Merge branch 'edac-for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras.git

# Conflicts:
#	drivers/edac/i10nm_base.c

MAINTAINERS

@@ -13612,6 +13612,12 @@ S:	Maintained
 F:	Documentation/devicetree/bindings/thermal/loongson,ls2k-thermal.yaml
 F:	drivers/thermal/loongson2_thermal.c
 
+LOONGSON EDAC DRIVER
+M:	Zhao Qunqin <zhaoqunqin@loongson.cn>
+L:	linux-edac@vger.kernel.org
+S:	Maintained
+F:	drivers/edac/loongson_edac.c
+
 LSILOGIC MPT FUSION DRIVERS (FC/SAS/SPI)
 M:	Sathya Prakash <sathya.prakash@broadcom.com>
 M:	Sreekanth Reddy <sreekanth.reddy@broadcom.com>

arch/loongarch/Kconfig

@@ -81,6 +81,7 @@ config LOONGARCH
 	select BUILDTIME_TABLE_SORT
 	select COMMON_CLK
 	select CPU_PM
+	select EDAC_SUPPORT
 	select EFI
 	select GENERIC_CLOCKEVENTS
 	select GENERIC_CMOS_UPDATE

drivers/edac/Kconfig

@@ -547,5 +547,13 @@ config EDAC_VERSAL
 	  Support injecting both correctable and uncorrectable errors
 	  for debugging purposes.
 
+config EDAC_LOONGSON
+	tristate "Loongson Memory Controller"
+	depends on LOONGARCH && ACPI
+	help
+	  Support for error detection and correction on the Loongson
+	  family memory controller. This driver reports single bit
+	  errors (CE) only. Loongson-3A5000/3C5000/3D5000/3A6000/3C6000
+	  are compatible.
 
 endif # EDAC

drivers/edac/Makefile

@@ -86,3 +86,4 @@ obj-$(CONFIG_EDAC_DMC520)		+= dmc520_edac.o
 obj-$(CONFIG_EDAC_NPCM)			+= npcm_edac.o
 obj-$(CONFIG_EDAC_ZYNQMP)		+= zynqmp_edac.o
 obj-$(CONFIG_EDAC_VERSAL)		+= versal_edac.o
+obj-$(CONFIG_EDAC_LOONGSON)		+= loongson_edac.o

drivers/edac/edac_mc.c

@@ -214,7 +214,7 @@ static int edac_mc_alloc_csrows(struct mem_ctl_info *mci)
 	unsigned int row, chn;
 
 	/*
-	 * Alocate and fill the csrow/channels structs
+	 * Allocate and fill the csrow/channels structs
 	 */
 	mci->csrows = kcalloc(tot_csrows, sizeof(*mci->csrows), GFP_KERNEL);
 	if (!mci->csrows)

drivers/edac/edac_mc_sysfs.c

@@ -422,7 +422,7 @@ static inline int nr_pages_per_csrow(struct csrow_info *csrow)
 	return nr_pages;
 }
 
-/* Create a CSROW object under specifed edac_mc_device */
+/* Create a CSROW object under specified edac_mc_device */
 static int edac_create_csrow_object(struct mem_ctl_info *mci,
 				    struct csrow_info *csrow, int index)
 {
@@ -449,7 +449,7 @@ static int edac_create_csrow_object(struct mem_ctl_info *mci,
 	return 0;
 }
 
-/* Create a CSROW object under specifed edac_mc_device */
+/* Create a CSROW object under specified edac_mc_device */
 static int edac_create_csrow_objects(struct mem_ctl_info *mci)
 {
 	int err, i;
@@ -636,7 +636,7 @@ static void dimm_release(struct device *dev)
 	 */
 }
 
-/* Create a DIMM object under specifed memory controller device */
+/* Create a DIMM object under specified memory controller device */
 static int edac_create_dimm_object(struct mem_ctl_info *mci,
 				   struct dimm_info *dimm)
 {

drivers/edac/i10nm_base.c

@@ -949,6 +949,7 @@ static const struct x86_cpu_id i10nm_cpuids[] = {
 	X86_MATCH_VFM(INTEL_GRANITERAPIDS_X,  &gnr_cfg),
 	X86_MATCH_VFM(INTEL_ATOM_CRESTMONT_X, &gnr_cfg),
 	X86_MATCH_VFM(INTEL_ATOM_CRESTMONT,   &gnr_cfg),
+	X86_MATCH_VFM(INTEL_ATOM_DARKMONT_X,  &gnr_cfg),
 	{}
 };
 MODULE_DEVICE_TABLE(x86cpu, i10nm_cpuids);
@@ -1011,7 +1012,7 @@ static struct notifier_block i10nm_mce_dec = {
 
 static int __init i10nm_init(void)
 {
-	u8 mc = 0, src_id = 0, node_id = 0;
+	u8 mc = 0, src_id = 0;
 	const struct x86_cpu_id *id;
 	struct res_config *cfg;
 	const char *owner;
@@ -1071,19 +1072,14 @@ static int __init i10nm_init(void)
 		if (rc < 0)
 			goto fail;
 
-		rc = skx_get_node_id(d, &node_id);
+		edac_dbg(2, "src_id = %d\n", src_id);
-		if (rc < 0)
-			goto fail;
-
-		edac_dbg(2, "src_id = %d node_id = %d\n", src_id, node_id);
 		for (i = 0; i < imc_num; i++) {
 			if (!d->imc[i].mdev)
 				continue;
 
 			d->imc[i].mc  = mc++;
 			d->imc[i].lmc = i;
-			d->imc[i].src_id  = src_id;
+			d->imc[i].src_id = src_id;
-			d->imc[i].node_id = node_id;
 			if (d->imc[i].hbm_mc) {
 				d->imc[i].chan_mmio_sz = cfg->hbm_chan_mmio_sz;
 				d->imc[i].num_channels = cfg->hbm_chan_num;

drivers/edac/i5000_edac.c

@@ -338,11 +338,11 @@ struct i5000_pvt {
 
 	u16 mir0, mir1, mir2;
 
-	u16 b0_mtr[NUM_MTRS];	/* Memory Technlogy Reg */
+	u16 b0_mtr[NUM_MTRS];	/* Memory Technology Reg */
 	u16 b0_ambpresent0;	/* Branch 0, Channel 0 */
-	u16 b0_ambpresent1;	/* Brnach 0, Channel 1 */
+	u16 b0_ambpresent1;	/* Branch 0, Channel 1 */
 
-	u16 b1_mtr[NUM_MTRS];	/* Memory Technlogy Reg */
+	u16 b1_mtr[NUM_MTRS];	/* Memory Technology Reg */
 	u16 b1_ambpresent0;	/* Branch 1, Channel 8 */
 	u16 b1_ambpresent1;	/* Branch 1, Channel 1 */
 
@@ -1210,7 +1210,7 @@ static void i5000_get_mc_regs(struct mem_ctl_info *mci)
 			&pvt->b0_ambpresent1);
 	edac_dbg(2, "\t\tAMB-Branch 0-present1 0x%x:\n", pvt->b0_ambpresent1);
 
-	/* Only if we have 2 branchs (4 channels) */
+	/* Only if we have 2 branches (4 channels) */
 	if (pvt->maxch < CHANNELS_PER_BRANCH) {
 		pvt->b1_ambpresent0 = 0;
 		pvt->b1_ambpresent1 = 0;

drivers/edac/loongson_edac.c

@@ -0,0 +1,157 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2024 Loongson Technology Corporation Limited.
+ */
+
+#include <linux/acpi.h>
+#include <linux/edac.h>
+#include <linux/init.h>
+#include <linux/io-64-nonatomic-lo-hi.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include "edac_module.h"
+
+#define ECC_CS_COUNT_REG	0x18
+
+struct loongson_edac_pvt {
+	void __iomem *ecc_base;
+
+	/*
+	 * The ECC register in this controller records the number of errors
+	 * encountered since reset and cannot be zeroed so in order to be able
+	 * to report the error count at each check, this records the previous
+	 * register state.
+	 */
+	int last_ce_count;
+};
+
+static int read_ecc(struct mem_ctl_info *mci)
+{
+	struct loongson_edac_pvt *pvt = mci->pvt_info;
+	u64 ecc;
+	int cs;
+
+	ecc = readq(pvt->ecc_base + ECC_CS_COUNT_REG);
+	/* cs0 -- cs3 */
+	cs = ecc & 0xff;
+	cs += (ecc >> 8) & 0xff;
+	cs += (ecc >> 16) & 0xff;
+	cs += (ecc >> 24) & 0xff;
+
+	return cs;
+}
+
+static void edac_check(struct mem_ctl_info *mci)
+{
+	struct loongson_edac_pvt *pvt = mci->pvt_info;
+	int new, add;
+
+	new = read_ecc(mci);
+	add = new - pvt->last_ce_count;
+	pvt->last_ce_count = new;
+	if (add <= 0)
+		return;
+
+	edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, add,
+			     0, 0, 0, 0, 0, -1, "error", "");
+}
+
+static void dimm_config_init(struct mem_ctl_info *mci)
+{
+	struct dimm_info *dimm;
+	u32 size, npages;
+
+	/* size not used */
+	size = -1;
+	npages = MiB_TO_PAGES(size);
+
+	dimm = edac_get_dimm(mci, 0, 0, 0);
+	dimm->nr_pages = npages;
+	snprintf(dimm->label, sizeof(dimm->label),
+		 "MC#%uChannel#%u_DIMM#%u", mci->mc_idx, 0, 0);
+	dimm->grain = 8;
+}
+
+static void pvt_init(struct mem_ctl_info *mci, void __iomem *vbase)
+{
+	struct loongson_edac_pvt *pvt = mci->pvt_info;
+
+	pvt->ecc_base = vbase;
+	pvt->last_ce_count = read_ecc(mci);
+}
+
+static int edac_probe(struct platform_device *pdev)
+{
+	struct edac_mc_layer layers[2];
+	struct mem_ctl_info *mci;
+	void __iomem *vbase;
+	int ret;
+
+	vbase = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(vbase))
+		return PTR_ERR(vbase);
+
+	layers[0].type = EDAC_MC_LAYER_CHANNEL;
+	layers[0].size = 1;
+	layers[0].is_virt_csrow = false;
+	layers[1].type = EDAC_MC_LAYER_SLOT;
+	layers[1].size = 1;
+	layers[1].is_virt_csrow = true;
+	mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers,
+			    sizeof(struct loongson_edac_pvt));
+	if (mci == NULL)
+		return -ENOMEM;
+
+	mci->mc_idx = edac_device_alloc_index();
+	mci->mtype_cap = MEM_FLAG_RDDR4;
+	mci->edac_ctl_cap = EDAC_FLAG_NONE;
+	mci->edac_cap = EDAC_FLAG_NONE;
+	mci->mod_name = "loongson_edac.c";
+	mci->ctl_name = "loongson_edac_ctl";
+	mci->dev_name = "loongson_edac_dev";
+	mci->ctl_page_to_phys = NULL;
+	mci->pdev = &pdev->dev;
+	mci->error_desc.grain = 8;
+	mci->edac_check = edac_check;
+
+	pvt_init(mci, vbase);
+	dimm_config_init(mci);
+
+	ret = edac_mc_add_mc(mci);
+	if (ret) {
+		edac_dbg(0, "MC: failed edac_mc_add_mc()\n");
+		edac_mc_free(mci);
+		return ret;
+	}
+	edac_op_state = EDAC_OPSTATE_POLL;
+
+	return 0;
+}
+
+static void edac_remove(struct platform_device *pdev)
+{
+	struct mem_ctl_info *mci = edac_mc_del_mc(&pdev->dev);
+
+	if (mci)
+		edac_mc_free(mci);
+}
+
+static const struct acpi_device_id loongson_edac_acpi_match[] = {
+	{"LOON0010", 0},
+	{}
+};
+MODULE_DEVICE_TABLE(acpi, loongson_edac_acpi_match);
+
+static struct platform_driver loongson_edac_driver = {
+	.probe		= edac_probe,
+	.remove		= edac_remove,
+	.driver		= {
+		.name	= "loongson-mc-edac",
+		.acpi_match_table = loongson_edac_acpi_match,
+	},
+};
+module_platform_driver(loongson_edac_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Zhao Qunqin <zhaoqunqin@loongson.cn>");
+MODULE_DESCRIPTION("EDAC driver for loongson memory controller");

drivers/edac/skx_base.c

@@ -600,7 +600,7 @@ static int __init skx_init(void)
 	const struct munit *m;
 	const char *owner;
 	int rc = 0, i, off[3] = {0xd0, 0xd4, 0xd8};
-	u8 mc = 0, src_id, node_id;
+	u8 mc = 0, src_id;
 	struct skx_dev *d;
 
 	edac_dbg(2, "\n");
@@ -650,15 +650,12 @@ static int __init skx_init(void)
 		rc = skx_get_src_id(d, 0xf0, &src_id);
 		if (rc < 0)
 			goto fail;
-		rc = skx_get_node_id(d, &node_id);
+
-		if (rc < 0)
+		edac_dbg(2, "src_id = %d\n", src_id);
-			goto fail;
-		edac_dbg(2, "src_id=%d node_id=%d\n", src_id, node_id);
 		for (i = 0; i < SKX_NUM_IMC; i++) {
 			d->imc[i].mc = mc++;
 			d->imc[i].lmc = i;
 			d->imc[i].src_id = src_id;
-			d->imc[i].node_id = node_id;
 			rc = skx_register_mci(&d->imc[i], d->imc[i].chan[0].cdev,
 					      "Skylake Socket", EDAC_MOD_STR,
 					      skx_get_dimm_config, cfg);

drivers/edac/skx_common.c

@@ -19,6 +19,7 @@
 #include <linux/adxl.h>
 #include <acpi/nfit.h>
 #include <asm/mce.h>
+#include <asm/uv/uv.h>
 #include "edac_module.h"
 #include "skx_common.h"
 
@@ -221,10 +222,42 @@ void skx_set_decode(skx_decode_f decode, skx_show_retry_log_f show_retry_log)
 }
 EXPORT_SYMBOL_GPL(skx_set_decode);
 
+static int skx_get_pkg_id(struct skx_dev *d, u8 *id)
+{
+	int node;
+	int cpu;
+
+	node = pcibus_to_node(d->util_all->bus);
+	if (numa_valid_node(node)) {
+		for_each_cpu(cpu, cpumask_of_pcibus(d->util_all->bus)) {
+			struct cpuinfo_x86 *c = &cpu_data(cpu);
+
+			if (c->initialized && cpu_to_node(cpu) == node) {
+				*id = c->topo.pkg_id;
+				return 0;
+			}
+		}
+	}
+
+	skx_printk(KERN_ERR, "Failed to get package ID from NUMA information\n");
+	return -ENODEV;
+}
+
 int skx_get_src_id(struct skx_dev *d, int off, u8 *id)
 {
 	u32 reg;
 
+	/*
+	 * The 3-bit source IDs in PCI configuration space registers are limited
+	 * to 8 unique IDs, and each ID is local to a UPI/QPI domain.
+	 *
+	 * Source IDs cannot be used to map devices to sockets on UV systems
+	 * because they can exceed 8 sockets and have multiple UPI/QPI domains
+	 * with identical, repeating source IDs.
+	 */
+	if (is_uv_system())
+		return skx_get_pkg_id(d, id);
+
 	if (pci_read_config_dword(d->util_all, off, &reg)) {
 		skx_printk(KERN_ERR, "Failed to read src id\n");
 		return -ENODEV;
@@ -235,20 +268,6 @@ int skx_get_src_id(struct skx_dev *d, int off, u8 *id)
 }
 EXPORT_SYMBOL_GPL(skx_get_src_id);
 
-int skx_get_node_id(struct skx_dev *d, u8 *id)
-{
-	u32 reg;
-
-	if (pci_read_config_dword(d->util_all, 0xf4, &reg)) {
-		skx_printk(KERN_ERR, "Failed to read node id\n");
-		return -ENODEV;
-	}
-
-	*id = GET_BITFIELD(reg, 0, 2);
-	return 0;
-}
-EXPORT_SYMBOL_GPL(skx_get_node_id);
-
 static int get_width(u32 mtr)
 {
 	switch (GET_BITFIELD(mtr, 8, 9)) {
@@ -507,7 +526,7 @@ int skx_register_mci(struct skx_imc *imc, struct pci_dev *pdev,
 	pvt->imc = imc;
 
 	mci->ctl_name = kasprintf(GFP_KERNEL, "%s#%d IMC#%d", ctl_name,
-				  imc->node_id, imc->lmc);
+				  imc->src_id, imc->lmc);
 	if (!mci->ctl_name) {
 		rc = -ENOMEM;
 		goto fail0;

drivers/edac/skx_common.h

@@ -103,7 +103,7 @@ struct skx_dev {
 		bool hbm_mc;
 		u8 mc;	/* system wide mc# */
 		u8 lmc;	/* socket relative mc# */
-		u8 src_id, node_id;
+		u8 src_id;
 		struct skx_channel {
 			struct pci_dev	*cdev;
 			struct pci_dev	*edev;
@@ -244,7 +244,6 @@ void skx_set_mem_cfg(bool mem_cfg_2lm);
 void skx_set_res_cfg(struct res_config *cfg);
 
 int skx_get_src_id(struct skx_dev *d, int off, u8 *id);
-int skx_get_node_id(struct skx_dev *d, u8 *id);
 
 int skx_get_all_bus_mappings(struct res_config *cfg, struct list_head **list);