diff options
author | Yunhong Jiang <yunhong.jiang@intel.com> | 2015-08-04 12:17:53 -0700 |
---|---|---|
committer | Yunhong Jiang <yunhong.jiang@intel.com> | 2015-08-04 15:44:42 -0700 |
commit | 9ca8dbcc65cfc63d6f5ef3312a33184e1d726e00 (patch) | |
tree | 1c9cafbcd35f783a87880a10f85d1a060db1a563 /kernel/drivers/edac/i7300_edac.c | |
parent | 98260f3884f4a202f9ca5eabed40b1354c489b29 (diff) |
Add the rt linux 4.1.3-rt3 as base
Import the rt linux 4.1.3-rt3 as OPNFV kvm base.
It's from git://git.kernel.org/pub/scm/linux/kernel/git/rt/linux-rt-devel.git linux-4.1.y-rt and
the base is:
commit 0917f823c59692d751951bf5ea699a2d1e2f26a2
Author: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Date: Sat Jul 25 12:13:34 2015 +0200
Prepare v4.1.3-rt3
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
We lose all the git history this way and it's not good. We
should apply another opnfv project repo in future.
Change-Id: I87543d81c9df70d99c5001fbdf646b202c19f423
Signed-off-by: Yunhong Jiang <yunhong.jiang@intel.com>
Diffstat (limited to 'kernel/drivers/edac/i7300_edac.c')
-rw-r--r-- | kernel/drivers/edac/i7300_edac.c | 1218 |
1 files changed, 1218 insertions, 0 deletions
diff --git a/kernel/drivers/edac/i7300_edac.c b/kernel/drivers/edac/i7300_edac.c new file mode 100644 index 000000000..dcac982fd --- /dev/null +++ b/kernel/drivers/edac/i7300_edac.c @@ -0,0 +1,1218 @@ +/* + * Intel 7300 class Memory Controllers kernel module (Clarksboro) + * + * This file may be distributed under the terms of the + * GNU General Public License version 2 only. + * + * Copyright (c) 2010 by: + * Mauro Carvalho Chehab + * + * Red Hat Inc. http://www.redhat.com + * + * Intel 7300 Chipset Memory Controller Hub (MCH) - Datasheet + * http://www.intel.com/Assets/PDF/datasheet/318082.pdf + * + * TODO: The chipset allow checking for PCI Express errors also. Currently, + * the driver covers only memory error errors + * + * This driver uses "csrows" EDAC attribute to represent DIMM slot# + */ + +#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 I7300 module when modifications are made + */ +#define I7300_REVISION " Ver: 1.0.0" + +#define EDAC_MOD_STR "i7300_edac" + +#define i7300_printk(level, fmt, arg...) \ + edac_printk(level, "i7300", fmt, ##arg) + +#define i7300_mc_printk(mci, level, fmt, arg...) \ + edac_mc_chipset_printk(mci, level, "i7300", fmt, ##arg) + +/*********************************************** + * i7300 Limit constants Structs and static vars + ***********************************************/ + +/* + * Memory topology is organized as: + * Branch 0 - 2 channels: channels 0 and 1 (FDB0 PCI dev 21.0) + * Branch 1 - 2 channels: channels 2 and 3 (FDB1 PCI dev 22.0) + * Each channel can have to 8 DIMM sets (called as SLOTS) + * Slots should generally be filled in pairs + * Except on Single Channel mode of operation + * just slot 0/channel0 filled on this mode + * On normal operation mode, the two channels on a branch should be + * filled together for the same SLOT# + * When in mirrored mode, Branch 1 replicate memory at Branch 0, so, the four + * channels on both branches should be filled + */ + +/* Limits for i7300 */ +#define MAX_SLOTS 8 +#define MAX_BRANCHES 2 +#define MAX_CH_PER_BRANCH 2 +#define MAX_CHANNELS (MAX_CH_PER_BRANCH * MAX_BRANCHES) +#define MAX_MIR 3 + +#define to_channel(ch, branch) ((((branch)) << 1) | (ch)) + +#define to_csrow(slot, ch, branch) \ + (to_channel(ch, branch) | ((slot) << 2)) + +/* Device name and register DID (Device ID) */ +struct i7300_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 i7300_dev_info i7300_devs[] = { + { + .ctl_name = "I7300", + .fsb_mapping_errors = PCI_DEVICE_ID_INTEL_I7300_MCH_ERR, + }, +}; + +struct i7300_dimm_info { + int megabytes; /* size, 0 means not present */ +}; + +/* driver private data structure */ +struct i7300_pvt { + struct pci_dev *pci_dev_16_0_fsb_ctlr; /* 16.0 */ + struct pci_dev *pci_dev_16_1_fsb_addr_map; /* 16.1 */ + struct pci_dev *pci_dev_16_2_fsb_err_regs; /* 16.2 */ + struct pci_dev *pci_dev_2x_0_fbd_branch[MAX_BRANCHES]; /* 21.0 and 22.0 */ + + u16 tolm; /* top of low memory */ + u64 ambase; /* AMB BAR */ + + u32 mc_settings; /* Report several settings */ + u32 mc_settings_a; + + u16 mir[MAX_MIR]; /* Memory Interleave Reg*/ + + u16 mtr[MAX_SLOTS][MAX_BRANCHES]; /* Memory Technlogy Reg */ + u16 ambpresent[MAX_CHANNELS]; /* AMB present regs */ + + /* DIMM information matrix, allocating architecture maximums */ + struct i7300_dimm_info dimm_info[MAX_SLOTS][MAX_CHANNELS]; + + /* Temporary buffer for use when preparing error messages */ + char *tmp_prt_buffer; +}; + +/* FIXME: Why do we need to have this static? */ +static struct edac_pci_ctl_info *i7300_pci; + +/*************************************************** + * i7300 Register definitions for memory enumeration + ***************************************************/ + +/* + * Device 16, + * Function 0: System Address (not documented) + * Function 1: Memory Branch Map, Control, Errors Register + */ + + /* 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 MC_SETTINGS 0x40 + #define IS_MIRRORED(mc) ((mc) & (1 << 16)) + #define IS_ECC_ENABLED(mc) ((mc) & (1 << 5)) + #define IS_RETRY_ENABLED(mc) ((mc) & (1 << 31)) + #define IS_SCRBALGO_ENHANCED(mc) ((mc) & (1 << 8)) + +#define MC_SETTINGS_A 0x58 + #define IS_SINGLE_MODE(mca) ((mca) & (1 << 14)) + +#define TOLM 0x6C + +#define MIR0 0x80 +#define MIR1 0x84 +#define MIR2 0x88 + +/* + * Note: Other Intel EDAC drivers use AMBPRESENT to identify if the available + * memory. From datasheet item 7.3.1 (FB-DIMM technology & organization), it + * seems that we cannot use this information directly for the same usage. + * Each memory slot may have up to 2 AMB interfaces, one for income and another + * for outcome interface to the next slot. + * For now, the driver just stores the AMB present registers, but rely only at + * the MTR info to detect memory. + * Datasheet is also not clear about how to map each AMBPRESENT registers to + * one of the 4 available channels. + */ +#define AMBPRESENT_0 0x64 +#define AMBPRESENT_1 0x66 + +static const u16 mtr_regs[MAX_SLOTS] = { + 0x80, 0x84, 0x88, 0x8c, + 0x82, 0x86, 0x8a, 0x8e +}; + +/* + * Defines to extract the vaious fields from the + * MTRx - Memory Technology Registers + */ +#define MTR_DIMMS_PRESENT(mtr) ((mtr) & (1 << 8)) +#define MTR_DIMMS_ETHROTTLE(mtr) ((mtr) & (1 << 7)) +#define MTR_DRAM_WIDTH(mtr) (((mtr) & (1 << 6)) ? 8 : 4) +#define MTR_DRAM_BANKS(mtr) (((mtr) & (1 << 5)) ? 8 : 4) +#define MTR_DIMM_RANKS(mtr) (((mtr) & (1 << 4)) ? 1 : 0) +#define MTR_DIMM_ROWS(mtr) (((mtr) >> 2) & 0x3) +#define MTR_DRAM_BANKS_ADDR_BITS 2 +#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) + +/************************************************ + * i7300 Register definitions for error detection + ************************************************/ + +/* + * Device 16.1: FBD Error Registers + */ +#define FERR_FAT_FBD 0x98 +static const char *ferr_fat_fbd_name[] = { + [22] = "Non-Redundant Fast Reset Timeout", + [2] = ">Tmid Thermal event with intelligent throttling disabled", + [1] = "Memory or FBD configuration CRC read error", + [0] = "Memory Write error on non-redundant retry or " + "FBD configuration Write error on retry", +}; +#define GET_FBD_FAT_IDX(fbderr) (((fbderr) >> 28) & 3) +#define FERR_FAT_FBD_ERR_MASK ((1 << 0) | (1 << 1) | (1 << 2) | (1 << 22)) + +#define FERR_NF_FBD 0xa0 +static const char *ferr_nf_fbd_name[] = { + [24] = "DIMM-Spare Copy Completed", + [23] = "DIMM-Spare Copy Initiated", + [22] = "Redundant Fast Reset Timeout", + [21] = "Memory Write error on redundant retry", + [18] = "SPD protocol Error", + [17] = "FBD Northbound parity error on FBD Sync Status", + [16] = "Correctable Patrol Data ECC", + [15] = "Correctable Resilver- or Spare-Copy Data ECC", + [14] = "Correctable Mirrored Demand Data ECC", + [13] = "Correctable Non-Mirrored Demand Data ECC", + [11] = "Memory or FBD configuration CRC read error", + [10] = "FBD Configuration Write error on first attempt", + [9] = "Memory Write error on first attempt", + [8] = "Non-Aliased Uncorrectable Patrol Data ECC", + [7] = "Non-Aliased Uncorrectable Resilver- or Spare-Copy Data ECC", + [6] = "Non-Aliased Uncorrectable Mirrored Demand Data ECC", + [5] = "Non-Aliased Uncorrectable Non-Mirrored Demand Data ECC", + [4] = "Aliased Uncorrectable Patrol Data ECC", + [3] = "Aliased Uncorrectable Resilver- or Spare-Copy Data ECC", + [2] = "Aliased Uncorrectable Mirrored Demand Data ECC", + [1] = "Aliased Uncorrectable Non-Mirrored Demand Data ECC", + [0] = "Uncorrectable Data ECC on Replay", +}; +#define GET_FBD_NF_IDX(fbderr) (((fbderr) >> 28) & 3) +#define FERR_NF_FBD_ERR_MASK ((1 << 24) | (1 << 23) | (1 << 22) | (1 << 21) |\ + (1 << 18) | (1 << 17) | (1 << 16) | (1 << 15) |\ + (1 << 14) | (1 << 13) | (1 << 11) | (1 << 10) |\ + (1 << 9) | (1 << 8) | (1 << 7) | (1 << 6) |\ + (1 << 5) | (1 << 4) | (1 << 3) | (1 << 2) |\ + (1 << 1) | (1 << 0)) + +#define EMASK_FBD 0xa8 +#define EMASK_FBD_ERR_MASK ((1 << 27) | (1 << 26) | (1 << 25) | (1 << 24) |\ + (1 << 22) | (1 << 21) | (1 << 20) | (1 << 19) |\ + (1 << 18) | (1 << 17) | (1 << 16) | (1 << 14) |\ + (1 << 13) | (1 << 12) | (1 << 11) | (1 << 10) |\ + (1 << 9) | (1 << 8) | (1 << 7) | (1 << 6) |\ + (1 << 5) | (1 << 4) | (1 << 3) | (1 << 2) |\ + (1 << 1) | (1 << 0)) + +/* + * Device 16.2: Global Error Registers + */ + +#define FERR_GLOBAL_HI 0x48 +static const char *ferr_global_hi_name[] = { + [3] = "FSB 3 Fatal Error", + [2] = "FSB 2 Fatal Error", + [1] = "FSB 1 Fatal Error", + [0] = "FSB 0 Fatal Error", +}; +#define ferr_global_hi_is_fatal(errno) 1 + +#define FERR_GLOBAL_LO 0x40 +static const char *ferr_global_lo_name[] = { + [31] = "Internal MCH Fatal Error", + [30] = "Intel QuickData Technology Device Fatal Error", + [29] = "FSB1 Fatal Error", + [28] = "FSB0 Fatal Error", + [27] = "FBD Channel 3 Fatal Error", + [26] = "FBD Channel 2 Fatal Error", + [25] = "FBD Channel 1 Fatal Error", + [24] = "FBD Channel 0 Fatal Error", + [23] = "PCI Express Device 7Fatal Error", + [22] = "PCI Express Device 6 Fatal Error", + [21] = "PCI Express Device 5 Fatal Error", + [20] = "PCI Express Device 4 Fatal Error", + [19] = "PCI Express Device 3 Fatal Error", + [18] = "PCI Express Device 2 Fatal Error", + [17] = "PCI Express Device 1 Fatal Error", + [16] = "ESI Fatal Error", + [15] = "Internal MCH Non-Fatal Error", + [14] = "Intel QuickData Technology Device Non Fatal Error", + [13] = "FSB1 Non-Fatal Error", + [12] = "FSB 0 Non-Fatal Error", + [11] = "FBD Channel 3 Non-Fatal Error", + [10] = "FBD Channel 2 Non-Fatal Error", + [9] = "FBD Channel 1 Non-Fatal Error", + [8] = "FBD Channel 0 Non-Fatal Error", + [7] = "PCI Express Device 7 Non-Fatal Error", + [6] = "PCI Express Device 6 Non-Fatal Error", + [5] = "PCI Express Device 5 Non-Fatal Error", + [4] = "PCI Express Device 4 Non-Fatal Error", + [3] = "PCI Express Device 3 Non-Fatal Error", + [2] = "PCI Express Device 2 Non-Fatal Error", + [1] = "PCI Express Device 1 Non-Fatal Error", + [0] = "ESI Non-Fatal Error", +}; +#define ferr_global_lo_is_fatal(errno) ((errno < 16) ? 0 : 1) + +#define NRECMEMA 0xbe + #define NRECMEMA_BANK(v) (((v) >> 12) & 7) + #define NRECMEMA_RANK(v) (((v) >> 8) & 15) + +#define NRECMEMB 0xc0 + #define NRECMEMB_IS_WR(v) ((v) & (1 << 31)) + #define NRECMEMB_CAS(v) (((v) >> 16) & 0x1fff) + #define NRECMEMB_RAS(v) ((v) & 0xffff) + +#define REDMEMA 0xdc + +#define REDMEMB 0x7c + #define IS_SECOND_CH(v) ((v) * (1 << 17)) + +#define RECMEMA 0xe0 + #define RECMEMA_BANK(v) (((v) >> 12) & 7) + #define RECMEMA_RANK(v) (((v) >> 8) & 15) + +#define RECMEMB 0xe4 + #define RECMEMB_IS_WR(v) ((v) & (1 << 31)) + #define RECMEMB_CAS(v) (((v) >> 16) & 0x1fff) + #define RECMEMB_RAS(v) ((v) & 0xffff) + +/******************************************** + * i7300 Functions related to error detection + ********************************************/ + +/** + * get_err_from_table() - Gets the error message from a table + * @table: table name (array of char *) + * @size: number of elements at the table + * @pos: position of the element to be returned + * + * This is a small routine that gets the pos-th element of a table. If the + * element doesn't exist (or it is empty), it returns "reserved". + * Instead of calling it directly, the better is to call via the macro + * GET_ERR_FROM_TABLE(), that automatically checks the table size via + * ARRAY_SIZE() macro + */ +static const char *get_err_from_table(const char *table[], int size, int pos) +{ + if (unlikely(pos >= size)) + return "Reserved"; + + if (unlikely(!table[pos])) + return "Reserved"; + + return table[pos]; +} + +#define GET_ERR_FROM_TABLE(table, pos) \ + get_err_from_table(table, ARRAY_SIZE(table), pos) + +/** + * i7300_process_error_global() - Retrieve the hardware error information from + * the hardware global error registers and + * sends it to dmesg + * @mci: struct mem_ctl_info pointer + */ +static void i7300_process_error_global(struct mem_ctl_info *mci) +{ + struct i7300_pvt *pvt; + u32 errnum, error_reg; + unsigned long errors; + const char *specific; + bool is_fatal; + + pvt = mci->pvt_info; + + /* read in the 1st FATAL error register */ + pci_read_config_dword(pvt->pci_dev_16_2_fsb_err_regs, + FERR_GLOBAL_HI, &error_reg); + if (unlikely(error_reg)) { + errors = error_reg; + errnum = find_first_bit(&errors, + ARRAY_SIZE(ferr_global_hi_name)); + specific = GET_ERR_FROM_TABLE(ferr_global_hi_name, errnum); + is_fatal = ferr_global_hi_is_fatal(errnum); + + /* Clear the error bit */ + pci_write_config_dword(pvt->pci_dev_16_2_fsb_err_regs, + FERR_GLOBAL_HI, error_reg); + + goto error_global; + } + + pci_read_config_dword(pvt->pci_dev_16_2_fsb_err_regs, + FERR_GLOBAL_LO, &error_reg); + if (unlikely(error_reg)) { + errors = error_reg; + errnum = find_first_bit(&errors, + ARRAY_SIZE(ferr_global_lo_name)); + specific = GET_ERR_FROM_TABLE(ferr_global_lo_name, errnum); + is_fatal = ferr_global_lo_is_fatal(errnum); + + /* Clear the error bit */ + pci_write_config_dword(pvt->pci_dev_16_2_fsb_err_regs, + FERR_GLOBAL_LO, error_reg); + + goto error_global; + } + return; + +error_global: + i7300_mc_printk(mci, KERN_EMERG, "%s misc error: %s\n", + is_fatal ? "Fatal" : "NOT fatal", specific); +} + +/** + * i7300_process_fbd_error() - Retrieve the hardware error information from + * the FBD error registers and sends it via + * EDAC error API calls + * @mci: struct mem_ctl_info pointer + */ +static void i7300_process_fbd_error(struct mem_ctl_info *mci) +{ + struct i7300_pvt *pvt; + u32 errnum, value, error_reg; + u16 val16; + unsigned branch, channel, bank, rank, cas, ras; + u32 syndrome; + + unsigned long errors; + const char *specific; + bool is_wr; + + pvt = mci->pvt_info; + + /* read in the 1st FATAL error register */ + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + FERR_FAT_FBD, &error_reg); + if (unlikely(error_reg & FERR_FAT_FBD_ERR_MASK)) { + errors = error_reg & FERR_FAT_FBD_ERR_MASK ; + errnum = find_first_bit(&errors, + ARRAY_SIZE(ferr_fat_fbd_name)); + specific = GET_ERR_FROM_TABLE(ferr_fat_fbd_name, errnum); + branch = (GET_FBD_FAT_IDX(error_reg) == 2) ? 1 : 0; + + pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map, + NRECMEMA, &val16); + bank = NRECMEMA_BANK(val16); + rank = NRECMEMA_RANK(val16); + + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + NRECMEMB, &value); + is_wr = NRECMEMB_IS_WR(value); + cas = NRECMEMB_CAS(value); + ras = NRECMEMB_RAS(value); + + /* Clean the error register */ + pci_write_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + FERR_FAT_FBD, error_reg); + + snprintf(pvt->tmp_prt_buffer, PAGE_SIZE, + "Bank=%d RAS=%d CAS=%d Err=0x%lx (%s))", + bank, ras, cas, errors, specific); + + edac_mc_handle_error(HW_EVENT_ERR_FATAL, mci, 1, 0, 0, 0, + branch, -1, rank, + is_wr ? "Write error" : "Read error", + pvt->tmp_prt_buffer); + + } + + /* read in the 1st NON-FATAL error register */ + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + FERR_NF_FBD, &error_reg); + if (unlikely(error_reg & FERR_NF_FBD_ERR_MASK)) { + errors = error_reg & FERR_NF_FBD_ERR_MASK; + errnum = find_first_bit(&errors, + ARRAY_SIZE(ferr_nf_fbd_name)); + specific = GET_ERR_FROM_TABLE(ferr_nf_fbd_name, errnum); + branch = (GET_FBD_NF_IDX(error_reg) == 2) ? 1 : 0; + + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + REDMEMA, &syndrome); + + pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map, + RECMEMA, &val16); + bank = RECMEMA_BANK(val16); + rank = RECMEMA_RANK(val16); + + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + RECMEMB, &value); + is_wr = RECMEMB_IS_WR(value); + cas = RECMEMB_CAS(value); + ras = RECMEMB_RAS(value); + + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + REDMEMB, &value); + channel = (branch << 1); + if (IS_SECOND_CH(value)) + channel++; + + /* Clear the error bit */ + pci_write_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + FERR_NF_FBD, error_reg); + + /* Form out message */ + snprintf(pvt->tmp_prt_buffer, PAGE_SIZE, + "DRAM-Bank=%d RAS=%d CAS=%d, Err=0x%lx (%s))", + bank, ras, cas, errors, specific); + + edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, 0, 0, + syndrome, + branch >> 1, channel % 2, rank, + is_wr ? "Write error" : "Read error", + pvt->tmp_prt_buffer); + } + return; +} + +/** + * i7300_check_error() - Calls the error checking subroutines + * @mci: struct mem_ctl_info pointer + */ +static void i7300_check_error(struct mem_ctl_info *mci) +{ + i7300_process_error_global(mci); + i7300_process_fbd_error(mci); +}; + +/** + * i7300_clear_error() - Clears the error registers + * @mci: struct mem_ctl_info pointer + */ +static void i7300_clear_error(struct mem_ctl_info *mci) +{ + struct i7300_pvt *pvt = mci->pvt_info; + u32 value; + /* + * All error values are RWC - we need to read and write 1 to the + * bit that we want to cleanup + */ + + /* Clear global error registers */ + pci_read_config_dword(pvt->pci_dev_16_2_fsb_err_regs, + FERR_GLOBAL_HI, &value); + pci_write_config_dword(pvt->pci_dev_16_2_fsb_err_regs, + FERR_GLOBAL_HI, value); + + pci_read_config_dword(pvt->pci_dev_16_2_fsb_err_regs, + FERR_GLOBAL_LO, &value); + pci_write_config_dword(pvt->pci_dev_16_2_fsb_err_regs, + FERR_GLOBAL_LO, value); + + /* Clear FBD error registers */ + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + FERR_FAT_FBD, &value); + pci_write_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + FERR_FAT_FBD, value); + + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + FERR_NF_FBD, &value); + pci_write_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + FERR_NF_FBD, value); +} + +/** + * i7300_enable_error_reporting() - Enable the memory reporting logic at the + * hardware + * @mci: struct mem_ctl_info pointer + */ +static void i7300_enable_error_reporting(struct mem_ctl_info *mci) +{ + struct i7300_pvt *pvt = mci->pvt_info; + u32 fbd_error_mask; + + /* Read the FBD Error Mask Register */ + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + EMASK_FBD, &fbd_error_mask); + + /* Enable with a '0' */ + fbd_error_mask &= ~(EMASK_FBD_ERR_MASK); + + pci_write_config_dword(pvt->pci_dev_16_1_fsb_addr_map, + EMASK_FBD, fbd_error_mask); +} + +/************************************************ + * i7300 Functions related to memory enumberation + ************************************************/ + +/** + * decode_mtr() - Decodes the MTR descriptor, filling the edac structs + * @pvt: pointer to the private data struct used by i7300 driver + * @slot: DIMM slot (0 to 7) + * @ch: Channel number within the branch (0 or 1) + * @branch: Branch number (0 or 1) + * @dinfo: Pointer to DIMM info where dimm size is stored + * @p_csrow: Pointer to the struct csrow_info that corresponds to that element + */ +static int decode_mtr(struct i7300_pvt *pvt, + int slot, int ch, int branch, + struct i7300_dimm_info *dinfo, + struct dimm_info *dimm) +{ + int mtr, ans, addrBits, channel; + + channel = to_channel(ch, branch); + + mtr = pvt->mtr[slot][branch]; + ans = MTR_DIMMS_PRESENT(mtr) ? 1 : 0; + + edac_dbg(2, "\tMTR%d CH%d: DIMMs are %sPresent (mtr)\n", + slot, channel, ans ? "" : "NOT "); + + /* Determine if there is a DIMM present in this DIMM slot */ + if (!ans) + return 0; + + /* Start with the number of bits for a Bank + * on the DRAM */ + addrBits = MTR_DRAM_BANKS_ADDR_BITS; + /* 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); + /* add the number of RANK bits */ + addrBits += MTR_DIMM_RANKS(mtr); + + addrBits += 6; /* add 64 bits per DIMM */ + addrBits -= 20; /* divide by 2^^20 */ + addrBits -= 3; /* 8 bits per bytes */ + + dinfo->megabytes = 1 << addrBits; + + edac_dbg(2, "\t\tWIDTH: x%d\n", MTR_DRAM_WIDTH(mtr)); + + edac_dbg(2, "\t\tELECTRICAL THROTTLING is %s\n", + MTR_DIMMS_ETHROTTLE(mtr) ? "enabled" : "disabled"); + + edac_dbg(2, "\t\tNUMBANK: %d bank(s)\n", MTR_DRAM_BANKS(mtr)); + edac_dbg(2, "\t\tNUMRANK: %s\n", + MTR_DIMM_RANKS(mtr) ? "double" : "single"); + edac_dbg(2, "\t\tNUMROW: %s\n", + MTR_DIMM_ROWS(mtr) == 0 ? "8,192 - 13 rows" : + MTR_DIMM_ROWS(mtr) == 1 ? "16,384 - 14 rows" : + MTR_DIMM_ROWS(mtr) == 2 ? "32,768 - 15 rows" : + "65,536 - 16 rows"); + edac_dbg(2, "\t\tNUMCOL: %s\n", + MTR_DIMM_COLS(mtr) == 0 ? "1,024 - 10 columns" : + MTR_DIMM_COLS(mtr) == 1 ? "2,048 - 11 columns" : + MTR_DIMM_COLS(mtr) == 2 ? "4,096 - 12 columns" : + "reserved"); + edac_dbg(2, "\t\tSIZE: %d MB\n", dinfo->megabytes); + + /* + * The type of error detection actually depends of the + * mode of operation. When it is just one single memory chip, at + * socket 0, channel 0, it uses 8-byte-over-32-byte SECDED+ code. + * In normal or mirrored mode, it uses Lockstep mode, + * with the possibility of using an extended algorithm for x8 memories + * See datasheet Sections 7.3.6 to 7.3.8 + */ + + dimm->nr_pages = MiB_TO_PAGES(dinfo->megabytes); + dimm->grain = 8; + dimm->mtype = MEM_FB_DDR2; + if (IS_SINGLE_MODE(pvt->mc_settings_a)) { + dimm->edac_mode = EDAC_SECDED; + edac_dbg(2, "\t\tECC code is 8-byte-over-32-byte SECDED+ code\n"); + } else { + edac_dbg(2, "\t\tECC code is on Lockstep mode\n"); + if (MTR_DRAM_WIDTH(mtr) == 8) + dimm->edac_mode = EDAC_S8ECD8ED; + else + dimm->edac_mode = EDAC_S4ECD4ED; + } + + /* ask what device type on this row */ + if (MTR_DRAM_WIDTH(mtr) == 8) { + edac_dbg(2, "\t\tScrub algorithm for x8 is on %s mode\n", + IS_SCRBALGO_ENHANCED(pvt->mc_settings) ? + "enhanced" : "normal"); + + dimm->dtype = DEV_X8; + } else + dimm->dtype = DEV_X4; + + return mtr; +} + +/** + * print_dimm_size() - Prints dump of the memory organization + * @pvt: pointer to the private data struct used by i7300 driver + * + * Useful for debug. If debug is disabled, this routine do nothing + */ +static void print_dimm_size(struct i7300_pvt *pvt) +{ +#ifdef CONFIG_EDAC_DEBUG + struct i7300_dimm_info *dinfo; + char *p; + int space, n; + int channel, slot; + + space = PAGE_SIZE; + p = pvt->tmp_prt_buffer; + + n = snprintf(p, space, " "); + p += n; + space -= n; + for (channel = 0; channel < MAX_CHANNELS; channel++) { + n = snprintf(p, space, "channel %d | ", channel); + p += n; + space -= n; + } + edac_dbg(2, "%s\n", pvt->tmp_prt_buffer); + p = pvt->tmp_prt_buffer; + space = PAGE_SIZE; + n = snprintf(p, space, "-------------------------------" + "------------------------------"); + p += n; + space -= n; + edac_dbg(2, "%s\n", pvt->tmp_prt_buffer); + p = pvt->tmp_prt_buffer; + space = PAGE_SIZE; + + for (slot = 0; slot < MAX_SLOTS; slot++) { + n = snprintf(p, space, "csrow/SLOT %d ", slot); + p += n; + space -= n; + + for (channel = 0; channel < MAX_CHANNELS; channel++) { + dinfo = &pvt->dimm_info[slot][channel]; + n = snprintf(p, space, "%4d MB | ", dinfo->megabytes); + p += n; + space -= n; + } + + edac_dbg(2, "%s\n", pvt->tmp_prt_buffer); + p = pvt->tmp_prt_buffer; + space = PAGE_SIZE; + } + + n = snprintf(p, space, "-------------------------------" + "------------------------------"); + p += n; + space -= n; + edac_dbg(2, "%s\n", pvt->tmp_prt_buffer); + p = pvt->tmp_prt_buffer; + space = PAGE_SIZE; +#endif +} + +/** + * i7300_init_csrows() - Initialize the 'csrows' table within + * the mci control structure with the + * addressing of memory. + * @mci: struct mem_ctl_info pointer + */ +static int i7300_init_csrows(struct mem_ctl_info *mci) +{ + struct i7300_pvt *pvt; + struct i7300_dimm_info *dinfo; + int rc = -ENODEV; + int mtr; + int ch, branch, slot, channel, max_channel, max_branch; + struct dimm_info *dimm; + + pvt = mci->pvt_info; + + edac_dbg(2, "Memory Technology Registers:\n"); + + if (IS_SINGLE_MODE(pvt->mc_settings_a)) { + max_branch = 1; + max_channel = 1; + } else { + max_branch = MAX_BRANCHES; + max_channel = MAX_CH_PER_BRANCH; + } + + /* Get the AMB present registers for the four channels */ + for (branch = 0; branch < max_branch; branch++) { + /* Read and dump branch 0's MTRs */ + channel = to_channel(0, branch); + pci_read_config_word(pvt->pci_dev_2x_0_fbd_branch[branch], + AMBPRESENT_0, + &pvt->ambpresent[channel]); + edac_dbg(2, "\t\tAMB-present CH%d = 0x%x:\n", + channel, pvt->ambpresent[channel]); + + if (max_channel == 1) + continue; + + channel = to_channel(1, branch); + pci_read_config_word(pvt->pci_dev_2x_0_fbd_branch[branch], + AMBPRESENT_1, + &pvt->ambpresent[channel]); + edac_dbg(2, "\t\tAMB-present CH%d = 0x%x:\n", + channel, pvt->ambpresent[channel]); + } + + /* Get the set of MTR[0-7] regs by each branch */ + for (slot = 0; slot < MAX_SLOTS; slot++) { + int where = mtr_regs[slot]; + for (branch = 0; branch < max_branch; branch++) { + pci_read_config_word(pvt->pci_dev_2x_0_fbd_branch[branch], + where, + &pvt->mtr[slot][branch]); + for (ch = 0; ch < max_channel; ch++) { + int channel = to_channel(ch, branch); + + dimm = EDAC_DIMM_PTR(mci->layers, mci->dimms, + mci->n_layers, branch, ch, slot); + + dinfo = &pvt->dimm_info[slot][channel]; + + mtr = decode_mtr(pvt, slot, ch, branch, + dinfo, dimm); + + /* if no DIMMS on this row, continue */ + if (!MTR_DIMMS_PRESENT(mtr)) + continue; + + rc = 0; + + } + } + } + + return rc; +} + +/** + * decode_mir() - Decodes Memory Interleave Register (MIR) info + * @int mir_no: number of the MIR register to decode + * @mir: array with the MIR data cached on the driver + */ +static void decode_mir(int mir_no, u16 mir[MAX_MIR]) +{ + if (mir[mir_no] & 3) + edac_dbg(2, "MIR%d: limit= 0x%x Branch(es) that participate: %s %s\n", + mir_no, + (mir[mir_no] >> 4) & 0xfff, + (mir[mir_no] & 1) ? "B0" : "", + (mir[mir_no] & 2) ? "B1" : ""); +} + +/** + * i7300_get_mc_regs() - Get the contents of the MC enumeration registers + * @mci: struct mem_ctl_info pointer + * + * Data read is cached internally for its usage when needed + */ +static int i7300_get_mc_regs(struct mem_ctl_info *mci) +{ + struct i7300_pvt *pvt; + u32 actual_tolm; + int i, rc; + + pvt = mci->pvt_info; + + pci_read_config_dword(pvt->pci_dev_16_0_fsb_ctlr, AMBASE, + (u32 *) &pvt->ambase); + + edac_dbg(2, "AMBASE= 0x%lx\n", (long unsigned int)pvt->ambase); + + /* Get the Branch Map regs */ + pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map, TOLM, &pvt->tolm); + pvt->tolm >>= 12; + edac_dbg(2, "TOLM (number of 256M regions) =%u (0x%x)\n", + pvt->tolm, pvt->tolm); + + actual_tolm = (u32) ((1000l * pvt->tolm) >> (30 - 28)); + edac_dbg(2, "Actual TOLM byte addr=%u.%03u GB (0x%x)\n", + actual_tolm/1000, actual_tolm % 1000, pvt->tolm << 28); + + /* Get memory controller settings */ + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, MC_SETTINGS, + &pvt->mc_settings); + pci_read_config_dword(pvt->pci_dev_16_1_fsb_addr_map, MC_SETTINGS_A, + &pvt->mc_settings_a); + + if (IS_SINGLE_MODE(pvt->mc_settings_a)) + edac_dbg(0, "Memory controller operating on single mode\n"); + else + edac_dbg(0, "Memory controller operating on %smirrored mode\n", + IS_MIRRORED(pvt->mc_settings) ? "" : "non-"); + + edac_dbg(0, "Error detection is %s\n", + IS_ECC_ENABLED(pvt->mc_settings) ? "enabled" : "disabled"); + edac_dbg(0, "Retry is %s\n", + IS_RETRY_ENABLED(pvt->mc_settings) ? "enabled" : "disabled"); + + /* Get Memory Interleave Range registers */ + pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map, MIR0, + &pvt->mir[0]); + pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map, MIR1, + &pvt->mir[1]); + pci_read_config_word(pvt->pci_dev_16_1_fsb_addr_map, MIR2, + &pvt->mir[2]); + + /* Decode the MIR regs */ + for (i = 0; i < MAX_MIR; i++) + decode_mir(i, pvt->mir); + + rc = i7300_init_csrows(mci); + if (rc < 0) + return rc; + + /* Go and determine the size of each DIMM and place in an + * orderly matrix */ + print_dimm_size(pvt); + + return 0; +} + +/************************************************* + * i7300 Functions related to device probe/release + *************************************************/ + +/** + * i7300_put_devices() - Release the PCI devices + * @mci: struct mem_ctl_info pointer + */ +static void i7300_put_devices(struct mem_ctl_info *mci) +{ + struct i7300_pvt *pvt; + int branch; + + pvt = mci->pvt_info; + + /* Decrement usage count for devices */ + for (branch = 0; branch < MAX_CH_PER_BRANCH; branch++) + pci_dev_put(pvt->pci_dev_2x_0_fbd_branch[branch]); + pci_dev_put(pvt->pci_dev_16_2_fsb_err_regs); + pci_dev_put(pvt->pci_dev_16_1_fsb_addr_map); +} + +/** + * i7300_get_devices() - Find and perform 'get' operation on the MCH's + * device/functions we want to reference for this driver + * @mci: struct mem_ctl_info pointer + * + * Access and prepare the several devices for usage: + * I7300 devices used by this driver: + * Device 16, functions 0,1 and 2: PCI_DEVICE_ID_INTEL_I7300_MCH_ERR + * Device 21 function 0: PCI_DEVICE_ID_INTEL_I7300_MCH_FB0 + * Device 22 function 0: PCI_DEVICE_ID_INTEL_I7300_MCH_FB1 + */ +static int i7300_get_devices(struct mem_ctl_info *mci) +{ + struct i7300_pvt *pvt; + struct pci_dev *pdev; + + pvt = mci->pvt_info; + + /* Attempt to 'get' the MCH register we want */ + pdev = NULL; + while ((pdev = pci_get_device(PCI_VENDOR_ID_INTEL, + PCI_DEVICE_ID_INTEL_I7300_MCH_ERR, + pdev))) { + /* Store device 16 funcs 1 and 2 */ + switch (PCI_FUNC(pdev->devfn)) { + case 1: + if (!pvt->pci_dev_16_1_fsb_addr_map) + pvt->pci_dev_16_1_fsb_addr_map = + pci_dev_get(pdev); + break; + case 2: + if (!pvt->pci_dev_16_2_fsb_err_regs) + pvt->pci_dev_16_2_fsb_err_regs = + pci_dev_get(pdev); + break; + } + } + + if (!pvt->pci_dev_16_1_fsb_addr_map || + !pvt->pci_dev_16_2_fsb_err_regs) { + /* At least one device was not found */ + i7300_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_I7300_MCH_ERR); + goto error; + } + + edac_dbg(1, "System Address, processor bus- PCI Bus ID: %s %x:%x\n", + pci_name(pvt->pci_dev_16_0_fsb_ctlr), + pvt->pci_dev_16_0_fsb_ctlr->vendor, + pvt->pci_dev_16_0_fsb_ctlr->device); + edac_dbg(1, "Branchmap, control and errors - PCI Bus ID: %s %x:%x\n", + pci_name(pvt->pci_dev_16_1_fsb_addr_map), + pvt->pci_dev_16_1_fsb_addr_map->vendor, + pvt->pci_dev_16_1_fsb_addr_map->device); + edac_dbg(1, "FSB Error Regs - PCI Bus ID: %s %x:%x\n", + pci_name(pvt->pci_dev_16_2_fsb_err_regs), + pvt->pci_dev_16_2_fsb_err_regs->vendor, + pvt->pci_dev_16_2_fsb_err_regs->device); + + pvt->pci_dev_2x_0_fbd_branch[0] = pci_get_device(PCI_VENDOR_ID_INTEL, + PCI_DEVICE_ID_INTEL_I7300_MCH_FB0, + NULL); + if (!pvt->pci_dev_2x_0_fbd_branch[0]) { + i7300_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_I7300_MCH_FB0); + goto error; + } + + pvt->pci_dev_2x_0_fbd_branch[1] = pci_get_device(PCI_VENDOR_ID_INTEL, + PCI_DEVICE_ID_INTEL_I7300_MCH_FB1, + NULL); + if (!pvt->pci_dev_2x_0_fbd_branch[1]) { + i7300_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_I7300_MCH_FB1); + goto error; + } + + return 0; + +error: + i7300_put_devices(mci); + return -ENODEV; +} + +/** + * i7300_init_one() - Probe for one instance of the device + * @pdev: struct pci_dev pointer + * @id: struct pci_device_id pointer - currently unused + */ +static int i7300_init_one(struct pci_dev *pdev, const struct pci_device_id *id) +{ + struct mem_ctl_info *mci; + struct edac_mc_layer layers[3]; + struct i7300_pvt *pvt; + int rc; + + /* wake up device */ + rc = pci_enable_device(pdev); + if (rc == -EIO) + return rc; + + edac_dbg(0, "MC: pdev bus %u dev=0x%x fn=0x%x\n", + 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; + + /* allocate a new MC control structure */ + layers[0].type = EDAC_MC_LAYER_BRANCH; + layers[0].size = MAX_BRANCHES; + layers[0].is_virt_csrow = false; + layers[1].type = EDAC_MC_LAYER_CHANNEL; + layers[1].size = MAX_CH_PER_BRANCH; + layers[1].is_virt_csrow = true; + layers[2].type = EDAC_MC_LAYER_SLOT; + layers[2].size = MAX_SLOTS; + layers[2].is_virt_csrow = true; + mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(*pvt)); + if (mci == NULL) + return -ENOMEM; + + edac_dbg(0, "MC: mci = %p\n", mci); + + mci->pdev = &pdev->dev; /* record ptr to the generic device */ + + pvt = mci->pvt_info; + pvt->pci_dev_16_0_fsb_ctlr = pdev; /* Record this device in our private */ + + pvt->tmp_prt_buffer = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (!pvt->tmp_prt_buffer) { + edac_mc_free(mci); + return -ENOMEM; + } + + /* 'get' the pci devices we want to reserve for our use */ + if (i7300_get_devices(mci)) + goto fail0; + + 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 = "i7300_edac.c"; + mci->mod_ver = I7300_REVISION; + mci->ctl_name = i7300_devs[0].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 = i7300_check_error; + + /* initialize the MC control structure 'csrows' table + * with the mapping and control information */ + if (i7300_get_mc_regs(mci)) { + edac_dbg(0, "MC: Setting mci->edac_cap to EDAC_FLAG_NONE because i7300_init_csrows() returned nonzero value\n"); + mci->edac_cap = EDAC_FLAG_NONE; /* no csrows found */ + } else { + edac_dbg(1, "MC: Enable error reporting now\n"); + i7300_enable_error_reporting(mci); + } + + /* add this new MC control structure to EDAC's list of MCs */ + if (edac_mc_add_mc(mci)) { + edac_dbg(0, "MC: failed edac_mc_add_mc()\n"); + /* FIXME: perhaps some code should go here that disables error + * reporting if we just enabled it + */ + goto fail1; + } + + i7300_clear_error(mci); + + /* allocating generic PCI control info */ + i7300_pci = edac_pci_create_generic_ctl(&pdev->dev, EDAC_MOD_STR); + if (!i7300_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: + + i7300_put_devices(mci); + +fail0: + kfree(pvt->tmp_prt_buffer); + edac_mc_free(mci); + return -ENODEV; +} + +/** + * i7300_remove_one() - Remove the driver + * @pdev: struct pci_dev pointer + */ +static void i7300_remove_one(struct pci_dev *pdev) +{ + struct mem_ctl_info *mci; + char *tmp; + + edac_dbg(0, "\n"); + + if (i7300_pci) + edac_pci_release_generic_ctl(i7300_pci); + + mci = edac_mc_del_mc(&pdev->dev); + if (!mci) + return; + + tmp = ((struct i7300_pvt *)mci->pvt_info)->tmp_prt_buffer; + + /* retrieve references to resources, and free those resources */ + i7300_put_devices(mci); + + kfree(tmp); + edac_mc_free(mci); +} + +/* + * pci_device_id: table for which devices we are looking for + * + * Has only 8086:360c PCI ID + */ +static const struct pci_device_id i7300_pci_tbl[] = { + {PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_I7300_MCH_ERR)}, + {0,} /* 0 terminated list. */ +}; + +MODULE_DEVICE_TABLE(pci, i7300_pci_tbl); + +/* + * i7300_driver: pci_driver structure for this module + */ +static struct pci_driver i7300_driver = { + .name = "i7300_edac", + .probe = i7300_init_one, + .remove = i7300_remove_one, + .id_table = i7300_pci_tbl, +}; + +/** + * i7300_init() - Registers the driver + */ +static int __init i7300_init(void) +{ + int pci_rc; + + edac_dbg(2, "\n"); + + /* Ensure that the OPSTATE is set correctly for POLL or NMI */ + opstate_init(); + + pci_rc = pci_register_driver(&i7300_driver); + + return (pci_rc < 0) ? pci_rc : 0; +} + +/** + * i7300_init() - Unregisters the driver + */ +static void __exit i7300_exit(void) +{ + edac_dbg(2, "\n"); + pci_unregister_driver(&i7300_driver); +} + +module_init(i7300_init); +module_exit(i7300_exit); + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("Mauro Carvalho Chehab"); +MODULE_AUTHOR("Red Hat Inc. (http://www.redhat.com)"); +MODULE_DESCRIPTION("MC Driver for Intel I7300 memory controllers - " + I7300_REVISION); + +module_param(edac_op_state, int, 0444); +MODULE_PARM_DESC(edac_op_state, "EDAC Error Reporting state: 0=Poll,1=NMI"); |