EDAC driver for i5400 MCH (Seaburg)
This driver adds support for i5400 MCH chipset.
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
index 5a11e3c..e5d1fcb 100644
--- a/drivers/edac/Kconfig
+++ b/drivers/edac/Kconfig
@@ -102,6 +102,13 @@ config EDAC_I3000
Support for error detection and correction on the Intel
3000 and 3010 server chipsets.
+config EDAC_I5400
+ tristate "Intel 5400 (Seaburg) chipsets"
+ depends on EDAC_MM_EDAC && PCI && X86
+ help
+ Support for error detection and correction the Intel
+ i5400 MCH chipset (Seaburg).
+
config EDAC_I82860
tristate "Intel 82860"
depends on EDAC_MM_EDAC && PCI && X86_32
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
index e5e9104..4e093e5 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -20,6 +20,7 @@ endif
obj-$(CONFIG_EDAC_AMD76X) += amd76x_edac.o
obj-$(CONFIG_EDAC_I5000) += i5000_edac.o
obj-$(CONFIG_EDAC_I5100) += i5100_edac.o
+obj-$(CONFIG_EDAC_I5400) += i5400_edac.o
obj-$(CONFIG_EDAC_E7XXX) += e7xxx_edac.o
obj-$(CONFIG_EDAC_E752X) += e752x_edac.o
obj-$(CONFIG_EDAC_I82443BXGX) += i82443bxgx_edac.o
diff --git a/drivers/edac/i5400_edac.c b/drivers/edac/i5400_edac.c
new file mode 100644
index 0000000..8ec3eca
--- /dev/null
+++ b/drivers/edac/i5400_edac.c
@@ -0,0 +1,1471 @@
+/*
+ * Intel 5400 class Memory Controllers kernel module
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Copyright (c) 2008 by:
+ * Ben Woodard <[email protected]>
+ * Mauro Carvalho Chehab <[email protected]>
+ *
+ * Red Hat Inc. http://www.redhat.com
+ *
+ * Forked and adapted from the i5000_edac driver which was
+ * written by Douglas Thompson Linux Networx <[email protected]>
+ *
+ * This module is based on the following document:
+ *
+ * Intel 5400 Chipset Memory Controller Hub (MCH) - Datasheet
+ * http://developer.intel.com/design/chipsets/datashts/313070.htm
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/edac.h>
+#include <linux/mmzone.h>
+
+#include "edac_core.h"
+
+/*
+ * Alter this version for the I5400 module when modifications are made
+ */
+#define I5400_REVISION " Ver: 1.0.0 " __DATE__
+
+#define EDAC_MOD_STR "i5400_edac"
+
+#define i5400_printk(level, fmt, arg...) \
+ edac_printk(level, "i5400", fmt, ##arg)
+
+#define i5400_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "i5400", fmt, ##arg)
+
+/* Limits for i5400 */
+#define NUM_MTRS_PER_BRANCH 4
+#define CHANNELS_PER_BRANCH 2
+#define MAX_CHANNELS 4
+#define MAX_DIMMS (MAX_CHANNELS * 4) /* Up to 4 DIMM's per channel */
+#define MAX_CSROWS (MAX_DIMMS * 2) /* max possible csrows per channel */
+
+/* Device 16,
+ * Function 0: System Address
+ * Function 1: Memory Branch Map, Control, Errors Register
+ * Function 2: FSB Error Registers
+ *
+ * All 3 functions of Device 16 (0,1,2) share the SAME DID
+ */
+#ifndef PCI_DEVICE_ID_INTEL_5400_ERR
+#define PCI_DEVICE_ID_INTEL_5400_ERR 0x4030 /* Device 16 (0,1,2) */
+#define PCI_DEVICE_ID_INTEL_5400_FBD0 0x4035 /* Device 21 (0,1) */
+#define PCI_DEVICE_ID_INTEL_5400_FBD1 0x4036 /* Device 21 (0,1) */
+#endif
+
+ /* OFFSETS for Function 0 */
+#define AMBASE 0x48 /* AMB Mem Mapped Reg Region Base */
+#define MAXCH 0x56 /* Max Channel Number */
+#define MAXDIMMPERCH 0x57 /* Max DIMM PER Channel Number */
+
+ /* OFFSETS for Function 1 */
+#define TOLM 0x6C
+#define REDMEMB 0x7C
+#define REC_ECC_LOCATOR_ODD(x) ((x) & 0x3fe00) /* bits [17:9] indicate ODD, [8:0] indicate EVEN */
+#define MIR0 0x80
+#define MIR1 0x84
+#define AMIR0 0x8c
+#define AMIR1 0x90
+
+ /* Fatal error registers */
+#define FERR_FAT_FBD 0x98 /* also called as FERR_FAT_FB_DIMM at datasheet */
+#define FERR_FAT_FBDCHAN (3<<28) /* channel index where the highest-order error occurred */
+
+#define NERR_FAT_FBD 0x9c
+#define FERR_NF_FBD 0xa0 /* also called as FERR_NFAT_FB_DIMM at datasheet */
+
+ /* Non-fatal error register */
+#define NERR_NF_FBD 0xa4
+
+ /* Enable error mask */
+#define EMASK_FBD 0xa8
+
+#define ERR0_FBD 0xac
+#define ERR1_FBD 0xb0
+#define ERR2_FBD 0xb4
+#define MCERR_FBD 0xb8
+
+ /* No OFFSETS for Device 16 Function 2 */
+
+/*
+ * Device 21,
+ * Function 0: Memory Map Branch 0
+ *
+ * Device 22,
+ * Function 0: Memory Map Branch 1
+ */
+
+ /* OFFSETS for Function 0 */
+#define AMBPRESENT_0 0x64
+#define AMBPRESENT_1 0x66
+#define MTR0 0x80
+#define MTR1 0x82
+#define MTR2 0x84
+#define MTR3 0x86
+
+ /* OFFSETS for Function 1 */
+#define NRECFGLOG 0x74
+#define RECFGLOG 0x78
+#define NRECMEMA 0xbe
+#define NRECMEMB 0xc0
+#define NRECFB_DIMMA 0xc4
+#define NRECFB_DIMMB 0xc8
+#define NRECFB_DIMMC 0xcc
+#define NRECFB_DIMMD 0xd0
+#define NRECFB_DIMME 0xd4
+#define NRECFB_DIMMF 0xd8
+#define REDMEMA 0xdC
+#define RECMEMA 0xf0
+#define RECMEMB 0xf4
+#define RECFB_DIMMA 0xf8
+#define RECFB_DIMMB 0xec
+#define RECFB_DIMMC 0xf0
+#define RECFB_DIMMD 0xf4
+#define RECFB_DIMME 0xf8
+#define RECFB_DIMMF 0xfC
+
+/*
+ * Error indicator bits and masks
+ * Error masks are according with Table 5-17 of i5400 datasheet
+ */
+
+enum error_mask {
+ EMASK_M1 = 1<<0, /* Memory Write error on non-redundant retry */
+ EMASK_M2 = 1<<1, /* Memory or FB-DIMM configuration CRC read error */
+ EMASK_M3 = 1<<2, /* Reserved */
+ EMASK_M4 = 1<<3, /* Uncorrectable Data ECC on Replay */
+ EMASK_M5 = 1<<4, /* Aliased Uncorrectable Non-Mirrored Demand Data ECC */
+ EMASK_M6 = 1<<5, /* Unsupported on i5400 */
+ EMASK_M7 = 1<<6, /* Aliased Uncorrectable Resilver- or Spare-Copy Data ECC */
+ EMASK_M8 = 1<<7, /* Aliased Uncorrectable Patrol Data ECC */
+ EMASK_M9 = 1<<8, /* Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC */
+ EMASK_M10 = 1<<9, /* Unsupported on i5400 */
+ EMASK_M11 = 1<<10, /* Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC */
+ EMASK_M12 = 1<<11, /* Non-Aliased Uncorrectable Patrol Data ECC */
+ EMASK_M13 = 1<<12, /* Memory Write error on first attempt */
+ EMASK_M14 = 1<<13, /* FB-DIMM Configuration Write error on first attempt */
+ EMASK_M15 = 1<<14, /* Memory or FB-DIMM configuration CRC read error */
+ EMASK_M16 = 1<<15, /* Channel Failed-Over Occurred */
+ EMASK_M17 = 1<<16, /* Correctable Non-Mirrored Demand Data ECC */
+ EMASK_M18 = 1<<17, /* Unsupported on i5400 */
+ EMASK_M19 = 1<<18, /* Correctable Resilver- or Spare-Copy Data ECC */
+ EMASK_M20 = 1<<19, /* Correctable Patrol Data ECC */
+ EMASK_M21 = 1<<20, /* FB-DIMM Northbound parity error on FB-DIMM Sync Status */
+ EMASK_M22 = 1<<21, /* SPD protocol Error */
+ EMASK_M23 = 1<<22, /* Non-Redundant Fast Reset Timeout */
+ EMASK_M24 = 1<<23, /* Refresh error */
+ EMASK_M25 = 1<<24, /* Memory Write error on redundant retry */
+ EMASK_M26 = 1<<25, /* Redundant Fast Reset Timeout */
+ EMASK_M27 = 1<<26, /* Correctable Counter Threshold Exceeded */
+ EMASK_M28 = 1<<27, /* DIMM-Spare Copy Completed */
+ EMASK_M29 = 1<<28, /* DIMM-Isolation Completed */
+};
+
+/*
+ * Names to translate bit error into something useful
+ */
+char *error_name[] = {
+ [0] = "Memory Write error on non-redundant retry",
+ [1] = "Memory or FB-DIMM configuration CRC read error",
+ /* Reserved */
+ [3] = "Uncorrectable Data ECC on Replay",
+ [4] = "Aliased Uncorrectable Non-Mirrored Demand Data ECC",
+ /* Unsupported on i5400 */
+ [6] = "Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
+ [7] = "Aliased Uncorrectable Patrol Data ECC",
+ [8] = "Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC",
+ /* Unsupported */
+ [10] = "Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
+ [11] = "Non-Aliased Uncorrectable Patrol Data ECC",
+ [12] = "Memory Write error on first attempt",
+ [13] = "FB-DIMM Configuration Write error on first attempt",
+ [14] = "Memory or FB-DIMM configuration CRC read error",
+ [15] = "Channel Failed-Over Occurred",
+ [16] = "Correctable Non-Mirrored Demand Data ECC",
+ /* Unsupported */
+ [18] = "Correctable Resilver- or Spare-Copy Data ECC",
+ [19] = "Correctable Patrol Data ECC",
+ [20] = "FB-DIMM Northbound parity error on FB-DIMM Sync Status",
+ [21] = "SPD protocol Error",
+ [22] = "Non-Redundant Fast Reset Timeout",
+ [23] = "Refresh error",
+ [24] = "Memory Write error on redundant retry",
+ [25] = "Redundant Fast Reset Timeout",
+ [26] = "Correctable Counter Threshold Exceeded",
+ [27] = "DIMM-Spare Copy Completed",
+ [28] = "DIMM-Isolation Completed",
+};
+
+/* Fatal errors */
+#define ERROR_FAT_MASK (EMASK_M1 | \
+ EMASK_M2 | \
+ EMASK_M23)
+
+/* Correctable errors */
+#define ERROR_NF_CORRECTABLE (EMASK_M27 | \
+ EMASK_M20 | \
+ EMASK_M19 | \
+ EMASK_M18 | \
+ EMASK_M17 | \
+ EMASK_M16)
+#define ERROR_NF_DIMM_SPARE (EMASK_M29 | \
+ EMASK_M28)
+#define ERROR_NF_SPD_PROTOCOL (EMASK_M22)
+#define ERROR_NF_NORTH_CRC (EMASK_M21)
+
+/* Recoverable errors */
+#define ERROR_NF_RECOVERABLE (EMASK_M26 | \
+ EMASK_M25 | \
+ EMASK_M24 | \
+ EMASK_M15 | \
+ EMASK_M14 | \
+ EMASK_M13 | \
+ EMASK_M12 | \
+ EMASK_M11 | \
+ EMASK_M9 | \
+ EMASK_M8 | \
+ EMASK_M7 | \
+ EMASK_M5)
+
+/* uncorrectable errors */
+#define ERROR_NF_UNCORRECTABLE (EMASK_M4)
+
+/* mask to all non-fatal errors */
+#define ERROR_NF_MASK (ERROR_NF_CORRECTABLE | \
+ ERROR_NF_UNCORRECTABLE | \
+ ERROR_NF_RECOVERABLE | \
+ ERROR_NF_DIMM_SPARE | \
+ ERROR_NF_SPD_PROTOCOL | \
+ ERROR_NF_NORTH_CRC)
+
+/*
+ * Define error masks for the several registers
+ */
+
+/* Enable all fatal and non fatal errors */
+#define ENABLE_EMASK_ALL (ERROR_FAT_MASK | ERROR_NF_MASK)
+
+/* mask for fatal error registers */
+#define FERR_FAT_MASK ERROR_FAT_MASK
+
+/* masks for non-fatal error register */
+#define TO_NF_MASK(a) (((a) & EMASK_M29) | ((a) >> 3))
+#define FROM_NF_FERR(a) (((a) & EMASK_M29) | (((a) << 3) & ((1 << 30)-1)))
+
+#define FERR_NF_MASK TO_NF_MASK(ERROR_NF_MASK)
+#define FERR_NF_CORRECTABLE TO_NF_MASK(ERROR_NF_CORRECTABLE)
+#define FERR_NF_DIMM_SPARE TO_NF_MASK(ERROR_NF_DIMM_SPARE)
+#define FERR_NF_SPD_PROTOCOL TO_NF_MASK(ERROR_NF_SPD_PROTOCOL)
+#define FERR_NF_NORTH_CRC TO_NF_MASK(ERROR_NF_NORTH_CRC)
+#define FERR_NF_RECOVERABLE TO_NF_MASK(ERROR_NF_RECOVERABLE)
+#define FERR_NF_UNCORRECTABLE TO_NF_MASK(ERROR_NF_UNCORRECTABLE)
+
+/* Defines to extract the vaious fields from the
+ * MTRx - Memory Technology Registers
+ */
+#define MTR_DIMMS_PRESENT(mtr) ((mtr) & (1 << 10))
+#define MTR_DIMMS_ETHROTTLE(mtr) ((mtr) & (1 << 9))
+#define MTR_DRAM_WIDTH(mtr) (((mtr) & (1<< 8)) ? 8 : 4)
+#define MTR_DRAM_BANKS(mtr) (((mtr) & (1<< 6)) ? 8 : 4)
+#define MTR_DRAM_BANKS_ADDR_BITS(mtr) ((MTR_DRAM_BANKS(mtr) == 8) ? 3 : 2)
+#define MTR_DIMM_RANK(mtr) (((mtr) >> 5) & 0x1)
+#define MTR_DIMM_RANK_ADDR_BITS(mtr) (MTR_DIMM_RANK(mtr) ? 2 : 1)
+#define MTR_DIMM_ROWS(mtr) (((mtr) >> 2) & 0x3)
+#define MTR_DIMM_ROWS_ADDR_BITS(mtr) (MTR_DIMM_ROWS(mtr) + 13)
+#define MTR_DIMM_COLS(mtr) ((mtr) & 0x3)
+#define MTR_DIMM_COLS_ADDR_BITS(mtr) (MTR_DIMM_COLS(mtr) + 10)
+
+/* This applies to FERR_NF_FB-DIMM as well as FERR_FAT_FB-DIMM */
+static inline int extract_fbdchan_indx(u32 x)
+{
+ return (x>>28) & 0x3;
+}
+
+#ifdef CONFIG_EDAC_DEBUG
+/* MTR NUMROW */
+static char *numrow_toString[] = {
+ "8,192 - 13 rows",
+ "16,384 - 14 rows",
+ "32,768 - 15 rows",
+ "65,536 - 16 rows"
+};
+
+/* MTR NUMCOL */
+static char *numcol_toString[] = {
+ "1,024 - 10 columns",
+ "2,048 - 11 columns",
+ "4,096 - 12 columns",
+ "reserved"
+};
+#endif
+
+/* Device name and register DID (Device ID) */
+struct i5400_dev_info {
+ const char *ctl_name; /* name for this device */
+ u16 fsb_mapping_errors; /* DID for the branchmap,control */
+};
+
+/* Table of devices attributes supported by this driver */
+static const struct i5400_dev_info i5400_devs[] = {
+ {
+ .ctl_name = "I5400",
+ .fsb_mapping_errors = PCI_DEVICE_ID_INTEL_5400_ERR,
+ },
+};
+
+struct i5400_dimm_info {
+ int megabytes; /* size, 0 means not present */
+ int dual_rank;
+};
+
+/* driver private data structure */
+struct i5400_pvt {
+ struct pci_dev *system_address; /* 16.0 */
+ struct pci_dev *branchmap_werrors; /* 16.1 */
+ struct pci_dev *fsb_error_regs; /* 16.2 */
+ struct pci_dev *branch_0; /* 21.0 */
+ struct pci_dev *branch_1; /* 22.0 */
+
+ u16 tolm; /* top of low memory */
+ u64 ambase; /* AMB BAR */
+
+ u16 mir0, mir1;
+
+ u16 b0_mtr[NUM_MTRS_PER_BRANCH]; /* Memory Technlogy Reg */
+ u16 b0_ambpresent0; /* Branch 0, Channel 0 */
+ u16 b0_ambpresent1; /* Brnach 0, Channel 1 */
+
+ u16 b1_mtr[NUM_MTRS_PER_BRANCH]; /* Memory Technlogy Reg */
+ u16 b1_ambpresent0; /* Branch 1, Channel 8 */
+ u16 b1_ambpresent1; /* Branch 1, Channel 1 */
+
+ /* DIMM information matrix, allocating architecture maximums */
+ struct i5400_dimm_info dimm_info[MAX_CSROWS][MAX_CHANNELS];
+
+ /* Actual values for this controller */
+ int maxch; /* Max channels */
+ int maxdimmperch; /* Max DIMMs per channel */
+};
+
+/* I5400 MCH error information retrieved from Hardware */
+struct i5400_error_info {
+ /* These registers are always read from the MC */
+ u32 ferr_fat_fbd; /* First Errors Fatal */
+ u32 nerr_fat_fbd; /* Next Errors Fatal */
+ u32 ferr_nf_fbd; /* First Errors Non-Fatal */
+ u32 nerr_nf_fbd; /* Next Errors Non-Fatal */
+
+ /* These registers are input ONLY if there was a Recoverable Error */
+ u32 redmemb; /* Recoverable Mem Data Error log B */
+ u16 recmema; /* Recoverable Mem Error log A */
+ u32 recmemb; /* Recoverable Mem Error log B */
+
+ /* These registers are input ONLY if there was a Non-Recoverable Error */
+ u16 nrecmema; /* Non-Recoverable Mem log A */
+ u16 nrecmemb; /* Non-Recoverable Mem log B */
+
+};
+
+/* note that nrec_rdwr changed from NRECMEMA to NRECMEMB between the 5000 and
+ 5400 better to use an inline function than a macro in this case */
+static inline int nrec_bank(struct i5400_error_info *info)
+{
+ return ((info->nrecmema) >> 12) & 0x7;
+}
+static inline int nrec_rank(struct i5400_error_info *info)
+{
+ return ((info->nrecmema) >> 8) & 0xf;
+}
+static inline int nrec_buf_id(struct i5400_error_info *info)
+{
+ return ((info->nrecmema)) & 0xff;
+}
+static inline int nrec_rdwr(struct i5400_error_info *info)
+{
+ return (info->nrecmemb) >> 31;
+}
+/* This applies to both NREC and REC string so it can be used with nrec_rdwr
+ and rec_rdwr */
+static inline const char *rdwr_str(int rdwr)
+{
+ return rdwr ? "Write" : "Read";
+}
+static inline int nrec_cas(struct i5400_error_info *info)
+{
+ return ((info->nrecmemb) >> 16) & 0x1fff;
+}
+static inline int nrec_ras(struct i5400_error_info *info)
+{
+ return (info->nrecmemb) & 0xffff;
+}
+static inline int rec_bank(struct i5400_error_info *info)
+{
+ return ((info->recmema) >> 12) & 0x7;
+}
+static inline int rec_rank(struct i5400_error_info *info)
+{
+ return ((info->recmema) >> 8) & 0xf;
+}
+static inline int rec_rdwr(struct i5400_error_info *info)
+{
+ return (info->recmemb) >> 31;
+}
+static inline int rec_cas(struct i5400_error_info *info)
+{
+ return ((info->recmemb) >> 16) & 0x1fff;
+}
+static inline int rec_ras(struct i5400_error_info *info)
+{
+ return (info->recmemb) & 0xffff;
+}
+
+static struct edac_pci_ctl_info *i5400_pci;
+
+/*
+ * i5400_get_error_info Retrieve the hardware error information from
+ * the hardware and cache it in the 'info'
+ * structure
+ */
+static void i5400_get_error_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info)
+{
+ struct i5400_pvt *pvt;
+ u32 value;
+
+ pvt = mci->pvt_info;
+
+ /* read in the 1st FATAL error register */
+ pci_read_config_dword(pvt->branchmap_werrors, FERR_FAT_FBD, &value);
+
+ /* Mask only the bits that the doc says are valid
+ */
+ value &= (FERR_FAT_FBDCHAN | FERR_FAT_MASK);
+
+ /* If there is an error, then read in the
+ NEXT FATAL error register and the Memory Error Log Register A
+ */
+ if (value & FERR_FAT_MASK) {
+ info->ferr_fat_fbd = value;
+
+ /* harvest the various error data we need */
+ pci_read_config_dword(pvt->branchmap_werrors,
+ NERR_FAT_FBD, &info->nerr_fat_fbd);
+ pci_read_config_word(pvt->branchmap_werrors,
+ NRECMEMA, &info->nrecmema);
+ pci_read_config_word(pvt->branchmap_werrors,
+ NRECMEMB, &info->nrecmemb);
+
+ /* Clear the error bits, by writing them back */
+ pci_write_config_dword(pvt->branchmap_werrors,
+ FERR_FAT_FBD, value);
+ } else {
+ info->ferr_fat_fbd = 0;
+ info->nerr_fat_fbd = 0;
+ info->nrecmema = 0;
+ info->nrecmemb = 0;
+ }
+
+ /* read in the 1st NON-FATAL error register */
+ pci_read_config_dword(pvt->branchmap_werrors, FERR_NF_FBD, &value);
+
+ /* If there is an error, then read in the 1st NON-FATAL error
+ * register as well */
+ if (value & FERR_NF_MASK) {
+ info->ferr_nf_fbd = value;
+
+ /* harvest the various error data we need */
+ pci_read_config_dword(pvt->branchmap_werrors,
+ NERR_NF_FBD, &info->nerr_nf_fbd);
+ pci_read_config_word(pvt->branchmap_werrors,
+ RECMEMA, &info->recmema);
+ pci_read_config_dword(pvt->branchmap_werrors,
+ RECMEMB, &info->recmemb);
+ pci_read_config_dword(pvt->branchmap_werrors,
+ REDMEMB, &info->redmemb);
+
+ /* Clear the error bits, by writing them back */
+ pci_write_config_dword(pvt->branchmap_werrors,
+ FERR_NF_FBD, value);
+ } else {
+ info->ferr_nf_fbd = 0;
+ info->nerr_nf_fbd = 0;
+ info->recmema = 0;
+ info->recmemb = 0;
+ info->redmemb = 0;
+ }
+}
+
+/*
+ * i5400_proccess_non_recoverable_info(struct mem_ctl_info *mci,
+ * struct i5400_error_info *info,
+ * int handle_errors);
+ *
+ * handle the Intel FATAL and unrecoverable errors, if any
+ */
+static void i5400_proccess_non_recoverable_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info,
+ unsigned long allErrors)
+{
+ char msg[EDAC_MC_LABEL_LEN + 1 + 90 + 80];
+ int branch;
+ int channel;
+ int bank;
+ int buf_id;
+ int rank;
+ int rdwr;
+ int ras, cas;
+ int errnum;
+ char *type = NULL;
+
+ if (!allErrors)
+ return; /* if no error, return now */
+
+ if (allErrors & ERROR_FAT_MASK)
+ type = "FATAL";
+ else if (allErrors & FERR_NF_UNCORRECTABLE)
+ type = "NON-FATAL uncorrected";
+ else
+ type = "NON-FATAL recoverable";
+
+ /* ONLY ONE of the possible error bits will be set, as per the docs */
+
+ branch = extract_fbdchan_indx(info->ferr_fat_fbd);
+ channel = branch;
+
+ /* Use the NON-Recoverable macros to extract data */
+ bank = nrec_bank(info);
+ rank = nrec_rank(info);
+ buf_id = nrec_buf_id(info);
+ rdwr = nrec_rdwr(info);
+ ras = nrec_ras(info);
+ cas = nrec_cas(info);
+
+ debugf0("\t\tCSROW= %d Channels= %d,%d (Branch= %d "
+ "DRAM Bank= %d Buffer ID = %d rdwr= %s ras= %d cas= %d)\n",
+ rank, channel, channel + 1, branch >> 1, bank,
+ buf_id, rdwr_str(rdwr), ras, cas);
+
+ /* Only 1 bit will be on */
+ errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
+
+ /* Form out message */
+ snprintf(msg, sizeof(msg),
+ "%s (Branch=%d DRAM-Bank=%d Buffer ID = %d RDWR=%s RAS=%d CAS=%d "
+ "%s Err=0x%lx (%s))",
+ type, branch >> 1, bank, buf_id, rdwr_str(rdwr), ras, cas, type,
+ allErrors, error_name[errnum]);
+
+ /* Call the helper to output message */
+ edac_mc_handle_fbd_ue(mci, rank, channel, channel + 1, msg);
+}
+
+/*
+ * i5400_process_fatal_error_info(struct mem_ctl_info *mci,
+ * struct i5400_error_info *info,
+ * int handle_errors);
+ *
+ * handle the Intel NON-FATAL errors, if any
+ */
+static void i5400_process_nonfatal_error_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info)
+{
+ char msg[EDAC_MC_LABEL_LEN + 1 + 90 + 80];
+ unsigned long allErrors;
+ int branch;
+ int channel;
+ int bank;
+ int rank;
+ int rdwr;
+ int ras, cas;
+ int errnum;
+
+ /* mask off the Error bits that are possible */
+ allErrors = FROM_NF_FERR(info->ferr_nf_fbd & FERR_NF_MASK);
+ if (!allErrors)
+ return; /* if no error, return now */
+
+ /* ONLY ONE of the possible error bits will be set, as per the docs */
+
+ if (allErrors & (ERROR_NF_UNCORRECTABLE | ERROR_NF_RECOVERABLE)) {
+ i5400_proccess_non_recoverable_info(mci, info, allErrors);
+ return;
+ }
+
+ /* Correctable errors */
+ if (allErrors & ERROR_NF_CORRECTABLE) {
+ debugf0("\tCorrected bits= 0x%lx\n", allErrors);
+
+ branch = extract_fbdchan_indx(info->ferr_nf_fbd);
+
+ channel = 0;
+ if (REC_ECC_LOCATOR_ODD(info->redmemb))
+ channel = 1;
+
+ /* Convert channel to be based from zero, instead of
+ * from branch base of 0 */
+ channel += branch;
+
+ bank = rec_bank(info);
+ rank = rec_rank(info);
+ rdwr = rec_rdwr(info);
+ ras = rec_ras(info);
+ cas = rec_cas(info);
+
+ /* Only 1 bit will be on */
+ errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
+
+ debugf0("\t\tCSROW= %d Channel= %d (Branch %d "
+ "DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
+ rank, channel, branch >> 1, bank,
+ rdwr_str(rdwr), ras, cas);
+
+ /* Form out message */
+ snprintf(msg, sizeof(msg),
+ "Corrected error (Branch=%d DRAM-Bank=%d RDWR=%s RAS=%d "
+ "CAS=%d, CE Err=0x%lx (%s))", branch >> 1, bank,
+ rdwr_str(rdwr), ras, cas, allErrors,
+ error_name[errnum]);
+
+ /* Call the helper to output message */
+ edac_mc_handle_fbd_ce(mci, rank, channel, msg);
+
+ return;
+ }
+
+ /* Miscelaneous errors */
+ errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
+
+ branch = extract_fbdchan_indx(info->ferr_nf_fbd);
+
+ i5400_mc_printk(mci, KERN_EMERG,
+ "Non-Fatal misc error (Branch=%d Err=%#lx (%s))",
+ branch >> 1, allErrors, error_name[errnum]);
+}
+
+/*
+ * i5400_process_error_info Process the error info that is
+ * in the 'info' structure, previously retrieved from hardware
+ */
+static void i5400_process_error_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info)
+{ u32 allErrors;
+
+ /* First handle any fatal errors that occurred */
+ allErrors = (info->ferr_fat_fbd & FERR_FAT_MASK);
+ i5400_proccess_non_recoverable_info(mci, info, allErrors);
+
+ /* now handle any non-fatal errors that occurred */
+ i5400_process_nonfatal_error_info(mci, info);
+}
+
+/*
+ * i5400_clear_error Retrieve any error from the hardware
+ * but do NOT process that error.
+ * Used for 'clearing' out of previous errors
+ * Called by the Core module.
+ */
+static void i5400_clear_error(struct mem_ctl_info *mci)
+{
+ struct i5400_error_info info;
+
+ i5400_get_error_info(mci, &info);
+}
+
+/*
+ * i5400_check_error Retrieve and process errors reported by the
+ * hardware. Called by the Core module.
+ */
+static void i5400_check_error(struct mem_ctl_info *mci)
+{
+ struct i5400_error_info info;
+ debugf4("MC%d: " __FILE__ ": %s()\n", mci->mc_idx, __func__);
+ i5400_get_error_info(mci, &info);
+ i5400_process_error_info(mci, &info);
+}
+
+/*
+ * i5400_put_devices 'put' all the devices that we have
+ * reserved via 'get'
+ */
+static void i5400_put_devices(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+
+ pvt = mci->pvt_info;
+
+ /* Decrement usage count for devices */
+ if (pvt->branch_1)
+ pci_dev_put(pvt->branch_1);
+
+ if (pvt->branch_0)
+ pci_dev_put(pvt->branch_0);
+
+ if (pvt->fsb_error_regs)
+ pci_dev_put(pvt->fsb_error_regs);
+
+ if (pvt->branchmap_werrors)
+ pci_dev_put(pvt->branchmap_werrors);
+}
+
+/*
+ * i5400_get_devices Find and perform 'get' operation on the MCH's
+ * device/functions we want to reference for this driver
+ *
+ * Need to 'get' device 16 func 1 and func 2
+ */
+static int i5400_get_devices(struct mem_ctl_info *mci, int dev_idx)
+{
+ struct i5400_pvt *pvt;
+ struct pci_dev *pdev;
+
+ pvt = mci->pvt_info;
+ pvt->branchmap_werrors = NULL;
+ pvt->fsb_error_regs = NULL;
+ pvt->branch_0 = NULL;
+ pvt->branch_1 = NULL;
+
+ /* Attempt to 'get' the MCH register we want */
+ pdev = NULL;
+ while (!pvt->branchmap_werrors || !pvt->fsb_error_regs) {
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_ERR, pdev);
+ if (!pdev) {
+ /* End of list, leave */
+ i5400_printk(KERN_ERR,
+ "'system address,Process Bus' "
+ "device not found:"
+ "vendor 0x%x device 0x%x ERR funcs "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_ERR);
+ goto error;
+ }
+
+ /* Store device 16 funcs 1 and 2 */
+ switch (PCI_FUNC(pdev->devfn)) {
+ case 1:
+ pvt->branchmap_werrors = pdev;
+ break;
+ case 2:
+ pvt->fsb_error_regs = pdev;
+ break;
+ }
+ }
+
+ debugf1("System Address, processor bus- PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->system_address),
+ pvt->system_address->vendor, pvt->system_address->device);
+ debugf1("Branchmap, control and errors - PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->branchmap_werrors),
+ pvt->branchmap_werrors->vendor, pvt->branchmap_werrors->device);
+ debugf1("FSB Error Regs - PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->fsb_error_regs),
+ pvt->fsb_error_regs->vendor, pvt->fsb_error_regs->device);
+
+ pvt->branch_0 = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_FBD0, NULL);
+ if (!pvt->branch_0) {
+ i5400_printk(KERN_ERR,
+ "MC: 'BRANCH 0' device not found:"
+ "vendor 0x%x device 0x%x Func 0 (broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_5400_FBD0);
+ goto error;
+ }
+
+ /* If this device claims to have more than 2 channels then
+ * fetch Branch 1's information
+ */
+ if (pvt->maxch < CHANNELS_PER_BRANCH)
+ return 0;
+
+ pvt->branch_1 = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_FBD1, NULL);
+ if (!pvt->branch_1) {
+ i5400_printk(KERN_ERR,
+ "MC: 'BRANCH 1' device not found:"
+ "vendor 0x%x device 0x%x Func 0 "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_FBD1);
+ goto error;
+ }
+
+ return 0;
+
+error:
+ i5400_put_devices(mci);
+ return -ENODEV;
+}
+
+/*
+ * determine_amb_present
+ *
+ * the information is contained in NUM_MTRS_PER_BRANCH different registers
+ * determining which of the NUM_MTRS_PER_BRANCH requires knowing
+ * which channel is in question
+ *
+ * 2 branches, each with 2 channels
+ * b0_ambpresent0 for channel '0'
+ * b0_ambpresent1 for channel '1'
+ * b1_ambpresent0 for channel '2'
+ * b1_ambpresent1 for channel '3'
+ */
+static int determine_amb_present_reg(struct i5400_pvt *pvt, int channel)
+{
+ int amb_present;
+
+ if (channel < CHANNELS_PER_BRANCH) {
+ if (channel & 0x1)
+ amb_present = pvt->b0_ambpresent1;
+ else
+ amb_present = pvt->b0_ambpresent0;
+ } else {
+ if (channel & 0x1)
+ amb_present = pvt->b1_ambpresent1;
+ else
+ amb_present = pvt->b1_ambpresent0;
+ }
+
+ return amb_present;
+}
+
+/*
+ * determine_mtr(pvt, csrow, channel)
+ *
+ * return the proper MTR register as determine by the csrow and channel desired
+ */
+static int determine_mtr(struct i5400_pvt *pvt, int csrow, int channel)
+{
+ int mtr;
+ int n;
+
+ /* There is one MTR for each slot pair of FB-DIMMs,
+ Each slot may have one or two ranks (2 csrows),
+ Each slot pair may be at branch 0 or branch 1.
+ So, csrow should be divided by eight
+ */
+ n = csrow >> 3;
+
+ if (n >= NUM_MTRS_PER_BRANCH) {
+ debugf0("ERROR: trying to access an invalid csrow: %d\n", csrow);
+ return 0;
+ }
+
+ if (channel < CHANNELS_PER_BRANCH)
+ mtr = pvt->b0_mtr[n];
+ else
+ mtr = pvt->b1_mtr[n];
+
+ return mtr;
+}
+
+/*
+ */
+static void decode_mtr(int slot_row, u16 mtr)
+{
+ int ans;
+
+ ans = MTR_DIMMS_PRESENT(mtr);
+
+ debugf2("\tMTR%d=0x%x: DIMMs are %s\n", slot_row, mtr,
+ ans ? "Present" : "NOT Present");
+ if (!ans)
+ return;
+
+ debugf2("\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr));
+
+ debugf2("\t\tELECTRICAL THROTTLING is %s\n",
+ MTR_DIMMS_ETHROTTLE(mtr) ? "enabled": "disabled");
+
+ debugf2("\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr));
+ debugf2("\t\tNUMRANK: %s\n", MTR_DIMM_RANK(mtr) ? "double" : "single");
+ debugf2("\t\tNUMROW: %s\n", numrow_toString[MTR_DIMM_ROWS(mtr)]);
+ debugf2("\t\tNUMCOL: %s\n", numcol_toString[MTR_DIMM_COLS(mtr)]);
+}
+
+static void handle_channel(struct i5400_pvt *pvt, int csrow, int channel,
+ struct i5400_dimm_info *dinfo)
+{
+ int mtr;
+ int amb_present_reg;
+ int addrBits;
+
+ mtr = determine_mtr(pvt, csrow, channel);
+ if (MTR_DIMMS_PRESENT(mtr)) {
+ amb_present_reg = determine_amb_present_reg(pvt, channel);
+
+ /* Determine if there is a DIMM present in this DIMM slot */
+ if (amb_present_reg & (1 << (csrow >> 1))) {
+ dinfo->dual_rank = MTR_DIMM_RANK(mtr);
+
+ if (!((dinfo->dual_rank == 0) &&
+ ((csrow & 0x1) == 0x1))) {
+ /* Start with the number of bits for a Bank
+ * on the DRAM */
+ addrBits = MTR_DRAM_BANKS_ADDR_BITS(mtr);
+ /* Add thenumber of ROW bits */
+ addrBits += MTR_DIMM_ROWS_ADDR_BITS(mtr);
+ /* add the number of COLUMN bits */
+ addrBits += MTR_DIMM_COLS_ADDR_BITS(mtr);
+
+ addrBits += 6; /* add 64 bits per DIMM */
+ addrBits -= 20; /* divide by 2^^20 */
+ addrBits -= 3; /* 8 bits per bytes */
+
+ dinfo->megabytes = 1 << addrBits;
+ }
+ }
+ }
+}
+
+/*
+ * calculate_dimm_size
+ *
+ * also will output a DIMM matrix map, if debug is enabled, for viewing
+ * how the DIMMs are populated
+ */
+static void calculate_dimm_size(struct i5400_pvt *pvt)
+{
+ struct i5400_dimm_info *dinfo;
+ int csrow, max_csrows;
+ char *p, *mem_buffer;
+ int space, n;
+ int channel;
+
+ /* ================= Generate some debug output ================= */
+ space = PAGE_SIZE;
+ mem_buffer = p = kmalloc(space, GFP_KERNEL);
+ if (p == NULL) {
+ i5400_printk(KERN_ERR, "MC: %s:%s() kmalloc() failed\n",
+ __FILE__, __func__);
+ return;
+ }
+
+ /* Scan all the actual CSROWS (which is # of DIMMS * 2)
+ * and calculate the information for each DIMM
+ * Start with the highest csrow first, to display it first
+ * and work toward the 0th csrow
+ */
+ max_csrows = pvt->maxdimmperch * 2;
+ for (csrow = max_csrows - 1; csrow >= 0; csrow--) {
+
+ /* on an odd csrow, first output a 'boundary' marker,
+ * then reset the message buffer */
+ if (csrow & 0x1) {
+ n = snprintf(p, space, "---------------------------"
+ "--------------------------------");
+ p += n;
+ space -= n;
+ debugf2("%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+ }
+ n = snprintf(p, space, "csrow %2d ", csrow);
+ p += n;
+ space -= n;
+
+ for (channel = 0; channel < pvt->maxch; channel++) {
+ dinfo = &pvt->dimm_info[csrow][channel];
+ handle_channel(pvt, csrow, channel, dinfo);
+ n = snprintf(p, space, "%4d MB | ", dinfo->megabytes);
+ p += n;
+ space -= n;
+ }
+ debugf2("%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+ }
+
+ /* Output the last bottom 'boundary' marker */
+ n = snprintf(p, space, "---------------------------"
+ "--------------------------------");
+ p += n;
+ space -= n;
+ debugf2("%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+
+ /* now output the 'channel' labels */
+ n = snprintf(p, space, " ");
+ p += n;
+ space -= n;
+ for (channel = 0; channel < pvt->maxch; channel++) {
+ n = snprintf(p, space, "channel %d | ", channel);
+ p += n;
+ space -= n;
+ }
+
+ /* output the last message and free buffer */
+ debugf2("%s\n", mem_buffer);
+ kfree(mem_buffer);
+}
+
+/*
+ * i5400_get_mc_regs read in the necessary registers and
+ * cache locally
+ *
+ * Fills in the private data members
+ */
+static void i5400_get_mc_regs(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+ u32 actual_tolm;
+ u16 limit;
+ int slot_row;
+ int maxch;
+ int maxdimmperch;
+ int way0, way1;
+
+ pvt = mci->pvt_info;
+
+ pci_read_config_dword(pvt->system_address, AMBASE,
+ (u32 *) &pvt->ambase);
+ pci_read_config_dword(pvt->system_address, AMBASE + sizeof(u32),
+ ((u32 *) &pvt->ambase) + sizeof(u32));
+
+ maxdimmperch = pvt->maxdimmperch;
+ maxch = pvt->maxch;
+
+ debugf2("AMBASE= 0x%lx MAXCH= %d MAX-DIMM-Per-CH= %d\n",
+ (long unsigned int)pvt->ambase, pvt->maxch, pvt->maxdimmperch);
+
+ /* Get the Branch Map regs */
+ pci_read_config_word(pvt->branchmap_werrors, TOLM, &pvt->tolm);
+ pvt->tolm >>= 12;
+ debugf2("\nTOLM (number of 256M regions) =%u (0x%x)\n", pvt->tolm,
+ pvt->tolm);
+
+ actual_tolm = (u32) ((1000l * pvt->tolm) >> (30 - 28));
+ debugf2("Actual TOLM byte addr=%u.%03u GB (0x%x)\n",
+ actual_tolm/1000, actual_tolm % 1000, pvt->tolm << 28);
+
+ pci_read_config_word(pvt->branchmap_werrors, MIR0, &pvt->mir0);
+ pci_read_config_word(pvt->branchmap_werrors, MIR1, &pvt->mir1);
+
+ /* Get the MIR[0-1] regs */
+ limit = (pvt->mir0 >> 4) & 0x0fff;
+ way0 = pvt->mir0 & 0x1;
+ way1 = pvt->mir0 & 0x2;
+ debugf2("MIR0: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
+ limit = (pvt->mir1 >> 4) & 0xfff;
+ way0 = pvt->mir1 & 0x1;
+ way1 = pvt->mir1 & 0x2;
+ debugf2("MIR1: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
+
+ /* Get the set of MTR[0-3] regs by each branch */
+ for (slot_row = 0; slot_row < NUM_MTRS_PER_BRANCH; slot_row++) {
+ int where = MTR0 + (slot_row * sizeof(u32));
+
+ /* Branch 0 set of MTR registers */
+ pci_read_config_word(pvt->branch_0, where,
+ &pvt->b0_mtr[slot_row]);
+
+ debugf2("MTR%d where=0x%x B0 value=0x%x\n", slot_row, where,
+ pvt->b0_mtr[slot_row]);
+
+ if (pvt->maxch < CHANNELS_PER_BRANCH) {
+ pvt->b1_mtr[slot_row] = 0;
+ continue;
+ }
+
+ /* Branch 1 set of MTR registers */
+ pci_read_config_word(pvt->branch_1, where,
+ &pvt->b1_mtr[slot_row]);
+ debugf2("MTR%d where=0x%x B1 value=0x%x\n", slot_row, where,
+ pvt->b1_mtr[slot_row]);
+ }
+
+ /* Read and dump branch 0's MTRs */
+ debugf2("\nMemory Technology Registers:\n");
+ debugf2(" Branch 0:\n");
+ for (slot_row = 0; slot_row < NUM_MTRS_PER_BRANCH; slot_row++)
+ decode_mtr(slot_row, pvt->b0_mtr[slot_row]);
+
+ pci_read_config_word(pvt->branch_0, AMBPRESENT_0,
+ &pvt->b0_ambpresent0);
+ debugf2("\t\tAMB-Branch 0-present0 0x%x:\n", pvt->b0_ambpresent0);
+ pci_read_config_word(pvt->branch_0, AMBPRESENT_1,
+ &pvt->b0_ambpresent1);
+ debugf2("\t\tAMB-Branch 0-present1 0x%x:\n", pvt->b0_ambpresent1);
+
+ /* Only if we have 2 branchs (4 channels) */
+ if (pvt->maxch < CHANNELS_PER_BRANCH) {
+ pvt->b1_ambpresent0 = 0;
+ pvt->b1_ambpresent1 = 0;
+ } else {
+ /* Read and dump branch 1's MTRs */
+ debugf2(" Branch 1:\n");
+ for (slot_row = 0; slot_row < NUM_MTRS_PER_BRANCH; slot_row++)
+ decode_mtr(slot_row, pvt->b1_mtr[slot_row]);
+
+ pci_read_config_word(pvt->branch_1, AMBPRESENT_0,
+ &pvt->b1_ambpresent0);
+ debugf2("\t\tAMB-Branch 1-present0 0x%x:\n",
+ pvt->b1_ambpresent0);
+ pci_read_config_word(pvt->branch_1, AMBPRESENT_1,
+ &pvt->b1_ambpresent1);
+ debugf2("\t\tAMB-Branch 1-present1 0x%x:\n",
+ pvt->b1_ambpresent1);
+ }
+
+ /* Go and determine the size of each DIMM and place in an
+ * orderly matrix */
+ calculate_dimm_size(pvt);
+}
+
+/*
+ * i5400_init_csrows Initialize the 'csrows' table within
+ * the mci control structure with the
+ * addressing of memory.
+ *
+ * return:
+ * 0 success
+ * 1 no actual memory found on this MC
+ */
+static int i5400_init_csrows(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+ struct csrow_info *p_csrow;
+ int empty, channel_count;
+ int max_csrows;
+ int mtr;
+ int csrow_megs;
+ int channel;
+ int csrow;
+
+ pvt = mci->pvt_info;
+
+ channel_count = pvt->maxch;
+ max_csrows = pvt->maxdimmperch * 2;
+
+ empty = 1; /* Assume NO memory */
+
+ for (csrow = 0; csrow < max_csrows; csrow++) {
+ p_csrow = &mci->csrows[csrow];
+
+ p_csrow->csrow_idx = csrow;
+
+ /* use branch 0 for the basis */
+ mtr = determine_mtr(pvt, csrow, 0);
+
+ /* if no DIMMS on this row, continue */
+ if (!MTR_DIMMS_PRESENT(mtr))
+ continue;
+
+ /* FAKE OUT VALUES, FIXME */
+ p_csrow->first_page = 0 + csrow * 20;
+ p_csrow->last_page = 9 + csrow * 20;
+ p_csrow->page_mask = 0xFFF;
+
+ p_csrow->grain = 8;
+
+ csrow_megs = 0;
+ for (channel = 0; channel < pvt->maxch; channel++)
+ csrow_megs += pvt->dimm_info[csrow][channel].megabytes;
+
+ p_csrow->nr_pages = csrow_megs << 8;
+
+ /* Assume DDR2 for now */
+ p_csrow->mtype = MEM_FB_DDR2;
+
+ /* ask what device type on this row */
+ if (MTR_DRAM_WIDTH(mtr))
+ p_csrow->dtype = DEV_X8;
+ else
+ p_csrow->dtype = DEV_X4;
+
+ p_csrow->edac_mode = EDAC_S8ECD8ED;
+
+ empty = 0;
+ }
+
+ return empty;
+}
+
+/*
+ * i5400_enable_error_reporting
+ * Turn on the memory reporting features of the hardware
+ */
+static void i5400_enable_error_reporting(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+ u32 fbd_error_mask;
+
+ pvt = mci->pvt_info;
+
+ /* Read the FBD Error Mask Register */
+ pci_read_config_dword(pvt->branchmap_werrors, EMASK_FBD,
+ &fbd_error_mask);
+
+ /* Enable with a '0' */
+ fbd_error_mask &= ~(ENABLE_EMASK_ALL);
+
+ pci_write_config_dword(pvt->branchmap_werrors, EMASK_FBD,
+ fbd_error_mask);
+}
+
+/*
+ * i5400_get_dimm_and_channel_counts(pdev, &num_csrows, &num_channels)
+ *
+ * ask the device how many channels are present and how many CSROWS
+ * as well
+ */
+static void i5400_get_dimm_and_channel_counts(struct pci_dev *pdev,
+ int *num_dimms_per_channel,
+ int *num_channels)
+{
+ u8 value;
+
+ /* Need to retrieve just how many channels and dimms per channel are
+ * supported on this memory controller
+ */
+ pci_read_config_byte(pdev, MAXDIMMPERCH, &value);
+ *num_dimms_per_channel = (int)value * 2;
+
+ pci_read_config_byte(pdev, MAXCH, &value);
+ *num_channels = (int)value;
+}
+
+/*
+ * i5400_probe1 Probe for ONE instance of device to see if it is
+ * present.
+ * return:
+ * 0 for FOUND a device
+ * < 0 for error code
+ */
+static int i5400_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ struct mem_ctl_info *mci;
+ struct i5400_pvt *pvt;
+ int num_channels;
+ int num_dimms_per_channel;
+ int num_csrows;
+
+ debugf0("MC: " __FILE__ ": %s(), pdev bus %u dev=0x%x fn=0x%x\n",
+ __func__,
+ pdev->bus->number,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+
+ /* We only are looking for func 0 of the set */
+ if (PCI_FUNC(pdev->devfn) != 0)
+ return -ENODEV;
+
+ /* Ask the devices for the number of CSROWS and CHANNELS so
+ * that we can calculate the memory resources, etc
+ *
+ * The Chipset will report what it can handle which will be greater
+ * or equal to what the motherboard manufacturer will implement.
+ *
+ * As we don't have a motherboard identification routine to determine
+ * actual number of slots/dimms per channel, we thus utilize the
+ * resource as specified by the chipset. Thus, we might have
+ * have more DIMMs per channel than actually on the mobo, but this
+ * allows the driver to support upto the chipset max, without
+ * some fancy mobo determination.
+ */
+ i5400_get_dimm_and_channel_counts(pdev, &num_dimms_per_channel,
+ &num_channels);
+ num_csrows = num_dimms_per_channel * 2;
+
+ debugf0("MC: %s(): Number of - Channels= %d DIMMS= %d CSROWS= %d\n",
+ __func__, num_channels, num_dimms_per_channel, num_csrows);
+
+ /* allocate a new MC control structure */
+ mci = edac_mc_alloc(sizeof(*pvt), num_csrows, num_channels, 0);
+
+ if (mci == NULL)
+ return -ENOMEM;
+
+ debugf0("MC: " __FILE__ ": %s(): mci = %p\n", __func__, mci);
+
+ mci->dev = &pdev->dev; /* record ptr to the generic device */
+
+ pvt = mci->pvt_info;
+ pvt->system_address = pdev; /* Record this device in our private */
+ pvt->maxch = num_channels;
+ pvt->maxdimmperch = num_dimms_per_channel;
+
+ /* 'get' the pci devices we want to reserve for our use */
+ if (i5400_get_devices(mci, dev_idx))
+ goto fail0;
+
+ /* Time to get serious */
+ i5400_get_mc_regs(mci); /* retrieve the hardware registers */
+
+ mci->mc_idx = 0;
+ mci->mtype_cap = MEM_FLAG_FB_DDR2;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+ mci->edac_cap = EDAC_FLAG_NONE;
+ mci->mod_name = "i5400_edac.c";
+ mci->mod_ver = I5400_REVISION;
+ mci->ctl_name = i5400_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->ctl_page_to_phys = NULL;
+
+ /* Set the function pointer to an actual operation function */
+ mci->edac_check = i5400_check_error;
+
+ /* initialize the MC control structure 'csrows' table
+ * with the mapping and control information */
+ if (i5400_init_csrows(mci)) {
+ debugf0("MC: Setting mci->edac_cap to EDAC_FLAG_NONE\n"
+ " because i5400_init_csrows() returned nonzero "
+ "value\n");
+ mci->edac_cap = EDAC_FLAG_NONE; /* no csrows found */
+ } else {
+ debugf1("MC: Enable error reporting now\n");
+ i5400_enable_error_reporting(mci);
+ }
+
+ /* add this new MC control structure to EDAC's list of MCs */
+ if (edac_mc_add_mc(mci)) {
+ debugf0("MC: " __FILE__
+ ": %s(): failed edac_mc_add_mc()\n", __func__);
+ /* FIXME: perhaps some code should go here that disables error
+ * reporting if we just enabled it
+ */
+ goto fail1;
+ }
+
+ i5400_clear_error(mci);
+
+ /* allocating generic PCI control info */
+ i5400_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!i5400_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ return 0;
+
+ /* Error exit unwinding stack */
+fail1:
+
+ i5400_put_devices(mci);
+
+fail0:
+ edac_mc_free(mci);
+ return -ENODEV;
+}
+
+/*
+ * i5400_init_one constructor for one instance of device
+ *
+ * returns:
+ * negative on error
+ * count (>= 0)
+ */
+static int __devinit i5400_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *id)
+{
+ int rc;
+
+ debugf0("MC: " __FILE__ ": %s()\n", __func__);
+
+ /* wake up device */
+ rc = pci_enable_device(pdev);
+ if (rc == -EIO)
+ return rc;
+
+ /* now probe and enable the device */
+ return i5400_probe1(pdev, id->driver_data);
+}
+
+/*
+ * i5400_remove_one destructor for one instance of device
+ *
+ */
+static void __devexit i5400_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ debugf0(__FILE__ ": %s()\n", __func__);
+
+ if (i5400_pci)
+ edac_pci_release_generic_ctl(i5400_pci);
+
+ mci = edac_mc_del_mc(&pdev->dev);
+ if (!mci)
+ return;
+
+ /* retrieve references to resources, and free those resources */
+ i5400_put_devices(mci);
+
+ edac_mc_free(mci);
+}
+
+/*
+ * pci_device_id table for which devices we are looking for
+ *
+ * The "E500P" device is the first device supported.
+ */
+static const struct pci_device_id i5400_pci_tbl[] __devinitdata = {
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_5400_ERR)},
+ {0,} /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i5400_pci_tbl);
+
+/*
+ * i5400_driver pci_driver structure for this module
+ *
+ */
+static struct pci_driver i5400_driver = {
+ .name = KBUILD_BASENAME,
+ .probe = i5400_init_one,
+ .remove = __devexit_p(i5400_remove_one),
+ .id_table = i5400_pci_tbl,
+};
+
+/*
+ * i5400_init Module entry function
+ * Try to initialize this module for its devices
+ */
+static int __init i5400_init(void)
+{
+ int pci_rc;
+
+ debugf2("MC: " __FILE__ ": %s()\n", __func__);
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&i5400_driver);
+
+ return (pci_rc < 0) ? pci_rc : 0;
+}
+
+/*
+ * i5400_exit() Module exit function
+ * Unregister the driver
+ */
+static void __exit i5400_exit(void)
+{
+ debugf2("MC: " __FILE__ ": %s()\n", __func__);
+ pci_unregister_driver(&i5400_driver);
+}
+
+module_init(i5400_init);
+module_exit(i5400_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Ben Woodard <[email protected]> Red Hat Inc. (http://www.redhat.com)");
+MODULE_AUTHOR("Mauro Carvalho Chehab <[email protected]> Red Hat Inc. (http://www.redhat.com)");
+MODULE_DESCRIPTION("MC Driver for Intel I5400 memory controllers - " I5400_REVISION);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");
On Mon, 1 Dec 2008 19:46:23 -0200 Mauro Carvalho Chehab <[email protected]> wrote:
> EDAC driver for i5400 MCH (Seaburg)
>
> This driver adds support for i5400 MCH chipset.
checkpatch has fun with this one.
> Signed-off-by: Mauro Carvalho Chehab <[email protected]>
Both you and Ben Woodard are listed under MODULE_AUTHOR. Should this
patch have Ben's Signed-off-by:? And/or his From:?
>
> ...
>
> +/* Device 16,
> + * Function 0: System Address
> + * Function 1: Memory Branch Map, Control, Errors Register
> + * Function 2: FSB Error Registers
> + *
> + * All 3 functions of Device 16 (0,1,2) share the SAME DID
> + */
> +#ifndef PCI_DEVICE_ID_INTEL_5400_ERR
This ifndef is unneeded: PCI_DEVICE_ID_INTEL_5400_ERR _is_ defined.
What's more, it is actually bad, because if pci_ids.h were to #define
PCI_DEVICE_ID_INTEL_5400_ERR to something other than 0x4030, the ifndef
would prevent the compiler from informing us of this potential problem.
> +#define PCI_DEVICE_ID_INTEL_5400_ERR 0x4030 /* Device 16 (0,1,2) */
> +#define PCI_DEVICE_ID_INTEL_5400_FBD0 0x4035 /* Device 21 (0,1) */
> +#define PCI_DEVICE_ID_INTEL_5400_FBD1 0x4036 /* Device 21 (0,1) */
> +#endif
> +
>
> ...
>
> +/*
> + * Names to translate bit error into something useful
> + */
> +char *error_name[] = {
error_name is a global symbol.
> + [0] = "Memory Write error on non-redundant retry",
> + [1] = "Memory or FB-DIMM configuration CRC read error",
> + /* Reserved */
> + [3] = "Uncorrectable Data ECC on Replay",
> + [4] = "Aliased Uncorrectable Non-Mirrored Demand Data ECC",
> + /* Unsupported on i5400 */
> + [6] = "Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
> + [7] = "Aliased Uncorrectable Patrol Data ECC",
> + [8] = "Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC",
> + /* Unsupported */
> + [10] = "Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
> + [11] = "Non-Aliased Uncorrectable Patrol Data ECC",
> + [12] = "Memory Write error on first attempt",
> + [13] = "FB-DIMM Configuration Write error on first attempt",
> + [14] = "Memory or FB-DIMM configuration CRC read error",
> + [15] = "Channel Failed-Over Occurred",
> + [16] = "Correctable Non-Mirrored Demand Data ECC",
> + /* Unsupported */
> + [18] = "Correctable Resilver- or Spare-Copy Data ECC",
> + [19] = "Correctable Patrol Data ECC",
> + [20] = "FB-DIMM Northbound parity error on FB-DIMM Sync Status",
> + [21] = "SPD protocol Error",
> + [22] = "Non-Redundant Fast Reset Timeout",
> + [23] = "Refresh error",
> + [24] = "Memory Write error on redundant retry",
> + [25] = "Redundant Fast Reset Timeout",
> + [26] = "Correctable Counter Threshold Exceeded",
> + [27] = "DIMM-Spare Copy Completed",
> + [28] = "DIMM-Isolation Completed",
> +};
> +
>
> ...
>
> +/* masks for non-fatal error register */
> +#define TO_NF_MASK(a) (((a) & EMASK_M29) | ((a) >> 3))
> +#define FROM_NF_FERR(a) (((a) & EMASK_M29) | (((a) << 3) & ((1 << 30)-1)))
These macros will misbehave if passed an expression with side-effects.
Not really a problem as they're only used for constructing constants.
I guess if one was being paranoid, one could #undef them again as soon
as possible.
> +#define FERR_NF_MASK TO_NF_MASK(ERROR_NF_MASK)
> +#define FERR_NF_CORRECTABLE TO_NF_MASK(ERROR_NF_CORRECTABLE)
> +#define FERR_NF_DIMM_SPARE TO_NF_MASK(ERROR_NF_DIMM_SPARE)
> +#define FERR_NF_SPD_PROTOCOL TO_NF_MASK(ERROR_NF_SPD_PROTOCOL)
> +#define FERR_NF_NORTH_CRC TO_NF_MASK(ERROR_NF_NORTH_CRC)
> +#define FERR_NF_RECOVERABLE TO_NF_MASK(ERROR_NF_RECOVERABLE)
> +#define FERR_NF_UNCORRECTABLE TO_NF_MASK(ERROR_NF_UNCORRECTABLE)
> +
>
> ...
>
> +#ifdef CONFIG_EDAC_DEBUG
> +/* MTR NUMROW */
> +static char *numrow_toString[] = {
> + "8,192 - 13 rows",
> + "16,384 - 14 rows",
> + "32,768 - 15 rows",
> + "65,536 - 16 rows"
> +};
> +
> +/* MTR NUMCOL */
> +static char *numcol_toString[] = {
> + "1,024 - 10 columns",
> + "2,048 - 11 columns",
> + "4,096 - 12 columns",
> + "reserved"
> +};
These could possibly be made const. We don't modify them at runtime...
> +#endif
> +
>
> ...
>
> +static void i5400_proccess_non_recoverable_info(struct mem_ctl_info *mci,
> + struct i5400_error_info *info,
> + unsigned long allErrors)
> +{
> + char msg[EDAC_MC_LABEL_LEN + 1 + 90 + 80];
202 bytes. Tolerable, I guess.
Heaven knows why and how this dimension was chosen, and under which
circumstances it will overflow :(
It looks like such a hack that perhaps you should give up and use
kasprintf()+kfree()?
> + int branch;
> + int channel;
> + int bank;
> + int buf_id;
> + int rank;
> + int rdwr;
> + int ras, cas;
> + int errnum;
> + char *type = NULL;
> +
> + if (!allErrors)
> + return; /* if no error, return now */
> +
> + if (allErrors & ERROR_FAT_MASK)
> + type = "FATAL";
> + else if (allErrors & FERR_NF_UNCORRECTABLE)
> + type = "NON-FATAL uncorrected";
> + else
> + type = "NON-FATAL recoverable";
> +
> + /* ONLY ONE of the possible error bits will be set, as per the docs */
> +
> + branch = extract_fbdchan_indx(info->ferr_fat_fbd);
> + channel = branch;
> +
> + /* Use the NON-Recoverable macros to extract data */
> + bank = nrec_bank(info);
> + rank = nrec_rank(info);
> + buf_id = nrec_buf_id(info);
> + rdwr = nrec_rdwr(info);
> + ras = nrec_ras(info);
> + cas = nrec_cas(info);
> +
> + debugf0("\t\tCSROW= %d Channels= %d,%d (Branch= %d "
> + "DRAM Bank= %d Buffer ID = %d rdwr= %s ras= %d cas= %d)\n",
> + rank, channel, channel + 1, branch >> 1, bank,
> + buf_id, rdwr_str(rdwr), ras, cas);
> +
> + /* Only 1 bit will be on */
> + errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
> +
> + /* Form out message */
> + snprintf(msg, sizeof(msg),
> + "%s (Branch=%d DRAM-Bank=%d Buffer ID = %d RDWR=%s RAS=%d CAS=%d "
> + "%s Err=0x%lx (%s))",
> + type, branch >> 1, bank, buf_id, rdwr_str(rdwr), ras, cas, type,
> + allErrors, error_name[errnum]);
> +
> + /* Call the helper to output message */
> + edac_mc_handle_fbd_ue(mci, rank, channel, channel + 1, msg);
> +}
> +
> +/*
> + * i5400_process_fatal_error_info(struct mem_ctl_info *mci,
> + * struct i5400_error_info *info,
> + * int handle_errors);
> + *
> + * handle the Intel NON-FATAL errors, if any
> + */
> +static void i5400_process_nonfatal_error_info(struct mem_ctl_info *mci,
> + struct i5400_error_info *info)
> +{
> + char msg[EDAC_MC_LABEL_LEN + 1 + 90 + 80];
dittoes.
> + unsigned long allErrors;
> + int branch;
> + int channel;
> + int bank;
> + int rank;
> + int rdwr;
> + int ras, cas;
> + int errnum;
> +
> + /* mask off the Error bits that are possible */
> + allErrors = FROM_NF_FERR(info->ferr_nf_fbd & FERR_NF_MASK);
> + if (!allErrors)
> + return; /* if no error, return now */
> +
> + /* ONLY ONE of the possible error bits will be set, as per the docs */
> +
> + if (allErrors & (ERROR_NF_UNCORRECTABLE | ERROR_NF_RECOVERABLE)) {
> + i5400_proccess_non_recoverable_info(mci, info, allErrors);
> + return;
> + }
> +
> + /* Correctable errors */
> + if (allErrors & ERROR_NF_CORRECTABLE) {
> + debugf0("\tCorrected bits= 0x%lx\n", allErrors);
> +
> + branch = extract_fbdchan_indx(info->ferr_nf_fbd);
> +
> + channel = 0;
> + if (REC_ECC_LOCATOR_ODD(info->redmemb))
> + channel = 1;
> +
> + /* Convert channel to be based from zero, instead of
> + * from branch base of 0 */
> + channel += branch;
> +
> + bank = rec_bank(info);
> + rank = rec_rank(info);
> + rdwr = rec_rdwr(info);
> + ras = rec_ras(info);
> + cas = rec_cas(info);
> +
> + /* Only 1 bit will be on */
> + errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
> +
> + debugf0("\t\tCSROW= %d Channel= %d (Branch %d "
> + "DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
> + rank, channel, branch >> 1, bank,
> + rdwr_str(rdwr), ras, cas);
> +
> + /* Form out message */
> + snprintf(msg, sizeof(msg),
> + "Corrected error (Branch=%d DRAM-Bank=%d RDWR=%s RAS=%d "
> + "CAS=%d, CE Err=0x%lx (%s))", branch >> 1, bank,
> + rdwr_str(rdwr), ras, cas, allErrors,
> + error_name[errnum]);
> +
> + /* Call the helper to output message */
> + edac_mc_handle_fbd_ce(mci, rank, channel, msg);
> +
> + return;
> + }
> +
> + /* Miscelaneous errors */
> + errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
> +
> + branch = extract_fbdchan_indx(info->ferr_nf_fbd);
> +
> + i5400_mc_printk(mci, KERN_EMERG,
> + "Non-Fatal misc error (Branch=%d Err=%#lx (%s))",
> + branch >> 1, allErrors, error_name[errnum]);
> +}
> +
>
> ...
>
> +static void i5400_put_devices(struct mem_ctl_info *mci)
> +{
> + struct i5400_pvt *pvt;
> +
> + pvt = mci->pvt_info;
> +
> + /* Decrement usage count for devices */
> + if (pvt->branch_1)
> + pci_dev_put(pvt->branch_1);
> +
> + if (pvt->branch_0)
> + pci_dev_put(pvt->branch_0);
> +
> + if (pvt->fsb_error_regs)
> + pci_dev_put(pvt->fsb_error_regs);
> +
> + if (pvt->branchmap_werrors)
> + pci_dev_put(pvt->branchmap_werrors);
pci_dev_put(NULL) is legal.
> +}
> +
>
> ...
>
> +static int i5400_probe1(struct pci_dev *pdev, int dev_idx)
> +{
> + struct mem_ctl_info *mci;
> + struct i5400_pvt *pvt;
> + int num_channels;
> + int num_dimms_per_channel;
> + int num_csrows;
> +
> + debugf0("MC: " __FILE__ ": %s(), pdev bus %u dev=0x%x fn=0x%x\n",
> + __func__,
> + pdev->bus->number,
> + PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
> +
> + /* We only are looking for func 0 of the set */
> + if (PCI_FUNC(pdev->devfn) != 0)
> + return -ENODEV;
> +
> + /* Ask the devices for the number of CSROWS and CHANNELS so
> + * that we can calculate the memory resources, etc
> + *
> + * The Chipset will report what it can handle which will be greater
> + * or equal to what the motherboard manufacturer will implement.
> + *
> + * As we don't have a motherboard identification routine to determine
> + * actual number of slots/dimms per channel, we thus utilize the
> + * resource as specified by the chipset. Thus, we might have
> + * have more DIMMs per channel than actually on the mobo, but this
> + * allows the driver to support upto the chipset max, without
> + * some fancy mobo determination.
> + */
> + i5400_get_dimm_and_channel_counts(pdev, &num_dimms_per_channel,
> + &num_channels);
> + num_csrows = num_dimms_per_channel * 2;
> +
> + debugf0("MC: %s(): Number of - Channels= %d DIMMS= %d CSROWS= %d\n",
> + __func__, num_channels, num_dimms_per_channel, num_csrows);
> +
> + /* allocate a new MC control structure */
> + mci = edac_mc_alloc(sizeof(*pvt), num_csrows, num_channels, 0);
> +
> + if (mci == NULL)
> + return -ENOMEM;
> +
> + debugf0("MC: " __FILE__ ": %s(): mci = %p\n", __func__, mci);
> +
> + mci->dev = &pdev->dev; /* record ptr to the generic device */
> +
> + pvt = mci->pvt_info;
> + pvt->system_address = pdev; /* Record this device in our private */
> + pvt->maxch = num_channels;
> + pvt->maxdimmperch = num_dimms_per_channel;
> +
> + /* 'get' the pci devices we want to reserve for our use */
> + if (i5400_get_devices(mci, dev_idx))
> + goto fail0;
> +
> + /* Time to get serious */
> + i5400_get_mc_regs(mci); /* retrieve the hardware registers */
> +
> + mci->mc_idx = 0;
> + mci->mtype_cap = MEM_FLAG_FB_DDR2;
> + mci->edac_ctl_cap = EDAC_FLAG_NONE;
> + mci->edac_cap = EDAC_FLAG_NONE;
> + mci->mod_name = "i5400_edac.c";
> + mci->mod_ver = I5400_REVISION;
> + mci->ctl_name = i5400_devs[dev_idx].ctl_name;
This is a bit strange. i5400_devs[] has a single entry, so what's the
point in indexing it with `dev_idx', which MUST have a value of zero
anyway?
> + mci->dev_name = pci_name(pdev);
> + mci->ctl_page_to_phys = NULL;
> +
> + /* Set the function pointer to an actual operation function */
> + mci->edac_check = i5400_check_error;
> +
> + /* initialize the MC control structure 'csrows' table
> + * with the mapping and control information */
> + if (i5400_init_csrows(mci)) {
> + debugf0("MC: Setting mci->edac_cap to EDAC_FLAG_NONE\n"
> + " because i5400_init_csrows() returned nonzero "
> + "value\n");
> + mci->edac_cap = EDAC_FLAG_NONE; /* no csrows found */
> + } else {
> + debugf1("MC: Enable error reporting now\n");
> + i5400_enable_error_reporting(mci);
> + }
> +
> + /* add this new MC control structure to EDAC's list of MCs */
> + if (edac_mc_add_mc(mci)) {
> + debugf0("MC: " __FILE__
> + ": %s(): failed edac_mc_add_mc()\n", __func__);
> + /* FIXME: perhaps some code should go here that disables error
> + * reporting if we just enabled it
> + */
> + goto fail1;
> + }
> +
> + i5400_clear_error(mci);
> +
> + /* allocating generic PCI control info */
> + i5400_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
> + if (!i5400_pci) {
> + printk(KERN_WARNING
> + "%s(): Unable to create PCI control\n",
> + __func__);
> + printk(KERN_WARNING
> + "%s(): PCI error report via EDAC not setup\n",
> + __func__);
> + }
> +
> + return 0;
> +
> + /* Error exit unwinding stack */
> +fail1:
> +
> + i5400_put_devices(mci);
> +
> +fail0:
> + edac_mc_free(mci);
> + return -ENODEV;
> +}
> +
>
> ...
>
> +static struct pci_driver i5400_driver = {
> + .name = KBUILD_BASENAME,
> + .probe = i5400_init_one,
> + .remove = __devexit_p(i5400_remove_one),
> + .id_table = i5400_pci_tbl,
> +};
Only one other driver in the tree uses KBUILD_BASENAME, and that is
drivers/edac/i5100_edac.c, from which I suspect the above was copied.
I suspect that something is being done wrongly here.
>
> ...
>
On Tue, 2 Dec 2008 22:33:47 -0800
Andrew Morton <[email protected]> wrote:
> On Mon, 1 Dec 2008 19:46:23 -0200 Mauro Carvalho Chehab <[email protected]> wrote:
>
> > EDAC driver for i5400 MCH (Seaburg)
> >
> > This driver adds support for i5400 MCH chipset.
>
> checkpatch has fun with this one.
Ok. I fixed the remaining errors.
I'll just keep the 80 char warns on lines with comments explaining each field, like:
WARNING: line over 80 characters
#200: FILE: drivers/edac/i5400_edac.c:151:
+ EMASK_M9 = 1<<8, /* Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC */
Where breaking the line wouldn't be that fun.
> > Signed-off-by: Mauro Carvalho Chehab <[email protected]>
>
> Both you and Ben Woodard are listed under MODULE_AUTHOR. Should this
> patch have Ben's Signed-off-by:? And/or his From:?
I'll ask his SOB.
> > +#ifndef PCI_DEVICE_ID_INTEL_5400_ERR
>
> This ifndef is unneeded: PCI_DEVICE_ID_INTEL_5400_ERR _is_ defined.
>
> What's more, it is actually bad, because if pci_ids.h were to #define
> PCI_DEVICE_ID_INTEL_5400_ERR to something other than 0x4030, the ifndef
> would prevent the compiler from informing us of this potential problem.
Fixed.
> > +char *error_name[] = {
>
> error_name is a global symbol.
Changed. Also changed it to const.
> > +/* masks for non-fatal error register */
> > +#define TO_NF_MASK(a) (((a) & EMASK_M29) | ((a) >> 3))
> > +#define FROM_NF_FERR(a) (((a) & EMASK_M29) | (((a) << 3) & ((1 << 30)-1)))
>
> These macros will misbehave if passed an expression with side-effects.
> Not really a problem as they're only used for constructing constants.
> I guess if one was being paranoid, one could #undef them again as soon
> as possible.
Good point. I've converted both macros into inline functions, since gcc will likely optimize
the code into a constant, instead of calculating at runtime.
> > +#ifdef CONFIG_EDAC_DEBUG
> > +/* MTR NUMROW */
> > +static char *numrow_toString[] = {
> > + "8,192 - 13 rows",
> > + "16,384 - 14 rows",
> > + "32,768 - 15 rows",
> > + "65,536 - 16 rows"
> > +};
> > +
> > +/* MTR NUMCOL */
> > +static char *numcol_toString[] = {
> > + "1,024 - 10 columns",
> > + "2,048 - 11 columns",
> > + "4,096 - 12 columns",
> > + "reserved"
> > +};
>
> These could possibly be made const. We don't modify them at runtime...
Fixed.
> 202 bytes. Tolerable, I guess.
>
> Heaven knows why and how this dimension was chosen, and under which
> circumstances it will overflow :(
>
> It looks like such a hack that perhaps you should give up and use
> kasprintf()+kfree()?
I also didn't like that code. It is present also on other edac drivers.
I'll fix this on a later patch. I'll probably propose some patches to the other
edac drivers that use this kind of construction.
> > + /* Decrement usage count for devices */
> > + if (pvt->branch_1)
> > + pci_dev_put(pvt->branch_1);
> > +
> > + if (pvt->branch_0)
> > + pci_dev_put(pvt->branch_0);
> > +
> > + if (pvt->fsb_error_regs)
> > + pci_dev_put(pvt->fsb_error_regs);
> > +
> > + if (pvt->branchmap_werrors)
> > + pci_dev_put(pvt->branchmap_werrors);
>
> pci_dev_put(NULL) is legal.
Fixed.
> > + mci->ctl_name = i5400_devs[dev_idx].ctl_name;
>
> This is a bit strange. i5400_devs[] has a single entry, so what's the
> point in indexing it with `dev_idx', which MUST have a value of zero
> anyway?
I'm not sure if we covered all PCI ID's for i5400 (maybe some i5400 variant may
be released?). That's why I preferred to keep a table: makes easier to add
other entries there, but I see your point.
I'll add a code to prevent dev_idx values equal or higher than
ARRAY_SIZE, to avoid the risk of trying to acess something outside the buffer.
> > +static struct pci_driver i5400_driver = {
> > + .name = KBUILD_BASENAME,
> > + .probe = i5400_init_one,
> > + .remove = __devexit_p(i5400_remove_one),
> > + .id_table = i5400_pci_tbl,
> > +};
>
> Only one other driver in the tree uses KBUILD_BASENAME, and that is
> drivers/edac/i5100_edac.c, from which I suspect the above was copied.
>
> I suspect that something is being done wrongly here.
Fixed.
I'll run some tests here, and submit a version 2 of the patch soon.
Cheers,
Mauro
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Andrew Morton wrote:
>
>> Signed-off-by: Mauro Carvalho Chehab <[email protected]>
>
> Both you and Ben Woodard are listed under MODULE_AUTHOR. Should
> this patch have Ben's Signed-off-by:? And/or his From:?
Go ahead and add a:
Signed-off-by: Ben Woodard <[email protected]>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)
Comment: Using GnuPG with Red Hat - http://enigmail.mozdev.org
iD8DBQFJNufYYuGAPpw3tAARAlA5AJ4hNh6Mzzs60atqL06MkoD7RulsbwCdGSP7
DIQiX8F4+1eQXYeZPSQkc0M=
=anBc
-----END PGP SIGNATURE-----
From: Mauro Carvalho Chehab <[email protected]>
This patch adds a driver for support errors reported by i5400 MCH chipset (Seaburg)
Signed-off-by: Ben Woodard <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
index e0dbd38..c1fcbed 100644
--- a/drivers/edac/Kconfig
+++ b/drivers/edac/Kconfig
@@ -109,6 +109,13 @@ config EDAC_X38
Support for error detection and correction on the Intel
X38 server chipsets.
+config EDAC_I5400
+ tristate "Intel 5400 (Seaburg) chipsets"
+ depends on EDAC_MM_EDAC && PCI && X86
+ help
+ Support for error detection and correction the Intel
+ i5400 MCH chipset (Seaburg).
+
config EDAC_I82860
tristate "Intel 82860"
depends on EDAC_MM_EDAC && PCI && X86_32
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
index 62c2d9b..b751969 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -20,6 +20,7 @@ endif
obj-$(CONFIG_EDAC_AMD76X) += amd76x_edac.o
obj-$(CONFIG_EDAC_I5000) += i5000_edac.o
obj-$(CONFIG_EDAC_I5100) += i5100_edac.o
+obj-$(CONFIG_EDAC_I5400) += i5400_edac.o
obj-$(CONFIG_EDAC_E7XXX) += e7xxx_edac.o
obj-$(CONFIG_EDAC_E752X) += e752x_edac.o
obj-$(CONFIG_EDAC_I82443BXGX) += i82443bxgx_edac.o
diff --git a/drivers/edac/i5400_edac.c b/drivers/edac/i5400_edac.c
new file mode 100644
index 0000000..4568bc9
--- /dev/null
+++ b/drivers/edac/i5400_edac.c
@@ -0,0 +1,1476 @@
+/*
+ * Intel 5400 class Memory Controllers kernel module (Seaburg)
+ *
+ * This file may be distributed under the terms of the
+ * GNU General Public License.
+ *
+ * Copyright (c) 2008 by:
+ * Ben Woodard <[email protected]>
+ * Mauro Carvalho Chehab <[email protected]>
+ *
+ * Red Hat Inc. http://www.redhat.com
+ *
+ * Forked and adapted from the i5000_edac driver which was
+ * written by Douglas Thompson Linux Networx <[email protected]>
+ *
+ * This module is based on the following document:
+ *
+ * Intel 5400 Chipset Memory Controller Hub (MCH) - Datasheet
+ * http://developer.intel.com/design/chipsets/datashts/313070.htm
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/pci_ids.h>
+#include <linux/slab.h>
+#include <linux/edac.h>
+#include <linux/mmzone.h>
+
+#include "edac_core.h"
+
+/*
+ * Alter this version for the I5400 module when modifications are made
+ */
+#define I5400_REVISION " Ver: 1.0.0 " __DATE__
+
+#define EDAC_MOD_STR "i5400_edac"
+
+#define i5400_printk(level, fmt, arg...) \
+ edac_printk(level, "i5400", fmt, ##arg)
+
+#define i5400_mc_printk(mci, level, fmt, arg...) \
+ edac_mc_chipset_printk(mci, level, "i5400", fmt, ##arg)
+
+/* Limits for i5400 */
+#define NUM_MTRS_PER_BRANCH 4
+#define CHANNELS_PER_BRANCH 2
+#define MAX_CHANNELS 4
+#define MAX_DIMMS (MAX_CHANNELS * 4) /* Up to 4 DIMM's per channel */
+#define MAX_CSROWS (MAX_DIMMS * 2) /* max possible csrows per channel */
+
+/* Device 16,
+ * Function 0: System Address
+ * Function 1: Memory Branch Map, Control, Errors Register
+ * Function 2: FSB Error Registers
+ *
+ * All 3 functions of Device 16 (0,1,2) share the SAME DID and
+ * uses PCI_DEVICE_ID_INTEL_5400_ERR for device 16 (0,1,2),
+ * PCI_DEVICE_ID_INTEL_5400_FBD0 and PCI_DEVICE_ID_INTEL_5400_FBD1
+ * for device 21 (0,1).
+ */
+
+ /* OFFSETS for Function 0 */
+#define AMBASE 0x48 /* AMB Mem Mapped Reg Region Base */
+#define MAXCH 0x56 /* Max Channel Number */
+#define MAXDIMMPERCH 0x57 /* Max DIMM PER Channel Number */
+
+ /* OFFSETS for Function 1 */
+#define TOLM 0x6C
+#define REDMEMB 0x7C
+#define REC_ECC_LOCATOR_ODD(x) ((x) & 0x3fe00) /* bits [17:9] indicate ODD, [8:0] indicate EVEN */
+#define MIR0 0x80
+#define MIR1 0x84
+#define AMIR0 0x8c
+#define AMIR1 0x90
+
+ /* Fatal error registers */
+#define FERR_FAT_FBD 0x98 /* also called as FERR_FAT_FB_DIMM at datasheet */
+#define FERR_FAT_FBDCHAN (3<<28) /* channel index where the highest-order error occurred */
+
+#define NERR_FAT_FBD 0x9c
+#define FERR_NF_FBD 0xa0 /* also called as FERR_NFAT_FB_DIMM at datasheet */
+
+ /* Non-fatal error register */
+#define NERR_NF_FBD 0xa4
+
+ /* Enable error mask */
+#define EMASK_FBD 0xa8
+
+#define ERR0_FBD 0xac
+#define ERR1_FBD 0xb0
+#define ERR2_FBD 0xb4
+#define MCERR_FBD 0xb8
+
+ /* No OFFSETS for Device 16 Function 2 */
+
+/*
+ * Device 21,
+ * Function 0: Memory Map Branch 0
+ *
+ * Device 22,
+ * Function 0: Memory Map Branch 1
+ */
+
+ /* OFFSETS for Function 0 */
+#define AMBPRESENT_0 0x64
+#define AMBPRESENT_1 0x66
+#define MTR0 0x80
+#define MTR1 0x82
+#define MTR2 0x84
+#define MTR3 0x86
+
+ /* OFFSETS for Function 1 */
+#define NRECFGLOG 0x74
+#define RECFGLOG 0x78
+#define NRECMEMA 0xbe
+#define NRECMEMB 0xc0
+#define NRECFB_DIMMA 0xc4
+#define NRECFB_DIMMB 0xc8
+#define NRECFB_DIMMC 0xcc
+#define NRECFB_DIMMD 0xd0
+#define NRECFB_DIMME 0xd4
+#define NRECFB_DIMMF 0xd8
+#define REDMEMA 0xdC
+#define RECMEMA 0xf0
+#define RECMEMB 0xf4
+#define RECFB_DIMMA 0xf8
+#define RECFB_DIMMB 0xec
+#define RECFB_DIMMC 0xf0
+#define RECFB_DIMMD 0xf4
+#define RECFB_DIMME 0xf8
+#define RECFB_DIMMF 0xfC
+
+/*
+ * Error indicator bits and masks
+ * Error masks are according with Table 5-17 of i5400 datasheet
+ */
+
+enum error_mask {
+ EMASK_M1 = 1<<0, /* Memory Write error on non-redundant retry */
+ EMASK_M2 = 1<<1, /* Memory or FB-DIMM configuration CRC read error */
+ EMASK_M3 = 1<<2, /* Reserved */
+ EMASK_M4 = 1<<3, /* Uncorrectable Data ECC on Replay */
+ EMASK_M5 = 1<<4, /* Aliased Uncorrectable Non-Mirrored Demand Data ECC */
+ EMASK_M6 = 1<<5, /* Unsupported on i5400 */
+ EMASK_M7 = 1<<6, /* Aliased Uncorrectable Resilver- or Spare-Copy Data ECC */
+ EMASK_M8 = 1<<7, /* Aliased Uncorrectable Patrol Data ECC */
+ EMASK_M9 = 1<<8, /* Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC */
+ EMASK_M10 = 1<<9, /* Unsupported on i5400 */
+ EMASK_M11 = 1<<10, /* Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC */
+ EMASK_M12 = 1<<11, /* Non-Aliased Uncorrectable Patrol Data ECC */
+ EMASK_M13 = 1<<12, /* Memory Write error on first attempt */
+ EMASK_M14 = 1<<13, /* FB-DIMM Configuration Write error on first attempt */
+ EMASK_M15 = 1<<14, /* Memory or FB-DIMM configuration CRC read error */
+ EMASK_M16 = 1<<15, /* Channel Failed-Over Occurred */
+ EMASK_M17 = 1<<16, /* Correctable Non-Mirrored Demand Data ECC */
+ EMASK_M18 = 1<<17, /* Unsupported on i5400 */
+ EMASK_M19 = 1<<18, /* Correctable Resilver- or Spare-Copy Data ECC */
+ EMASK_M20 = 1<<19, /* Correctable Patrol Data ECC */
+ EMASK_M21 = 1<<20, /* FB-DIMM Northbound parity error on FB-DIMM Sync Status */
+ EMASK_M22 = 1<<21, /* SPD protocol Error */
+ EMASK_M23 = 1<<22, /* Non-Redundant Fast Reset Timeout */
+ EMASK_M24 = 1<<23, /* Refresh error */
+ EMASK_M25 = 1<<24, /* Memory Write error on redundant retry */
+ EMASK_M26 = 1<<25, /* Redundant Fast Reset Timeout */
+ EMASK_M27 = 1<<26, /* Correctable Counter Threshold Exceeded */
+ EMASK_M28 = 1<<27, /* DIMM-Spare Copy Completed */
+ EMASK_M29 = 1<<28, /* DIMM-Isolation Completed */
+};
+
+/*
+ * Names to translate bit error into something useful
+ */
+static const char *error_name[] = {
+ [0] = "Memory Write error on non-redundant retry",
+ [1] = "Memory or FB-DIMM configuration CRC read error",
+ /* Reserved */
+ [3] = "Uncorrectable Data ECC on Replay",
+ [4] = "Aliased Uncorrectable Non-Mirrored Demand Data ECC",
+ /* M6 Unsupported on i5400 */
+ [6] = "Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
+ [7] = "Aliased Uncorrectable Patrol Data ECC",
+ [8] = "Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC",
+ /* M10 Unsupported on i5400 */
+ [10] = "Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC",
+ [11] = "Non-Aliased Uncorrectable Patrol Data ECC",
+ [12] = "Memory Write error on first attempt",
+ [13] = "FB-DIMM Configuration Write error on first attempt",
+ [14] = "Memory or FB-DIMM configuration CRC read error",
+ [15] = "Channel Failed-Over Occurred",
+ [16] = "Correctable Non-Mirrored Demand Data ECC",
+ /* M18 Unsupported on i5400 */
+ [18] = "Correctable Resilver- or Spare-Copy Data ECC",
+ [19] = "Correctable Patrol Data ECC",
+ [20] = "FB-DIMM Northbound parity error on FB-DIMM Sync Status",
+ [21] = "SPD protocol Error",
+ [22] = "Non-Redundant Fast Reset Timeout",
+ [23] = "Refresh error",
+ [24] = "Memory Write error on redundant retry",
+ [25] = "Redundant Fast Reset Timeout",
+ [26] = "Correctable Counter Threshold Exceeded",
+ [27] = "DIMM-Spare Copy Completed",
+ [28] = "DIMM-Isolation Completed",
+};
+
+/* Fatal errors */
+#define ERROR_FAT_MASK (EMASK_M1 | \
+ EMASK_M2 | \
+ EMASK_M23)
+
+/* Correctable errors */
+#define ERROR_NF_CORRECTABLE (EMASK_M27 | \
+ EMASK_M20 | \
+ EMASK_M19 | \
+ EMASK_M18 | \
+ EMASK_M17 | \
+ EMASK_M16)
+#define ERROR_NF_DIMM_SPARE (EMASK_M29 | \
+ EMASK_M28)
+#define ERROR_NF_SPD_PROTOCOL (EMASK_M22)
+#define ERROR_NF_NORTH_CRC (EMASK_M21)
+
+/* Recoverable errors */
+#define ERROR_NF_RECOVERABLE (EMASK_M26 | \
+ EMASK_M25 | \
+ EMASK_M24 | \
+ EMASK_M15 | \
+ EMASK_M14 | \
+ EMASK_M13 | \
+ EMASK_M12 | \
+ EMASK_M11 | \
+ EMASK_M9 | \
+ EMASK_M8 | \
+ EMASK_M7 | \
+ EMASK_M5)
+
+/* uncorrectable errors */
+#define ERROR_NF_UNCORRECTABLE (EMASK_M4)
+
+/* mask to all non-fatal errors */
+#define ERROR_NF_MASK (ERROR_NF_CORRECTABLE | \
+ ERROR_NF_UNCORRECTABLE | \
+ ERROR_NF_RECOVERABLE | \
+ ERROR_NF_DIMM_SPARE | \
+ ERROR_NF_SPD_PROTOCOL | \
+ ERROR_NF_NORTH_CRC)
+
+/*
+ * Define error masks for the several registers
+ */
+
+/* Enable all fatal and non fatal errors */
+#define ENABLE_EMASK_ALL (ERROR_FAT_MASK | ERROR_NF_MASK)
+
+/* mask for fatal error registers */
+#define FERR_FAT_MASK ERROR_FAT_MASK
+
+/* masks for non-fatal error register */
+static inline int to_nf_mask(unsigned int mask)
+{
+ return (mask & EMASK_M29) | (mask >> 3);
+};
+
+static inline int from_nf_ferr(unsigned int mask)
+{
+ return (mask & EMASK_M29) | /* Bit 28 */
+ (mask & ((1 << 28) - 1) << 3); /* Bits 0 to 27 */
+};
+
+#define FERR_NF_MASK to_nf_mask(ERROR_NF_MASK)
+#define FERR_NF_CORRECTABLE to_nf_mask(ERROR_NF_CORRECTABLE)
+#define FERR_NF_DIMM_SPARE to_nf_mask(ERROR_NF_DIMM_SPARE)
+#define FERR_NF_SPD_PROTOCOL to_nf_mask(ERROR_NF_SPD_PROTOCOL)
+#define FERR_NF_NORTH_CRC to_nf_mask(ERROR_NF_NORTH_CRC)
+#define FERR_NF_RECOVERABLE to_nf_mask(ERROR_NF_RECOVERABLE)
+#define FERR_NF_UNCORRECTABLE to_nf_mask(ERROR_NF_UNCORRECTABLE)
+
+/* Defines to extract the vaious fields from the
+ * MTRx - Memory Technology Registers
+ */
+#define MTR_DIMMS_PRESENT(mtr) ((mtr) & (1 << 10))
+#define MTR_DIMMS_ETHROTTLE(mtr) ((mtr) & (1 << 9))
+#define MTR_DRAM_WIDTH(mtr) (((mtr) & (1 << 8)) ? 8 : 4)
+#define MTR_DRAM_BANKS(mtr) (((mtr) & (1 << 6)) ? 8 : 4)
+#define MTR_DRAM_BANKS_ADDR_BITS(mtr) ((MTR_DRAM_BANKS(mtr) == 8) ? 3 : 2)
+#define MTR_DIMM_RANK(mtr) (((mtr) >> 5) & 0x1)
+#define MTR_DIMM_RANK_ADDR_BITS(mtr) (MTR_DIMM_RANK(mtr) ? 2 : 1)
+#define MTR_DIMM_ROWS(mtr) (((mtr) >> 2) & 0x3)
+#define MTR_DIMM_ROWS_ADDR_BITS(mtr) (MTR_DIMM_ROWS(mtr) + 13)
+#define MTR_DIMM_COLS(mtr) ((mtr) & 0x3)
+#define MTR_DIMM_COLS_ADDR_BITS(mtr) (MTR_DIMM_COLS(mtr) + 10)
+
+/* This applies to FERR_NF_FB-DIMM as well as FERR_FAT_FB-DIMM */
+static inline int extract_fbdchan_indx(u32 x)
+{
+ return (x>>28) & 0x3;
+}
+
+#ifdef CONFIG_EDAC_DEBUG
+/* MTR NUMROW */
+static const char *numrow_toString[] = {
+ "8,192 - 13 rows",
+ "16,384 - 14 rows",
+ "32,768 - 15 rows",
+ "65,536 - 16 rows"
+};
+
+/* MTR NUMCOL */
+static const char *numcol_toString[] = {
+ "1,024 - 10 columns",
+ "2,048 - 11 columns",
+ "4,096 - 12 columns",
+ "reserved"
+};
+#endif
+
+/* Device name and register DID (Device ID) */
+struct i5400_dev_info {
+ const char *ctl_name; /* name for this device */
+ u16 fsb_mapping_errors; /* DID for the branchmap,control */
+};
+
+/* Table of devices attributes supported by this driver */
+static const struct i5400_dev_info i5400_devs[] = {
+ {
+ .ctl_name = "I5400",
+ .fsb_mapping_errors = PCI_DEVICE_ID_INTEL_5400_ERR,
+ },
+};
+
+struct i5400_dimm_info {
+ int megabytes; /* size, 0 means not present */
+ int dual_rank;
+};
+
+/* driver private data structure */
+struct i5400_pvt {
+ struct pci_dev *system_address; /* 16.0 */
+ struct pci_dev *branchmap_werrors; /* 16.1 */
+ struct pci_dev *fsb_error_regs; /* 16.2 */
+ struct pci_dev *branch_0; /* 21.0 */
+ struct pci_dev *branch_1; /* 22.0 */
+
+ u16 tolm; /* top of low memory */
+ u64 ambase; /* AMB BAR */
+
+ u16 mir0, mir1;
+
+ u16 b0_mtr[NUM_MTRS_PER_BRANCH]; /* Memory Technlogy Reg */
+ u16 b0_ambpresent0; /* Branch 0, Channel 0 */
+ u16 b0_ambpresent1; /* Brnach 0, Channel 1 */
+
+ u16 b1_mtr[NUM_MTRS_PER_BRANCH]; /* Memory Technlogy Reg */
+ u16 b1_ambpresent0; /* Branch 1, Channel 8 */
+ u16 b1_ambpresent1; /* Branch 1, Channel 1 */
+
+ /* DIMM information matrix, allocating architecture maximums */
+ struct i5400_dimm_info dimm_info[MAX_CSROWS][MAX_CHANNELS];
+
+ /* Actual values for this controller */
+ int maxch; /* Max channels */
+ int maxdimmperch; /* Max DIMMs per channel */
+};
+
+/* I5400 MCH error information retrieved from Hardware */
+struct i5400_error_info {
+ /* These registers are always read from the MC */
+ u32 ferr_fat_fbd; /* First Errors Fatal */
+ u32 nerr_fat_fbd; /* Next Errors Fatal */
+ u32 ferr_nf_fbd; /* First Errors Non-Fatal */
+ u32 nerr_nf_fbd; /* Next Errors Non-Fatal */
+
+ /* These registers are input ONLY if there was a Recoverable Error */
+ u32 redmemb; /* Recoverable Mem Data Error log B */
+ u16 recmema; /* Recoverable Mem Error log A */
+ u32 recmemb; /* Recoverable Mem Error log B */
+
+ /* These registers are input ONLY if there was a Non-Rec Error */
+ u16 nrecmema; /* Non-Recoverable Mem log A */
+ u16 nrecmemb; /* Non-Recoverable Mem log B */
+
+};
+
+/* note that nrec_rdwr changed from NRECMEMA to NRECMEMB between the 5000 and
+ 5400 better to use an inline function than a macro in this case */
+static inline int nrec_bank(struct i5400_error_info *info)
+{
+ return ((info->nrecmema) >> 12) & 0x7;
+}
+static inline int nrec_rank(struct i5400_error_info *info)
+{
+ return ((info->nrecmema) >> 8) & 0xf;
+}
+static inline int nrec_buf_id(struct i5400_error_info *info)
+{
+ return ((info->nrecmema)) & 0xff;
+}
+static inline int nrec_rdwr(struct i5400_error_info *info)
+{
+ return (info->nrecmemb) >> 31;
+}
+/* This applies to both NREC and REC string so it can be used with nrec_rdwr
+ and rec_rdwr */
+static inline const char *rdwr_str(int rdwr)
+{
+ return rdwr ? "Write" : "Read";
+}
+static inline int nrec_cas(struct i5400_error_info *info)
+{
+ return ((info->nrecmemb) >> 16) & 0x1fff;
+}
+static inline int nrec_ras(struct i5400_error_info *info)
+{
+ return (info->nrecmemb) & 0xffff;
+}
+static inline int rec_bank(struct i5400_error_info *info)
+{
+ return ((info->recmema) >> 12) & 0x7;
+}
+static inline int rec_rank(struct i5400_error_info *info)
+{
+ return ((info->recmema) >> 8) & 0xf;
+}
+static inline int rec_rdwr(struct i5400_error_info *info)
+{
+ return (info->recmemb) >> 31;
+}
+static inline int rec_cas(struct i5400_error_info *info)
+{
+ return ((info->recmemb) >> 16) & 0x1fff;
+}
+static inline int rec_ras(struct i5400_error_info *info)
+{
+ return (info->recmemb) & 0xffff;
+}
+
+static struct edac_pci_ctl_info *i5400_pci;
+
+/*
+ * i5400_get_error_info Retrieve the hardware error information from
+ * the hardware and cache it in the 'info'
+ * structure
+ */
+static void i5400_get_error_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info)
+{
+ struct i5400_pvt *pvt;
+ u32 value;
+
+ pvt = mci->pvt_info;
+
+ /* read in the 1st FATAL error register */
+ pci_read_config_dword(pvt->branchmap_werrors, FERR_FAT_FBD, &value);
+
+ /* Mask only the bits that the doc says are valid
+ */
+ value &= (FERR_FAT_FBDCHAN | FERR_FAT_MASK);
+
+ /* If there is an error, then read in the
+ NEXT FATAL error register and the Memory Error Log Register A
+ */
+ if (value & FERR_FAT_MASK) {
+ info->ferr_fat_fbd = value;
+
+ /* harvest the various error data we need */
+ pci_read_config_dword(pvt->branchmap_werrors,
+ NERR_FAT_FBD, &info->nerr_fat_fbd);
+ pci_read_config_word(pvt->branchmap_werrors,
+ NRECMEMA, &info->nrecmema);
+ pci_read_config_word(pvt->branchmap_werrors,
+ NRECMEMB, &info->nrecmemb);
+
+ /* Clear the error bits, by writing them back */
+ pci_write_config_dword(pvt->branchmap_werrors,
+ FERR_FAT_FBD, value);
+ } else {
+ info->ferr_fat_fbd = 0;
+ info->nerr_fat_fbd = 0;
+ info->nrecmema = 0;
+ info->nrecmemb = 0;
+ }
+
+ /* read in the 1st NON-FATAL error register */
+ pci_read_config_dword(pvt->branchmap_werrors, FERR_NF_FBD, &value);
+
+ /* If there is an error, then read in the 1st NON-FATAL error
+ * register as well */
+ if (value & FERR_NF_MASK) {
+ info->ferr_nf_fbd = value;
+
+ /* harvest the various error data we need */
+ pci_read_config_dword(pvt->branchmap_werrors,
+ NERR_NF_FBD, &info->nerr_nf_fbd);
+ pci_read_config_word(pvt->branchmap_werrors,
+ RECMEMA, &info->recmema);
+ pci_read_config_dword(pvt->branchmap_werrors,
+ RECMEMB, &info->recmemb);
+ pci_read_config_dword(pvt->branchmap_werrors,
+ REDMEMB, &info->redmemb);
+
+ /* Clear the error bits, by writing them back */
+ pci_write_config_dword(pvt->branchmap_werrors,
+ FERR_NF_FBD, value);
+ } else {
+ info->ferr_nf_fbd = 0;
+ info->nerr_nf_fbd = 0;
+ info->recmema = 0;
+ info->recmemb = 0;
+ info->redmemb = 0;
+ }
+}
+
+/*
+ * i5400_proccess_non_recoverable_info(struct mem_ctl_info *mci,
+ * struct i5400_error_info *info,
+ * int handle_errors);
+ *
+ * handle the Intel FATAL and unrecoverable errors, if any
+ */
+static void i5400_proccess_non_recoverable_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info,
+ unsigned long allErrors)
+{
+ char msg[EDAC_MC_LABEL_LEN + 1 + 90 + 80];
+ int branch;
+ int channel;
+ int bank;
+ int buf_id;
+ int rank;
+ int rdwr;
+ int ras, cas;
+ int errnum;
+ char *type = NULL;
+
+ if (!allErrors)
+ return; /* if no error, return now */
+
+ if (allErrors & ERROR_FAT_MASK)
+ type = "FATAL";
+ else if (allErrors & FERR_NF_UNCORRECTABLE)
+ type = "NON-FATAL uncorrected";
+ else
+ type = "NON-FATAL recoverable";
+
+ /* ONLY ONE of the possible error bits will be set, as per the docs */
+
+ branch = extract_fbdchan_indx(info->ferr_fat_fbd);
+ channel = branch;
+
+ /* Use the NON-Recoverable macros to extract data */
+ bank = nrec_bank(info);
+ rank = nrec_rank(info);
+ buf_id = nrec_buf_id(info);
+ rdwr = nrec_rdwr(info);
+ ras = nrec_ras(info);
+ cas = nrec_cas(info);
+
+ debugf0("\t\tCSROW= %d Channels= %d,%d (Branch= %d "
+ "DRAM Bank= %d Buffer ID = %d rdwr= %s ras= %d cas= %d)\n",
+ rank, channel, channel + 1, branch >> 1, bank,
+ buf_id, rdwr_str(rdwr), ras, cas);
+
+ /* Only 1 bit will be on */
+ errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
+
+ /* Form out message */
+ snprintf(msg, sizeof(msg),
+ "%s (Branch=%d DRAM-Bank=%d Buffer ID = %d RDWR=%s "
+ "RAS=%d CAS=%d %s Err=0x%lx (%s))",
+ type, branch >> 1, bank, buf_id, rdwr_str(rdwr), ras, cas,
+ type, allErrors, error_name[errnum]);
+
+ /* Call the helper to output message */
+ edac_mc_handle_fbd_ue(mci, rank, channel, channel + 1, msg);
+}
+
+/*
+ * i5400_process_fatal_error_info(struct mem_ctl_info *mci,
+ * struct i5400_error_info *info,
+ * int handle_errors);
+ *
+ * handle the Intel NON-FATAL errors, if any
+ */
+static void i5400_process_nonfatal_error_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info)
+{
+ char msg[EDAC_MC_LABEL_LEN + 1 + 90 + 80];
+ unsigned long allErrors;
+ int branch;
+ int channel;
+ int bank;
+ int rank;
+ int rdwr;
+ int ras, cas;
+ int errnum;
+
+ /* mask off the Error bits that are possible */
+ allErrors = from_nf_ferr(info->ferr_nf_fbd & FERR_NF_MASK);
+ if (!allErrors)
+ return; /* if no error, return now */
+
+ /* ONLY ONE of the possible error bits will be set, as per the docs */
+
+ if (allErrors & (ERROR_NF_UNCORRECTABLE | ERROR_NF_RECOVERABLE)) {
+ i5400_proccess_non_recoverable_info(mci, info, allErrors);
+ return;
+ }
+
+ /* Correctable errors */
+ if (allErrors & ERROR_NF_CORRECTABLE) {
+ debugf0("\tCorrected bits= 0x%lx\n", allErrors);
+
+ branch = extract_fbdchan_indx(info->ferr_nf_fbd);
+
+ channel = 0;
+ if (REC_ECC_LOCATOR_ODD(info->redmemb))
+ channel = 1;
+
+ /* Convert channel to be based from zero, instead of
+ * from branch base of 0 */
+ channel += branch;
+
+ bank = rec_bank(info);
+ rank = rec_rank(info);
+ rdwr = rec_rdwr(info);
+ ras = rec_ras(info);
+ cas = rec_cas(info);
+
+ /* Only 1 bit will be on */
+ errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
+
+ debugf0("\t\tCSROW= %d Channel= %d (Branch %d "
+ "DRAM Bank= %d rdwr= %s ras= %d cas= %d)\n",
+ rank, channel, branch >> 1, bank,
+ rdwr_str(rdwr), ras, cas);
+
+ /* Form out message */
+ snprintf(msg, sizeof(msg),
+ "Corrected error (Branch=%d DRAM-Bank=%d RDWR=%s "
+ "RAS=%d CAS=%d, CE Err=0x%lx (%s))",
+ branch >> 1, bank, rdwr_str(rdwr), ras, cas,
+ allErrors, error_name[errnum]);
+
+ /* Call the helper to output message */
+ edac_mc_handle_fbd_ce(mci, rank, channel, msg);
+
+ return;
+ }
+
+ /* Miscelaneous errors */
+ errnum = find_first_bit(&allErrors, ARRAY_SIZE(error_name));
+
+ branch = extract_fbdchan_indx(info->ferr_nf_fbd);
+
+ i5400_mc_printk(mci, KERN_EMERG,
+ "Non-Fatal misc error (Branch=%d Err=%#lx (%s))",
+ branch >> 1, allErrors, error_name[errnum]);
+}
+
+/*
+ * i5400_process_error_info Process the error info that is
+ * in the 'info' structure, previously retrieved from hardware
+ */
+static void i5400_process_error_info(struct mem_ctl_info *mci,
+ struct i5400_error_info *info)
+{ u32 allErrors;
+
+ /* First handle any fatal errors that occurred */
+ allErrors = (info->ferr_fat_fbd & FERR_FAT_MASK);
+ i5400_proccess_non_recoverable_info(mci, info, allErrors);
+
+ /* now handle any non-fatal errors that occurred */
+ i5400_process_nonfatal_error_info(mci, info);
+}
+
+/*
+ * i5400_clear_error Retrieve any error from the hardware
+ * but do NOT process that error.
+ * Used for 'clearing' out of previous errors
+ * Called by the Core module.
+ */
+static void i5400_clear_error(struct mem_ctl_info *mci)
+{
+ struct i5400_error_info info;
+
+ i5400_get_error_info(mci, &info);
+}
+
+/*
+ * i5400_check_error Retrieve and process errors reported by the
+ * hardware. Called by the Core module.
+ */
+static void i5400_check_error(struct mem_ctl_info *mci)
+{
+ struct i5400_error_info info;
+ debugf4("MC%d: " __FILE__ ": %s()\n", mci->mc_idx, __func__);
+ i5400_get_error_info(mci, &info);
+ i5400_process_error_info(mci, &info);
+}
+
+/*
+ * i5400_put_devices 'put' all the devices that we have
+ * reserved via 'get'
+ */
+static void i5400_put_devices(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+
+ pvt = mci->pvt_info;
+
+ /* Decrement usage count for devices */
+ pci_dev_put(pvt->branch_1);
+ pci_dev_put(pvt->branch_0);
+ pci_dev_put(pvt->fsb_error_regs);
+ pci_dev_put(pvt->branchmap_werrors);
+}
+
+/*
+ * i5400_get_devices Find and perform 'get' operation on the MCH's
+ * device/functions we want to reference for this driver
+ *
+ * Need to 'get' device 16 func 1 and func 2
+ */
+static int i5400_get_devices(struct mem_ctl_info *mci, int dev_idx)
+{
+ struct i5400_pvt *pvt;
+ struct pci_dev *pdev;
+
+ pvt = mci->pvt_info;
+ pvt->branchmap_werrors = NULL;
+ pvt->fsb_error_regs = NULL;
+ pvt->branch_0 = NULL;
+ pvt->branch_1 = NULL;
+
+ /* Attempt to 'get' the MCH register we want */
+ pdev = NULL;
+ while (!pvt->branchmap_werrors || !pvt->fsb_error_regs) {
+ pdev = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_ERR, pdev);
+ if (!pdev) {
+ /* End of list, leave */
+ i5400_printk(KERN_ERR,
+ "'system address,Process Bus' "
+ "device not found:"
+ "vendor 0x%x device 0x%x ERR funcs "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_ERR);
+ goto error;
+ }
+
+ /* Store device 16 funcs 1 and 2 */
+ switch (PCI_FUNC(pdev->devfn)) {
+ case 1:
+ pvt->branchmap_werrors = pdev;
+ break;
+ case 2:
+ pvt->fsb_error_regs = pdev;
+ break;
+ }
+ }
+
+ debugf1("System Address, processor bus- PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->system_address),
+ pvt->system_address->vendor, pvt->system_address->device);
+ debugf1("Branchmap, control and errors - PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->branchmap_werrors),
+ pvt->branchmap_werrors->vendor, pvt->branchmap_werrors->device);
+ debugf1("FSB Error Regs - PCI Bus ID: %s %x:%x\n",
+ pci_name(pvt->fsb_error_regs),
+ pvt->fsb_error_regs->vendor, pvt->fsb_error_regs->device);
+
+ pvt->branch_0 = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_FBD0, NULL);
+ if (!pvt->branch_0) {
+ i5400_printk(KERN_ERR,
+ "MC: 'BRANCH 0' device not found:"
+ "vendor 0x%x device 0x%x Func 0 (broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_5400_FBD0);
+ goto error;
+ }
+
+ /* If this device claims to have more than 2 channels then
+ * fetch Branch 1's information
+ */
+ if (pvt->maxch < CHANNELS_PER_BRANCH)
+ return 0;
+
+ pvt->branch_1 = pci_get_device(PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_FBD1, NULL);
+ if (!pvt->branch_1) {
+ i5400_printk(KERN_ERR,
+ "MC: 'BRANCH 1' device not found:"
+ "vendor 0x%x device 0x%x Func 0 "
+ "(broken BIOS?)\n",
+ PCI_VENDOR_ID_INTEL,
+ PCI_DEVICE_ID_INTEL_5400_FBD1);
+ goto error;
+ }
+
+ return 0;
+
+error:
+ i5400_put_devices(mci);
+ return -ENODEV;
+}
+
+/*
+ * determine_amb_present
+ *
+ * the information is contained in NUM_MTRS_PER_BRANCH different
+ * registers determining which of the NUM_MTRS_PER_BRANCH requires
+ * knowing which channel is in question
+ *
+ * 2 branches, each with 2 channels
+ * b0_ambpresent0 for channel '0'
+ * b0_ambpresent1 for channel '1'
+ * b1_ambpresent0 for channel '2'
+ * b1_ambpresent1 for channel '3'
+ */
+static int determine_amb_present_reg(struct i5400_pvt *pvt, int channel)
+{
+ int amb_present;
+
+ if (channel < CHANNELS_PER_BRANCH) {
+ if (channel & 0x1)
+ amb_present = pvt->b0_ambpresent1;
+ else
+ amb_present = pvt->b0_ambpresent0;
+ } else {
+ if (channel & 0x1)
+ amb_present = pvt->b1_ambpresent1;
+ else
+ amb_present = pvt->b1_ambpresent0;
+ }
+
+ return amb_present;
+}
+
+/*
+ * determine_mtr(pvt, csrow, channel)
+ *
+ * return the proper MTR register as determine by the csrow and desired channel
+ */
+static int determine_mtr(struct i5400_pvt *pvt, int csrow, int channel)
+{
+ int mtr;
+ int n;
+
+ /* There is one MTR for each slot pair of FB-DIMMs,
+ Each slot may have one or two ranks (2 csrows),
+ Each slot pair may be at branch 0 or branch 1.
+ So, csrow should be divided by eight
+ */
+ n = csrow >> 3;
+
+ if (n >= NUM_MTRS_PER_BRANCH) {
+ debugf0("ERROR: trying to access an invalid csrow: %d\n",
+ csrow);
+ return 0;
+ }
+
+ if (channel < CHANNELS_PER_BRANCH)
+ mtr = pvt->b0_mtr[n];
+ else
+ mtr = pvt->b1_mtr[n];
+
+ return mtr;
+}
+
+/*
+ */
+static void decode_mtr(int slot_row, u16 mtr)
+{
+ int ans;
+
+ ans = MTR_DIMMS_PRESENT(mtr);
+
+ debugf2("\tMTR%d=0x%x: DIMMs are %s\n", slot_row, mtr,
+ ans ? "Present" : "NOT Present");
+ if (!ans)
+ return;
+
+ debugf2("\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr));
+
+ debugf2("\t\tELECTRICAL THROTTLING is %s\n",
+ MTR_DIMMS_ETHROTTLE(mtr) ? "enabled" : "disabled");
+
+ debugf2("\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr));
+ debugf2("\t\tNUMRANK: %s\n", MTR_DIMM_RANK(mtr) ? "double" : "single");
+ debugf2("\t\tNUMROW: %s\n", numrow_toString[MTR_DIMM_ROWS(mtr)]);
+ debugf2("\t\tNUMCOL: %s\n", numcol_toString[MTR_DIMM_COLS(mtr)]);
+}
+
+static void handle_channel(struct i5400_pvt *pvt, int csrow, int channel,
+ struct i5400_dimm_info *dinfo)
+{
+ int mtr;
+ int amb_present_reg;
+ int addrBits;
+
+ mtr = determine_mtr(pvt, csrow, channel);
+ if (MTR_DIMMS_PRESENT(mtr)) {
+ amb_present_reg = determine_amb_present_reg(pvt, channel);
+
+ /* Determine if there is a DIMM present in this DIMM slot */
+ if (amb_present_reg & (1 << (csrow >> 1))) {
+ dinfo->dual_rank = MTR_DIMM_RANK(mtr);
+
+ if (!((dinfo->dual_rank == 0) &&
+ ((csrow & 0x1) == 0x1))) {
+ /* Start with the number of bits for a Bank
+ * on the DRAM */
+ addrBits = MTR_DRAM_BANKS_ADDR_BITS(mtr);
+ /* Add thenumber of ROW bits */
+ addrBits += MTR_DIMM_ROWS_ADDR_BITS(mtr);
+ /* add the number of COLUMN bits */
+ addrBits += MTR_DIMM_COLS_ADDR_BITS(mtr);
+
+ addrBits += 6; /* add 64 bits per DIMM */
+ addrBits -= 20; /* divide by 2^^20 */
+ addrBits -= 3; /* 8 bits per bytes */
+
+ dinfo->megabytes = 1 << addrBits;
+ }
+ }
+ }
+}
+
+/*
+ * calculate_dimm_size
+ *
+ * also will output a DIMM matrix map, if debug is enabled, for viewing
+ * how the DIMMs are populated
+ */
+static void calculate_dimm_size(struct i5400_pvt *pvt)
+{
+ struct i5400_dimm_info *dinfo;
+ int csrow, max_csrows;
+ char *p, *mem_buffer;
+ int space, n;
+ int channel;
+
+ /* ================= Generate some debug output ================= */
+ space = PAGE_SIZE;
+ mem_buffer = p = kmalloc(space, GFP_KERNEL);
+ if (p == NULL) {
+ i5400_printk(KERN_ERR, "MC: %s:%s() kmalloc() failed\n",
+ __FILE__, __func__);
+ return;
+ }
+
+ /* Scan all the actual CSROWS (which is # of DIMMS * 2)
+ * and calculate the information for each DIMM
+ * Start with the highest csrow first, to display it first
+ * and work toward the 0th csrow
+ */
+ max_csrows = pvt->maxdimmperch * 2;
+ for (csrow = max_csrows - 1; csrow >= 0; csrow--) {
+
+ /* on an odd csrow, first output a 'boundary' marker,
+ * then reset the message buffer */
+ if (csrow & 0x1) {
+ n = snprintf(p, space, "---------------------------"
+ "--------------------------------");
+ p += n;
+ space -= n;
+ debugf2("%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+ }
+ n = snprintf(p, space, "csrow %2d ", csrow);
+ p += n;
+ space -= n;
+
+ for (channel = 0; channel < pvt->maxch; channel++) {
+ dinfo = &pvt->dimm_info[csrow][channel];
+ handle_channel(pvt, csrow, channel, dinfo);
+ n = snprintf(p, space, "%4d MB | ", dinfo->megabytes);
+ p += n;
+ space -= n;
+ }
+ debugf2("%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+ }
+
+ /* Output the last bottom 'boundary' marker */
+ n = snprintf(p, space, "---------------------------"
+ "--------------------------------");
+ p += n;
+ space -= n;
+ debugf2("%s\n", mem_buffer);
+ p = mem_buffer;
+ space = PAGE_SIZE;
+
+ /* now output the 'channel' labels */
+ n = snprintf(p, space, " ");
+ p += n;
+ space -= n;
+ for (channel = 0; channel < pvt->maxch; channel++) {
+ n = snprintf(p, space, "channel %d | ", channel);
+ p += n;
+ space -= n;
+ }
+
+ /* output the last message and free buffer */
+ debugf2("%s\n", mem_buffer);
+ kfree(mem_buffer);
+}
+
+/*
+ * i5400_get_mc_regs read in the necessary registers and
+ * cache locally
+ *
+ * Fills in the private data members
+ */
+static void i5400_get_mc_regs(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+ u32 actual_tolm;
+ u16 limit;
+ int slot_row;
+ int maxch;
+ int maxdimmperch;
+ int way0, way1;
+
+ pvt = mci->pvt_info;
+
+ pci_read_config_dword(pvt->system_address, AMBASE,
+ (u32 *) &pvt->ambase);
+ pci_read_config_dword(pvt->system_address, AMBASE + sizeof(u32),
+ ((u32 *) &pvt->ambase) + sizeof(u32));
+
+ maxdimmperch = pvt->maxdimmperch;
+ maxch = pvt->maxch;
+
+ debugf2("AMBASE= 0x%lx MAXCH= %d MAX-DIMM-Per-CH= %d\n",
+ (long unsigned int)pvt->ambase, pvt->maxch, pvt->maxdimmperch);
+
+ /* Get the Branch Map regs */
+ pci_read_config_word(pvt->branchmap_werrors, TOLM, &pvt->tolm);
+ pvt->tolm >>= 12;
+ debugf2("\nTOLM (number of 256M regions) =%u (0x%x)\n", pvt->tolm,
+ pvt->tolm);
+
+ actual_tolm = (u32) ((1000l * pvt->tolm) >> (30 - 28));
+ debugf2("Actual TOLM byte addr=%u.%03u GB (0x%x)\n",
+ actual_tolm/1000, actual_tolm % 1000, pvt->tolm << 28);
+
+ pci_read_config_word(pvt->branchmap_werrors, MIR0, &pvt->mir0);
+ pci_read_config_word(pvt->branchmap_werrors, MIR1, &pvt->mir1);
+
+ /* Get the MIR[0-1] regs */
+ limit = (pvt->mir0 >> 4) & 0x0fff;
+ way0 = pvt->mir0 & 0x1;
+ way1 = pvt->mir0 & 0x2;
+ debugf2("MIR0: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
+ limit = (pvt->mir1 >> 4) & 0xfff;
+ way0 = pvt->mir1 & 0x1;
+ way1 = pvt->mir1 & 0x2;
+ debugf2("MIR1: limit= 0x%x WAY1= %u WAY0= %x\n", limit, way1, way0);
+
+ /* Get the set of MTR[0-3] regs by each branch */
+ for (slot_row = 0; slot_row < NUM_MTRS_PER_BRANCH; slot_row++) {
+ int where = MTR0 + (slot_row * sizeof(u32));
+
+ /* Branch 0 set of MTR registers */
+ pci_read_config_word(pvt->branch_0, where,
+ &pvt->b0_mtr[slot_row]);
+
+ debugf2("MTR%d where=0x%x B0 value=0x%x\n", slot_row, where,
+ pvt->b0_mtr[slot_row]);
+
+ if (pvt->maxch < CHANNELS_PER_BRANCH) {
+ pvt->b1_mtr[slot_row] = 0;
+ continue;
+ }
+
+ /* Branch 1 set of MTR registers */
+ pci_read_config_word(pvt->branch_1, where,
+ &pvt->b1_mtr[slot_row]);
+ debugf2("MTR%d where=0x%x B1 value=0x%x\n", slot_row, where,
+ pvt->b1_mtr[slot_row]);
+ }
+
+ /* Read and dump branch 0's MTRs */
+ debugf2("\nMemory Technology Registers:\n");
+ debugf2(" Branch 0:\n");
+ for (slot_row = 0; slot_row < NUM_MTRS_PER_BRANCH; slot_row++)
+ decode_mtr(slot_row, pvt->b0_mtr[slot_row]);
+
+ pci_read_config_word(pvt->branch_0, AMBPRESENT_0,
+ &pvt->b0_ambpresent0);
+ debugf2("\t\tAMB-Branch 0-present0 0x%x:\n", pvt->b0_ambpresent0);
+ pci_read_config_word(pvt->branch_0, AMBPRESENT_1,
+ &pvt->b0_ambpresent1);
+ debugf2("\t\tAMB-Branch 0-present1 0x%x:\n", pvt->b0_ambpresent1);
+
+ /* Only if we have 2 branchs (4 channels) */
+ if (pvt->maxch < CHANNELS_PER_BRANCH) {
+ pvt->b1_ambpresent0 = 0;
+ pvt->b1_ambpresent1 = 0;
+ } else {
+ /* Read and dump branch 1's MTRs */
+ debugf2(" Branch 1:\n");
+ for (slot_row = 0; slot_row < NUM_MTRS_PER_BRANCH; slot_row++)
+ decode_mtr(slot_row, pvt->b1_mtr[slot_row]);
+
+ pci_read_config_word(pvt->branch_1, AMBPRESENT_0,
+ &pvt->b1_ambpresent0);
+ debugf2("\t\tAMB-Branch 1-present0 0x%x:\n",
+ pvt->b1_ambpresent0);
+ pci_read_config_word(pvt->branch_1, AMBPRESENT_1,
+ &pvt->b1_ambpresent1);
+ debugf2("\t\tAMB-Branch 1-present1 0x%x:\n",
+ pvt->b1_ambpresent1);
+ }
+
+ /* Go and determine the size of each DIMM and place in an
+ * orderly matrix */
+ calculate_dimm_size(pvt);
+}
+
+/*
+ * i5400_init_csrows Initialize the 'csrows' table within
+ * the mci control structure with the
+ * addressing of memory.
+ *
+ * return:
+ * 0 success
+ * 1 no actual memory found on this MC
+ */
+static int i5400_init_csrows(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+ struct csrow_info *p_csrow;
+ int empty, channel_count;
+ int max_csrows;
+ int mtr;
+ int csrow_megs;
+ int channel;
+ int csrow;
+
+ pvt = mci->pvt_info;
+
+ channel_count = pvt->maxch;
+ max_csrows = pvt->maxdimmperch * 2;
+
+ empty = 1; /* Assume NO memory */
+
+ for (csrow = 0; csrow < max_csrows; csrow++) {
+ p_csrow = &mci->csrows[csrow];
+
+ p_csrow->csrow_idx = csrow;
+
+ /* use branch 0 for the basis */
+ mtr = determine_mtr(pvt, csrow, 0);
+
+ /* if no DIMMS on this row, continue */
+ if (!MTR_DIMMS_PRESENT(mtr))
+ continue;
+
+ /* FAKE OUT VALUES, FIXME */
+ p_csrow->first_page = 0 + csrow * 20;
+ p_csrow->last_page = 9 + csrow * 20;
+ p_csrow->page_mask = 0xFFF;
+
+ p_csrow->grain = 8;
+
+ csrow_megs = 0;
+ for (channel = 0; channel < pvt->maxch; channel++)
+ csrow_megs += pvt->dimm_info[csrow][channel].megabytes;
+
+ p_csrow->nr_pages = csrow_megs << 8;
+
+ /* Assume DDR2 for now */
+ p_csrow->mtype = MEM_FB_DDR2;
+
+ /* ask what device type on this row */
+ if (MTR_DRAM_WIDTH(mtr))
+ p_csrow->dtype = DEV_X8;
+ else
+ p_csrow->dtype = DEV_X4;
+
+ p_csrow->edac_mode = EDAC_S8ECD8ED;
+
+ empty = 0;
+ }
+
+ return empty;
+}
+
+/*
+ * i5400_enable_error_reporting
+ * Turn on the memory reporting features of the hardware
+ */
+static void i5400_enable_error_reporting(struct mem_ctl_info *mci)
+{
+ struct i5400_pvt *pvt;
+ u32 fbd_error_mask;
+
+ pvt = mci->pvt_info;
+
+ /* Read the FBD Error Mask Register */
+ pci_read_config_dword(pvt->branchmap_werrors, EMASK_FBD,
+ &fbd_error_mask);
+
+ /* Enable with a '0' */
+ fbd_error_mask &= ~(ENABLE_EMASK_ALL);
+
+ pci_write_config_dword(pvt->branchmap_werrors, EMASK_FBD,
+ fbd_error_mask);
+}
+
+/*
+ * i5400_get_dimm_and_channel_counts(pdev, &num_csrows, &num_channels)
+ *
+ * ask the device how many channels are present and how many CSROWS
+ * as well
+ */
+static void i5400_get_dimm_and_channel_counts(struct pci_dev *pdev,
+ int *num_dimms_per_channel,
+ int *num_channels)
+{
+ u8 value;
+
+ /* Need to retrieve just how many channels and dimms per channel are
+ * supported on this memory controller
+ */
+ pci_read_config_byte(pdev, MAXDIMMPERCH, &value);
+ *num_dimms_per_channel = (int)value * 2;
+
+ pci_read_config_byte(pdev, MAXCH, &value);
+ *num_channels = (int)value;
+}
+
+/*
+ * i5400_probe1 Probe for ONE instance of device to see if it is
+ * present.
+ * return:
+ * 0 for FOUND a device
+ * < 0 for error code
+ */
+static int i5400_probe1(struct pci_dev *pdev, int dev_idx)
+{
+ struct mem_ctl_info *mci;
+ struct i5400_pvt *pvt;
+ int num_channels;
+ int num_dimms_per_channel;
+ int num_csrows;
+
+ if (dev_idx >= ARRAY_SIZE(i5400_devs))
+ return -EINVAL;
+
+ debugf0("MC: " __FILE__ ": %s(), pdev bus %u dev=0x%x fn=0x%x\n",
+ __func__,
+ pdev->bus->number,
+ PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
+
+ /* We only are looking for func 0 of the set */
+ if (PCI_FUNC(pdev->devfn) != 0)
+ return -ENODEV;
+
+ /* Ask the devices for the number of CSROWS and CHANNELS so
+ * that we can calculate the memory resources, etc
+ *
+ * The Chipset will report what it can handle which will be greater
+ * or equal to what the motherboard manufacturer will implement.
+ *
+ * As we don't have a motherboard identification routine to determine
+ * actual number of slots/dimms per channel, we thus utilize the
+ * resource as specified by the chipset. Thus, we might have
+ * have more DIMMs per channel than actually on the mobo, but this
+ * allows the driver to support upto the chipset max, without
+ * some fancy mobo determination.
+ */
+ i5400_get_dimm_and_channel_counts(pdev, &num_dimms_per_channel,
+ &num_channels);
+ num_csrows = num_dimms_per_channel * 2;
+
+ debugf0("MC: %s(): Number of - Channels= %d DIMMS= %d CSROWS= %d\n",
+ __func__, num_channels, num_dimms_per_channel, num_csrows);
+
+ /* allocate a new MC control structure */
+ mci = edac_mc_alloc(sizeof(*pvt), num_csrows, num_channels, 0);
+
+ if (mci == NULL)
+ return -ENOMEM;
+
+ debugf0("MC: " __FILE__ ": %s(): mci = %p\n", __func__, mci);
+
+ mci->dev = &pdev->dev; /* record ptr to the generic device */
+
+ pvt = mci->pvt_info;
+ pvt->system_address = pdev; /* Record this device in our private */
+ pvt->maxch = num_channels;
+ pvt->maxdimmperch = num_dimms_per_channel;
+
+ /* 'get' the pci devices we want to reserve for our use */
+ if (i5400_get_devices(mci, dev_idx))
+ goto fail0;
+
+ /* Time to get serious */
+ i5400_get_mc_regs(mci); /* retrieve the hardware registers */
+
+ mci->mc_idx = 0;
+ mci->mtype_cap = MEM_FLAG_FB_DDR2;
+ mci->edac_ctl_cap = EDAC_FLAG_NONE;
+ mci->edac_cap = EDAC_FLAG_NONE;
+ mci->mod_name = "i5400_edac.c";
+ mci->mod_ver = I5400_REVISION;
+ mci->ctl_name = i5400_devs[dev_idx].ctl_name;
+ mci->dev_name = pci_name(pdev);
+ mci->ctl_page_to_phys = NULL;
+
+ /* Set the function pointer to an actual operation function */
+ mci->edac_check = i5400_check_error;
+
+ /* initialize the MC control structure 'csrows' table
+ * with the mapping and control information */
+ if (i5400_init_csrows(mci)) {
+ debugf0("MC: Setting mci->edac_cap to EDAC_FLAG_NONE\n"
+ " because i5400_init_csrows() returned nonzero "
+ "value\n");
+ mci->edac_cap = EDAC_FLAG_NONE; /* no csrows found */
+ } else {
+ debugf1("MC: Enable error reporting now\n");
+ i5400_enable_error_reporting(mci);
+ }
+
+ /* add this new MC control structure to EDAC's list of MCs */
+ if (edac_mc_add_mc(mci)) {
+ debugf0("MC: " __FILE__
+ ": %s(): failed edac_mc_add_mc()\n", __func__);
+ /* FIXME: perhaps some code should go here that disables error
+ * reporting if we just enabled it
+ */
+ goto fail1;
+ }
+
+ i5400_clear_error(mci);
+
+ /* allocating generic PCI control info */
+ i5400_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR);
+ if (!i5400_pci) {
+ printk(KERN_WARNING
+ "%s(): Unable to create PCI control\n",
+ __func__);
+ printk(KERN_WARNING
+ "%s(): PCI error report via EDAC not setup\n",
+ __func__);
+ }
+
+ return 0;
+
+ /* Error exit unwinding stack */
+fail1:
+
+ i5400_put_devices(mci);
+
+fail0:
+ edac_mc_free(mci);
+ return -ENODEV;
+}
+
+/*
+ * i5400_init_one constructor for one instance of device
+ *
+ * returns:
+ * negative on error
+ * count (>= 0)
+ */
+static int __devinit i5400_init_one(struct pci_dev *pdev,
+ const struct pci_device_id *id)
+{
+ int rc;
+
+ debugf0("MC: " __FILE__ ": %s()\n", __func__);
+
+ /* wake up device */
+ rc = pci_enable_device(pdev);
+ if (rc == -EIO)
+ return rc;
+
+ /* now probe and enable the device */
+ return i5400_probe1(pdev, id->driver_data);
+}
+
+/*
+ * i5400_remove_one destructor for one instance of device
+ *
+ */
+static void __devexit i5400_remove_one(struct pci_dev *pdev)
+{
+ struct mem_ctl_info *mci;
+
+ debugf0(__FILE__ ": %s()\n", __func__);
+
+ if (i5400_pci)
+ edac_pci_release_generic_ctl(i5400_pci);
+
+ mci = edac_mc_del_mc(&pdev->dev);
+ if (!mci)
+ return;
+
+ /* retrieve references to resources, and free those resources */
+ i5400_put_devices(mci);
+
+ edac_mc_free(mci);
+}
+
+/*
+ * pci_device_id table for which devices we are looking for
+ *
+ * The "E500P" device is the first device supported.
+ */
+static const struct pci_device_id i5400_pci_tbl[] __devinitdata = {
+ {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_5400_ERR)},
+ {0,} /* 0 terminated list. */
+};
+
+MODULE_DEVICE_TABLE(pci, i5400_pci_tbl);
+
+/*
+ * i5400_driver pci_driver structure for this module
+ *
+ */
+static struct pci_driver i5400_driver = {
+ .name = "i5400_edac",
+ .probe = i5400_init_one,
+ .remove = __devexit_p(i5400_remove_one),
+ .id_table = i5400_pci_tbl,
+};
+
+/*
+ * i5400_init Module entry function
+ * Try to initialize this module for its devices
+ */
+static int __init i5400_init(void)
+{
+ int pci_rc;
+
+ debugf2("MC: " __FILE__ ": %s()\n", __func__);
+
+ /* Ensure that the OPSTATE is set correctly for POLL or NMI */
+ opstate_init();
+
+ pci_rc = pci_register_driver(&i5400_driver);
+
+ return (pci_rc < 0) ? pci_rc : 0;
+}
+
+/*
+ * i5400_exit() Module exit function
+ * Unregister the driver
+ */
+static void __exit i5400_exit(void)
+{
+ debugf2("MC: " __FILE__ ": %s()\n", __func__);
+ pci_unregister_driver(&i5400_driver);
+}
+
+module_init(i5400_init);
+module_exit(i5400_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Ben Woodard <[email protected]>");
+MODULE_AUTHOR("Mauro Carvalho Chehab <[email protected]>");
+MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)");
+MODULE_DESCRIPTION("MC Driver for Intel I5400 memory controllers - "
+ I5400_REVISION);
+
+module_param(edac_op_state, int, 0444);
+MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI");