Message-Id: <1552934340-10103-1-git-send-email-Junhan@mellanox.com>
Date:   Mon, 18 Mar 2019 14:39:00 -0400
From:   Junhan Zhou <Junhan@...lanox.com>
To:     Borislav Petkov <bp@...en8.de>,
        Mauro Carvalho Chehab <mchehab@...nel.org>
Cc:     Junhan Zhou <Junhan@...lanox.com>, Liming Sun <lsun@...lanox.com>,
        linux-edac@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: [PATCH v2] EDAC, mellanox: Add ECC support for BlueField DDR4

Add ECC support for Mellanox BlueField SoC DDR controller.
This requires SMC to the running Arm Trusted Firmware to report
what is the current memory configuration.

Signed-off-by: Junhan Zhou <Junhan@...lanox.com>
---
 MAINTAINERS                   |   5 +
 drivers/edac/Kconfig          |   7 +
 drivers/edac/Makefile         |   1 +
 drivers/edac/bluefield_edac.c | 396 ++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 409 insertions(+)
 create mode 100644 drivers/edac/bluefield_edac.c

diff --git a/MAINTAINERS b/MAINTAINERS
index e17ebf7..394d6c9 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -5539,6 +5539,11 @@ S:	Supported
 F:	drivers/edac/aspeed_edac.c
 F:	Documentation/devicetree/bindings/edac/aspeed-sdram-edac.txt
 
+EDAC-BLUEFIELD
+M:	Junhan Zhou <Junhan@...lanox.com>
+S:	Supported
+F:	drivers/edac/bluefield_edac.c
+
 EDAC-CALXEDA
 M:	Robert Richter <rric@...nel.org>
 L:	linux-edac@...r.kernel.org
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
index 47eb4d1..dfb36f2 100644
--- a/drivers/edac/Kconfig
+++ b/drivers/edac/Kconfig
@@ -504,4 +504,11 @@ config EDAC_ASPEED
 	  First, ECC must be configured in the bootloader. Then, this driver
 	  will expose error counters via the EDAC kernel framework.
 
+config EDAC_BLUEFIELD
+	tristate "Mellanox BlueField Memory ECC"
+	depends on (MELLANOX_PLATFORM && ARM64 && ACPI) || COMPILE_TEST
+	help
+	  Support for error detection and correction on the
+	  Mellanox BlueField SoCs.
+
 endif # EDAC
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
index 89ad4a84..0294a67 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -84,3 +84,4 @@ obj-$(CONFIG_EDAC_XGENE)		+= xgene_edac.o
 obj-$(CONFIG_EDAC_TI)			+= ti_edac.o
 obj-$(CONFIG_EDAC_QCOM)			+= qcom_edac.o
 obj-$(CONFIG_EDAC_ASPEED)		+= aspeed_edac.o
+obj-$(CONFIG_EDAC_BLUEFIELD)		+= bluefield_edac.o
diff --git a/drivers/edac/bluefield_edac.c b/drivers/edac/bluefield_edac.c
new file mode 100644
index 0000000..981f419
--- /dev/null
+++ b/drivers/edac/bluefield_edac.c
@@ -0,0 +1,396 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Bluefield-specific EDAC driver.
+ *
+ * Copyright (c) 2019 Mellanox Technologies.
+ */
+
+#include <linux/acpi.h>
+#include <linux/arm-smccc.h>
+#include <linux/edac.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+
+#include "edac_module.h"
+
+#define DRIVER_NAME		"bluefield-edac"
+
+/*
+ * Mellanox BlueField EMI (External Memory Interface) register definitions.
+ * Registers which have individual bitfields have a union defining the
+ * bitfields following the address define.
+ */
+
+#define MLXBF_EMI_DRAM_ECC_COUNT 0x340
+
+union mlxbf_emi_dram_ecc_count {
+	struct {
+		u32 single_error_count : 16;
+		u32 double_error_count : 16;
+	};
+
+	u32 word;
+};
+
+#define MLXBF_EMI_DRAM_ECC_ERROR 0x348
+
+union mlxbf_emi_dram_ecc_error {
+	struct {
+		u32 dram_ecc_single : 1;
+		u32 __reserved_0    : 15;
+		u32 dram_ecc_double : 1;
+		u32 __reserved_1    : 15;
+	};
+
+	u32 word;
+};
+
+#define MLXBF_EMI_DRAM_ECC_LATCH_SELECT 0x354
+
+union mlxbf_emi_dram_ecc_latch_select {
+	struct {
+		u32 dram_ecc_first : 1;
+		u32 __reserved_0   : 15;
+		u32 edge_sel       : 4;
+		u32 __reserved_1   : 4;
+		u32 start          : 1;
+		u32 __reserved_2   : 7;
+	};
+
+	u32 word;
+};
+
+#define MLXBF_EMI_DRAM_ERR_ADDR_0 0x358
+
+#define MLXBF_EMI_DRAM_ERR_ADDR_1 0x37c
+
+#define MLXBF_EMI_DRAM_SYNDROM 0x35c
+
+union mlxbf_emi_dram_syndrom {
+	struct {
+		u32 derr         : 1;
+		u32 serr         : 1;
+		u32 __reserved_0 : 14;
+		/* ECC syndrome (error bit according to the Hamming code). */
+		u32 syndrom      : 10;
+		u32 __reserved_1 : 6;
+	};
+
+	u32 word;
+};
+
+#define MLXBF_EMI_DRAM_ADDITIONAL_INFO_0 0x364
+
+union mlxbf_emi_dram_additional_info_0 {
+	struct {
+		u32 err_bank     : 4;
+		u32 err_lrank    : 2;
+		u32 __reserved_0 : 2;
+		u32 err_prank    : 2;
+		u32 __reserved_1 : 6;
+		u32 err_edge     : 8;
+		u32 __reserved_2 : 8;
+	};
+
+	u32 word;
+};
+
+#define MLXBF_EDAC_DIMM_PER_MC		2
+#define MLXBF_EDAC_ERROR_GRAIN		8
+
+/*
+ * Request MLNX_SIP_GET_DIMM_INFO
+ *
+ * Retrieve information about DIMM on a certain slot.
+ *
+ * Call register usage:
+ * a0: MLNX_SIP_GET_DIMM_INFO
+ * a1: (Memory controller index) << 16 | (Dimm index in memory controller)
+ * a2-7: not used.
+ *
+ * Return status:
+ * a0: dimm_info_smc union defined below describing the DIMM.
+ * a1-3: not used.
+ */
+#define MLNX_SIP_GET_DIMM_INFO		0x82000008
+
+/* Format for the SMC response about the memory information */
+union dimm_info_smc {
+	struct {
+		unsigned long size_GB : 16;
+		unsigned long is_rdimm : 1;
+		unsigned long is_lrdimm : 1;
+		unsigned long is_nvdimm : 1;
+		unsigned long __reserved0 : 2;
+		unsigned long ranks : 3;
+		unsigned long package_X : 8;	/* Bits per memory package */
+		unsigned long __reserved1 : 32;
+	};
+	unsigned long val;
+};
+
+struct bluefield_edac_priv {
+	int dimm_ranks[MLXBF_EDAC_DIMM_PER_MC];
+	void __iomem *emi_base;
+};
+
+static unsigned long smc_call1(unsigned long smc_op, unsigned long smc_arg)
+{
+	struct arm_smccc_res res;
+
+	arm_smccc_smc(smc_op, smc_arg, 0, 0, 0, 0, 0, 0, &res);
+
+	return res.a0;
+}
+
+/*
+ * Gather the ECC information from the External Memory Interface registers
+ * and report it to the edac handler.
+ */
+static void bluefield_gather_report_ecc(struct mem_ctl_info *mci,
+					int error_cnt,
+					int is_single_ecc)
+{
+	struct bluefield_edac_priv *priv = mci->pvt_info;
+	union mlxbf_emi_dram_additional_info_0 edai0;
+	union mlxbf_emi_dram_ecc_latch_select edels;
+	union mlxbf_emi_dram_syndrom eds;
+	enum hw_event_mc_err_type ecc_type;
+	unsigned long ecc_dimm_addr;
+	int ecc_dimm;
+	u32 edea0;
+	u32 edea1;
+
+	ecc_type = is_single_ecc ? HW_EVENT_ERR_CORRECTED :
+				   HW_EVENT_ERR_UNCORRECTED;
+
+	/*
+	 * Tell the External Memory Interface to populate the relevant
+	 * registers with information about the last ECC error occurrence.
+	 */
+	edels.word = 0;
+	edels.start = 1;
+	writel(edels.word, priv->emi_base + MLXBF_EMI_DRAM_ECC_LATCH_SELECT);
+
+	/*
+	 * Verify that the ECC reported info in the registers is of the
+	 * same type as the one asked to report. If not, just report the
+	 * error without the detailed information.
+	 */
+	eds.word = readl(priv->emi_base + MLXBF_EMI_DRAM_SYNDROM);
+
+	if ((is_single_ecc && !eds.serr) || (!is_single_ecc && !eds.derr)) {
+		edac_mc_handle_error(ecc_type, mci, error_cnt, 0, 0, 0,
+				     0, 0, -1, mci->ctl_name, "");
+		return;
+	}
+
+	edai0.word = readl(priv->emi_base + MLXBF_EMI_DRAM_ADDITIONAL_INFO_0);
+
+	ecc_dimm = edai0.err_prank >= 2 && priv->dimm_ranks[0] <= 2;
+
+	edea0 = readl(priv->emi_base + MLXBF_EMI_DRAM_ERR_ADDR_0);
+	edea1 = readl(priv->emi_base + MLXBF_EMI_DRAM_ERR_ADDR_1);
+
+	ecc_dimm_addr = ((unsigned long)edea1 << 32) | edea0;
+
+	edac_mc_handle_error(ecc_type, mci, error_cnt,
+			     ecc_dimm_addr / 0x1000, ecc_dimm_addr & 0xfff,
+			     eds.syndrom, ecc_dimm, 0, 0, mci->ctl_name, "");
+}
+
+static void bluefield_edac_check(struct mem_ctl_info *mci)
+{
+	union mlxbf_emi_dram_ecc_error edee = { .word = 0 };
+	struct bluefield_edac_priv *priv = mci->pvt_info;
+	union mlxbf_emi_dram_ecc_count edec;
+
+	/*
+	 * The memory controller might not be initialized by the firmware
+	 * when there isn't memory, which may lead to bad register readings.
+	 */
+	if (mci->edac_cap == EDAC_FLAG_NONE)
+		return;
+
+	edec.word = readl(priv->emi_base + MLXBF_EMI_DRAM_ECC_COUNT);
+
+	if (edec.single_error_count) {
+		edee.dram_ecc_single = 1;
+
+		bluefield_gather_report_ecc(mci, edec.single_error_count, 1);
+	}
+
+	if (edec.double_error_count) {
+		edee.dram_ecc_double = 1;
+
+		bluefield_gather_report_ecc(mci, edec.double_error_count, 0);
+	}
+
+	/* Write to clear reported errors. */
+	if (edec.word)
+		writel(edee.word, priv->emi_base + MLXBF_EMI_DRAM_ECC_ERROR);
+}
+
+/* Initialize the DIMMs information for the given memory controller. */
+static void bluefield_edac_init_dimms(struct mem_ctl_info *mci)
+{
+	struct bluefield_edac_priv *priv = mci->pvt_info;
+	int mem_ctrl_idx = mci->mc_idx;
+	union dimm_info_smc smc_info;
+	struct dimm_info *dimm;
+	unsigned long smc_arg;
+	int is_empty = 1;
+	int i;
+
+	for (i = 0; i < MLXBF_EDAC_DIMM_PER_MC; i++) {
+		dimm = mci->dimms[i];
+
+		smc_arg = mem_ctrl_idx << 16 | i;
+		smc_info = (union dimm_info_smc) {
+			.val = smc_call1(MLNX_SIP_GET_DIMM_INFO, smc_arg),
+		};
+
+		if (!smc_info.size_GB) {
+			dimm->mtype = MEM_EMPTY;
+			continue;
+		}
+
+		is_empty = 0;
+
+		dimm->edac_mode = EDAC_SECDED;
+
+		if (smc_info.is_nvdimm)
+			dimm->mtype = MEM_NVDIMM;
+		else if (smc_info.is_lrdimm)
+			dimm->mtype = MEM_LRDDR4;
+		else if (smc_info.is_rdimm)
+			dimm->mtype = MEM_RDDR4;
+		else
+			dimm->mtype = MEM_DDR4;
+
+		dimm->nr_pages = smc_info.size_GB * 256 * 1024;
+		dimm->grain = MLXBF_EDAC_ERROR_GRAIN;
+
+		/* Mem controller for BlueField only supports x4, x8 and x16 */
+		if (smc_info.package_X == 4)
+			dimm->dtype = DEV_X4;
+		else if (smc_info.package_X == 8)
+			dimm->dtype = DEV_X8;
+		else if (smc_info.package_X == 16)
+			dimm->dtype = DEV_X16;
+		else
+			dimm->dtype = DEV_UNKNOWN;
+
+		priv->dimm_ranks[i] = smc_info.ranks;
+	}
+
+	if (is_empty)
+		mci->edac_cap = EDAC_FLAG_NONE;
+	else
+		mci->edac_cap = EDAC_FLAG_SECDED;
+}
+
+static int bluefield_edac_mc_probe(struct platform_device *pdev)
+{
+	struct bluefield_edac_priv *priv;
+	struct device *dev = &pdev->dev;
+	struct edac_mc_layer layers[1];
+	struct mem_ctl_info *mci;
+	struct resource *emi_res;
+	unsigned int mc_idx;
+	int rc, ret;
+
+	/* Read the MSS (Memory SubSystem) index from ACPI table. */
+	if (device_property_read_u32(dev, "mss_number", &mc_idx)) {
+		dev_warn(dev, "bf_edac: MSS number unknown\n");
+		return -EINVAL;
+	}
+
+	emi_res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!emi_res)
+		return -EINVAL;
+
+	layers[0].type = EDAC_MC_LAYER_SLOT;
+	layers[0].size = MLXBF_EDAC_DIMM_PER_MC;
+	layers[0].is_virt_csrow = true;
+
+	mci = edac_mc_alloc(mc_idx, ARRAY_SIZE(layers), layers, sizeof(*priv));
+	if (!mci)
+		return -ENOMEM;
+
+	priv = mci->pvt_info;
+
+	priv->emi_base = devm_ioremap_resource(dev, emi_res);
+	if (IS_ERR(priv->emi_base)) {
+		dev_err(dev, "failed to map EMI IO resource\n");
+		ret = PTR_ERR(priv->emi_base);
+		goto err;
+	}
+
+	mci->pdev = dev;
+	mci->mtype_cap = MEM_FLAG_DDR4 | MEM_FLAG_RDDR4 |
+			 MEM_FLAG_LRDDR4 | MEM_FLAG_NVDIMM;
+	mci->edac_ctl_cap = EDAC_FLAG_SECDED;
+
+	mci->mod_name = DRIVER_NAME;
+	mci->ctl_name = "BlueField_Memory_Controller";
+	mci->dev_name = dev_name(dev);
+	mci->edac_check = bluefield_edac_check;
+
+	/* Initialize mci with the actual populated DIMM information. */
+	bluefield_edac_init_dimms(mci);
+
+	platform_set_drvdata(pdev, mci);
+
+	/* Register with EDAC core */
+	rc = edac_mc_add_mc(mci);
+	if (rc) {
+		dev_err(dev, "failed to register with EDAC core\n");
+		ret = rc;
+		goto err;
+	}
+
+	/* Only POLL mode supported so far. */
+	edac_op_state = EDAC_OPSTATE_POLL;
+
+	return 0;
+
+err:
+	edac_mc_free(mci);
+
+	return ret;
+
+}
+
+static int bluefield_edac_mc_remove(struct platform_device *pdev)
+{
+	struct mem_ctl_info *mci = platform_get_drvdata(pdev);
+
+	edac_mc_del_mc(&pdev->dev);
+	edac_mc_free(mci);
+
+	return 0;
+}
+
+static const struct acpi_device_id bluefield_mc_acpi_ids[] = {
+	{"MLNXBF08", 0},
+	{}
+};
+
+MODULE_DEVICE_TABLE(acpi, bluefield_mc_acpi_ids);
+
+static struct platform_driver bluefield_edac_mc_driver = {
+	.driver = {
+		.name = DRIVER_NAME,
+		.acpi_match_table = bluefield_mc_acpi_ids,
+	},
+	.probe = bluefield_edac_mc_probe,
+	.remove = bluefield_edac_mc_remove,
+};
+
+module_platform_driver(bluefield_edac_mc_driver);
+
+MODULE_DESCRIPTION("Mellanox BlueField memory edac driver");
+MODULE_AUTHOR("Mellanox Technologies");
+MODULE_LICENSE("GPL v2");
-- 
2.1.2

Powered by blists - more mailing lists