SoC instance add function, allocate EDAC context after all
>   initalization successed
> * Support raw or detail info for SoC EDAC error reporting
>
> v1:
> * Add L3/SoC support to the APM X-Gene SoC EDAC driver
> ---
> Loc Ho (4):
>   edac: Export edac_debugfs
>   Documentation: Update the APM X-Gene SoC EDAC DTS binding for L3/SoC
>     subnodes
>   edac: Add L3/SoC EDAC support to the APM X-Gene SoC EDAC driver
>   arm64: Add L3/SoC DT subnodes to the APM X-Gene SoC EDAC node
>
>  .../devicetree/bindings/edac/apm-xgene-edac.txt    |   23 +
>  arch/arm64/boot/dts/apm/apm-storm.dtsi             |   10 +
>  drivers/edac/edac_core.h                           |    1 +
>  drivers/edac/edac_mc_sysfs.c                       |    5 +-
>  drivers/edac/xgene_edac.c                          | 1169 ++++++++++++++++----
>  5 files changed, 1012 insertions(+), 196 deletions(-)
>

[PATCH v4 0/4] edac: Add L3/SoC support to the APM X-Gene SoC EDAC driver

[Borislav Petkov]

On Mon, Aug 31, 2015 at 02:26:11PM -0700, Loc Ho wrote:
> Are you okay with this version to pull in for 4.3 release?

It was already too late for 4.3 when you sent them. I'll take a look at
them after the merge window is over and also someone (Arnd?) needs to
check the DT changes.

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--

[PATCH v4 1/4] edac: Export edac_debugfs

[Borislav Petkov]

On Fri, Aug 14, 2015 at 12:46:06AM -0600, Loc Ho wrote:
> This patch exports and expose the edac_debugfs file node. This allows
> EDAC driver to create debugfs node under the EDAC debugfs node.

...

> -#ifdef CONFIG_EDAC_DEBUG
> -static struct dentry *edac_debugfs;
> +struct dentry *edac_debugfs;
> +EXPORT_SYMBOL_GPL(edac_debugfs);

Hmm, so Arnd was right about this being the wrong direction we're
chasing. We should rather be exporting a bunch of EDAC debugfs wrappers
which EDAC drivers can call instead of exposing edac_debugfs itself.

And this is much better and the right(tm) approach IMO. So I went
and hacked it in, it builds on those !x86 targets (as far as my
cross-compilers are concerned, build succeeds :-)).

So here it is as a follow-up to this message.

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.

[PATCH 1/5] EDAC: Carve out debugfs functionality

[Borislav Petkov]

From: Borislav Petkov <bp@suse.de>

... into a separate compilation unit and drop a couple of
CONFIG_EDAC_DEBUG ifdefferies. Rename edac_create_debug_nodes() to
edac_create_debugfs_nodes(), while at it.

No functionality change.

Cc: <linux-edac@vger.kernel.org>
Signed-off-by: Borislav Petkov <bp@suse.de>
---
 drivers/edac/Makefile        |   2 +
 drivers/edac/debugfs.c       | 103 ++++++++++++++++++++++++++++++++++++++++
 drivers/edac/edac_core.h     |   2 +
 drivers/edac/edac_mc_sysfs.c | 110 +------------------------------------------
 drivers/edac/edac_module.h   |   2 +
 include/linux/edac.h         |   2 -
 6 files changed, 110 insertions(+), 111 deletions(-)
 create mode 100644 drivers/edac/debugfs.c

diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
index ae3c5f3ce405..dbf53e08bdd1 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -12,6 +12,8 @@ obj-$(CONFIG_EDAC_MM_EDAC)		+= edac_core.o
 edac_core-y	:= edac_mc.o edac_device.o edac_mc_sysfs.o
 edac_core-y	+= edac_module.o edac_device_sysfs.o
 
+edac_core-$(CONFIG_EDAC_DEBUG)		+= debugfs.o
+
 ifdef CONFIG_PCI
 edac_core-y	+= edac_pci.o edac_pci_sysfs.o
 endif
diff --git a/drivers/edac/debugfs.c b/drivers/edac/debugfs.c
new file mode 100644
index 000000000000..bcd558d5cb48
--- /dev/null
+++ b/drivers/edac/debugfs.c
@@ -0,0 +1,103 @@
+#include "edac_module.h"
+
+static struct dentry *edac_debugfs;
+
+static ssize_t edac_fake_inject_write(struct file *file,
+				      const char __user *data,
+				      size_t count, loff_t *ppos)
+{
+	struct device *dev = file->private_data;
+	struct mem_ctl_info *mci = to_mci(dev);
+	static enum hw_event_mc_err_type type;
+	u16 errcount = mci->fake_inject_count;
+
+	if (!errcount)
+		errcount = 1;
+
+	type = mci->fake_inject_ue ? HW_EVENT_ERR_UNCORRECTED
+				   : HW_EVENT_ERR_CORRECTED;
+
+	printk(KERN_DEBUG
+	       "Generating %d %s fake error%s to %d.%d.%d to test core handling. NOTE: this won't test the driver-specific decoding logic.\n",
+		errcount,
+		(type == HW_EVENT_ERR_UNCORRECTED) ? "UE" : "CE",
+		errcount > 1 ? "s" : "",
+		mci->fake_inject_layer[0],
+		mci->fake_inject_layer[1],
+		mci->fake_inject_layer[2]
+	       );
+	edac_mc_handle_error(type, mci, errcount, 0, 0, 0,
+			     mci->fake_inject_layer[0],
+			     mci->fake_inject_layer[1],
+			     mci->fake_inject_layer[2],
+			     "FAKE ERROR", "for EDAC testing only");
+
+	return count;
+}
+
+static const struct file_operations debug_fake_inject_fops = {
+	.open = simple_open,
+	.write = edac_fake_inject_write,
+	.llseek = generic_file_llseek,
+};
+
+int __init edac_debugfs_init(void)
+{
+	edac_debugfs = debugfs_create_dir("edac", NULL);
+	if (IS_ERR(edac_debugfs)) {
+		edac_debugfs = NULL;
+		return -ENOMEM;
+	}
+	return 0;
+}
+
+void edac_debugfs_exit(void)
+{
+	debugfs_remove(edac_debugfs);
+}
+
+int edac_create_debugfs_nodes(struct mem_ctl_info *mci)
+{
+	struct dentry *d, *parent;
+	char name[80];
+	int i;
+
+	if (!edac_debugfs)
+		return -ENODEV;
+
+	d = debugfs_create_dir(mci->dev.kobj.name, edac_debugfs);
+	if (!d)
+		return -ENOMEM;
+	parent = d;
+
+	for (i = 0; i < mci->n_layers; i++) {
+		sprintf(name, "fake_inject_%s",
+			     edac_layer_name[mci->layers[i].type]);
+		d = debugfs_create_u8(name, S_IRUGO | S_IWUSR, parent,
+				      &mci->fake_inject_layer[i]);
+		if (!d)
+			goto nomem;
+	}
+
+	d = debugfs_create_bool("fake_inject_ue", S_IRUGO | S_IWUSR, parent,
+				&mci->fake_inject_ue);
+	if (!d)
+		goto nomem;
+
+	d = debugfs_create_u16("fake_inject_count", S_IRUGO | S_IWUSR, parent,
+				&mci->fake_inject_count);
+	if (!d)
+		goto nomem;
+
+	d = debugfs_create_file("fake_inject", S_IWUSR, parent,
+				&mci->dev,
+				&debug_fake_inject_fops);
+	if (!d)
+		goto nomem;
+
+	mci->debugfs = parent;
+	return 0;
+nomem:
+	debugfs_remove(mci->debugfs);
+	return -ENOMEM;
+}
diff --git a/drivers/edac/edac_core.h b/drivers/edac/edac_core.h
index ad42587c3f4d..4861542163d7 100644
--- a/drivers/edac/edac_core.h
+++ b/drivers/edac/edac_core.h
@@ -94,6 +94,8 @@ do {									\
 
 #define edac_dev_name(dev) (dev)->dev_name
 
+#define to_mci(k) container_of(k, struct mem_ctl_info, dev)
+
 /*
  * The following are the structures to provide for a generic
  * or abstract 'edac_device'. This set of structures and the
diff --git a/drivers/edac/edac_mc_sysfs.c b/drivers/edac/edac_mc_sysfs.c
index 33df7d93c857..071f7fca5ff5 100644
--- a/drivers/edac/edac_mc_sysfs.c
+++ b/drivers/edac/edac_mc_sysfs.c
@@ -785,47 +785,6 @@ static ssize_t mci_max_location_show(struct device *dev,
 	return p - data;
 }
 
-#ifdef CONFIG_EDAC_DEBUG
-static ssize_t edac_fake_inject_write(struct file *file,
-				      const char __user *data,
-				      size_t count, loff_t *ppos)
-{
-	struct device *dev = file->private_data;
-	struct mem_ctl_info *mci = to_mci(dev);
-	static enum hw_event_mc_err_type type;
-	u16 errcount = mci->fake_inject_count;
-
-	if (!errcount)
-		errcount = 1;
-
-	type = mci->fake_inject_ue ? HW_EVENT_ERR_UNCORRECTED
-				   : HW_EVENT_ERR_CORRECTED;
-
-	printk(KERN_DEBUG
-	       "Generating %d %s fake error%s to %d.%d.%d to test core handling. NOTE: this won't test the driver-specific decoding logic.\n",
-		errcount,
-		(type == HW_EVENT_ERR_UNCORRECTED) ? "UE" : "CE",
-		errcount > 1 ? "s" : "",
-		mci->fake_inject_layer[0],
-		mci->fake_inject_layer[1],
-		mci->fake_inject_layer[2]
-	       );
-	edac_mc_handle_error(type, mci, errcount, 0, 0, 0,
-			     mci->fake_inject_layer[0],
-			     mci->fake_inject_layer[1],
-			     mci->fake_inject_layer[2],
-			     "FAKE ERROR", "for EDAC testing only");
-
-	return count;
-}
-
-static const struct file_operations debug_fake_inject_fops = {
-	.open = simple_open,
-	.write = edac_fake_inject_write,
-	.llseek = generic_file_llseek,
-};
-#endif
-
 /* default Control file */
 static DEVICE_ATTR(reset_counters, S_IWUSR, NULL, mci_reset_counters_store);
 
@@ -896,71 +855,6 @@ static struct device_type mci_attr_type = {
 	.release	= mci_attr_release,
 };
 
-#ifdef CONFIG_EDAC_DEBUG
-static struct dentry *edac_debugfs;
-
-int __init edac_debugfs_init(void)
-{
-	edac_debugfs = debugfs_create_dir("edac", NULL);
-	if (IS_ERR(edac_debugfs)) {
-		edac_debugfs = NULL;
-		return -ENOMEM;
-	}
-	return 0;
-}
-
-void edac_debugfs_exit(void)
-{
-	debugfs_remove(edac_debugfs);
-}
-
-static int edac_create_debug_nodes(struct mem_ctl_info *mci)
-{
-	struct dentry *d, *parent;
-	char name[80];
-	int i;
-
-	if (!edac_debugfs)
-		return -ENODEV;
-
-	d = debugfs_create_dir(mci->dev.kobj.name, edac_debugfs);
-	if (!d)
-		return -ENOMEM;
-	parent = d;
-
-	for (i = 0; i < mci->n_layers; i++) {
-		sprintf(name, "fake_inject_%s",
-			     edac_layer_name[mci->layers[i].type]);
-		d = debugfs_create_u8(name, S_IRUGO | S_IWUSR, parent,
-				      &mci->fake_inject_layer[i]);
-		if (!d)
-			goto nomem;
-	}
-
-	d = debugfs_create_bool("fake_inject_ue", S_IRUGO | S_IWUSR, parent,
-				&mci->fake_inject_ue);
-	if (!d)
-		goto nomem;
-
-	d = debugfs_create_u16("fake_inject_count", S_IRUGO | S_IWUSR, parent,
-				&mci->fake_inject_count);
-	if (!d)
-		goto nomem;
-
-	d = debugfs_create_file("fake_inject", S_IWUSR, parent,
-				&mci->dev,
-				&debug_fake_inject_fops);
-	if (!d)
-		goto nomem;
-
-	mci->debugfs = parent;
-	return 0;
-nomem:
-	debugfs_remove(mci->debugfs);
-	return -ENOMEM;
-}
-#endif
-
 /*
  * Create a new Memory Controller kobject instance,
  *	mc<id> under the 'mc' directory
@@ -1039,9 +933,7 @@ int edac_create_sysfs_mci_device(struct mem_ctl_info *mci,
 		goto fail_unregister_dimm;
 #endif
 
-#ifdef CONFIG_EDAC_DEBUG
-	edac_create_debug_nodes(mci);
-#endif
+	edac_create_debugfs_nodes(mci);
 	return 0;
 
 fail_unregister_dimm:
diff --git a/drivers/edac/edac_module.h b/drivers/edac/edac_module.h
index 26ecc52e073d..79a6c6e20819 100644
--- a/drivers/edac/edac_module.h
+++ b/drivers/edac/edac_module.h
@@ -63,12 +63,14 @@ extern void *edac_align_ptr(void **p, unsigned size, int n_elems);
 #ifdef CONFIG_EDAC_DEBUG
 int edac_debugfs_init(void);
 void edac_debugfs_exit(void);
+int edac_create_debugfs_nodes(struct mem_ctl_info *mci);
 #else
 static inline int edac_debugfs_init(void)
 {
 	return -ENODEV;
 }
 static inline void edac_debugfs_exit(void) {}
+static inline int edac_create_debugfs_nodes(struct mem_ctl_info *mci) { return 0; }
 #endif
 
 /*
diff --git a/include/linux/edac.h b/include/linux/edac.h
index da3b72e95db3..b3d87e5822f8 100644
--- a/include/linux/edac.h
+++ b/include/linux/edac.h
@@ -769,12 +769,10 @@ struct mem_ctl_info {
 	/* the internal state of this controller instance */
 	int op_state;
 
-#ifdef CONFIG_EDAC_DEBUG
 	struct dentry *debugfs;
 	u8 fake_inject_layer[EDAC_MAX_LAYERS];
 	u32 fake_inject_ue;
 	u16 fake_inject_count;
-#endif
 };
 
 /*
-- 
2.3.5

[PATCH 2/5] EDAC: Add debugfs wrappers

[Borislav Petkov]

From: Borislav Petkov <bp@suse.de>

Later patches will convert EDAC users to those.

Signed-off-by: Borislav Petkov <bp@suse.de>
---
 drivers/edac/debugfs.c     | 60 ++++++++++++++++++++++++++++++++++++++++++++++
 drivers/edac/edac_module.h | 34 +++++++++++++++++++++-----
 2 files changed, 88 insertions(+), 6 deletions(-)

diff --git a/drivers/edac/debugfs.c b/drivers/edac/debugfs.c
index bcd558d5cb48..4864703115cc 100644
--- a/drivers/edac/debugfs.c
+++ b/drivers/edac/debugfs.c
@@ -101,3 +101,63 @@ nomem:
 	debugfs_remove(mci->debugfs);
 	return -ENOMEM;
 }
+
+/* Create a toplevel dir under EDAC's debugfs hierarchy */
+struct dentry *edac_debugfs_create_dir(const char *dirname)
+{
+	if (!edac_debugfs)
+		return NULL;
+
+	return debugfs_create_dir(dirname, edac_debugfs);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_dir);
+
+/* Create a toplevel dir under EDAC's debugfs hierarchy with parent @parent */
+struct dentry *
+edac_debugfs_create_dir_at(const char *dirname, struct dentry *parent)
+{
+	return debugfs_create_dir(dirname, parent);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_dir_at);
+
+/*
+ * Create a file under EDAC's hierarchy or a sub-hierarchy:
+ *
+ * @name: file name
+ * @mode: file permissions
+ * @parent: parent dentry. If NULL, it becomes the toplevel EDAC dir
+ * @data: private data of caller
+ * @fops: file operations of this file
+ */
+struct dentry *
+edac_debugfs_create_file(const char *name, umode_t mode, struct dentry *parent,
+			 void *data, const struct file_operations *fops)
+{
+	if (!parent)
+		parent = edac_debugfs;
+
+	return debugfs_create_file(name, mode, parent, data, fops);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_file);
+
+/* Wrapper for debugfs_create_x8() */
+struct dentry *edac_debugfs_create_x8(const char *name, umode_t mode,
+				       struct dentry *parent, u8 *value)
+{
+	if (!parent)
+		parent = edac_debugfs;
+
+	return debugfs_create_x8(name, mode, parent, value);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_x8);
+
+/* Wrapper for debugfs_create_x16() */
+struct dentry *edac_debugfs_create_x16(const char *name, umode_t mode,
+				       struct dentry *parent, u16 *value)
+{
+	if (!parent)
+		parent = edac_debugfs;
+
+	return debugfs_create_x16(name, mode, parent, value);
+}
+EXPORT_SYMBOL_GPL(edac_debugfs_create_x16);
diff --git a/drivers/edac/edac_module.h b/drivers/edac/edac_module.h
index 79a6c6e20819..b95a48fc723d 100644
--- a/drivers/edac/edac_module.h
+++ b/drivers/edac/edac_module.h
@@ -60,17 +60,39 @@ extern void *edac_align_ptr(void **p, unsigned size, int n_elems);
 /*
  * EDAC debugfs functions
  */
+
+#define edac_debugfs_remove_recursive debugfs_remove_recursive
+#define edac_debugfs_remove debugfs_remove
 #ifdef CONFIG_EDAC_DEBUG
 int edac_debugfs_init(void);
 void edac_debugfs_exit(void);
 int edac_create_debugfs_nodes(struct mem_ctl_info *mci);
+struct dentry *edac_debugfs_create_dir(const char *dirname);
+struct dentry *
+edac_debugfs_create_dir_at(const char *dirname, struct dentry *parent);
+struct dentry *
+edac_debugfs_create_file(const char *name, umode_t mode, struct dentry *parent,
+			 void *data, const struct file_operations *fops);
+struct dentry *
+edac_debugfs_create_x8(const char *name, umode_t mode, struct dentry *parent, u8 *value);
+struct dentry *
+edac_debugfs_create_x16(const char *name, umode_t mode, struct dentry *parent, u16 *value);
 #else
-static inline int edac_debugfs_init(void)
-{
-	return -ENODEV;
-}
-static inline void edac_debugfs_exit(void) {}
-static inline int edac_create_debugfs_nodes(struct mem_ctl_info *mci) { return 0; }
+static inline int edac_debugfs_init(void)					{ return -ENODEV; }
+static inline void edac_debugfs_exit(void)					{ }
+static inline int edac_create_debugfs_nodes(struct mem_ctl_info *mci)		{ return 0; }
+static inline struct dentry *edac_debugfs_create_dir(const char *dirname)	{ return NULL; }
+static inline struct dentry *
+edac_debugfs_create_dir_at(const char *dirname, struct dentry *parent)		{ return NULL; }
+static inline struct dentry *
+edac_debugfs_create_file(const char *name, umode_t mode, struct dentry *parent,
+			 void *data, const struct file_operations *fops)	{ return NULL; }
+static inline struct dentry *
+edac_debugfs_create_x8(const char *name, umode_t mode,
+		       struct dentry *parent, u8 *value)			{ return NULL; }
+static inline struct dentry *
+edac_debugfs_create_x16(const char *name, umode_t mode,
+		       struct dentry *parent, u16 *value)			{ return NULL; }
 #endif
 
 /*
-- 
2.3.5

[PATCH 3/5] EDAC, altera: Convert to debugfs wrappers

[Borislav Petkov]

From: Borislav Petkov <bp@suse.de>

Use the EDAC-specific wrappers. Drop CONFIG_EDAC_DEBUG ifdeffery.

Cc: Thor Thayer <tthayer@opensource.altera.com>
Cc: <linux-edac@vger.kernel.org>
Signed-off-by: Borislav Petkov <bp@suse.de>
---
 drivers/edac/altera_edac.c | 20 ++++++++------------
 drivers/edac/altera_edac.h |  2 --
 2 files changed, 8 insertions(+), 14 deletions(-)

diff --git a/drivers/edac/altera_edac.c b/drivers/edac/altera_edac.c
index 23ef0917483c..929640981d8a 100644
--- a/drivers/edac/altera_edac.c
+++ b/drivers/edac/altera_edac.c
@@ -51,11 +51,9 @@ static const struct altr_sdram_prv_data c5_data = {
 	.ecc_irq_clr_mask   = (CV_DRAMINTR_INTRCLR | CV_DRAMINTR_INTREN),
 	.ecc_cnt_rst_offset = CV_DRAMINTR_OFST,
 	.ecc_cnt_rst_mask   = CV_DRAMINTR_INTRCLR,
-#ifdef CONFIG_EDAC_DEBUG
 	.ce_ue_trgr_offset  = CV_CTLCFG_OFST,
 	.ce_set_mask        = CV_CTLCFG_GEN_SB_ERR,
 	.ue_set_mask        = CV_CTLCFG_GEN_DB_ERR,
-#endif
 };
 
 static const struct altr_sdram_prv_data a10_data = {
@@ -72,11 +70,9 @@ static const struct altr_sdram_prv_data a10_data = {
 	.ecc_irq_clr_mask   = (A10_INTSTAT_SBEERR | A10_INTSTAT_DBEERR),
 	.ecc_cnt_rst_offset = A10_ECCCTRL1_OFST,
 	.ecc_cnt_rst_mask   = A10_ECC_CNT_RESET_MASK,
-#ifdef CONFIG_EDAC_DEBUG
 	.ce_ue_trgr_offset  = A10_DIAGINTTEST_OFST,
 	.ce_set_mask        = A10_DIAGINT_TSERRA_MASK,
 	.ue_set_mask        = A10_DIAGINT_TDERRA_MASK,
-#endif
 };
 
 static irqreturn_t altr_sdram_mc_err_handler(int irq, void *dev_id)
@@ -116,7 +112,6 @@ static irqreturn_t altr_sdram_mc_err_handler(int irq, void *dev_id)
 	return IRQ_NONE;
 }
 
-#ifdef CONFIG_EDAC_DEBUG
 static ssize_t altr_sdr_mc_err_inject_write(struct file *file,
 					    const char __user *data,
 					    size_t count, loff_t *ppos)
@@ -191,14 +186,15 @@ static const struct file_operations altr_sdr_mc_debug_inject_fops = {
 
 static void altr_sdr_mc_create_debugfs_nodes(struct mem_ctl_info *mci)
 {
-	if (mci->debugfs)
-		debugfs_create_file("inject_ctrl", S_IWUSR, mci->debugfs, mci,
-				    &altr_sdr_mc_debug_inject_fops);
+	if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
+		return;
+
+	if (!mci->debugfs)
+		return;
+
+	edac_debugfs_create_file("inject_ctrl", S_IWUSR, mci->debugfs, mci,
+				 &altr_sdr_mc_debug_inject_fops);
 }
-#else
-static void altr_sdr_mc_create_debugfs_nodes(struct mem_ctl_info *mci)
-{}
-#endif
 
 /* Get total memory size from Open Firmware DTB */
 static unsigned long get_total_mem(void)
diff --git a/drivers/edac/altera_edac.h b/drivers/edac/altera_edac.h
index 7b64dc7c4eb7..4ef4fff463e1 100644
--- a/drivers/edac/altera_edac.h
+++ b/drivers/edac/altera_edac.h
@@ -181,13 +181,11 @@ struct altr_sdram_prv_data {
 	int ecc_irq_clr_mask;
 	int ecc_cnt_rst_offset;
 	int ecc_cnt_rst_mask;
-#ifdef CONFIG_EDAC_DEBUG
 	struct edac_dev_sysfs_attribute *eccmgr_sysfs_attr;
 	int ecc_enable_mask;
 	int ce_set_mask;
 	int ue_set_mask;
 	int ce_ue_trgr_offset;
-#endif
 };
 
 /* Altera SDRAM Memory Controller data */
-- 
2.3.5

[PATCH 4/5] EDAC, i5100: Convert to debugfs wrappers

[Borislav Petkov]

From: Borislav Petkov <bp@suse.de>

This driver creates its debugfs hierarchy under the toplevel debugfs dir
- see i5100_init() - so make it use edac_debugfs_create_dir_at( , NULL)
because we're not breaking userspace. Oh well.

Signed-off-by: Borislav Petkov <bp@suse.de>
---
 drivers/edac/i5100_edac.c | 37 +++++++++++++++++++------------------
 1 file changed, 19 insertions(+), 18 deletions(-)

diff --git a/drivers/edac/i5100_edac.c b/drivers/edac/i5100_edac.c
index e9f8a393915a..40917775dca1 100644
--- a/drivers/edac/i5100_edac.c
+++ b/drivers/edac/i5100_edac.c
@@ -30,6 +30,7 @@
 #include <linux/debugfs.h>
 
 #include "edac_core.h"
+#include "edac_module.h"
 
 /* register addresses */
 
@@ -966,25 +967,25 @@ static int i5100_setup_debugfs(struct mem_ctl_info *mci)
 	if (!i5100_debugfs)
 		return -ENODEV;
 
-	priv->debugfs = debugfs_create_dir(mci->bus->name, i5100_debugfs);
+	priv->debugfs = edac_debugfs_create_dir_at(mci->bus->name, i5100_debugfs);
 
 	if (!priv->debugfs)
 		return -ENOMEM;
 
-	debugfs_create_x8("inject_channel", S_IRUGO | S_IWUSR, priv->debugfs,
-			&priv->inject_channel);
-	debugfs_create_x8("inject_hlinesel", S_IRUGO | S_IWUSR, priv->debugfs,
-			&priv->inject_hlinesel);
-	debugfs_create_x8("inject_deviceptr1", S_IRUGO | S_IWUSR, priv->debugfs,
-			&priv->inject_deviceptr1);
-	debugfs_create_x8("inject_deviceptr2", S_IRUGO | S_IWUSR, priv->debugfs,
-			&priv->inject_deviceptr2);
-	debugfs_create_x16("inject_eccmask1", S_IRUGO | S_IWUSR, priv->debugfs,
-			&priv->inject_eccmask1);
-	debugfs_create_x16("inject_eccmask2", S_IRUGO | S_IWUSR, priv->debugfs,
-			&priv->inject_eccmask2);
-	debugfs_create_file("inject_enable", S_IWUSR, priv->debugfs,
-			&mci->dev, &i5100_inject_enable_fops);
+	edac_debugfs_create_x8("inject_channel", S_IRUGO | S_IWUSR, priv->debugfs,
+				&priv->inject_channel);
+	edac_debugfs_create_x8("inject_hlinesel", S_IRUGO | S_IWUSR, priv->debugfs,
+				&priv->inject_hlinesel);
+	edac_debugfs_create_x8("inject_deviceptr1", S_IRUGO | S_IWUSR, priv->debugfs,
+				&priv->inject_deviceptr1);
+	edac_debugfs_create_x8("inject_deviceptr2", S_IRUGO | S_IWUSR, priv->debugfs,
+				&priv->inject_deviceptr2);
+	edac_debugfs_create_x16("inject_eccmask1", S_IRUGO | S_IWUSR, priv->debugfs,
+				&priv->inject_eccmask1);
+	edac_debugfs_create_x16("inject_eccmask2", S_IRUGO | S_IWUSR, priv->debugfs,
+				&priv->inject_eccmask2);
+	edac_debugfs_create_file("inject_enable", S_IWUSR, priv->debugfs,
+				&mci->dev, &i5100_inject_enable_fops);
 
 	return 0;
 
@@ -1189,7 +1190,7 @@ static void i5100_remove_one(struct pci_dev *pdev)
 
 	priv = mci->pvt_info;
 
-	debugfs_remove_recursive(priv->debugfs);
+	edac_debugfs_remove_recursive(priv->debugfs);
 
 	priv->scrub_enable = 0;
 	cancel_delayed_work_sync(&(priv->i5100_scrubbing));
@@ -1223,7 +1224,7 @@ static int __init i5100_init(void)
 {
 	int pci_rc;
 
-	i5100_debugfs = debugfs_create_dir("i5100_edac", NULL);
+	i5100_debugfs = edac_debugfs_create_dir_at("i5100_edac", NULL);
 
 	pci_rc = pci_register_driver(&i5100_driver);
 	return (pci_rc < 0) ? pci_rc : 0;
@@ -1231,7 +1232,7 @@ static int __init i5100_init(void)
 
 static void __exit i5100_exit(void)
 {
-	debugfs_remove(i5100_debugfs);
+	edac_debugfs_remove(i5100_debugfs);
 
 	pci_unregister_driver(&i5100_driver);
 }
-- 
2.3.5

[PATCH 5/5] EDAC, xgene: Convert to debugfs wrappers

[Borislav Petkov]

From: Borislav Petkov <bp@suse.de>

Drop CONFIG_EDAC_DEBUG ifdeffery too, while at it.

Cc: Loc Ho <lho@apm.com>
Cc: linux-edac at vger.kernel.org
Signed-off-by: Borislav Petkov <bp@suse.de>
---
 drivers/edac/xgene_edac.c | 26 +++++++++++++-------------
 1 file changed, 13 insertions(+), 13 deletions(-)

diff --git a/drivers/edac/xgene_edac.c b/drivers/edac/xgene_edac.c
index ba06904af2e1..5ff42d5b019d 100644
--- a/drivers/edac/xgene_edac.c
+++ b/drivers/edac/xgene_edac.c
@@ -29,6 +29,7 @@
 #include <linux/regmap.h>
 
 #include "edac_core.h"
+#include "edac_module.h"
 
 #define EDAC_MOD_STR			"xgene_edac"
 
@@ -172,12 +173,12 @@ static void xgene_edac_mc_create_debugfs_node(struct mem_ctl_info *mci)
 {
 	if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
 		return;
-#ifdef CONFIG_EDAC_DEBUG
+
 	if (!mci->debugfs)
 		return;
-	debugfs_create_file("inject_ctrl", S_IWUSR, mci->debugfs, mci,
-			    &xgene_edac_mc_debug_inject_fops);
-#endif
+
+	edac_debugfs_create_file("inject_ctrl", S_IWUSR, mci->debugfs, mci,
+				 &xgene_edac_mc_debug_inject_fops);
 }
 
 static void xgene_edac_mc_check(struct mem_ctl_info *mci)
@@ -881,7 +882,7 @@ static void xgene_edac_pmd_create_debugfs_nodes(
 	struct edac_device_ctl_info *edac_dev)
 {
 	struct xgene_edac_pmd_ctx *ctx = edac_dev->pvt_info;
-	struct dentry *edac_debugfs;
+	struct dentry *dbgfs_dir;
 	char name[30];
 
 	if (!IS_ENABLED(CONFIG_EDAC_DEBUG))
@@ -892,20 +893,19 @@ static void xgene_edac_pmd_create_debugfs_nodes(
 	 *       when available.
 	 */
 	if (!ctx->edac->dfs) {
-		ctx->edac->dfs = debugfs_create_dir(edac_dev->dev->kobj.name,
-						    NULL);
+		ctx->edac->dfs = edac_debugfs_create_dir(edac_dev->dev->kobj.name);
 		if (!ctx->edac->dfs)
 			return;
 	}
 	sprintf(name, "PMD%d", ctx->pmd);
-	edac_debugfs = debugfs_create_dir(name, ctx->edac->dfs);
-	if (!edac_debugfs)
+	dbgfs_dir = edac_debugfs_create_dir_at(name, ctx->edac->dfs);
+	if (!dbgfs_dir)
 		return;
 
-	debugfs_create_file("l1_inject_ctrl", S_IWUSR, edac_debugfs, edac_dev,
-			    &xgene_edac_pmd_debug_inject_fops[0]);
-	debugfs_create_file("l2_inject_ctrl", S_IWUSR, edac_debugfs, edac_dev,
-			    &xgene_edac_pmd_debug_inject_fops[1]);
+	edac_debugfs_create_file("l1_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev,
+				 &xgene_edac_pmd_debug_inject_fops[0]);
+	edac_debugfs_create_file("l2_inject_ctrl", S_IWUSR, dbgfs_dir, edac_dev,
+				 &xgene_edac_pmd_debug_inject_fops[1]);
 }
 
 static int xgene_edac_pmd_available(u32 efuse, int pmd)
-- 
2.3.5

[PATCH v4 2/4] Documentation: Update the APM X-Gene SoC EDAC DTS binding for L3/SoC subnodes

[Borislav Petkov]

On Fri, Aug 14, 2015 at 12:46:07AM -0600, Loc Ho wrote:
> This patch updates documentation for the APM X-Gene SoC EDAC DTS binding

No need to start the commit message with "This patch" - we know it is
this patch. :)

> for L3/SoC subnodes.
> 
> Signed-off-by: Loc Ho <lho@apm.com>
> ---
>  .../devicetree/bindings/edac/apm-xgene-edac.txt    |   23 ++++++++++++++++++++
>  1 files changed, 23 insertions(+), 0 deletions(-)

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.

[PATCH v4 3/4] edac: Add L3/SoC EDAC support to the APM X-Gene SoC EDAC driver

[Borislav Petkov]

On Fri, Aug 14, 2015 at 12:46:08AM -0600, Loc Ho wrote:
> This patch adds EDAC support for the L3 and SoC components.
> 
> Signed-off-by: Loc Ho <lho@apm.com>
> ---
>  drivers/edac/xgene_edac.c | 1169 +++++++++++++++++++++++++++++++++++++--------
>  1 files changed, 975 insertions(+), 194 deletions(-)

...

> +/* L3 Error device */
> +#define L3C_ESR				(0x0A * 4)
> +#define  L3C_ESR_DATATAG_MASK		BIT(9)
> +#define  L3C_ESR_MULTIHIT_MASK		BIT(8)
> +#define  L3C_ESR_UCEVICT_MASK		BIT(6)
> +#define  L3C_ESR_MULTIUCERR_MASK	BIT(5)
> +#define  L3C_ESR_MULTICERR_MASK		BIT(4)
> +#define  L3C_ESR_UCERR_MASK		BIT(3)
> +#define  L3C_ESR_CERR_MASK		BIT(2)
> +#define  L3C_ESR_UCERRINTR_MASK		BIT(1)
> +#define  L3C_ESR_CERRINTR_MASK		BIT(0)
> +#define L3C_ECR				(0x0B * 4)
> +#define  L3C_ECR_UCINTREN		BIT(3)
> +#define  L3C_ECR_CINTREN		BIT(2)
> +#define  L3C_UCERREN			BIT(1)
> +#define  L3C_CERREN			BIT(0)
> +#define L3C_ELR				(0x0C * 4)
> +#define  L3C_ELR_ERRSYN(src)		((src & 0xFF800000) >> 23)
> +#define  L3C_ELR_ERRWAY(src)		((src & 0x007E0000) >> 17)
> +#define  L3C_ELR_AGENTID(src)		((src & 0x0001E000) >> 13)
> +#define  L3C_ELR_ERRGRP(src)		((src & 0x00000F00) >> 8)
> +#define  L3C_ELR_OPTYPE(src)		((src & 0x000000F0) >> 4)
> +#define  L3C_ELR_PADDRHIGH(src)		(src & 0x0000000F)
> +#define L3C_AELR			(0x0D * 4)
> +#define L3C_BELR			(0x0E * 4)
> +#define  L3C_BELR_BANK(src)		(src & 0x0000000F)
> +
> +struct xgene_edac_dev_ctx {
> +	struct list_head	next;
> +	struct device		ddev;
> +	char			*name;
> +	struct xgene_edac	*edac;
> +	struct edac_device_ctl_info *edac_dev;
> +	int			edac_idx;
> +	void __iomem		*dev_csr;
> +	int			version;
> +};
> +

Put the comment ontop of the function. Also, I'd fixup the comment like this:

/*
 * Version 1 of the L3 controller has broken single bit correctable logic for
 * certain error syndromes. Log them as uncorrectable in that case.
 */

and then you can drop the comment in the function itself and at the call site.

> +static bool xgene_edac_l3_v1_errata_chk(u32 l3cesr, u32 l3celr)

This function name should be something like:

should_promote_error_to_uc()

or so because it is what it does.

> +{
> +	/*
> +	 * L3 version 1 has certain conditions in which correctable error
> +	 * needs to be flagged as un-correctable error. This function
> +	 * check for such conditions.
> +	 */
> +	if (l3cesr & L3C_ESR_DATATAG_MASK) {
> +		switch (L3C_ELR_ERRSYN(l3celr)) {
> +		case 0x13C:
> +		case 0x0B4:
> +		case 0x007:
> +		case 0x00D:
> +		case 0x00E:
> +		case 0x019:
> +		case 0x01A:
> +		case 0x01C:
> +		case 0x04E:
> +		case 0x041:
> +			return true;
> +		}
> +	} else if (L3C_ELR_ERRSYN(l3celr) == 9) {
> +		return true;
> +	}

No need for {} around a single statement.

...

> +static ssize_t xgene_edac_l3_inject_ctrl_write(struct file *file,
> +					       const char __user *data,
> +					       size_t count, loff_t *ppos)
> +{
> +	struct edac_device_ctl_info *edac_dev = file->private_data;
> +	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
> +
> +	/* Generate all errors */
> +	writel(0xFFFFFFFF, ctx->dev_csr + L3C_ESR);
> +	return count;
> +}
> +
> +static const struct file_operations xgene_edac_l3_debug_inject_fops = {
> +	.open = simple_open,
> +	.write = xgene_edac_l3_inject_ctrl_write,
> +	.llseek = generic_file_llseek
> +};
> +
> +static void xgene_edac_l3_create_debugfs_nodes(
> +	struct edac_device_ctl_info *edac_dev)
> +{

This way of formatting the function name should be more readable:

static void
xgene_edac_l3_create_debugfs_nodes(struct edac_device_ctl_info *edac_dev)
{

> +	struct xgene_edac_dev_ctx *ctx = edac_dev->pvt_info;
> +	struct dentry *edac_debugfs;
> +	char name[30];

That's a 30-chars array on the stack ...

> +
> +	if (!ctx->edac->dfs)
> +		return;
> +	sprintf(name, "l3c%d", ctx->edac_idx);

... but you're not clearing or NULL-terminating the rest of it here,
should the string be smaller.

> +	edac_debugfs = debugfs_create_dir(name, ctx->edac->dfs);
> +	if (!edac_debugfs)
> +		return;
> +
> +	debugfs_create_file("l3_inject_ctrl", S_IWUSR, edac_debugfs, edac_dev,
> +			    &xgene_edac_l3_debug_inject_fops);
> +}
> +
> +static int xgene_edac_l3_add(struct xgene_edac *edac, struct device_node *np,
> +			     int version)
> +{
> +	struct edac_device_ctl_info *edac_dev;
> +	struct xgene_edac_dev_ctx *ctx;
> +	struct resource res;
> +	void __iomem *dev_csr;
> +	int edac_idx;
> +	int rc = 0;
> +
> +	if (!devres_open_group(edac->dev, xgene_edac_l3_add, GFP_KERNEL))
> +		return -ENOMEM;
> +
> +	rc = of_address_to_resource(np, 0, &res);
> +	if (rc < 0) {
> +		dev_err(edac->dev, "no L3 resource address\n");
> +		goto err_release_group;
> +	}
> +	dev_csr = devm_ioremap_resource(edac->dev, &res);
> +	if (IS_ERR(dev_csr)) {
> +		dev_err(edac->dev,
> +			"devm_ioremap_resource failed for L3 resource address\n");
> +		rc = PTR_ERR(dev_csr);
> +		goto err_release_group;
> +	}
> +
> +	edac_idx = edac_device_alloc_index();
> +	edac_dev = edac_device_alloc_ctl_info(sizeof(*ctx),
> +					      "l3c", 1, "l3c", 1, 0, NULL, 0,
> +					      edac_idx);
> +	if (!edac_dev) {
> +		rc = -ENOMEM;
> +		goto err_release_group;
> +	}
> +
> +	ctx = edac_dev->pvt_info;
> +	ctx->dev_csr = dev_csr;
> +	ctx->name = "xgene_l3_err";
> +	ctx->edac_idx = edac_idx;
> +	ctx->edac = edac;
> +	ctx->edac_dev = edac_dev;
> +	ctx->ddev = *edac->dev;
> +	ctx->version = version;
> +	edac_dev->dev = &ctx->ddev;
> +	edac_dev->ctl_name = ctx->name;
> +	edac_dev->dev_name = ctx->name;
> +	edac_dev->mod_name = EDAC_MOD_STR;
> +
> +	if (edac_op_state == EDAC_OPSTATE_POLL)
> +		edac_dev->edac_check = xgene_edac_l3_check;
> +
> +	xgene_edac_l3_create_debugfs_nodes(edac_dev);
> +
> +	rc = edac_device_add_device(edac_dev);
> +	if (rc > 0) {
> +		dev_err(edac->dev, "failed edac_device_add_device()\n");
> +		rc = -ENOMEM;
> +		goto err_ctl_free;
> +	}
> +
> +	if (edac_op_state == EDAC_OPSTATE_INT)
> +		edac_dev->op_state = OP_RUNNING_INTERRUPT;
> +
> +	list_add(&ctx->next, &edac->l3s);
> +
> +	xgene_edac_l3_hw_init(edac_dev, 1);
> +
> +	devres_remove_group(edac->dev, xgene_edac_l3_add);
> +
> +	dev_info(edac->dev, "X-Gene EDAC L3 registered\n");
> +	return 0;
> +
> +err_ctl_free:
> +	edac_device_free_ctl_info(edac_dev);
> +err_release_group:
> +	devres_release_group(edac->dev, xgene_edac_l3_add);
> +	return rc;
> +}
> +
> +static int xgene_edac_l3_remove(struct xgene_edac_dev_ctx *l3)
> +{
> +	struct edac_device_ctl_info *edac_dev = l3->edac_dev;
> +
> +	xgene_edac_l3_hw_init(edac_dev, 0);
> +	edac_device_del_device(l3->edac->dev);
> +	edac_device_free_ctl_info(edac_dev);
> +	return 0;
> +}

This is a natural point to stop. The rest should be in a separate patch
called something like "EDAC, xgene: Add SoC support...".

> +
> +/* SoC Error device */
> +#define IOBAXIS0TRANSERRINTSTS		0x0000
> +#define  IOBAXIS0_M_ILLEGAL_ACCESS_MASK	BIT(1)
> +#define  IOBAXIS0_ILLEGAL_ACCESS_MASK	BIT(0)
> +#define IOBAXIS0TRANSERRINTMSK		0x0004
> +#define IOBAXIS0TRANSERRREQINFOL	0x0008
> +#define IOBAXIS0TRANSERRREQINFOH	0x000c
> +#define  REQTYPE_RD(src)		(((src) & BIT(0)))
> +#define  ERRADDRH_RD(src)		(((src) & 0xffc00000) >> 22)
> +#define IOBAXIS1TRANSERRINTSTS		0x0010
> +#define IOBAXIS1TRANSERRINTMSK		0x0014
> +#define IOBAXIS1TRANSERRREQINFOL	0x0018
> +#define IOBAXIS1TRANSERRREQINFOH	0x001c
> +#define IOBPATRANSERRINTSTS		0x0020

...

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.

[PATCH v4 1/4] edac: Export edac_debugfs

[Loc Ho]

Hi,

>> This patch exports and expose the edac_debugfs file node. This allows
>> EDAC driver to create debugfs node under the EDAC debugfs node.
>
> ...
>
>> -#ifdef CONFIG_EDAC_DEBUG
>> -static struct dentry *edac_debugfs;
>> +struct dentry *edac_debugfs;
>> +EXPORT_SYMBOL_GPL(edac_debugfs);
>
> Hmm, so Arnd was right about this being the wrong direction we're
> chasing. We should rather be exporting a bunch of EDAC debugfs wrappers
> which EDAC drivers can call instead of exposing edac_debugfs itself.
>
> And this is much better and the right(tm) approach IMO. So I went
> and hacked it in, it builds on those !x86 targets (as far as my
> cross-compilers are concerned, build succeeds :-)).
>
> So here it is as a follow-up to this message.

Let's me sync and give it a try. Do you see any other issue besides this?

-Loc

[PATCH v4 1/4] edac: Export edac_debugfs

[Borislav Petkov]

On Tue, Sep 22, 2015 at 10:46:07AM -0700, Loc Ho wrote:
> Let's me sync and give it a try. Do you see any other issue besides this?

I replied to each message separately but they were all small nitpicks
only.

Btw, I have the debugfs pile here if you want to base your stuff ontop:

git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp.git#edac-debugfs

Thanks.

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.

[PATCH 5/5] EDAC, xgene: Convert to debugfs wrappers

[Loc Ho]

Hi,

>
> Drop CONFIG_EDAC_DEBUG ifdeffery too, while at it.
>
> Cc: Loc Ho <lho@apm.com>
> Cc: linux-edac at vger.kernel.org
> Signed-off-by: Borislav Petkov <bp@suse.de>

Tested by Loc Ho.

-Loc