From 6866b390562b2c948c6f6fc2a5c103f090a02e01 Mon Sep 17 00:00:00 2001 From: Fabian Frederick Date: Mon, 9 Jun 2014 21:20:18 +0200 Subject: EDAC, edac_module.c: Remove unnecessary test on unsigned value unsigned long value is never < 0. Cc: Doug Thompson Signed-off-by: Fabian Frederick Link: http://lkml.kernel.org/r/1402341618-10674-1-git-send-email-fabf@skynet.be Signed-off-by: Borislav Petkov --- drivers/edac/edac_module.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/edac/edac_module.c b/drivers/edac/edac_module.c index a66941fea5a4..e6d1691dfa45 100644 --- a/drivers/edac/edac_module.c +++ b/drivers/edac/edac_module.c @@ -28,7 +28,7 @@ static int edac_set_debug_level(const char *buf, struct kernel_param *kp) if (ret) return ret; - if (val < 0 || val > 4) + if (val > 4) return -EINVAL; return param_set_int(buf, kp); -- cgit v1.2.3 From 3a044178cccfeb8664423c2950c499c3a209ed9f Mon Sep 17 00:00:00 2001 From: Jason Baron Date: Fri, 4 Jul 2014 13:27:04 +0200 Subject: readq/writeq: Add explicit lo_hi_[read|write]_q and hi_lo_[read|write]_q Even on x86-64, I've found the need to break up a readq() into 2 readl() calls. According to the Intel datasheet for the E3-1200 processor: " Software must not access B0/D0/F0 32-bit memory-mapped registers with requests that cross a DW boundary. " (http://www.intel.com/content/www/us/en/processors/xeon/xeon-e3-1200-family-vol-2-datasheet.html p. 16) I can confirm this is true via several hard machine lockups. Thus, add explicit hi_lo_[readq|write]_q and lo_hi_[read|write]_q so that these uses are spelled out. Signed-off-by: Jason Baron Link: http://lkml.kernel.org/r/281f09da7ad01e5cea99737ec34d2399bdbbbf63.1403818526.git.jbaron@akamai.com Signed-off-by: Borislav Petkov --- include/asm-generic/io-64-nonatomic-hi-lo.h | 14 +++++++++----- include/asm-generic/io-64-nonatomic-lo-hi.h | 14 +++++++++----- 2 files changed, 18 insertions(+), 10 deletions(-) diff --git a/include/asm-generic/io-64-nonatomic-hi-lo.h b/include/asm-generic/io-64-nonatomic-hi-lo.h index a6806a94250d..2e29d13fc154 100644 --- a/include/asm-generic/io-64-nonatomic-hi-lo.h +++ b/include/asm-generic/io-64-nonatomic-hi-lo.h @@ -4,8 +4,7 @@ #include #include -#ifndef readq -static inline __u64 readq(const volatile void __iomem *addr) +static inline __u64 hi_lo_readq(const volatile void __iomem *addr) { const volatile u32 __iomem *p = addr; u32 low, high; @@ -15,14 +14,19 @@ static inline __u64 readq(const volatile void __iomem *addr) return low + ((u64)high << 32); } -#endif -#ifndef writeq -static inline void writeq(__u64 val, volatile void __iomem *addr) +static inline void hi_lo_writeq(__u64 val, volatile void __iomem *addr) { writel(val >> 32, addr + 4); writel(val, addr); } + +#ifndef readq +#define readq hi_lo_readq +#endif + +#ifndef writeq +#define writeq hi_lo_writeq #endif #endif /* _ASM_IO_64_NONATOMIC_HI_LO_H_ */ diff --git a/include/asm-generic/io-64-nonatomic-lo-hi.h b/include/asm-generic/io-64-nonatomic-lo-hi.h index ca546b1ff8b5..0efacff0a1ce 100644 --- a/include/asm-generic/io-64-nonatomic-lo-hi.h +++ b/include/asm-generic/io-64-nonatomic-lo-hi.h @@ -4,8 +4,7 @@ #include #include -#ifndef readq -static inline __u64 readq(const volatile void __iomem *addr) +static inline __u64 lo_hi_readq(const volatile void __iomem *addr) { const volatile u32 __iomem *p = addr; u32 low, high; @@ -15,14 +14,19 @@ static inline __u64 readq(const volatile void __iomem *addr) return low + ((u64)high << 32); } -#endif -#ifndef writeq -static inline void writeq(__u64 val, volatile void __iomem *addr) +static inline void lo_hi_writeq(__u64 val, volatile void __iomem *addr) { writel(val, addr); writel(val >> 32, addr + 4); } + +#ifndef readq +#define readq lo_hi_readq +#endif + +#ifndef writeq +#define writeq lo_hi_writeq #endif #endif /* _ASM_IO_64_NONATOMIC_LO_HI_H_ */ -- cgit v1.2.3 From a21e98ce1e0cafac30c1b11e41b9e690d8d93442 Mon Sep 17 00:00:00 2001 From: Jason Baron Date: Thu, 26 Jun 2014 21:58:32 +0000 Subject: x38_edac: make use of lo_hi_readq() Convert to the generic API. Signed-off-by: Jason Baron Link: http://lkml.kernel.org/r/bb9a4cbb980cc7b51be75cbfcf644553bf6a04cd.1403818526.git.jbaron@akamai.com Signed-off-by: Borislav Petkov --- drivers/edac/x38_edac.c | 15 ++++++--------- 1 file changed, 6 insertions(+), 9 deletions(-) diff --git a/drivers/edac/x38_edac.c b/drivers/edac/x38_edac.c index 4891b450830b..e644b52c287c 100644 --- a/drivers/edac/x38_edac.c +++ b/drivers/edac/x38_edac.c @@ -14,6 +14,8 @@ #include #include #include + +#include #include "edac_core.h" #define X38_REVISION "1.1" @@ -161,11 +163,6 @@ static void x38_clear_error_info(struct mem_ctl_info *mci) X38_ERRSTS_BITS); } -static u64 x38_readq(const void __iomem *addr) -{ - return readl(addr) | (((u64)readl(addr + 4)) << 32); -} - static void x38_get_and_clear_error_info(struct mem_ctl_info *mci, struct x38_error_info *info) { @@ -183,9 +180,9 @@ static void x38_get_and_clear_error_info(struct mem_ctl_info *mci, if (!(info->errsts & X38_ERRSTS_BITS)) return; - info->eccerrlog[0] = x38_readq(window + X38_C0ECCERRLOG); + info->eccerrlog[0] = lo_hi_readq(window + X38_C0ECCERRLOG); if (x38_channel_num == 2) - info->eccerrlog[1] = x38_readq(window + X38_C1ECCERRLOG); + info->eccerrlog[1] = lo_hi_readq(window + X38_C1ECCERRLOG); pci_read_config_word(pdev, X38_ERRSTS, &info->errsts2); @@ -196,10 +193,10 @@ static void x38_get_and_clear_error_info(struct mem_ctl_info *mci, * should be UE info. */ if ((info->errsts ^ info->errsts2) & X38_ERRSTS_BITS) { - info->eccerrlog[0] = x38_readq(window + X38_C0ECCERRLOG); + info->eccerrlog[0] = lo_hi_readq(window + X38_C0ECCERRLOG); if (x38_channel_num == 2) info->eccerrlog[1] = - x38_readq(window + X38_C1ECCERRLOG); + lo_hi_readq(window + X38_C1ECCERRLOG); } x38_clear_error_info(mci); -- cgit v1.2.3 From 7ee40b897d18ab03111eda9a6a0550e98166eada Mon Sep 17 00:00:00 2001 From: Jason Baron Date: Fri, 4 Jul 2014 13:48:32 +0200 Subject: ie31200_edac: Introduce the driver Add a driver for the E3-1200 series of Intel DRAM controllers, based on the following E3-1200 specs: http://www.intel.com/content/www/us/en/processors/xeon/xeon-e3-1200-family-vol-2-datasheet.html http://www.intel.com/content/www/us/en/processors/xeon/xeon-e3-1200v3-vol-2-datasheet.html I've tested this on bad memory hardware, and observed correlating bad reads and uncorrected memory errors as reported by the driver. Tested against: CPU E3-1270 v3 @ 3.50GHz : 8086:0c08 (haswell) CPU E3-1270 V2 @ 3.50GHz : 8086:0158 (ivy bridge) CPU E31270 @ 3.40GHz : 8086:0108 (sandy bridge) Signed-off-by: Jason Baron Link: http://lkml.kernel.org/r/95c83e80dd40b5377e8bb206285c5d95ac623872.1403818526.git.jbaron@akamai.com [ Boris: realign defines ] Signed-off-by: Borislav Petkov --- drivers/edac/Kconfig | 7 + drivers/edac/Makefile | 1 + drivers/edac/ie31200_edac.c | 541 ++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 549 insertions(+) create mode 100644 drivers/edac/ie31200_edac.c diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig index 878f09005fad..e339c6b91425 100644 --- a/drivers/edac/Kconfig +++ b/drivers/edac/Kconfig @@ -186,6 +186,13 @@ config EDAC_I3200 Support for error detection and correction on the Intel 3200 and 3210 server chipsets. +config EDAC_IE31200 + tristate "Intel e312xx" + depends on EDAC_MM_EDAC && PCI && X86 + help + Support for error detection and correction on the Intel + E3-1200 based DRAM controllers. + config EDAC_X38 tristate "Intel X38" depends on EDAC_MM_EDAC && PCI && X86 diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile index 4154ed6a02c6..c479a24d8f77 100644 --- a/drivers/edac/Makefile +++ b/drivers/edac/Makefile @@ -37,6 +37,7 @@ obj-$(CONFIG_EDAC_I82875P) += i82875p_edac.o obj-$(CONFIG_EDAC_I82975X) += i82975x_edac.o obj-$(CONFIG_EDAC_I3000) += i3000_edac.o obj-$(CONFIG_EDAC_I3200) += i3200_edac.o +obj-$(CONFIG_EDAC_IE31200) += ie31200_edac.o obj-$(CONFIG_EDAC_X38) += x38_edac.o obj-$(CONFIG_EDAC_I82860) += i82860_edac.o obj-$(CONFIG_EDAC_R82600) += r82600_edac.o diff --git a/drivers/edac/ie31200_edac.c b/drivers/edac/ie31200_edac.c new file mode 100644 index 000000000000..6d3d2c3cb6d6 --- /dev/null +++ b/drivers/edac/ie31200_edac.c @@ -0,0 +1,541 @@ +/* + * Intel E3-1200 + * Copyright (C) 2014 Jason Baron + * + * Support for the E3-1200 processor family. Heavily based on previous + * Intel EDAC drivers. + * + * Since the DRAM controller is on the cpu chip, we can use its PCI device + * id to identify these processors. + * + * PCI DRAM controller device ids (Taken from The PCI ID Repository - http://pci-ids.ucw.cz/) + * + * 0108: Xeon E3-1200 Processor Family DRAM Controller + * 010c: Xeon E3-1200/2nd Generation Core Processor Family DRAM Controller + * 0150: Xeon E3-1200 v2/3rd Gen Core processor DRAM Controller + * 0158: Xeon E3-1200 v2/Ivy Bridge DRAM Controller + * 015c: Xeon E3-1200 v2/3rd Gen Core processor DRAM Controller + * 0c04: Xeon E3-1200 v3/4th Gen Core Processor DRAM Controller + * 0c08: Xeon E3-1200 v3 Processor DRAM Controller + * + * Based on Intel specification: + * http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e3-1200v3-vol-2-datasheet.pdf + * http://www.intel.com/content/www/us/en/processors/xeon/xeon-e3-1200-family-vol-2-datasheet.html + * + * According to the above datasheet (p.16): + * " + * 6. Software must not access B0/D0/F0 32-bit memory-mapped registers with + * requests that cross a DW boundary. + * " + * + * Thus, we make use of the explicit: lo_hi_readq(), which breaks the readq into + * 2 readl() calls. This restriction may be lifted in subsequent chip releases, + * but lo_hi_readq() ensures that we are safe across all e3-1200 processors. + */ + +#include +#include +#include +#include +#include + +#include +#include "edac_core.h" + +#define IE31200_REVISION "1.0" +#define EDAC_MOD_STR "ie31200_edac" + +#define ie31200_printk(level, fmt, arg...) \ + edac_printk(level, "ie31200", fmt, ##arg) + +#define PCI_DEVICE_ID_INTEL_IE31200_HB_1 0x0108 +#define PCI_DEVICE_ID_INTEL_IE31200_HB_2 0x010c +#define PCI_DEVICE_ID_INTEL_IE31200_HB_3 0x0150 +#define PCI_DEVICE_ID_INTEL_IE31200_HB_4 0x0158 +#define PCI_DEVICE_ID_INTEL_IE31200_HB_5 0x015c +#define PCI_DEVICE_ID_INTEL_IE31200_HB_6 0x0c04 +#define PCI_DEVICE_ID_INTEL_IE31200_HB_7 0x0c08 + +#define IE31200_DIMMS 4 +#define IE31200_RANKS 8 +#define IE31200_RANKS_PER_CHANNEL 4 +#define IE31200_DIMMS_PER_CHANNEL 2 +#define IE31200_CHANNELS 2 + +/* Intel IE31200 register addresses - device 0 function 0 - DRAM Controller */ +#define IE31200_MCHBAR_LOW 0x48 +#define IE31200_MCHBAR_HIGH 0x4c +#define IE31200_MCHBAR_MASK GENMASK_ULL(38, 15) +#define IE31200_MMR_WINDOW_SIZE BIT(15) + +/* + * Error Status Register (16b) + * + * 15 reserved + * 14 Isochronous TBWRR Run Behind FIFO Full + * (ITCV) + * 13 Isochronous TBWRR Run Behind FIFO Put + * (ITSTV) + * 12 reserved + * 11 MCH Thermal Sensor Event + * for SMI/SCI/SERR (GTSE) + * 10 reserved + * 9 LOCK to non-DRAM Memory Flag (LCKF) + * 8 reserved + * 7 DRAM Throttle Flag (DTF) + * 6:2 reserved + * 1 Multi-bit DRAM ECC Error Flag (DMERR) + * 0 Single-bit DRAM ECC Error Flag (DSERR) + */ +#define IE31200_ERRSTS 0xc8 +#define IE31200_ERRSTS_UE BIT(1) +#define IE31200_ERRSTS_CE BIT(0) +#define IE31200_ERRSTS_BITS (IE31200_ERRSTS_UE | IE31200_ERRSTS_CE) + +/* + * Channel 0 ECC Error Log (64b) + * + * 63:48 Error Column Address (ERRCOL) + * 47:32 Error Row Address (ERRROW) + * 31:29 Error Bank Address (ERRBANK) + * 28:27 Error Rank Address (ERRRANK) + * 26:24 reserved + * 23:16 Error Syndrome (ERRSYND) + * 15: 2 reserved + * 1 Multiple Bit Error Status (MERRSTS) + * 0 Correctable Error Status (CERRSTS) + */ +#define IE31200_C0ECCERRLOG 0x40c8 +#define IE31200_C1ECCERRLOG 0x44c8 +#define IE31200_ECCERRLOG_CE BIT(0) +#define IE31200_ECCERRLOG_UE BIT(1) +#define IE31200_ECCERRLOG_RANK_BITS GENMASK_ULL(28, 27) +#define IE31200_ECCERRLOG_RANK_SHIFT 27 +#define IE31200_ECCERRLOG_SYNDROME_BITS GENMASK_ULL(23, 16) +#define IE31200_ECCERRLOG_SYNDROME_SHIFT 16 + +#define IE31200_ECCERRLOG_SYNDROME(log) \ + ((log & IE31200_ECCERRLOG_SYNDROME_BITS) >> \ + IE31200_ECCERRLOG_SYNDROME_SHIFT) + +#define IE31200_CAPID0 0xe4 +#define IE31200_CAPID0_PDCD BIT(4) +#define IE31200_CAPID0_DDPCD BIT(6) +#define IE31200_CAPID0_ECC BIT(1) + +#define IE31200_MAD_DIMM_0_OFFSET 0x5004 +#define IE31200_MAD_DIMM_SIZE GENMASK_ULL(7, 0) +#define IE31200_MAD_DIMM_A_RANK BIT(17) +#define IE31200_MAD_DIMM_A_WIDTH BIT(19) + +#define IE31200_PAGES(n) (n << (28 - PAGE_SHIFT)) + +static int nr_channels; + +struct ie31200_priv { + void __iomem *window; +}; + +enum ie31200_chips { + IE31200 = 0, +}; + +struct ie31200_dev_info { + const char *ctl_name; +}; + +struct ie31200_error_info { + u16 errsts; + u16 errsts2; + u64 eccerrlog[IE31200_CHANNELS]; +}; + +static const struct ie31200_dev_info ie31200_devs[] = { + [IE31200] = { + .ctl_name = "IE31200" + }, +}; + +struct dimm_data { + u8 size; /* in 256MB multiples */ + u8 dual_rank : 1, + x16_width : 1; /* 0 means x8 width */ +}; + +static int how_many_channels(struct pci_dev *pdev) +{ + int n_channels; + unsigned char capid0_2b; /* 2nd byte of CAPID0 */ + + pci_read_config_byte(pdev, IE31200_CAPID0 + 1, &capid0_2b); + + /* check PDCD: Dual Channel Disable */ + if (capid0_2b & IE31200_CAPID0_PDCD) { + edac_dbg(0, "In single channel mode\n"); + n_channels = 1; + } else { + edac_dbg(0, "In dual channel mode\n"); + n_channels = 2; + } + + /* check DDPCD - check if both channels are filled */ + if (capid0_2b & IE31200_CAPID0_DDPCD) + edac_dbg(0, "2 DIMMS per channel disabled\n"); + else + edac_dbg(0, "2 DIMMS per channel enabled\n"); + + return n_channels; +} + +static bool ecc_capable(struct pci_dev *pdev) +{ + unsigned char capid0_4b; /* 4th byte of CAPID0 */ + + pci_read_config_byte(pdev, IE31200_CAPID0 + 3, &capid0_4b); + if (capid0_4b & IE31200_CAPID0_ECC) + return false; + return true; +} + +static int eccerrlog_row(int channel, u64 log) +{ + int rank = ((log & IE31200_ECCERRLOG_RANK_BITS) >> + IE31200_ECCERRLOG_RANK_SHIFT); + return rank | (channel * IE31200_RANKS_PER_CHANNEL); +} + +static void ie31200_clear_error_info(struct mem_ctl_info *mci) +{ + /* + * Clear any error bits. + * (Yes, we really clear bits by writing 1 to them.) + */ + pci_write_bits16(to_pci_dev(mci->pdev), IE31200_ERRSTS, + IE31200_ERRSTS_BITS, IE31200_ERRSTS_BITS); +} + +static void ie31200_get_and_clear_error_info(struct mem_ctl_info *mci, + struct ie31200_error_info *info) +{ + struct pci_dev *pdev; + struct ie31200_priv *priv = mci->pvt_info; + void __iomem *window = priv->window; + + pdev = to_pci_dev(mci->pdev); + + /* + * This is a mess because there is no atomic way to read all the + * registers at once and the registers can transition from CE being + * overwritten by UE. + */ + pci_read_config_word(pdev, IE31200_ERRSTS, &info->errsts); + if (!(info->errsts & IE31200_ERRSTS_BITS)) + return; + + info->eccerrlog[0] = lo_hi_readq(window + IE31200_C0ECCERRLOG); + if (nr_channels == 2) + info->eccerrlog[1] = lo_hi_readq(window + IE31200_C1ECCERRLOG); + + pci_read_config_word(pdev, IE31200_ERRSTS, &info->errsts2); + + /* + * If the error is the same for both reads then the first set + * of reads is valid. If there is a change then there is a CE + * with no info and the second set of reads is valid and + * should be UE info. + */ + if ((info->errsts ^ info->errsts2) & IE31200_ERRSTS_BITS) { + info->eccerrlog[0] = lo_hi_readq(window + IE31200_C0ECCERRLOG); + if (nr_channels == 2) + info->eccerrlog[1] = + lo_hi_readq(window + IE31200_C1ECCERRLOG); + } + + ie31200_clear_error_info(mci); +} + +static void ie31200_process_error_info(struct mem_ctl_info *mci, + struct ie31200_error_info *info) +{ + int channel; + u64 log; + + if (!(info->errsts & IE31200_ERRSTS_BITS)) + return; + + if ((info->errsts ^ info->errsts2) & IE31200_ERRSTS_BITS) { + edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, 0, 0, 0, + -1, -1, -1, "UE overwrote CE", ""); + info->errsts = info->errsts2; + } + + for (channel = 0; channel < nr_channels; channel++) { + log = info->eccerrlog[channel]; + if (log & IE31200_ECCERRLOG_UE) { + edac_mc_handle_error(HW_EVENT_ERR_UNCORRECTED, mci, 1, + 0, 0, 0, + eccerrlog_row(channel, log), + channel, -1, + "ie31200 UE", ""); + } else if (log & IE31200_ECCERRLOG_CE) { + edac_mc_handle_error(HW_EVENT_ERR_CORRECTED, mci, 1, + 0, 0, + IE31200_ECCERRLOG_SYNDROME(log), + eccerrlog_row(channel, log), + channel, -1, + "ie31200 CE", ""); + } + } +} + +static void ie31200_check(struct mem_ctl_info *mci) +{ + struct ie31200_error_info info; + + edac_dbg(1, "MC%d\n", mci->mc_idx); + ie31200_get_and_clear_error_info(mci, &info); + ie31200_process_error_info(mci, &info); +} + +static void __iomem *ie31200_map_mchbar(struct pci_dev *pdev) +{ + union { + u64 mchbar; + struct { + u32 mchbar_low; + u32 mchbar_high; + }; + } u; + void __iomem *window; + + pci_read_config_dword(pdev, IE31200_MCHBAR_LOW, &u.mchbar_low); + pci_read_config_dword(pdev, IE31200_MCHBAR_HIGH, &u.mchbar_high); + u.mchbar &= IE31200_MCHBAR_MASK; + + if (u.mchbar != (resource_size_t)u.mchbar) { + ie31200_printk(KERN_ERR, "mmio space beyond accessible range (0x%llx)\n", + (unsigned long long)u.mchbar); + return NULL; + } + + window = ioremap_nocache(u.mchbar, IE31200_MMR_WINDOW_SIZE); + if (!window) + ie31200_printk(KERN_ERR, "Cannot map mmio space at 0x%llx\n", + (unsigned long long)u.mchbar); + + return window; +} + +static int ie31200_probe1(struct pci_dev *pdev, int dev_idx) +{ + int rc; + int i, j; + struct mem_ctl_info *mci = NULL; + struct edac_mc_layer layers[2]; + struct dimm_data dimm_info[IE31200_CHANNELS][IE31200_DIMMS_PER_CHANNEL]; + void __iomem *window; + struct ie31200_priv *priv; + u32 addr_decode; + + edac_dbg(0, "MC:\n"); + + if (!ecc_capable(pdev)) { + ie31200_printk(KERN_INFO, "No ECC support\n"); + return -ENODEV; + } + + window = ie31200_map_mchbar(pdev); + if (!window) + return -ENODEV; + + /* populate DIMM info */ + for (i = 0; i < IE31200_CHANNELS; i++) { + addr_decode = readl(window + IE31200_MAD_DIMM_0_OFFSET + + (i * 4)); + edac_dbg(0, "addr_decode: 0x%x\n", addr_decode); + for (j = 0; j < IE31200_DIMMS_PER_CHANNEL; j++) { + dimm_info[i][j].size = (addr_decode >> (j * 8)) & + IE31200_MAD_DIMM_SIZE; + dimm_info[i][j].dual_rank = (addr_decode & + (IE31200_MAD_DIMM_A_RANK << j)) ? 1 : 0; + dimm_info[i][j].x16_width = (addr_decode & + (IE31200_MAD_DIMM_A_WIDTH << j)) ? 1 : 0; + edac_dbg(0, "size: 0x%x, rank: %d, width: %d\n", + dimm_info[i][j].size, + dimm_info[i][j].dual_rank, + dimm_info[i][j].x16_width); + } + } + + nr_channels = how_many_channels(pdev); + + layers[0].type = EDAC_MC_LAYER_CHIP_SELECT; + layers[0].size = IE31200_DIMMS; + layers[0].is_virt_csrow = true; + layers[1].type = EDAC_MC_LAYER_CHANNEL; + layers[1].size = nr_channels; + layers[1].is_virt_csrow = false; + mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, + sizeof(struct ie31200_priv)); + + rc = -ENOMEM; + if (!mci) + goto fail_unmap; + + edac_dbg(3, "MC: init mci\n"); + + mci->pdev = &pdev->dev; + mci->mtype_cap = MEM_FLAG_DDR3; + + mci->edac_ctl_cap = EDAC_FLAG_SECDED; + mci->edac_cap = EDAC_FLAG_SECDED; + + mci->mod_name = EDAC_MOD_STR; + mci->mod_ver = IE31200_REVISION; + mci->ctl_name = ie31200_devs[dev_idx].ctl_name; + mci->dev_name = pci_name(pdev); + mci->edac_check = ie31200_check; + mci->ctl_page_to_phys = NULL; + priv = mci->pvt_info; + priv->window = window; + + /* + * The dram rank boundary (DRB) reg values are boundary addresses + * for each DRAM rank with a granularity of 64MB. DRB regs are + * cumulative; the last one will contain the total memory + * contained in all ranks. + */ + for (i = 0; i < IE31200_DIMMS_PER_CHANNEL; i++) { + for (j = 0; j < IE31200_CHANNELS; j++) { + struct dimm_info *dimm; + unsigned long nr_pages; + + nr_pages = IE31200_PAGES(dimm_info[j][i].size); + if (nr_pages == 0) + continue; + + if (dimm_info[j][i].dual_rank) { + nr_pages = nr_pages / 2; + dimm = EDAC_DIMM_PTR(mci->layers, mci->dimms, + mci->n_layers, (i * 2) + 1, + j, 0); + dimm->nr_pages = nr_pages; + edac_dbg(0, "set nr pages: 0x%lx\n", nr_pages); + dimm->grain = 8; /* just a guess */ + dimm->mtype = MEM_DDR3; + dimm->dtype = DEV_UNKNOWN; + dimm->edac_mode = EDAC_UNKNOWN; + } + dimm = EDAC_DIMM_PTR(mci->layers, mci->dimms, + mci->n_layers, i * 2, j, 0); + dimm->nr_pages = nr_pages; + edac_dbg(0, "set nr pages: 0x%lx\n", nr_pages); + dimm->grain = 8; /* same guess */ + dimm->mtype = MEM_DDR3; + dimm->dtype = DEV_UNKNOWN; + dimm->edac_mode = EDAC_UNKNOWN; + } + } + + ie31200_clear_error_info(mci); + + rc = -ENODEV; + if (edac_mc_add_mc(mci)) { + edac_dbg(3, "MC: failed edac_mc_add_mc()\n"); + goto fail_free; + } + + /* get this far and it's successful */ + edac_dbg(3, "MC: success\n"); + return 0; + +fail_free: + if (mci) + edac_mc_free(mci); +fail_unmap: + iounmap(window); + + return rc; +} + +static int ie31200_init_one(struct pci_dev *pdev, + const struct pci_device_id *ent) +{ + edac_dbg(0, "MC:\n"); + + if (pci_enable_device(pdev) < 0) + return -EIO; + + return ie31200_probe1(pdev, ent->driver_data); +} + +static void ie31200_remove_one(struct pci_dev *pdev) +{ + struct mem_ctl_info *mci; + struct ie31200_priv *priv; + + edac_dbg(0, "\n"); + mci = edac_mc_del_mc(&pdev->dev); + if (!mci) + return; + priv = mci->pvt_info; + iounmap(priv->window); + edac_mc_free(mci); +} + +static const struct pci_device_id ie31200_pci_tbl[] = { + { + PCI_VEND_DEV(INTEL, IE31200_HB_1), PCI_ANY_ID, PCI_ANY_ID, 0, 0, + IE31200}, + { + PCI_VEND_DEV(INTEL, IE31200_HB_2), PCI_ANY_ID, PCI_ANY_ID, 0, 0, + IE31200}, + { + PCI_VEND_DEV(INTEL, IE31200_HB_3), PCI_ANY_ID, PCI_ANY_ID, 0, 0, + IE31200}, + { + PCI_VEND_DEV(INTEL, IE31200_HB_4), PCI_ANY_ID, PCI_ANY_ID, 0, 0, + IE31200}, + { + PCI_VEND_DEV(INTEL, IE31200_HB_5), PCI_ANY_ID, PCI_ANY_ID, 0, 0, + IE31200}, + { + PCI_VEND_DEV(INTEL, IE31200_HB_6), PCI_ANY_ID, PCI_ANY_ID, 0, 0, + IE31200}, + { + PCI_VEND_DEV(INTEL, IE31200_HB_7), PCI_ANY_ID, PCI_ANY_ID, 0, 0, + IE31200}, + { + 0, + } /* 0 terminated list. */ +}; +MODULE_DEVICE_TABLE(pci, ie31200_pci_tbl); + +static struct pci_driver ie31200_driver = { + .name = EDAC_MOD_STR, + .probe = ie31200_init_one, + .remove = ie31200_remove_one, + .id_table = ie31200_pci_tbl, +}; + +static int __init ie31200_init(void) +{ + edac_dbg(3, "MC:\n"); + /* Ensure that the OPSTATE is set correctly for POLL or NMI */ + opstate_init(); + + return pci_register_driver(&ie31200_driver); +} + +static void __exit ie31200_exit(void) +{ + edac_dbg(3, "MC:\n"); + pci_unregister_driver(&ie31200_driver); +} + +module_init(ie31200_init); +module_exit(ie31200_exit); + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("Jason Baron "); +MODULE_DESCRIPTION("MC support for Intel Processor E31200 memory hub controllers"); -- cgit v1.2.3 From 78fd4d1242e88fbe5ea269087a47bd7e05bf84a1 Mon Sep 17 00:00:00 2001 From: Jason Baron Date: Wed, 9 Jul 2014 21:13:07 +0000 Subject: ie31200_edac: Allocate mci and map mchbar first Check for memory allocation and mchbar mapping failures before initializing the dimm info tables needlessly. Signed-off-by: Jason Baron Suggested-by: Borislav Petkov Link: http://lkml.kernel.org/r/ead8f53e699f1ce21c2e17f3cffb4685d4faf72a.1404939455.git.jbaron@akamai.com Signed-off-by: Borislav Petkov --- drivers/edac/ie31200_edac.c | 71 +++++++++++++++++++++------------------------ 1 file changed, 33 insertions(+), 38 deletions(-) diff --git a/drivers/edac/ie31200_edac.c b/drivers/edac/ie31200_edac.c index 6d3d2c3cb6d6..a981dc6fd88e 100644 --- a/drivers/edac/ie31200_edac.c +++ b/drivers/edac/ie31200_edac.c @@ -328,8 +328,7 @@ static void __iomem *ie31200_map_mchbar(struct pci_dev *pdev) static int ie31200_probe1(struct pci_dev *pdev, int dev_idx) { - int rc; - int i, j; + int i, j, ret; struct mem_ctl_info *mci = NULL; struct edac_mc_layer layers[2]; struct dimm_data dimm_info[IE31200_CHANNELS][IE31200_DIMMS_PER_CHANNEL]; @@ -344,31 +343,7 @@ static int ie31200_probe1(struct pci_dev *pdev, int dev_idx) return -ENODEV; } - window = ie31200_map_mchbar(pdev); - if (!window) - return -ENODEV; - - /* populate DIMM info */ - for (i = 0; i < IE31200_CHANNELS; i++) { - addr_decode = readl(window + IE31200_MAD_DIMM_0_OFFSET + - (i * 4)); - edac_dbg(0, "addr_decode: 0x%x\n", addr_decode); - for (j = 0; j < IE31200_DIMMS_PER_CHANNEL; j++) { - dimm_info[i][j].size = (addr_decode >> (j * 8)) & - IE31200_MAD_DIMM_SIZE; - dimm_info[i][j].dual_rank = (addr_decode & - (IE31200_MAD_DIMM_A_RANK << j)) ? 1 : 0; - dimm_info[i][j].x16_width = (addr_decode & - (IE31200_MAD_DIMM_A_WIDTH << j)) ? 1 : 0; - edac_dbg(0, "size: 0x%x, rank: %d, width: %d\n", - dimm_info[i][j].size, - dimm_info[i][j].dual_rank, - dimm_info[i][j].x16_width); - } - } - nr_channels = how_many_channels(pdev); - layers[0].type = EDAC_MC_LAYER_CHIP_SELECT; layers[0].size = IE31200_DIMMS; layers[0].is_virt_csrow = true; @@ -377,19 +352,20 @@ static int ie31200_probe1(struct pci_dev *pdev, int dev_idx) layers[1].is_virt_csrow = false; mci = edac_mc_alloc(0, ARRAY_SIZE(layers), layers, sizeof(struct ie31200_priv)); - - rc = -ENOMEM; if (!mci) - goto fail_unmap; + return -ENOMEM; - edac_dbg(3, "MC: init mci\n"); + window = ie31200_map_mchbar(pdev); + if (!window) { + ret = -ENODEV; + goto fail_free; + } + edac_dbg(3, "MC: init mci\n"); mci->pdev = &pdev->dev; mci->mtype_cap = MEM_FLAG_DDR3; - mci->edac_ctl_cap = EDAC_FLAG_SECDED; mci->edac_cap = EDAC_FLAG_SECDED; - mci->mod_name = EDAC_MOD_STR; mci->mod_ver = IE31200_REVISION; mci->ctl_name = ie31200_devs[dev_idx].ctl_name; @@ -399,6 +375,25 @@ static int ie31200_probe1(struct pci_dev *pdev, int dev_idx) priv = mci->pvt_info; priv->window = window; + /* populate DIMM info */ + for (i = 0; i < IE31200_CHANNELS; i++) { + addr_decode = readl(window + IE31200_MAD_DIMM_0_OFFSET + + (i * 4)); + edac_dbg(0, "addr_decode: 0x%x\n", addr_decode); + for (j = 0; j < IE31200_DIMMS_PER_CHANNEL; j++) { + dimm_info[i][j].size = (addr_decode >> (j * 8)) & + IE31200_MAD_DIMM_SIZE; + dimm_info[i][j].dual_rank = (addr_decode & + (IE31200_MAD_DIMM_A_RANK << j)) ? 1 : 0; + dimm_info[i][j].x16_width = (addr_decode & + (IE31200_MAD_DIMM_A_WIDTH << j)) ? 1 : 0; + edac_dbg(0, "size: 0x%x, rank: %d, width: %d\n", + dimm_info[i][j].size, + dimm_info[i][j].dual_rank, + dimm_info[i][j].x16_width); + } + } + /* * The dram rank boundary (DRB) reg values are boundary addresses * for each DRAM rank with a granularity of 64MB. DRB regs are @@ -439,23 +434,23 @@ static int ie31200_probe1(struct pci_dev *pdev, int dev_idx) ie31200_clear_error_info(mci); - rc = -ENODEV; if (edac_mc_add_mc(mci)) { edac_dbg(3, "MC: failed edac_mc_add_mc()\n"); - goto fail_free; + ret = -ENODEV; + goto fail_unmap; } /* get this far and it's successful */ edac_dbg(3, "MC: success\n"); return 0; -fail_free: - if (mci) - edac_mc_free(mci); fail_unmap: iounmap(window); - return rc; +fail_free: + edac_mc_free(mci); + + return ret; } static int ie31200_init_one(struct pci_dev *pdev, -- cgit v1.2.3 From 791b470684c1e99b45adabeda827545eeed042db Mon Sep 17 00:00:00 2001 From: Jason Baron Date: Wed, 9 Jul 2014 21:13:11 +0000 Subject: MAINTAINERS: add ie31200_edac entry Signed-off-by: Jason Baron Link: http://lkml.kernel.org/r/2b2b78c3831bd3c67960fe8247f94db4acb055db.1404939455.git.jbaron@akamai.com Signed-off-by: Borislav Petkov --- MAINTAINERS | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/MAINTAINERS b/MAINTAINERS index 3f2e171047b9..f3e9d00ff0c8 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -3324,6 +3324,13 @@ W: bluesmoke.sourceforge.net S: Maintained F: drivers/edac/i82975x_edac.c +EDAC-IE31200 +M: Jason Baron +L: linux-edac@vger.kernel.org +W: bluesmoke.sourceforge.net +S: Maintained +F: drivers/edac/ie31200_edac.c + EDAC-MPC85XX M: Johannes Thumshirn L: linux-edac@vger.kernel.org -- cgit v1.2.3 From eba4bfb34d45a2219d1d7534905c026eea6fcd49 Mon Sep 17 00:00:00 2001 From: Aravind Gopalakrishnan Date: Mon, 14 Jul 2014 16:58:19 +0200 Subject: EDAC, MCE, AMD: Add MCE decoding for F15h M60h Add decoding logic for new Fam15h model 60h. Tested using mce_amd_inj module and works fine. Signed-off-by: Aravind Gopalakrishnan Link: http://lkml.kernel.org/r/1405098795-4678-1-git-send-email-Aravind.Gopalakrishnan@amd.com [ Boris: simplify a bit. ] Signed-off-by: Borislav Petkov --- drivers/edac/mce_amd.c | 44 ++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 40 insertions(+), 4 deletions(-) diff --git a/drivers/edac/mce_amd.c b/drivers/edac/mce_amd.c index 5f43620d580a..f78c1c54dbd5 100644 --- a/drivers/edac/mce_amd.c +++ b/drivers/edac/mce_amd.c @@ -78,7 +78,8 @@ static const char * const f15h_mc1_mce_desc[] = { "uop queue", "insn buffer", "predecode buffer", - "fetch address FIFO" + "fetch address FIFO", + "dispatch uop queue" }; static const char * const f15h_mc2_mce_desc[] = { @@ -267,6 +268,12 @@ static bool f15h_mc0_mce(u16 ec, u8 xec) pr_cont("System Read Data Error.\n"); else pr_cont(" Internal error condition type %d.\n", xec); + } else if (INT_ERROR(ec)) { + if (xec <= 0x1f) + pr_cont("Hardware Assert.\n"); + else + ret = false; + } else ret = false; @@ -373,7 +380,7 @@ static bool f15h_mc1_mce(u16 ec, u8 xec) pr_cont("%s.\n", f15h_mc1_mce_desc[xec-4]); break; - case 0x11 ... 0x14: + case 0x11 ... 0x15: pr_cont("Decoder %s parity error.\n", f15h_mc1_mce_desc[xec-4]); break; @@ -397,10 +404,20 @@ static void decode_mc1_mce(struct mce *m) bool k8 = (boot_cpu_data.x86 == 0xf && (m->status & BIT_64(58))); pr_cont("during %s.\n", (k8 ? "system linefill" : "NB data read")); + } else if (INT_ERROR(ec)) { + if (xec <= 0x3f) + pr_cont("Hardware Assert.\n"); + else + goto wrong_mc1_mce; } else if (fam_ops->mc1_mce(ec, xec)) ; else - pr_emerg(HW_ERR "Corrupted MC1 MCE info?\n"); + goto wrong_mc1_mce; + + return; + +wrong_mc1_mce: + pr_emerg(HW_ERR "Corrupted MC1 MCE info?\n"); } static bool k8_mc2_mce(u16 ec, u8 xec) @@ -468,6 +485,11 @@ static bool f15h_mc2_mce(u16 ec, u8 xec) default: ret = false; } + } else if (INT_ERROR(ec)) { + if (xec <= 0x3f) + pr_cont("Hardware Assert.\n"); + else + ret = false; } return ret; @@ -615,6 +637,7 @@ static void decode_mc4_mce(struct mce *m) static void decode_mc5_mce(struct mce *m) { struct cpuinfo_x86 *c = &boot_cpu_data; + u16 ec = EC(m->status); u8 xec = XEC(m->status, xec_mask); if (c->x86 == 0xf || c->x86 == 0x11) @@ -622,6 +645,14 @@ static void decode_mc5_mce(struct mce *m) pr_emerg(HW_ERR "MC5 Error: "); + if (INT_ERROR(ec)) { + if (xec <= 0x1f) { + pr_cont("Hardware Assert.\n"); + return; + } else + goto wrong_mc5_mce; + } + if (xec == 0x0 || xec == 0xc) pr_cont("%s.\n", mc5_mce_desc[xec]); else if (xec <= 0xd) @@ -642,6 +673,10 @@ static void decode_mc6_mce(struct mce *m) pr_emerg(HW_ERR "MC6 Error: "); switch (xec) { + case 0x0: + pr_cont("Hardware Assertion"); + break; + case 0x1: pr_cont("Free List"); break; @@ -857,7 +892,8 @@ static int __init mce_amd_init(void) break; case 0x15: - xec_mask = 0x1f; + xec_mask = c->x86_model == 0x60 ? 0x3f : 0x1f; + fam_ops->mc0_mce = f15h_mc0_mce; fam_ops->mc1_mce = f15h_mc1_mce; fam_ops->mc2_mce = f15h_mc2_mce; -- cgit v1.2.3