2023-01-04 09:00:17

by Sai Krishna Potthuri

[permalink] [raw]
Subject: [PATCH v7 0/2] EDAC: Add support for Xilinx ZynqMP OCM EDAC

Add dt-binding and driver for Xilinx ZynqMP OCM controller.

changes in v7:
-> 2/2 - Renamed the driver file name to zynqmp_edac.c.
-> 2/2 - Update the UE error injection logic to use string separator,
also removed the inject_cebitpos() or inject_uebitpos() and handled in the
top level inject functions.
-> 2/2 - Changed the name of the debugfs directory to "ocm" to make it
simple.
-> 2/2 - Fixed few more comments like update commit description, used
IS_ENABLED(CONFIG_EDAC_DEBUG) instead of #ifdef, removing "controller"
string, local variable ordering.

changes in v6:
-> 2/2 - Updated subject prefix and commit description.
-> 2/2 - Used Debugfs instead of sysfs for error injection and
placed the injection logic under CONFIG_EDAC_DEBUG.
-> 2/2 - Dropped zynqmp_ocm prefix for all static APIs and structures.
-> 2/2 - Fixed few more comments related to using caps for acronyms,
dealing error info, UE logic simplification, using BIT() definitions.

changes in v5:
-> 1/2, 2/2 - Added 'Co-developed-by' tag.
-> 2/2 - Updated the driver hep text to be more clear about the hardware
this driver is targeted.
-> 2/2 - Fixed the warning reported by kernel test robot.

changes in v4:
-> 2/2 - Replaced \n\r with \n.

changes in v3:
-> 1/2 - Moved the binding from edac to memory-controllers directory.
-> 1/2 - Changed the file name to match with the compatible.
-> 1/2 - Used additionalProperties instead of unevaluatedProperties.
-> 1/2 - Used macro instead of constant value.

changes in v2:
-> 1/2 - Used define for interrupt flag.
-> 1/2 - Updated the description and title.
-> 2/2 - Removed Kernel doc for probe and remove.
-> 2/2 - Used COMPILE_TEST, used wrapper for get and ioremap resource.
-> 2/2 - Fixed few comments related to static variable declaration
and print statements.

Sai Krishna Potthuri (1):
EDAC/zynqmp: Add EDAC support for Xilinx ZynqMP OCM

Shubhrajyoti Datta (1):
dt-bindings: edac: Add bindings for Xilinx ZynqMP OCM

.../xlnx,zynqmp-ocmc-1.0.yaml | 45 ++
MAINTAINERS | 7 +
drivers/edac/Kconfig | 9 +
drivers/edac/Makefile | 1 +
drivers/edac/zynqmp_edac.c | 465 ++++++++++++++++++
5 files changed, 527 insertions(+)
create mode 100644 Documentation/devicetree/bindings/memory-controllers/xlnx,zynqmp-ocmc-1.0.yaml
create mode 100644 drivers/edac/zynqmp_edac.c

--
2.25.1


2023-01-04 09:01:16

by Sai Krishna Potthuri

[permalink] [raw]
Subject: [PATCH v7 2/2] EDAC/zynqmp: Add EDAC support for Xilinx ZynqMP OCM

Add EDAC support for Xilinx ZynqMP OCM Controller, so this driver
reports CE and UE errors upon interrupt generation, and also creates UE/CE
debugfs entries for error injection.
On Xilinx ZynqMP platform, both OCM Controller driver(zynqmp_edac) and
DDR Memory Controller driver(synopsys_edac) co-exist which means both
can be loaded at a time. This scenario is tested on Xilinx ZynqMP
platform.

Fix following issue reported by the robot.
"MAINTAINERS references a file that doesn't exist:
Documentation/devicetree/bindings/edac/xlnx,zynqmp-ocmc.yaml"

Co-developed-by: Shubhrajyoti Datta <[email protected]>
Signed-off-by: Shubhrajyoti Datta <[email protected]>
Signed-off-by: Sai Krishna Potthuri <[email protected]>
Reported-by: kernel test robot <[email protected]>
---
MAINTAINERS | 7 +
drivers/edac/Kconfig | 9 +
drivers/edac/Makefile | 1 +
drivers/edac/zynqmp_edac.c | 465 +++++++++++++++++++++++++++++++++++++
4 files changed, 482 insertions(+)
create mode 100644 drivers/edac/zynqmp_edac.c

diff --git a/MAINTAINERS b/MAINTAINERS
index edc96cdb85e8..2b6e74025549 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -21692,6 +21692,13 @@ F: Documentation/devicetree/bindings/dma/xilinx/xlnx,zynqmp-dpdma.yaml
F: drivers/dma/xilinx/xilinx_dpdma.c
F: include/dt-bindings/dma/xlnx-zynqmp-dpdma.h

+XILINX ZYNQMP OCM EDAC DRIVER
+M: Shubhrajyoti Datta <[email protected]>
+M: Sai Krishna Potthuri <[email protected]>
+S: Maintained
+F: Documentation/devicetree/bindings/memory-controllers/xlnx,zynqmp-ocmc-1.0.yaml
+F: drivers/edac/zynqmp_edac.c
+
XILINX ZYNQMP PSGTR PHY DRIVER
M: Anurag Kumar Vulisha <[email protected]>
M: Laurent Pinchart <[email protected]>
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
index 58ab63642e72..7944e40c67da 100644
--- a/drivers/edac/Kconfig
+++ b/drivers/edac/Kconfig
@@ -539,4 +539,13 @@ config EDAC_DMC520
Support for error detection and correction on the
SoCs with ARM DMC-520 DRAM controller.

+config EDAC_ZYNQMP_OCM
+ tristate "Xilinx ZynqMP OCM Controller"
+ depends on ARCH_ZYNQMP || COMPILE_TEST
+ help
+ This driver supports error detection and correction for the
+ Xilinx ZynqMP OCM (On Chip Memory) controller.
+ This driver can also be built as a module. If so, the module
+ will be called zynqmp_ocm_edac.
+
endif # EDAC
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
index 2d1641a27a28..47cbda06d7b0 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -84,3 +84,4 @@ obj-$(CONFIG_EDAC_QCOM) += qcom_edac.o
obj-$(CONFIG_EDAC_ASPEED) += aspeed_edac.o
obj-$(CONFIG_EDAC_BLUEFIELD) += bluefield_edac.o
obj-$(CONFIG_EDAC_DMC520) += dmc520_edac.o
+obj-$(CONFIG_EDAC_ZYNQMP_OCM) += zynqmp_edac.o
diff --git a/drivers/edac/zynqmp_edac.c b/drivers/edac/zynqmp_edac.c
new file mode 100644
index 000000000000..69069028457b
--- /dev/null
+++ b/drivers/edac/zynqmp_edac.c
@@ -0,0 +1,465 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Xilinx ZynqMP OCM ECC Driver
+ *
+ * Copyright (C) 2022 Advanced Micro Devices, Inc.
+ */
+
+#include <linux/edac.h>
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_platform.h>
+#include <linux/platform_device.h>
+
+#include "edac_module.h"
+
+#define ZYNQMP_OCM_EDAC_MSG_SIZE 256
+
+#define ZYNQMP_OCM_EDAC_STRING "zynqmp_ocm"
+
+/* Error/Interrupt registers */
+#define ERR_CTRL_OFST 0x0
+#define OCM_ISR_OFST 0x04
+#define OCM_IMR_OFST 0x08
+#define OCM_IEN_OFST 0x0C
+#define OCM_IDS_OFST 0x10
+
+/* ECC control register */
+#define ECC_CTRL_OFST 0x14
+
+/* Correctable error info registers */
+#define CE_FFA_OFST 0x1C
+#define CE_FFD0_OFST 0x20
+#define CE_FFD1_OFST 0x24
+#define CE_FFD2_OFST 0x28
+#define CE_FFD3_OFST 0x2C
+#define CE_FFE_OFST 0x30
+
+/* Uncorrectable error info registers */
+#define UE_FFA_OFST 0x34
+#define UE_FFD0_OFST 0x38
+#define UE_FFD1_OFST 0x3C
+#define UE_FFD2_OFST 0x40
+#define UE_FFD3_OFST 0x44
+#define UE_FFE_OFST 0x48
+
+/* ECC control register bit field definitions */
+#define ECC_CTRL_CLR_CE_ERR 0x40
+#define ECC_CTRL_CLR_UE_ERR 0x80
+
+/* Fault injection data and count registers */
+#define OCM_FID0_OFST 0x4C
+#define OCM_FID1_OFST 0x50
+#define OCM_FID2_OFST 0x54
+#define OCM_FID3_OFST 0x58
+#define OCM_FIC_OFST 0x74
+
+#define UE_MAX_BITPOS_LOWER 31
+#define UE_MIN_BITPOS_UPPER 32
+#define UE_MAX_BITPOS_UPPER 63
+
+/* Interrupt masks */
+#define OCM_CEINTR_MASK BIT(6)
+#define OCM_UEINTR_MASK BIT(7)
+#define OCM_ECC_ENABLE_MASK BIT(0)
+
+#define OCM_FICOUNT_MASK GENMASK(23, 0)
+#define OCM_NUM_UE_BITPOS 2
+#define OCM_BASEVAL 0xFFFC0000
+#define EDAC_DEVICE "ZynqMP-OCM"
+
+/**
+ * struct ecc_error_info - ECC error log information
+ * @addr: Fault generated at this address
+ * @fault_lo: Generated fault data (lower 32-bit)
+ * @fault_hi: Generated fault data (upper 32-bit)
+ */
+struct ecc_error_info {
+ u32 addr;
+ u32 fault_lo;
+ u32 fault_hi;
+};
+
+/**
+ * struct ecc_status - ECC status information to report
+ * @ce_cnt: Correctable error count
+ * @ue_cnt: Uncorrectable error count
+ * @ceinfo: Correctable error log information
+ * @ueinfo: Uncorrectable error log information
+ */
+struct ecc_status {
+ u32 ce_cnt;
+ u32 ue_cnt;
+ struct ecc_error_info ceinfo;
+ struct ecc_error_info ueinfo;
+};
+
+/**
+ * struct edac_priv - OCM private instance data
+ * @baseaddr: Base address of the OCM
+ * @message: Buffer for framing the event specific info
+ * @stat: ECC status information
+ * @ce_cnt: Correctable Error count
+ * @ue_cnt: Uncorrectable Error count
+ * @debugfs_dir: Directory entry for debugfs
+ * @ce_bitpos: Bit position for Correctable Error
+ * @ue_bitpos: Array to store UnCorrectable Error bit positions
+ * @fault_injection_cnt: Fault Injection Counter value
+ */
+struct edac_priv {
+ void __iomem *baseaddr;
+ char message[ZYNQMP_OCM_EDAC_MSG_SIZE];
+ struct ecc_status stat;
+ u32 ce_cnt;
+ u32 ue_cnt;
+#ifdef CONFIG_EDAC_DEBUG
+ struct dentry *debugfs_dir;
+ u8 ce_bitpos;
+ u8 ue_bitpos[OCM_NUM_UE_BITPOS];
+ u32 fault_injection_cnt;
+#endif
+};
+
+/**
+ * get_error_info - Get the current ECC error info
+ * @base: Pointer to the base address of the OCM
+ * @p: Pointer to the OCM ECC status structure
+ * @mask: Status register mask value
+ *
+ * Determines there is any ECC error or not
+ *
+ */
+static void get_error_info(void __iomem *base, struct ecc_status *p, int mask)
+{
+ if (mask & OCM_CEINTR_MASK) {
+ p->ce_cnt++;
+ p->ceinfo.fault_lo = readl(base + CE_FFD0_OFST);
+ p->ceinfo.fault_hi = readl(base + CE_FFD1_OFST);
+ p->ceinfo.addr = (OCM_BASEVAL | readl(base + CE_FFA_OFST));
+ writel(ECC_CTRL_CLR_CE_ERR, base + OCM_ISR_OFST);
+ } else {
+ p->ue_cnt++;
+ p->ueinfo.fault_lo = readl(base + UE_FFD0_OFST);
+ p->ueinfo.fault_hi = readl(base + UE_FFD1_OFST);
+ p->ueinfo.addr = (OCM_BASEVAL | readl(base + UE_FFA_OFST));
+ writel(ECC_CTRL_CLR_UE_ERR, base + OCM_ISR_OFST);
+ }
+}
+
+/**
+ * handle_error - Handle error types CE and UE
+ * @dci: Pointer to the EDAC device instance
+ * @p: Pointer to the OCM ECC status structure
+ *
+ * Handles the ECC correctable and uncorrectable error.
+ */
+static void handle_error(struct edac_device_ctl_info *dci, struct ecc_status *p)
+{
+ struct edac_priv *priv = dci->pvt_info;
+ struct ecc_error_info *pinf;
+
+ if (p->ce_cnt) {
+ pinf = &p->ceinfo;
+ snprintf(priv->message, ZYNQMP_OCM_EDAC_MSG_SIZE,
+ "\nOCM ECC error type :%s\nAddr: [0x%x]\nFault Data[0x%08x%08x]",
+ "CE", pinf->addr, pinf->fault_hi, pinf->fault_lo);
+ edac_device_handle_ce(dci, 0, 0, priv->message);
+ }
+
+ if (p->ue_cnt) {
+ pinf = &p->ueinfo;
+ snprintf(priv->message, ZYNQMP_OCM_EDAC_MSG_SIZE,
+ "\nOCM ECC error type :%s\nAddr: [0x%x]\nFault Data[0x%08x%08x]",
+ "UE", pinf->addr, pinf->fault_hi, pinf->fault_lo);
+ edac_device_handle_ue(dci, 0, 0, priv->message);
+ }
+
+ memset(p, 0, sizeof(*p));
+}
+
+/**
+ * intr_handler - ISR routine
+ * @irq: irq number
+ * @dev_id: device id pointer
+ *
+ * Return: IRQ_NONE, if CE/UE interrupt not set or IRQ_HANDLED otherwise
+ */
+static irqreturn_t intr_handler(int irq, void *dev_id)
+{
+ struct edac_device_ctl_info *dci = dev_id;
+ struct edac_priv *priv = dci->pvt_info;
+ int regval;
+
+ regval = readl(priv->baseaddr + OCM_ISR_OFST);
+ if (!(regval & (OCM_CEINTR_MASK | OCM_UEINTR_MASK))) {
+ WARN_ONCE(1, "Unhandled IRQ%d.", irq);
+ return IRQ_NONE;
+ }
+
+ get_error_info(priv->baseaddr, &priv->stat, regval);
+
+ priv->ce_cnt += priv->stat.ce_cnt;
+ priv->ue_cnt += priv->stat.ue_cnt;
+ handle_error(dci, &priv->stat);
+
+ return IRQ_HANDLED;
+}
+
+/**
+ * get_eccstate - Return the ECC status
+ * @base: Pointer to the OCM base address
+ *
+ * Get the ECC enable/disable status
+ *
+ * Return: ECC status 0/1.
+ */
+static bool get_eccstate(void __iomem *base)
+{
+ return readl(base + ECC_CTRL_OFST) & OCM_ECC_ENABLE_MASK;
+}
+
+#ifdef CONFIG_EDAC_DEBUG
+/**
+ * write_fault_count - write fault injection count
+ * @priv: Pointer to the EDAC private struct
+ *
+ * Update the fault injection count register, once the counter reaches
+ * zero, it injects errors
+ */
+static void write_fault_count(struct edac_priv *priv)
+{
+ u32 ficount = priv->fault_injection_cnt;
+
+ if (ficount & ~OCM_FICOUNT_MASK) {
+ ficount &= OCM_FICOUNT_MASK;
+ edac_printk(KERN_INFO, EDAC_DEVICE,
+ "Fault injection count value truncated to %d\n", ficount);
+ }
+
+ writel(ficount, priv->baseaddr + OCM_FIC_OFST);
+}
+
+/*
+ * To get the Correctable Error injected, the following steps are needed:
+ * - Setup the optional Fault Injection Count:
+ * echo <fault_count val> > /sys/kernel/debug/edac/ocm/inject_fault_count
+ * - Write the Correctable Error bit position value:
+ * echo <bit_pos val> > /sys/kernel/debug/edac/ocm/inject_ce_bitpos
+ */
+static ssize_t inject_ce_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 edac_priv *priv = edac_dev->pvt_info;
+ int ret;
+
+ if (!data)
+ return -EFAULT;
+
+ ret = kstrtou8_from_user(data, count, 0, &priv->ce_bitpos);
+ if (ret)
+ return ret;
+
+ if (priv->ce_bitpos > UE_MAX_BITPOS_UPPER)
+ return -EINVAL;
+
+ if (priv->ce_bitpos <= UE_MAX_BITPOS_LOWER) {
+ writel(BIT(priv->ce_bitpos), priv->baseaddr + OCM_FID0_OFST);
+ writel(0, priv->baseaddr + OCM_FID1_OFST);
+ } else {
+ writel(BIT(priv->ce_bitpos - UE_MIN_BITPOS_UPPER),
+ priv->baseaddr + OCM_FID1_OFST);
+ writel(0, priv->baseaddr + OCM_FID0_OFST);
+ }
+
+ write_fault_count(priv);
+
+ return count;
+}
+
+static const struct file_operations inject_ce_fops = {
+ .open = simple_open,
+ .write = inject_ce_write,
+ .llseek = generic_file_llseek,
+};
+
+/*
+ * To get the Uncorrectable Error injected, the following steps are needed:
+ * - Setup the optional Fault Injection Count:
+ * echo <fault_count val> > /sys/kernel/debug/edac/ocm/inject_fault_count
+ * - Write the Uncorrectable Error bit position values:
+ * echo <bit_pos0 val>,<bit_pos1 val> > /sys/kernel/debug/edac/ocm/inject_ue_bitpos
+ */
+static ssize_t inject_ue_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 edac_priv *priv = edac_dev->pvt_info;
+ char buf[6], *pbuf, *token[2];
+ u64 ue_bitpos;
+ int i, ret;
+ u8 len;
+
+ if (!data)
+ return -EFAULT;
+
+ len = min_t(size_t, count, sizeof(buf));
+ if (copy_from_user(buf, data, len))
+ return -EFAULT;
+
+ buf[len] = '\0';
+ pbuf = &buf[0];
+ for (i = 0; i < OCM_NUM_UE_BITPOS; i++)
+ token[i] = strsep(&pbuf, ",");
+
+ ret = kstrtou8(token[0], 0, &priv->ue_bitpos[0]);
+ if (ret)
+ return ret;
+
+ ret = kstrtou8(token[1], 0, &priv->ue_bitpos[1]);
+ if (ret)
+ return ret;
+
+ ue_bitpos = BIT(priv->ue_bitpos[0]) | BIT(priv->ue_bitpos[1]);
+
+ if (priv->ue_bitpos[0] == priv->ue_bitpos[1]) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "Bit positions should not be equal\n");
+ return -EINVAL;
+ }
+
+ if (priv->ue_bitpos[0] > UE_MAX_BITPOS_UPPER ||
+ priv->ue_bitpos[1] > UE_MAX_BITPOS_UPPER)
+ return -EINVAL;
+
+ writel((u32)ue_bitpos, priv->baseaddr + OCM_FID0_OFST);
+ writel((u32)(ue_bitpos >> 32), priv->baseaddr + OCM_FID1_OFST);
+
+ write_fault_count(priv);
+
+ return count;
+}
+
+static const struct file_operations inject_ue_fops = {
+ .open = simple_open,
+ .write = inject_ue_write,
+ .llseek = generic_file_llseek,
+};
+
+static void setup_debugfs(struct edac_device_ctl_info *edac_dev)
+{
+ struct edac_priv *priv = edac_dev->pvt_info;
+
+ priv->debugfs_dir = edac_debugfs_create_dir("ocm");
+ if (!priv->debugfs_dir)
+ return;
+
+ edac_debugfs_create_x32("inject_fault_count", 0644, priv->debugfs_dir,
+ &priv->fault_injection_cnt);
+ edac_debugfs_create_file("inject_ue_bitpos", 0644, priv->debugfs_dir,
+ edac_dev, &inject_ue_fops);
+ edac_debugfs_create_file("inject_ce_bitpos", 0644, priv->debugfs_dir,
+ edac_dev, &inject_ce_fops);
+}
+#endif
+
+static int edac_probe(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *dci;
+ struct edac_priv *priv;
+ void __iomem *baseaddr;
+ struct resource *res;
+ int irq, ret;
+
+ baseaddr = devm_platform_get_and_ioremap_resource(pdev, 0, &res);
+ if (IS_ERR(baseaddr))
+ return PTR_ERR(baseaddr);
+
+ if (!get_eccstate(baseaddr)) {
+ edac_printk(KERN_INFO, EDAC_DEVICE, "ECC not enabled\n");
+ return -ENXIO;
+ }
+
+ dci = edac_device_alloc_ctl_info(sizeof(*priv), ZYNQMP_OCM_EDAC_STRING,
+ 1, ZYNQMP_OCM_EDAC_STRING, 1, 0, NULL, 0,
+ edac_device_alloc_index());
+ if (!dci)
+ return -ENOMEM;
+
+ priv = dci->pvt_info;
+ platform_set_drvdata(pdev, dci);
+ dci->dev = &pdev->dev;
+ priv->baseaddr = baseaddr;
+ dci->mod_name = pdev->dev.driver->name;
+ dci->ctl_name = ZYNQMP_OCM_EDAC_STRING;
+ dci->dev_name = dev_name(&pdev->dev);
+
+ irq = platform_get_irq(pdev, 0);
+ if (irq < 0) {
+ ret = irq;
+ goto free_dev_ctl;
+ }
+
+ ret = devm_request_irq(&pdev->dev, irq, intr_handler, 0,
+ dev_name(&pdev->dev), dci);
+ if (ret) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "Failed to request Irq\n");
+ goto free_dev_ctl;
+ }
+
+ /* Enable UE, CE interrupts */
+ writel((OCM_CEINTR_MASK | OCM_UEINTR_MASK), priv->baseaddr + OCM_IEN_OFST);
+
+ if (IS_ENABLED(CONFIG_EDAC_DEBUG))
+ setup_debugfs(dci);
+
+ ret = edac_device_add_device(dci);
+ if (ret)
+ goto free_dev_ctl;
+
+ return 0;
+
+free_dev_ctl:
+ edac_device_free_ctl_info(dci);
+
+ return ret;
+}
+
+static int edac_remove(struct platform_device *pdev)
+{
+ struct edac_device_ctl_info *dci = platform_get_drvdata(pdev);
+ struct edac_priv *priv = dci->pvt_info;
+
+ /* Disable UE, CE interrupts */
+ writel((OCM_CEINTR_MASK | OCM_UEINTR_MASK), priv->baseaddr + OCM_IDS_OFST);
+
+ if (IS_ENABLED(CONFIG_EDAC_DEBUG))
+ debugfs_remove_recursive(priv->debugfs_dir);
+
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(dci);
+
+ return 0;
+}
+
+static const struct of_device_id zynqmp_ocm_edac_match[] = {
+ { .compatible = "xlnx,zynqmp-ocmc-1.0"},
+ { /* end of table */ }
+};
+
+MODULE_DEVICE_TABLE(of, zynqmp_ocm_edac_match);
+
+static struct platform_driver zynqmp_ocm_edac_driver = {
+ .driver = {
+ .name = "zynqmp-ocm-edac",
+ .of_match_table = zynqmp_ocm_edac_match,
+ },
+ .probe = edac_probe,
+ .remove = edac_remove,
+};
+
+module_platform_driver(zynqmp_ocm_edac_driver);
+
+MODULE_AUTHOR("Advanced Micro Devices, Inc");
+MODULE_DESCRIPTION("Xilinx ZynqMP OCM ECC driver");
+MODULE_LICENSE("GPL");
--
2.25.1

2023-01-04 09:36:58

by Sai Krishna Potthuri

[permalink] [raw]
Subject: [PATCH v7 1/2] dt-bindings: edac: Add bindings for Xilinx ZynqMP OCM

From: Shubhrajyoti Datta <[email protected]>

Add bindings for Xilinx ZynqMP OCM controller.

Signed-off-by: Shubhrajyoti Datta <[email protected]>
Co-developed-by: Sai Krishna Potthuri <[email protected]>
Signed-off-by: Sai Krishna Potthuri <[email protected]>
Reviewed-by: Krzysztof Kozlowski <[email protected]>
---
.../xlnx,zynqmp-ocmc-1.0.yaml | 45 +++++++++++++++++++
1 file changed, 45 insertions(+)
create mode 100644 Documentation/devicetree/bindings/memory-controllers/xlnx,zynqmp-ocmc-1.0.yaml

diff --git a/Documentation/devicetree/bindings/memory-controllers/xlnx,zynqmp-ocmc-1.0.yaml b/Documentation/devicetree/bindings/memory-controllers/xlnx,zynqmp-ocmc-1.0.yaml
new file mode 100644
index 000000000000..ca9fc747bf4f
--- /dev/null
+++ b/Documentation/devicetree/bindings/memory-controllers/xlnx,zynqmp-ocmc-1.0.yaml
@@ -0,0 +1,45 @@
+# SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/memory-controllers/xlnx,zynqmp-ocmc-1.0.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Xilinx Zynqmp OCM(On-Chip Memory) Controller
+
+maintainers:
+ - Shubhrajyoti Datta <[email protected]>
+ - Sai Krishna Potthuri <[email protected]>
+
+description: |
+ The OCM supports 64-bit wide ECC functionality to detect multi-bit errors
+ and recover from a single-bit memory fault.On a write, if all bytes are
+ being written, the ECC is generated and written into the ECC RAM along with
+ the write-data that is written into the data RAM. If one or more bytes are
+ not written, then the read operation results in an correctable error or
+ uncorrectable error.
+
+properties:
+ compatible:
+ const: xlnx,zynqmp-ocmc-1.0
+
+ reg:
+ maxItems: 1
+
+ interrupts:
+ maxItems: 1
+
+required:
+ - compatible
+ - reg
+ - interrupts
+
+additionalProperties: false
+
+examples:
+ - |
+ #include <dt-bindings/interrupt-controller/arm-gic.h>
+ memory-controller@ff960000 {
+ compatible = "xlnx,zynqmp-ocmc-1.0";
+ reg = <0xff960000 0x1000>;
+ interrupts = <GIC_SPI 10 IRQ_TYPE_LEVEL_HIGH>;
+ };
--
2.25.1

2023-01-08 17:01:00

by Borislav Petkov

[permalink] [raw]
Subject: Re: [PATCH v7 2/2] EDAC/zynqmp: Add EDAC support for Xilinx ZynqMP OCM

On Wed, Jan 04, 2023 at 02:15:12PM +0530, Sai Krishna Potthuri wrote:
> Add EDAC support for Xilinx ZynqMP OCM Controller, so this driver
> reports CE and UE errors upon interrupt generation, and also creates UE/CE
> debugfs entries for error injection.
> On Xilinx ZynqMP platform, both OCM Controller driver(zynqmp_edac) and
> DDR Memory Controller driver(synopsys_edac) co-exist which means both
> can be loaded at a time. This scenario is tested on Xilinx ZynqMP
> platform.
>
> Fix following issue reported by the robot.
> "MAINTAINERS references a file that doesn't exist:
> Documentation/devicetree/bindings/edac/xlnx,zynqmp-ocmc.yaml"
>
> Co-developed-by: Shubhrajyoti Datta <[email protected]>
> Signed-off-by: Shubhrajyoti Datta <[email protected]>
> Signed-off-by: Sai Krishna Potthuri <[email protected]>
> Reported-by: kernel test robot <[email protected]>
> ---
> MAINTAINERS | 7 +
> drivers/edac/Kconfig | 9 +
> drivers/edac/Makefile | 1 +
> drivers/edac/zynqmp_edac.c | 465 +++++++++++++++++++++++++++++++++++++
> 4 files changed, 482 insertions(+)
> create mode 100644 drivers/edac/zynqmp_edac.c

Some touchups ontop, see below.

I had to revert back to the #ifdeffery because IS_ENABLED doesn't prevent the
compiler from looking inside the conditional...

Anyway, inter-diff below. Holler if something's still amiss.

Thx.

---

diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
-index 58ab63642e72..7944e40c67da 100644
+index 4cfdefbd744d..68f576700911 100644
--- a/drivers/edac/Kconfig
+++ b/drivers/edac/Kconfig
-@@ -539,4 +539,13 @@ config EDAC_DMC520
+@@ -542,4 +542,12 @@ config EDAC_DMC520
Support for error detection and correction on the
SoCs with ARM DMC-520 DRAM controller.

-+config EDAC_ZYNQMP_OCM
++config EDAC_ZYNQMP
+ tristate "Xilinx ZynqMP OCM Controller"
+ depends on ARCH_ZYNQMP || COMPILE_TEST
+ help
+ This driver supports error detection and correction for the
-+ Xilinx ZynqMP OCM (On Chip Memory) controller.
-+ This driver can also be built as a module. If so, the module
-+ will be called zynqmp_ocm_edac.
++ Xilinx ZynqMP OCM (On Chip Memory) controller. It can also be
++ built as a module. In that case it will be called zynqmp_edac.
+
endif # EDAC
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
-index 2d1641a27a28..47cbda06d7b0 100644
+index 2d1641a27a28..9b025c5b3061 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -84,3 +84,4 @@ obj-$(CONFIG_EDAC_QCOM) += qcom_edac.o
obj-$(CONFIG_EDAC_ASPEED) += aspeed_edac.o
obj-$(CONFIG_EDAC_BLUEFIELD) += bluefield_edac.o
obj-$(CONFIG_EDAC_DMC520) += dmc520_edac.o
-+obj-$(CONFIG_EDAC_ZYNQMP_OCM) += zynqmp_edac.o
++obj-$(CONFIG_EDAC_ZYNQMP) += zynqmp_edac.o
diff --git a/drivers/edac/zynqmp_edac.c b/drivers/edac/zynqmp_edac.c
new file mode 100644
-index 000000000000..69069028457b
+index 000000000000..b11f1157d4bb
--- /dev/null
+++ b/drivers/edac/zynqmp_edac.c
-@@ -0,0 +1,465 @@
+@@ -0,0 +1,469 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Xilinx ZynqMP OCM ECC Driver
@@ -220,12 +224,14 @@ index 000000000000..69069028457b
+ p->ceinfo.fault_hi = readl(base + CE_FFD1_OFST);
+ p->ceinfo.addr = (OCM_BASEVAL | readl(base + CE_FFA_OFST));
+ writel(ECC_CTRL_CLR_CE_ERR, base + OCM_ISR_OFST);
-+ } else {
++ } else if (mask & OCM_UEINTR_MASK) {
+ p->ue_cnt++;
+ p->ueinfo.fault_lo = readl(base + UE_FFD0_OFST);
+ p->ueinfo.fault_hi = readl(base + UE_FFD1_OFST);
+ p->ueinfo.addr = (OCM_BASEVAL | readl(base + UE_FFA_OFST));
+ writel(ECC_CTRL_CLR_UE_ERR, base + OCM_ISR_OFST);
++ } else {
++ WARN_ON_ONCE(1);
+ }
+}
+
@@ -234,7 +240,7 @@ index 000000000000..69069028457b
+ * @dci: Pointer to the EDAC device instance
+ * @p: Pointer to the OCM ECC status structure
+ *
-+ * Handles the ECC correctable and uncorrectable error.
++ * Handles correctable and uncorrectable errors.
+ */
+static void handle_error(struct edac_device_ctl_info *dci, struct ecc_status *p)
+{
@@ -275,7 +281,7 @@ index 000000000000..69069028457b
+
+ regval = readl(priv->baseaddr + OCM_ISR_OFST);
+ if (!(regval & (OCM_CEINTR_MASK | OCM_UEINTR_MASK))) {
-+ WARN_ONCE(1, "Unhandled IRQ%d.", irq);
++ WARN_ONCE(1, "Unhandled IRQ%d, ISR: 0x%x", irq, regval);
+ return IRQ_NONE;
+ }
+
@@ -403,16 +409,16 @@ index 000000000000..69069028457b
+ if (ret)
+ return ret;
+
-+ ue_bitpos = BIT(priv->ue_bitpos[0]) | BIT(priv->ue_bitpos[1]);
++ if (priv->ue_bitpos[0] > UE_MAX_BITPOS_UPPER ||
++ priv->ue_bitpos[1] > UE_MAX_BITPOS_UPPER)
++ return -EINVAL;
+
+ if (priv->ue_bitpos[0] == priv->ue_bitpos[1]) {
+ edac_printk(KERN_ERR, EDAC_DEVICE, "Bit positions should not be equal\n");
+ return -EINVAL;
+ }
+
-+ if (priv->ue_bitpos[0] > UE_MAX_BITPOS_UPPER ||
-+ priv->ue_bitpos[1] > UE_MAX_BITPOS_UPPER)
-+ return -EINVAL;
++ ue_bitpos = BIT(priv->ue_bitpos[0]) | BIT(priv->ue_bitpos[1]);
+
+ writel((u32)ue_bitpos, priv->baseaddr + OCM_FID0_OFST);
+ writel((u32)(ue_bitpos >> 32), priv->baseaddr + OCM_FID1_OFST);
@@ -492,8 +498,9 @@ index 000000000000..69069028457b
+ /* Enable UE, CE interrupts */
+ writel((OCM_CEINTR_MASK | OCM_UEINTR_MASK), priv->baseaddr + OCM_IEN_OFST);
+
-+ if (IS_ENABLED(CONFIG_EDAC_DEBUG))
-+ setup_debugfs(dci);
++#ifdef CONFIG_EDAC_DEBUG
++ setup_debugfs(dci);
++#endif
+
+ ret = edac_device_add_device(dci);
+ if (ret)
@@ -515,8 +522,9 @@ index 000000000000..69069028457b
+ /* Disable UE, CE interrupts */
+ writel((OCM_CEINTR_MASK | OCM_UEINTR_MASK), priv->baseaddr + OCM_IDS_OFST);
+
-+ if (IS_ENABLED(CONFIG_EDAC_DEBUG))
-+ debugfs_remove_recursive(priv->debugfs_dir);
++#ifdef CONFIG_EDAC_DEBUG
++ debugfs_remove_recursive(priv->debugfs_dir);
++#endif
+
+ edac_device_del_device(&pdev->dev);
+ edac_device_free_ctl_info(dci);

--
Regards/Gruss,
Boris.

https://people.kernel.org/tglx/notes-about-netiquette

2023-01-09 06:31:00

by Sai Krishna Potthuri

[permalink] [raw]
Subject: RE: [PATCH v7 2/2] EDAC/zynqmp: Add EDAC support for Xilinx ZynqMP OCM

Hi Boris,

> -----Original Message-----
> From: Borislav Petkov <[email protected]>
> Sent: Sunday, January 8, 2023 10:28 PM
> To: Potthuri, Sai Krishna <[email protected]>
> Cc: Rob Herring <[email protected]>; Krzysztof Kozlowski
> <[email protected]>; Michal Simek
> <[email protected]>; Mauro Carvalho Chehab
> <[email protected]>; Tony Luck <[email protected]>; James Morse
> <[email protected]>; Robert Richter <[email protected]>;
> [email protected]; [email protected]; linux-
> [email protected]; [email protected];
> [email protected]; git (AMD-Xilinx) <[email protected]>; Datta,
> Shubhrajyoti <[email protected]>; kernel test robot
> <[email protected]>
> Subject: Re: [PATCH v7 2/2] EDAC/zynqmp: Add EDAC support for Xilinx
> ZynqMP OCM
>
> On Wed, Jan 04, 2023 at 02:15:12PM +0530, Sai Krishna Potthuri wrote:
> > Add EDAC support for Xilinx ZynqMP OCM Controller, so this driver
> > reports CE and UE errors upon interrupt generation, and also creates
> > UE/CE debugfs entries for error injection.
> > On Xilinx ZynqMP platform, both OCM Controller driver(zynqmp_edac) and
> > DDR Memory Controller driver(synopsys_edac) co-exist which means both
> > can be loaded at a time. This scenario is tested on Xilinx ZynqMP
> > platform.
> >
> > Fix following issue reported by the robot.
> > "MAINTAINERS references a file that doesn't exist:
> > Documentation/devicetree/bindings/edac/xlnx,zynqmp-ocmc.yaml"
> >
> > Co-developed-by: Shubhrajyoti Datta <[email protected]>
> > Signed-off-by: Shubhrajyoti Datta <[email protected]>
> > Signed-off-by: Sai Krishna Potthuri <[email protected]>
> > Reported-by: kernel test robot <[email protected]>
> > ---
> > MAINTAINERS | 7 +
> > drivers/edac/Kconfig | 9 +
> > drivers/edac/Makefile | 1 +
> > drivers/edac/zynqmp_edac.c | 465
> > +++++++++++++++++++++++++++++++++++++
> > 4 files changed, 482 insertions(+)
> > create mode 100644 drivers/edac/zynqmp_edac.c
>
> Some touchups ontop, see below.
>
> I had to revert back to the #ifdeffery because IS_ENABLED doesn't prevent
> the compiler from looking inside the conditional...
>
> Anyway, inter-diff below. Holler if something's still amiss.
>
> Thx.
>
> ---
>
> diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig -index
> 58ab63642e72..7944e40c67da 100644
> +index 4cfdefbd744d..68f576700911 100644
> --- a/drivers/edac/Kconfig
> +++ b/drivers/edac/Kconfig
> -@@ -539,4 +539,13 @@ config EDAC_DMC520
> +@@ -542,4 +542,12 @@ config EDAC_DMC520
> Support for error detection and correction on the
> SoCs with ARM DMC-520 DRAM controller.
>
> -+config EDAC_ZYNQMP_OCM
> ++config EDAC_ZYNQMP
> + tristate "Xilinx ZynqMP OCM Controller"
> + depends on ARCH_ZYNQMP || COMPILE_TEST
> + help
> + This driver supports error detection and correction for the
> -+ Xilinx ZynqMP OCM (On Chip Memory) controller.
> -+ This driver can also be built as a module. If so, the module
> -+ will be called zynqmp_ocm_edac.
> ++ Xilinx ZynqMP OCM (On Chip Memory) controller. It can also be
> ++ built as a module. In that case it will be called zynqmp_edac.
> +
> endif # EDAC
> diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile -index
> 2d1641a27a28..47cbda06d7b0 100644
> +index 2d1641a27a28..9b025c5b3061 100644
> --- a/drivers/edac/Makefile
> +++ b/drivers/edac/Makefile
> @@ -84,3 +84,4 @@ obj-$(CONFIG_EDAC_QCOM) +=
> qcom_edac.o
> obj-$(CONFIG_EDAC_ASPEED) += aspeed_edac.o
> obj-$(CONFIG_EDAC_BLUEFIELD) += bluefield_edac.o
> obj-$(CONFIG_EDAC_DMC520) += dmc520_edac.o
> -+obj-$(CONFIG_EDAC_ZYNQMP_OCM) += zynqmp_edac.o
> ++obj-$(CONFIG_EDAC_ZYNQMP) += zynqmp_edac.o
> diff --git a/drivers/edac/zynqmp_edac.c b/drivers/edac/zynqmp_edac.c
> new file mode 100644 -index 000000000000..69069028457b
> +index 000000000000..b11f1157d4bb
> --- /dev/null
> +++ b/drivers/edac/zynqmp_edac.c
> -@@ -0,0 +1,465 @@
> +@@ -0,0 +1,469 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Xilinx ZynqMP OCM ECC Driver
> @@ -220,12 +224,14 @@ index 000000000000..69069028457b
> + p->ceinfo.fault_hi = readl(base + CE_FFD1_OFST);
> + p->ceinfo.addr = (OCM_BASEVAL | readl(base +
> CE_FFA_OFST));
> + writel(ECC_CTRL_CLR_CE_ERR, base + OCM_ISR_OFST);
> -+ } else {
> ++ } else if (mask & OCM_UEINTR_MASK) {
> + p->ue_cnt++;
> + p->ueinfo.fault_lo = readl(base + UE_FFD0_OFST);
> + p->ueinfo.fault_hi = readl(base + UE_FFD1_OFST);
> + p->ueinfo.addr = (OCM_BASEVAL | readl(base +
> UE_FFA_OFST));
> + writel(ECC_CTRL_CLR_UE_ERR, base + OCM_ISR_OFST);
> ++ } else {
> ++ WARN_ON_ONCE(1);
> + }
As we are raising a warning message in intr_handler() if the flagged interrupt
is not UE or CE and we return from there, so do we really need else if{} and
WARN_ON_ONCE() in else{} here?

Regards
Sai Krishna

2023-01-09 10:47:14

by Borislav Petkov

[permalink] [raw]
Subject: Re: [PATCH v7 2/2] EDAC/zynqmp: Add EDAC support for Xilinx ZynqMP OCM

On Mon, Jan 09, 2023 at 06:09:02AM +0000, Potthuri, Sai Krishna wrote:
> As we are raising a warning message in intr_handler() if the flagged interrupt
> is not UE or CE and we return from there, so do we really need else if{} and
> WARN_ON_ONCE() in else{} here?

Yeah, ok, lemme remove it and queue the thing.

Thx.

--
Regards/Gruss,
Boris.

https://people.kernel.org/tglx/notes-about-netiquette