Merge tag 'edac_updates_for_v6.8' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras

Pull EDAC updates from Borislav Petkov:

 - The EDAC drivers part of the effort to make the ->remove() platform
   driver callback return void

 - Add support for AMD AI accelerators

 - Add support for a number of Intel SoCs: Alder Lake-N, Raptor Lake-P,
   Meteor Lake-{P,PS}

 - Random fixes and cleanups all over the place

* tag 'edac_updates_for_v6.8' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras: (39 commits)
  EDAC/skx_common: Filter out the invalid address
  EDAC, pnd2: Sort headers alphabetically
  EDAC, pnd2: Correct misleading error message in mk_region_mask()
  EDAC, pnd2: Apply bit macros and helpers where it makes sense
  EDAC, pnd2: Replace custom definition by one from sizes.h
  EDAC/igen6: Add Intel Meteor Lake-P SoCs support
  EDAC/igen6: Add Intel Meteor Lake-PS SoCs support
  EDAC/igen6: Add Intel Raptor Lake-P SoCs support
  EDAC/igen6: Add Intel Alder Lake-N SoCs support
  EDAC/igen6: Make get_mchbar() helper function
  EDAC/amd64: Add support for family 0x19, models 0x90-9f devices
  EDAC/mc: Add support for HBM3 memory type
  EDAC/{sb,i7core}_edac: Do not use a plain integer for a NULL pointer
  EDAC/armada_xp: Explicitly include correct DT includes
  EDAC/pci_sysfs: Use PCI_HEADER_TYPE_MASK instead of literals
  EDAC/thunderx: Fix possible out-of-bounds string access
  EDAC/fsl_ddr: Convert to platform remove callback returning void
  EDAC/zynqmp: Convert to platform remove callback returning void
  EDAC/xgene: Convert to platform remove callback returning void
  EDAC/ti: Convert to platform remove callback returning void
  ...

Author: torvalds

drivers/edac/altera_edac.c

@@ -22,6 +22,7 @@
 #include <linux/of_platform.h>
 #include <linux/panic_notifier.h>
 #include <linux/platform_device.h>
+#include <linux/property.h>
 #include <linux/regmap.h>
 #include <linux/types.h>
 #include <linux/uaccess.h>
@@ -279,7 +280,6 @@ static int a10_unmask_irq(struct platform_device *pdev, u32 mask)
 
 static int altr_sdram_probe(struct platform_device *pdev)
 {
-	const struct of_device_id *id;
 	struct edac_mc_layer layers[2];
 	struct mem_ctl_info *mci;
 	struct altr_sdram_mc_data *drvdata;
@@ -290,10 +290,6 @@ static int altr_sdram_probe(struct platform_device *pdev)
 	int irq, irq2, res = 0;
 	unsigned long mem_size, irqflags = 0;
 
-	id = of_match_device(altr_sdram_ctrl_of_match, &pdev->dev);
-	if (!id)
-		return -ENODEV;
-
 	/* Grab the register range from the sdr controller in device tree */
 	mc_vbase = syscon_regmap_lookup_by_phandle(pdev->dev.of_node,
 						   "altr,sdr-syscon");
@@ -304,8 +300,7 @@ static int altr_sdram_probe(struct platform_device *pdev)
 	}
 
 	/* Check specific dependencies for the module */
-	priv = of_match_node(altr_sdram_ctrl_of_match,
-			     pdev->dev.of_node)->data;
+	priv = device_get_match_data(&pdev->dev);
 
 	/* Validate the SDRAM controller has ECC enabled */
 	if (regmap_read(mc_vbase, priv->ecc_ctrl_offset, &read_reg) ||
@@ -459,15 +454,13 @@ static int altr_sdram_probe(struct platform_device *pdev)
 	return res;
 }
 
-static int altr_sdram_remove(struct platform_device *pdev)
+static void altr_sdram_remove(struct platform_device *pdev)
 {
 	struct mem_ctl_info *mci = platform_get_drvdata(pdev);
 
 	edac_mc_del_mc(&pdev->dev);
 	edac_mc_free(mci);
 	platform_set_drvdata(pdev, NULL);
-
-	return 0;
 }
 
 /*
@@ -489,7 +482,7 @@ static const struct dev_pm_ops altr_sdram_pm_ops = {
 
 static struct platform_driver altr_sdram_edac_driver = {
 	.probe = altr_sdram_probe,
-	.remove = altr_sdram_remove,
+	.remove_new = altr_sdram_remove,
 	.driver = {
 		.name = "altr_sdram_edac",
 #ifdef CONFIG_PM
@@ -812,21 +805,19 @@ static int altr_edac_device_probe(struct platform_device *pdev)
 	return res;
 }
 
-static int altr_edac_device_remove(struct platform_device *pdev)
+static void altr_edac_device_remove(struct platform_device *pdev)
 {
 	struct edac_device_ctl_info *dci = platform_get_drvdata(pdev);
 	struct altr_edac_device_dev *drvdata = dci->pvt_info;
 
 	debugfs_remove_recursive(drvdata->debugfs_dir);
 	edac_device_del_device(&pdev->dev);
 	edac_device_free_ctl_info(dci);
-
-	return 0;
 }
 
 static struct platform_driver altr_edac_device_driver = {
 	.probe =  altr_edac_device_probe,
-	.remove = altr_edac_device_remove,
+	.remove_new = altr_edac_device_remove,
 	.driver = {
 		.name = "altr_edac_device",
 		.of_match_table = altr_edac_device_of_match,

drivers/edac/amd64_edac.c

@@ -996,15 +996,23 @@ static struct local_node_map {
 #define LNTM_NODE_COUNT				GENMASK(27, 16)
 #define LNTM_BASE_NODE_ID			GENMASK(11, 0)
 
-static int gpu_get_node_map(void)
+static int gpu_get_node_map(struct amd64_pvt *pvt)
 {
 	struct pci_dev *pdev;
 	int ret;
 	u32 tmp;
 
 	/*
-	 * Node ID 0 is reserved for CPUs.
-	 * Therefore, a non-zero Node ID means we've already cached the values.
+	 * Mapping of nodes from hardware-provided AMD Node ID to a
+	 * Linux logical one is applicable for MI200 models. Therefore,
+	 * return early for other heterogeneous systems.
+	 */
+	if (pvt->F3->device != PCI_DEVICE_ID_AMD_MI200_DF_F3)
+		return 0;
+
+	/*
+	 * Node ID 0 is reserved for CPUs. Therefore, a non-zero Node ID
+	 * means the values have been already cached.
 	 */
 	if (gpu_node_map.base_node_id)
 		return 0;
@@ -3851,7 +3859,7 @@ static void gpu_init_csrows(struct mem_ctl_info *mci)
 
 			dimm->nr_pages = gpu_get_csrow_nr_pages(pvt, umc, cs);
 			dimm->edac_mode = EDAC_SECDED;
-			dimm->mtype = MEM_HBM2;
+			dimm->mtype = pvt->dram_type;
 			dimm->dtype = DEV_X16;
 			dimm->grain = 64;
 		}
@@ -3880,7 +3888,7 @@ static bool gpu_ecc_enabled(struct amd64_pvt *pvt)
 	return true;
 }
 
-static inline u32 gpu_get_umc_base(u8 umc, u8 channel)
+static inline u32 gpu_get_umc_base(struct amd64_pvt *pvt, u8 umc, u8 channel)
 {
 	/*
 	 * On CPUs, there is one channel per UMC, so UMC numbering equals
@@ -3893,13 +3901,16 @@ static inline u32 gpu_get_umc_base(u8 umc, u8 channel)
 	 * On GPU nodes channels are selected in 3rd nibble
 	 * HBM chX[3:0]= [Y  ]5X[3:0]000;
 	 * HBM chX[7:4]= [Y+1]5X[3:0]000
+	 *
+	 * On MI300 APU nodes, same as GPU nodes but channels are selected
+	 * in the base address of 0x90000
 	 */
 	umc *= 2;
 
 	if (channel >= 4)
 		umc++;
 
-	return 0x50000 + (umc << 20) + ((channel % 4) << 12);
+	return pvt->gpu_umc_base + (umc << 20) + ((channel % 4) << 12);
 }
 
 static void gpu_read_mc_regs(struct amd64_pvt *pvt)
@@ -3910,7 +3921,7 @@ static void gpu_read_mc_regs(struct amd64_pvt *pvt)
 
 	/* Read registers from each UMC */
 	for_each_umc(i) {
-		umc_base = gpu_get_umc_base(i, 0);
+		umc_base = gpu_get_umc_base(pvt, i, 0);
 		umc = &pvt->umc[i];
 
 		amd_smn_read(nid, umc_base + UMCCH_UMC_CFG, &umc->umc_cfg);
@@ -3927,15 +3938,15 @@ static void gpu_read_base_mask(struct amd64_pvt *pvt)
 
 	for_each_umc(umc) {
 		for_each_chip_select(cs, umc, pvt) {
-			base_reg = gpu_get_umc_base(umc, cs) + UMCCH_BASE_ADDR;
+			base_reg = gpu_get_umc_base(pvt, umc, cs) + UMCCH_BASE_ADDR;
 			base = &pvt->csels[umc].csbases[cs];
 
 			if (!amd_smn_read(pvt->mc_node_id, base_reg, base)) {
 				edac_dbg(0, "  DCSB%d[%d]=0x%08x reg: 0x%x\n",
 					 umc, cs, *base, base_reg);
 			}
 
-			mask_reg = gpu_get_umc_base(umc, cs) + UMCCH_ADDR_MASK;
+			mask_reg = gpu_get_umc_base(pvt, umc, cs) + UMCCH_ADDR_MASK;
 			mask = &pvt->csels[umc].csmasks[cs];
 
 			if (!amd_smn_read(pvt->mc_node_id, mask_reg, mask)) {
@@ -3960,7 +3971,7 @@ static int gpu_hw_info_get(struct amd64_pvt *pvt)
 {
 	int ret;
 
-	ret = gpu_get_node_map();
+	ret = gpu_get_node_map(pvt);
 	if (ret)
 		return ret;
 
@@ -4125,6 +4136,8 @@ static int per_family_init(struct amd64_pvt *pvt)
 			if (pvt->F3->device == PCI_DEVICE_ID_AMD_MI200_DF_F3) {
 				pvt->ctl_name		= "MI200";
 				pvt->max_mcs		= 4;
+				pvt->dram_type		= MEM_HBM2;
+				pvt->gpu_umc_base	= 0x50000;
 				pvt->ops		= &gpu_ops;
 			} else {
 				pvt->ctl_name		= "F19h_M30h";
@@ -4142,6 +4155,13 @@ static int per_family_init(struct amd64_pvt *pvt)
 			pvt->ctl_name			= "F19h_M70h";
 			pvt->flags.zn_regs_v2		= 1;
 			break;
+		case 0x90 ... 0x9f:
+			pvt->ctl_name			= "F19h_M90h";
+			pvt->max_mcs			= 4;
+			pvt->dram_type			= MEM_HBM3;
+			pvt->gpu_umc_base		= 0x90000;
+			pvt->ops			= &gpu_ops;
+			break;
 		case 0xa0 ... 0xaf:
 			pvt->ctl_name			= "F19h_MA0h";
 			pvt->max_mcs			= 12;
@@ -4180,23 +4200,33 @@ static const struct attribute_group *amd64_edac_attr_groups[] = {
 	NULL
 };
 
+/*
+ * For heterogeneous and APU models EDAC CHIP_SELECT and CHANNEL layers
+ * should be swapped to fit into the layers.
+ */
+static unsigned int get_layer_size(struct amd64_pvt *pvt, u8 layer)
+{
+	bool is_gpu = (pvt->ops == &gpu_ops);
+
+	if (!layer)
+		return is_gpu ? pvt->max_mcs
+			      : pvt->csels[0].b_cnt;
+	else
+		return is_gpu ? pvt->csels[0].b_cnt
+			      : pvt->max_mcs;
+}
+
 static int init_one_instance(struct amd64_pvt *pvt)
 {
 	struct mem_ctl_info *mci = NULL;
 	struct edac_mc_layer layers[2];
 	int ret = -ENOMEM;
 
-	/*
-	 * For Heterogeneous family EDAC CHIP_SELECT and CHANNEL layers should
-	 * be swapped to fit into the layers.
-	 */
 	layers[0].type = EDAC_MC_LAYER_CHIP_SELECT;
-	layers[0].size = (pvt->F3->device == PCI_DEVICE_ID_AMD_MI200_DF_F3) ?
-			 pvt->max_mcs : pvt->csels[0].b_cnt;
+	layers[0].size = get_layer_size(pvt, 0);
 	layers[0].is_virt_csrow = true;
 	layers[1].type = EDAC_MC_LAYER_CHANNEL;
-	layers[1].size = (pvt->F3->device == PCI_DEVICE_ID_AMD_MI200_DF_F3) ?
-			 pvt->csels[0].b_cnt : pvt->max_mcs;
+	layers[1].size = get_layer_size(pvt, 1);
 	layers[1].is_virt_csrow = false;
 
 	mci = edac_mc_alloc(pvt->mc_node_id, ARRAY_SIZE(layers), layers, 0);

drivers/edac/amd64_edac.h

@@ -362,6 +362,7 @@ struct amd64_pvt {
 	u32 dct_sel_lo;		/* DRAM Controller Select Low */
 	u32 dct_sel_hi;		/* DRAM Controller Select High */
 	u32 online_spare;	/* On-Line spare Reg */
+	u32 gpu_umc_base;	/* Base address used for channel selection on GPUs */
 
 	/* x4, x8, or x16 syndromes in use */
 	u8 ecc_sym_sz;

drivers/edac/armada_xp_edac.c

@@ -5,7 +5,9 @@
 
 #include <linux/kernel.h>
 #include <linux/edac.h>
-#include <linux/of_platform.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
 
 #include <asm/hardware/cache-l2x0.h>
 #include <asm/hardware/cache-aurora-l2.h>
@@ -351,20 +353,18 @@ static int axp_mc_probe(struct platform_device *pdev)
 	return 0;
 }
 
-static int axp_mc_remove(struct platform_device *pdev)
+static void axp_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);
 	platform_set_drvdata(pdev, NULL);
-
-	return 0;
 }
 
 static struct platform_driver axp_mc_driver = {
 	.probe = axp_mc_probe,
-	.remove = axp_mc_remove,
+	.remove_new = axp_mc_remove,
 	.driver = {
 		.name = "armada_xp_mc_edac",
 		.of_match_table = of_match_ptr(axp_mc_of_match),
@@ -564,7 +564,7 @@ static int aurora_l2_probe(struct platform_device *pdev)
 	return 0;
 }
 
-static int aurora_l2_remove(struct platform_device *pdev)
+static void aurora_l2_remove(struct platform_device *pdev)
 {
 	struct edac_device_ctl_info *dci = platform_get_drvdata(pdev);
 #ifdef CONFIG_EDAC_DEBUG
@@ -575,13 +575,11 @@ static int aurora_l2_remove(struct platform_device *pdev)
 	edac_device_del_device(&pdev->dev);
 	edac_device_free_ctl_info(dci);
 	platform_set_drvdata(pdev, NULL);
-
-	return 0;
 }
 
 static struct platform_driver aurora_l2_driver = {
 	.probe = aurora_l2_probe,
-	.remove = aurora_l2_remove,
+	.remove_new = aurora_l2_remove,
 	.driver = {
 		.name = "aurora_l2_edac",
 		.of_match_table = of_match_ptr(aurora_l2_of_match),

drivers/edac/aspeed_edac.c

@@ -357,7 +357,7 @@ static int aspeed_probe(struct platform_device *pdev)
 }
 
 
-static int aspeed_remove(struct platform_device *pdev)
+static void aspeed_remove(struct platform_device *pdev)
 {
 	struct mem_ctl_info *mci;
 
@@ -369,8 +369,6 @@ static int aspeed_remove(struct platform_device *pdev)
 	mci = edac_mc_del_mc(&pdev->dev);
 	if (mci)
 		edac_mc_free(mci);
-
-	return 0;
 }
 
 
@@ -389,7 +387,7 @@ static struct platform_driver aspeed_driver = {
 		.of_match_table = aspeed_of_match
 	},
 	.probe		= aspeed_probe,
-	.remove		= aspeed_remove
+	.remove_new	= aspeed_remove
 };
 module_platform_driver(aspeed_driver);
 

drivers/edac/bluefield_edac.c

@@ -323,14 +323,12 @@ static int bluefield_edac_mc_probe(struct platform_device *pdev)
 
 }
 
-static int bluefield_edac_mc_remove(struct platform_device *pdev)
+static void 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[] = {
@@ -346,7 +344,7 @@ static struct platform_driver bluefield_edac_mc_driver = {
 		.acpi_match_table = bluefield_mc_acpi_ids,
 	},
 	.probe = bluefield_edac_mc_probe,
-	.remove = bluefield_edac_mc_remove,
+	.remove_new = bluefield_edac_mc_remove,
 };
 
 module_platform_driver(bluefield_edac_mc_driver);

drivers/edac/cell_edac.c

@@ -234,20 +234,19 @@ static int cell_edac_probe(struct platform_device *pdev)
 	return 0;
 }
 
-static int cell_edac_remove(struct platform_device *pdev)
+static void cell_edac_remove(struct platform_device *pdev)
 {
 	struct mem_ctl_info *mci = edac_mc_del_mc(&pdev->dev);
 	if (mci)
 		edac_mc_free(mci);
-	return 0;
 }
 
 static struct platform_driver cell_edac_driver = {
 	.driver		= {
 		.name	= "cbe-mic",
 	},
 	.probe		= cell_edac_probe,
-	.remove		= cell_edac_remove,
+	.remove_new	= cell_edac_remove,
 };
 
 static int __init cell_edac_init(void)