diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2010-10-21 14:04:58 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2010-10-21 14:04:58 -0700 |
commit | c029e405bd3d4e92d09fbbc97c03952585986ebe (patch) | |
tree | 8c9b23481a22ca15240e6a4e0df9afc3492efe3f | |
parent | a9ccd80aadadef8c424142d41fddea7c0880f0e9 (diff) | |
parent | 525906bc898d712f21e5bfcfc85ab0e517e3d086 (diff) | |
download | blackbird-op-linux-c029e405bd3d4e92d09fbbc97c03952585986ebe.tar.gz blackbird-op-linux-c029e405bd3d4e92d09fbbc97c03952585986ebe.zip |
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp: (21 commits)
EDAC, MCE: Fix shift warning on 32-bit
EDAC, MCE: Add a BIT_64() macro
EDAC, MCE: Enable MCE decoding on F12h
EDAC, MCE: Add F12h NB MCE decoder
EDAC, MCE: Add F12h IC MCE decoder
EDAC, MCE: Add F12h DC MCE decoder
EDAC, MCE: Add support for F11h MCEs
EDAC, MCE: Enable MCE decoding on F14h
EDAC, MCE: Fix FR MCEs decoding
EDAC, MCE: Complete NB MCE decoders
EDAC, MCE: Warn about LS MCEs on F14h
EDAC, MCE: Adjust IC decoders to F14h
EDAC, MCE: Adjust DC decoders to F14h
EDAC, MCE: Rename files
EDAC, MCE: Rework MCE injection
EDAC: Export edac sysfs class to users.
EDAC, MCE: Pass complete MCE info to decoders
EDAC, MCE: Sanitize error codes
EDAC, MCE: Remove unused function parameter
EDAC, MCE: Add HW_ERR prefix
...
-rw-r--r-- | drivers/edac/Kconfig | 14 | ||||
-rw-r--r-- | drivers/edac/Makefile | 3 | ||||
-rw-r--r-- | drivers/edac/amd64_edac.c | 13 | ||||
-rw-r--r-- | drivers/edac/amd64_edac.h | 5 | ||||
-rw-r--r-- | drivers/edac/amd64_edac_dbg.c | 207 | ||||
-rw-r--r-- | drivers/edac/edac_device_sysfs.c | 16 | ||||
-rw-r--r-- | drivers/edac/edac_mc_sysfs.c | 11 | ||||
-rw-r--r-- | drivers/edac/edac_mce_amd.c | 452 | ||||
-rw-r--r-- | drivers/edac/edac_module.c | 79 | ||||
-rw-r--r-- | drivers/edac/edac_module.h | 1 | ||||
-rw-r--r-- | drivers/edac/edac_pci_sysfs.c | 10 | ||||
-rw-r--r-- | drivers/edac/edac_stub.c | 51 | ||||
-rw-r--r-- | drivers/edac/mce_amd.c | 680 | ||||
-rw-r--r-- | drivers/edac/mce_amd.h (renamed from drivers/edac/edac_mce_amd.h) | 59 | ||||
-rw-r--r-- | drivers/edac/mce_amd_inj.c | 171 | ||||
-rw-r--r-- | include/linux/edac.h | 4 |
16 files changed, 1018 insertions, 758 deletions
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig index 734e2e062374..9dbb28b9559f 100644 --- a/drivers/edac/Kconfig +++ b/drivers/edac/Kconfig @@ -39,7 +39,7 @@ config EDAC_DEBUG there're four debug levels (x=0,1,2,3 from low to high). Usually you should select 'N'. - config EDAC_DECODE_MCE +config EDAC_DECODE_MCE tristate "Decode MCEs in human-readable form (only on AMD for now)" depends on CPU_SUP_AMD && X86_MCE default y @@ -51,6 +51,16 @@ config EDAC_DEBUG which occur really early upon boot, before the module infrastructure has been initialized. +config EDAC_MCE_INJ + tristate "Simple MCE injection interface over /sysfs" + depends on EDAC_DECODE_MCE + default n + help + This is a simple interface to inject MCEs over /sysfs and test + the MCE decoding code in EDAC. + + This is currently AMD-only. + config EDAC_MM_EDAC tristate "Main Memory EDAC (Error Detection And Correction) reporting" help @@ -72,7 +82,7 @@ config EDAC_AMD64 Families of Memory Controllers (K8, F10h and F11h) config EDAC_AMD64_ERROR_INJECTION - bool "Sysfs Error Injection facilities" + bool "Sysfs HW Error injection facilities" depends on EDAC_AMD64 help Recent Opterons (Family 10h and later) provide for Memory Error diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile index ca6b1bb24ccc..32c7bc93c525 100644 --- a/drivers/edac/Makefile +++ b/drivers/edac/Makefile @@ -17,6 +17,9 @@ ifdef CONFIG_PCI edac_core-objs += edac_pci.o edac_pci_sysfs.o endif +obj-$(CONFIG_EDAC_MCE_INJ) += mce_amd_inj.o + +edac_mce_amd-objs := mce_amd.o obj-$(CONFIG_EDAC_DECODE_MCE) += edac_mce_amd.o obj-$(CONFIG_EDAC_AMD76X) += amd76x_edac.o diff --git a/drivers/edac/amd64_edac.c b/drivers/edac/amd64_edac.c index 09fcc5282327..8521401bbd75 100644 --- a/drivers/edac/amd64_edac.c +++ b/drivers/edac/amd64_edac.c @@ -2073,11 +2073,18 @@ static inline void __amd64_decode_bus_error(struct mem_ctl_info *mci, amd64_handle_ue(mci, info); } -void amd64_decode_bus_error(int node_id, struct err_regs *regs) +void amd64_decode_bus_error(int node_id, struct mce *m, u32 nbcfg) { struct mem_ctl_info *mci = mci_lookup[node_id]; + struct err_regs regs; - __amd64_decode_bus_error(mci, regs); + regs.nbsl = (u32) m->status; + regs.nbsh = (u32)(m->status >> 32); + regs.nbeal = (u32) m->addr; + regs.nbeah = (u32)(m->addr >> 32); + regs.nbcfg = nbcfg; + + __amd64_decode_bus_error(mci, ®s); /* * Check the UE bit of the NB status high register, if set generate some @@ -2086,7 +2093,7 @@ void amd64_decode_bus_error(int node_id, struct err_regs *regs) * * FIXME: this should go somewhere else, if at all. */ - if (regs->nbsh & K8_NBSH_UC_ERR && !report_gart_errors) + if (regs.nbsh & K8_NBSH_UC_ERR && !report_gart_errors) edac_mc_handle_ue_no_info(mci, "UE bit is set"); } diff --git a/drivers/edac/amd64_edac.h b/drivers/edac/amd64_edac.h index 613b9381e71a..044aee4f944d 100644 --- a/drivers/edac/amd64_edac.h +++ b/drivers/edac/amd64_edac.h @@ -72,7 +72,7 @@ #include <linux/edac.h> #include <asm/msr.h> #include "edac_core.h" -#include "edac_mce_amd.h" +#include "mce_amd.h" #define amd64_printk(level, fmt, arg...) \ edac_printk(level, "amd64", fmt, ##arg) @@ -482,11 +482,10 @@ extern const char *rrrr_msgs[16]; extern const char *to_msgs[2]; extern const char *pp_msgs[4]; extern const char *ii_msgs[4]; -extern const char *ext_msgs[32]; extern const char *htlink_msgs[8]; #ifdef CONFIG_EDAC_DEBUG -#define NUM_DBG_ATTRS 9 +#define NUM_DBG_ATTRS 5 #else #define NUM_DBG_ATTRS 0 #endif diff --git a/drivers/edac/amd64_edac_dbg.c b/drivers/edac/amd64_edac_dbg.c index 59cf2cf6e11e..e3562288f4ce 100644 --- a/drivers/edac/amd64_edac_dbg.c +++ b/drivers/edac/amd64_edac_dbg.c @@ -1,167 +1,16 @@ #include "amd64_edac.h" -/* - * accept a hex value and store it into the virtual error register file, field: - * nbeal and nbeah. Assume virtual error values have already been set for: NBSL, - * NBSH and NBCFG. Then proceed to map the error values to a MC, CSROW and - * CHANNEL - */ -static ssize_t amd64_nbea_store(struct mem_ctl_info *mci, const char *data, - size_t count) -{ - struct amd64_pvt *pvt = mci->pvt_info; - unsigned long long value; - int ret = 0; - - ret = strict_strtoull(data, 16, &value); - if (ret != -EINVAL) { - debugf0("received NBEA= 0x%llx\n", value); - - /* place the value into the virtual error packet */ - pvt->ctl_error_info.nbeal = (u32) value; - value >>= 32; - pvt->ctl_error_info.nbeah = (u32) value; - - /* Process the Mapping request */ - /* TODO: Add race prevention */ - amd_decode_nb_mce(pvt->mc_node_id, &pvt->ctl_error_info, 1); - - return count; - } - return ret; -} - -/* display back what the last NBEA (MCA NB Address (MC4_ADDR)) was written */ -static ssize_t amd64_nbea_show(struct mem_ctl_info *mci, char *data) -{ - struct amd64_pvt *pvt = mci->pvt_info; - u64 value; - - value = pvt->ctl_error_info.nbeah; - value <<= 32; - value |= pvt->ctl_error_info.nbeal; - - return sprintf(data, "%llx\n", value); -} - -/* store the NBSL (MCA NB Status Low (MC4_STATUS)) value user desires */ -static ssize_t amd64_nbsl_store(struct mem_ctl_info *mci, const char *data, - size_t count) -{ - struct amd64_pvt *pvt = mci->pvt_info; - unsigned long value; - int ret = 0; - - ret = strict_strtoul(data, 16, &value); - if (ret != -EINVAL) { - debugf0("received NBSL= 0x%lx\n", value); - - pvt->ctl_error_info.nbsl = (u32) value; - - return count; - } - return ret; -} - -/* display back what the last NBSL value written */ -static ssize_t amd64_nbsl_show(struct mem_ctl_info *mci, char *data) -{ - struct amd64_pvt *pvt = mci->pvt_info; - u32 value; - - value = pvt->ctl_error_info.nbsl; - - return sprintf(data, "%x\n", value); -} - -/* store the NBSH (MCA NB Status High) value user desires */ -static ssize_t amd64_nbsh_store(struct mem_ctl_info *mci, const char *data, - size_t count) -{ - struct amd64_pvt *pvt = mci->pvt_info; - unsigned long value; - int ret = 0; - - ret = strict_strtoul(data, 16, &value); - if (ret != -EINVAL) { - debugf0("received NBSH= 0x%lx\n", value); - - pvt->ctl_error_info.nbsh = (u32) value; - - return count; - } - return ret; -} - -/* display back what the last NBSH value written */ -static ssize_t amd64_nbsh_show(struct mem_ctl_info *mci, char *data) -{ - struct amd64_pvt *pvt = mci->pvt_info; - u32 value; - - value = pvt->ctl_error_info.nbsh; - - return sprintf(data, "%x\n", value); +#define EDAC_DCT_ATTR_SHOW(reg) \ +static ssize_t amd64_##reg##_show(struct mem_ctl_info *mci, char *data) \ +{ \ + struct amd64_pvt *pvt = mci->pvt_info; \ + return sprintf(data, "0x%016llx\n", (u64)pvt->reg); \ } -/* accept and store the NBCFG (MCA NB Configuration) value user desires */ -static ssize_t amd64_nbcfg_store(struct mem_ctl_info *mci, - const char *data, size_t count) -{ - struct amd64_pvt *pvt = mci->pvt_info; - unsigned long value; - int ret = 0; - - ret = strict_strtoul(data, 16, &value); - if (ret != -EINVAL) { - debugf0("received NBCFG= 0x%lx\n", value); - - pvt->ctl_error_info.nbcfg = (u32) value; - - return count; - } - return ret; -} - -/* various show routines for the controls of a MCI */ -static ssize_t amd64_nbcfg_show(struct mem_ctl_info *mci, char *data) -{ - struct amd64_pvt *pvt = mci->pvt_info; - - return sprintf(data, "%x\n", pvt->ctl_error_info.nbcfg); -} - - -static ssize_t amd64_dhar_show(struct mem_ctl_info *mci, char *data) -{ - struct amd64_pvt *pvt = mci->pvt_info; - - return sprintf(data, "%x\n", pvt->dhar); -} - - -static ssize_t amd64_dbam_show(struct mem_ctl_info *mci, char *data) -{ - struct amd64_pvt *pvt = mci->pvt_info; - - return sprintf(data, "%x\n", pvt->dbam0); -} - - -static ssize_t amd64_topmem_show(struct mem_ctl_info *mci, char *data) -{ - struct amd64_pvt *pvt = mci->pvt_info; - - return sprintf(data, "%llx\n", pvt->top_mem); -} - - -static ssize_t amd64_topmem2_show(struct mem_ctl_info *mci, char *data) -{ - struct amd64_pvt *pvt = mci->pvt_info; - - return sprintf(data, "%llx\n", pvt->top_mem2); -} +EDAC_DCT_ATTR_SHOW(dhar); +EDAC_DCT_ATTR_SHOW(dbam0); +EDAC_DCT_ATTR_SHOW(top_mem); +EDAC_DCT_ATTR_SHOW(top_mem2); static ssize_t amd64_hole_show(struct mem_ctl_info *mci, char *data) { @@ -182,38 +31,6 @@ struct mcidev_sysfs_attribute amd64_dbg_attrs[] = { { .attr = { - .name = "nbea_ctl", - .mode = (S_IRUGO | S_IWUSR) - }, - .show = amd64_nbea_show, - .store = amd64_nbea_store, - }, - { - .attr = { - .name = "nbsl_ctl", - .mode = (S_IRUGO | S_IWUSR) - }, - .show = amd64_nbsl_show, - .store = amd64_nbsl_store, - }, - { - .attr = { - .name = "nbsh_ctl", - .mode = (S_IRUGO | S_IWUSR) - }, - .show = amd64_nbsh_show, - .store = amd64_nbsh_store, - }, - { - .attr = { - .name = "nbcfg_ctl", - .mode = (S_IRUGO | S_IWUSR) - }, - .show = amd64_nbcfg_show, - .store = amd64_nbcfg_store, - }, - { - .attr = { .name = "dhar", .mode = (S_IRUGO) }, @@ -225,7 +42,7 @@ struct mcidev_sysfs_attribute amd64_dbg_attrs[] = { .name = "dbam", .mode = (S_IRUGO) }, - .show = amd64_dbam_show, + .show = amd64_dbam0_show, .store = NULL, }, { @@ -233,7 +50,7 @@ struct mcidev_sysfs_attribute amd64_dbg_attrs[] = { .name = "topmem", .mode = (S_IRUGO) }, - .show = amd64_topmem_show, + .show = amd64_top_mem_show, .store = NULL, }, { @@ -241,7 +58,7 @@ struct mcidev_sysfs_attribute amd64_dbg_attrs[] = { .name = "topmem2", .mode = (S_IRUGO) }, - .show = amd64_topmem2_show, + .show = amd64_top_mem2_show, .store = NULL, }, { diff --git a/drivers/edac/edac_device_sysfs.c b/drivers/edac/edac_device_sysfs.c index 070968178a24..2941dca91aae 100644 --- a/drivers/edac/edac_device_sysfs.c +++ b/drivers/edac/edac_device_sysfs.c @@ -13,6 +13,7 @@ #include <linux/ctype.h> #include <linux/module.h> #include <linux/slab.h> +#include <linux/edac.h> #include "edac_core.h" #include "edac_module.h" @@ -235,7 +236,7 @@ int edac_device_register_sysfs_main_kobj(struct edac_device_ctl_info *edac_dev) debugf1("%s()\n", __func__); /* get the /sys/devices/system/edac reference */ - edac_class = edac_get_edac_class(); + edac_class = edac_get_sysfs_class(); if (edac_class == NULL) { debugf1("%s() no edac_class error\n", __func__); err = -ENODEV; @@ -255,7 +256,7 @@ int edac_device_register_sysfs_main_kobj(struct edac_device_ctl_info *edac_dev) if (!try_module_get(edac_dev->owner)) { err = -ENODEV; - goto err_out; + goto err_mod_get; } /* register */ @@ -282,6 +283,9 @@ int edac_device_register_sysfs_main_kobj(struct edac_device_ctl_info *edac_dev) err_kobj_reg: module_put(edac_dev->owner); +err_mod_get: + edac_put_sysfs_class(); + err_out: return err; } @@ -290,12 +294,11 @@ err_out: * edac_device_unregister_sysfs_main_kobj: * the '..../edac/<name>' kobject */ -void edac_device_unregister_sysfs_main_kobj( - struct edac_device_ctl_info *edac_dev) +void edac_device_unregister_sysfs_main_kobj(struct edac_device_ctl_info *dev) { debugf0("%s()\n", __func__); debugf4("%s() name of kobject is: %s\n", - __func__, kobject_name(&edac_dev->kobj)); + __func__, kobject_name(&dev->kobj)); /* * Unregister the edac device's kobject and @@ -304,7 +307,8 @@ void edac_device_unregister_sysfs_main_kobj( * a) module_put() this module * b) 'kfree' the memory */ - kobject_put(&edac_dev->kobj); + kobject_put(&dev->kobj); + edac_put_sysfs_class(); } /* edac_dev -> instance information */ diff --git a/drivers/edac/edac_mc_sysfs.c b/drivers/edac/edac_mc_sysfs.c index 8aad94d10c0c..a4135860149b 100644 --- a/drivers/edac/edac_mc_sysfs.c +++ b/drivers/edac/edac_mc_sysfs.c @@ -11,6 +11,7 @@ #include <linux/ctype.h> #include <linux/slab.h> +#include <linux/edac.h> #include <linux/bug.h> #include "edac_core.h" @@ -1011,13 +1012,13 @@ void edac_remove_sysfs_mci_device(struct mem_ctl_info *mci) */ int edac_sysfs_setup_mc_kset(void) { - int err = 0; + int err = -EINVAL; struct sysdev_class *edac_class; debugf1("%s()\n", __func__); /* get the /sys/devices/system/edac class reference */ - edac_class = edac_get_edac_class(); + edac_class = edac_get_sysfs_class(); if (edac_class == NULL) { debugf1("%s() no edac_class error=%d\n", __func__, err); goto fail_out; @@ -1028,15 +1029,16 @@ int edac_sysfs_setup_mc_kset(void) if (!mc_kset) { err = -ENOMEM; debugf1("%s() Failed to register '.../edac/mc'\n", __func__); - goto fail_out; + goto fail_kset; } debugf1("%s() Registered '.../edac/mc' kobject\n", __func__); return 0; +fail_kset: + edac_put_sysfs_class(); - /* error unwind stack */ fail_out: return err; } @@ -1049,5 +1051,6 @@ fail_out: void edac_sysfs_teardown_mc_kset(void) { kset_unregister(mc_kset); + edac_put_sysfs_class(); } diff --git a/drivers/edac/edac_mce_amd.c b/drivers/edac/edac_mce_amd.c deleted file mode 100644 index 9014df6f605d..000000000000 --- a/drivers/edac/edac_mce_amd.c +++ /dev/null @@ -1,452 +0,0 @@ -#include <linux/module.h> -#include "edac_mce_amd.h" - -static bool report_gart_errors; -static void (*nb_bus_decoder)(int node_id, struct err_regs *regs); - -void amd_report_gart_errors(bool v) -{ - report_gart_errors = v; -} -EXPORT_SYMBOL_GPL(amd_report_gart_errors); - -void amd_register_ecc_decoder(void (*f)(int, struct err_regs *)) -{ - nb_bus_decoder = f; -} -EXPORT_SYMBOL_GPL(amd_register_ecc_decoder); - -void amd_unregister_ecc_decoder(void (*f)(int, struct err_regs *)) -{ - if (nb_bus_decoder) { - WARN_ON(nb_bus_decoder != f); - - nb_bus_decoder = NULL; - } -} -EXPORT_SYMBOL_GPL(amd_unregister_ecc_decoder); - -/* - * string representation for the different MCA reported error types, see F3x48 - * or MSR0000_0411. - */ -const char *tt_msgs[] = { /* transaction type */ - "instruction", - "data", - "generic", - "reserved" -}; -EXPORT_SYMBOL_GPL(tt_msgs); - -const char *ll_msgs[] = { /* cache level */ - "L0", - "L1", - "L2", - "L3/generic" -}; -EXPORT_SYMBOL_GPL(ll_msgs); - -const char *rrrr_msgs[] = { - "generic", - "generic read", - "generic write", - "data read", - "data write", - "inst fetch", - "prefetch", - "evict", - "snoop", - "reserved RRRR= 9", - "reserved RRRR= 10", - "reserved RRRR= 11", - "reserved RRRR= 12", - "reserved RRRR= 13", - "reserved RRRR= 14", - "reserved RRRR= 15" -}; -EXPORT_SYMBOL_GPL(rrrr_msgs); - -const char *pp_msgs[] = { /* participating processor */ - "local node originated (SRC)", - "local node responded to request (RES)", - "local node observed as 3rd party (OBS)", - "generic" -}; -EXPORT_SYMBOL_GPL(pp_msgs); - -const char *to_msgs[] = { - "no timeout", - "timed out" -}; -EXPORT_SYMBOL_GPL(to_msgs); - -const char *ii_msgs[] = { /* memory or i/o */ - "mem access", - "reserved", - "i/o access", - "generic" -}; -EXPORT_SYMBOL_GPL(ii_msgs); - -/* - * Map the 4 or 5 (family-specific) bits of Extended Error code to the - * string table. - */ -const char *ext_msgs[] = { - "K8 ECC error", /* 0_0000b */ - "CRC error on link", /* 0_0001b */ - "Sync error packets on link", /* 0_0010b */ - "Master Abort during link operation", /* 0_0011b */ - "Target Abort during link operation", /* 0_0100b */ - "Invalid GART PTE entry during table walk", /* 0_0101b */ - "Unsupported atomic RMW command received", /* 0_0110b */ - "WDT error: NB transaction timeout", /* 0_0111b */ - "ECC/ChipKill ECC error", /* 0_1000b */ - "SVM DEV Error", /* 0_1001b */ - "Link Data error", /* 0_1010b */ - "Link/L3/Probe Filter Protocol error", /* 0_1011b */ - "NB Internal Arrays Parity error", /* 0_1100b */ - "DRAM Address/Control Parity error", /* 0_1101b */ - "Link Transmission error", /* 0_1110b */ - "GART/DEV Table Walk Data error" /* 0_1111b */ - "Res 0x100 error", /* 1_0000b */ - "Res 0x101 error", /* 1_0001b */ - "Res 0x102 error", /* 1_0010b */ - "Res 0x103 error", /* 1_0011b */ - "Res 0x104 error", /* 1_0100b */ - "Res 0x105 error", /* 1_0101b */ - "Res 0x106 error", /* 1_0110b */ - "Res 0x107 error", /* 1_0111b */ - "Res 0x108 error", /* 1_1000b */ - "Res 0x109 error", /* 1_1001b */ - "Res 0x10A error", /* 1_1010b */ - "Res 0x10B error", /* 1_1011b */ - "ECC error in L3 Cache Data", /* 1_1100b */ - "L3 Cache Tag error", /* 1_1101b */ - "L3 Cache LRU Parity error", /* 1_1110b */ - "Probe Filter error" /* 1_1111b */ -}; -EXPORT_SYMBOL_GPL(ext_msgs); - -static void amd_decode_dc_mce(u64 mc0_status) -{ - u32 ec = mc0_status & 0xffff; - u32 xec = (mc0_status >> 16) & 0xf; - - pr_emerg("Data Cache Error"); - - if (xec == 1 && TLB_ERROR(ec)) - pr_cont(": %s TLB multimatch.\n", LL_MSG(ec)); - else if (xec == 0) { - if (mc0_status & (1ULL << 40)) - pr_cont(" during Data Scrub.\n"); - else if (TLB_ERROR(ec)) - pr_cont(": %s TLB parity error.\n", LL_MSG(ec)); - else if (MEM_ERROR(ec)) { - u8 ll = ec & 0x3; - u8 tt = (ec >> 2) & 0x3; - u8 rrrr = (ec >> 4) & 0xf; - - /* see F10h BKDG (31116), Table 92. */ - if (ll == 0x1) { - if (tt != 0x1) - goto wrong_dc_mce; - - pr_cont(": Data/Tag %s error.\n", RRRR_MSG(ec)); - - } else if (ll == 0x2 && rrrr == 0x3) - pr_cont(" during L1 linefill from L2.\n"); - else - goto wrong_dc_mce; - } else if (BUS_ERROR(ec) && boot_cpu_data.x86 == 0xf) - pr_cont(" during system linefill.\n"); - else - goto wrong_dc_mce; - } else - goto wrong_dc_mce; - - return; - -wrong_dc_mce: - pr_warning("Corrupted DC MCE info?\n"); -} - -static void amd_decode_ic_mce(u64 mc1_status) -{ - u32 ec = mc1_status & 0xffff; - u32 xec = (mc1_status >> 16) & 0xf; - - pr_emerg("Instruction Cache Error"); - - if (xec == 1 && TLB_ERROR(ec)) - pr_cont(": %s TLB multimatch.\n", LL_MSG(ec)); - else if (xec == 0) { - if (TLB_ERROR(ec)) - pr_cont(": %s TLB Parity error.\n", LL_MSG(ec)); - else if (BUS_ERROR(ec)) { - if (boot_cpu_data.x86 == 0xf && - (mc1_status & (1ULL << 58))) - pr_cont(" during system linefill.\n"); - else - pr_cont(" during attempted NB data read.\n"); - } else if (MEM_ERROR(ec)) { - u8 ll = ec & 0x3; - u8 rrrr = (ec >> 4) & 0xf; - - if (ll == 0x2) - pr_cont(" during a linefill from L2.\n"); - else if (ll == 0x1) { - - switch (rrrr) { - case 0x5: - pr_cont(": Parity error during " - "data load.\n"); - break; - - case 0x7: - pr_cont(": Copyback Parity/Victim" - " error.\n"); - break; - - case 0x8: - pr_cont(": Tag Snoop error.\n"); - break; - - default: - goto wrong_ic_mce; - break; - } - } - } else - goto wrong_ic_mce; - } else - goto wrong_ic_mce; - - return; - -wrong_ic_mce: - pr_warning("Corrupted IC MCE info?\n"); -} - -static void amd_decode_bu_mce(u64 mc2_status) -{ - u32 ec = mc2_status & 0xffff; - u32 xec = (mc2_status >> 16) & 0xf; - - pr_emerg("Bus Unit Error"); - - if (xec == 0x1) - pr_cont(" in the write data buffers.\n"); - else if (xec == 0x3) - pr_cont(" in the victim data buffers.\n"); - else if (xec == 0x2 && MEM_ERROR(ec)) - pr_cont(": %s error in the L2 cache tags.\n", RRRR_MSG(ec)); - else if (xec == 0x0) { - if (TLB_ERROR(ec)) - pr_cont(": %s error in a Page Descriptor Cache or " - "Guest TLB.\n", TT_MSG(ec)); - else if (BUS_ERROR(ec)) - pr_cont(": %s/ECC error in data read from NB: %s.\n", - RRRR_MSG(ec), PP_MSG(ec)); - else if (MEM_ERROR(ec)) { - u8 rrrr = (ec >> 4) & 0xf; - - if (rrrr >= 0x7) - pr_cont(": %s error during data copyback.\n", - RRRR_MSG(ec)); - else if (rrrr <= 0x1) - pr_cont(": %s parity/ECC error during data " - "access from L2.\n", RRRR_MSG(ec)); - else - goto wrong_bu_mce; - } else - goto wrong_bu_mce; - } else - goto wrong_bu_mce; - - return; - -wrong_bu_mce: - pr_warning("Corrupted BU MCE info?\n"); -} - -static void amd_decode_ls_mce(u64 mc3_status) -{ - u32 ec = mc3_status & 0xffff; - u32 xec = (mc3_status >> 16) & 0xf; - - pr_emerg("Load Store Error"); - - if (xec == 0x0) { - u8 rrrr = (ec >> 4) & 0xf; - - if (!BUS_ERROR(ec) || (rrrr != 0x3 && rrrr != 0x4)) - goto wrong_ls_mce; - - pr_cont(" during %s.\n", RRRR_MSG(ec)); - } - return; - -wrong_ls_mce: - pr_warning("Corrupted LS MCE info?\n"); -} - -void amd_decode_nb_mce(int node_id, struct err_regs *regs, int handle_errors) -{ - u32 ec = ERROR_CODE(regs->nbsl); - - if (!handle_errors) - return; - - /* - * GART TLB error reporting is disabled by default. Bail out early. - */ - if (TLB_ERROR(ec) && !report_gart_errors) - return; - - pr_emerg("Northbridge Error, node %d", node_id); - - /* - * F10h, revD can disable ErrCpu[3:0] so check that first and also the - * value encoding has changed so interpret those differently - */ - if ((boot_cpu_data.x86 == 0x10) && - (boot_cpu_data.x86_model > 7)) { - if (regs->nbsh & K8_NBSH_ERR_CPU_VAL) - pr_cont(", core: %u\n", (u8)(regs->nbsh & 0xf)); - } else { - u8 assoc_cpus = regs->nbsh & 0xf; - - if (assoc_cpus > 0) - pr_cont(", core: %d", fls(assoc_cpus) - 1); - - pr_cont("\n"); - } - - pr_emerg("%s.\n", EXT_ERR_MSG(regs->nbsl)); - - if (BUS_ERROR(ec) && nb_bus_decoder) - nb_bus_decoder(node_id, regs); -} -EXPORT_SYMBOL_GPL(amd_decode_nb_mce); - -static void amd_decode_fr_mce(u64 mc5_status) -{ - /* we have only one error signature so match all fields at once. */ - if ((mc5_status & 0xffff) == 0x0f0f) - pr_emerg(" FR Error: CPU Watchdog timer expire.\n"); - else - pr_warning("Corrupted FR MCE info?\n"); -} - -static inline void amd_decode_err_code(unsigned int ec) -{ - if (TLB_ERROR(ec)) { - pr_emerg("Transaction: %s, Cache Level %s\n", - TT_MSG(ec), LL_MSG(ec)); - } else if (MEM_ERROR(ec)) { - pr_emerg("Transaction: %s, Type: %s, Cache Level: %s", - RRRR_MSG(ec), TT_MSG(ec), LL_MSG(ec)); - } else if (BUS_ERROR(ec)) { - pr_emerg("Transaction type: %s(%s), %s, Cache Level: %s, " - "Participating Processor: %s\n", - RRRR_MSG(ec), II_MSG(ec), TO_MSG(ec), LL_MSG(ec), - PP_MSG(ec)); - } else - pr_warning("Huh? Unknown MCE error 0x%x\n", ec); -} - -static int amd_decode_mce(struct notifier_block *nb, unsigned long val, - void *data) -{ - struct mce *m = (struct mce *)data; - struct err_regs regs; - int node, ecc; - - pr_emerg("MC%d_STATUS: ", m->bank); - - pr_cont("%sorrected error, other errors lost: %s, " - "CPU context corrupt: %s", - ((m->status & MCI_STATUS_UC) ? "Unc" : "C"), - ((m->status & MCI_STATUS_OVER) ? "yes" : "no"), - ((m->status & MCI_STATUS_PCC) ? "yes" : "no")); - - /* do the two bits[14:13] together */ - ecc = (m->status >> 45) & 0x3; - if (ecc) - pr_cont(", %sECC Error", ((ecc == 2) ? "C" : "U")); - - pr_cont("\n"); - - switch (m->bank) { - case 0: - amd_decode_dc_mce(m->status); - break; - - case 1: - amd_decode_ic_mce(m->status); - break; - - case 2: - amd_decode_bu_mce(m->status); - break; - - case 3: - amd_decode_ls_mce(m->status); - break; - - case 4: - regs.nbsl = (u32) m->status; - regs.nbsh = (u32)(m->status >> 32); - regs.nbeal = (u32) m->addr; - regs.nbeah = (u32)(m->addr >> 32); - node = amd_get_nb_id(m->extcpu); - - amd_decode_nb_mce(node, ®s, 1); - break; - - case 5: - amd_decode_fr_mce(m->status); - break; - - default: - break; - } - - amd_decode_err_code(m->status & 0xffff); - - return NOTIFY_STOP; -} - -static struct notifier_block amd_mce_dec_nb = { - .notifier_call = amd_decode_mce, -}; - -static int __init mce_amd_init(void) -{ - /* - * We can decode MCEs for K8, F10h and F11h CPUs: - */ - if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD) - return 0; - - if (boot_cpu_data.x86 < 0xf || boot_cpu_data.x86 > 0x11) - return 0; - - atomic_notifier_chain_register(&x86_mce_decoder_chain, &amd_mce_dec_nb); - - return 0; -} -early_initcall(mce_amd_init); - -#ifdef MODULE -static void __exit mce_amd_exit(void) -{ - atomic_notifier_chain_unregister(&x86_mce_decoder_chain, &amd_mce_dec_nb); -} - -MODULE_DESCRIPTION("AMD MCE decoder"); -MODULE_ALIAS("edac-mce-amd"); -MODULE_LICENSE("GPL"); -module_exit(mce_amd_exit); -#endif diff --git a/drivers/edac/edac_module.c b/drivers/edac/edac_module.c index 7e1374afd967..be4b075c3098 100644 --- a/drivers/edac/edac_module.c +++ b/drivers/edac/edac_module.c @@ -27,15 +27,6 @@ EXPORT_SYMBOL_GPL(edac_debug_level); struct workqueue_struct *edac_workqueue; /* - * sysfs object: /sys/devices/system/edac - * need to export to other files in this modules - */ -static struct sysdev_class edac_class = { - .name = "edac", -}; -static int edac_class_valid; - -/* * edac_op_state_to_string() */ char *edac_op_state_to_string(int opstate) @@ -55,60 +46,6 @@ char *edac_op_state_to_string(int opstate) } /* - * edac_get_edac_class() - * - * return pointer to the edac class of 'edac' - */ -struct sysdev_class *edac_get_edac_class(void) -{ - struct sysdev_class *classptr = NULL; - - if (edac_class_valid) - classptr = &edac_class; - - return classptr; -} - -/* - * edac_register_sysfs_edac_name() - * - * register the 'edac' into /sys/devices/system - * - * return: - * 0 success - * !0 error - */ -static int edac_register_sysfs_edac_name(void) -{ - int err; - - /* create the /sys/devices/system/edac directory */ - err = sysdev_class_register(&edac_class); - - if (err) { - debugf1("%s() error=%d\n", __func__, err); - return err; - } - - edac_class_valid = 1; - return 0; -} - -/* - * sysdev_class_unregister() - * - * unregister the 'edac' from /sys/devices/system - */ -static void edac_unregister_sysfs_edac_name(void) -{ - /* only if currently registered, then unregister it */ - if (edac_class_valid) - sysdev_class_unregister(&edac_class); - - edac_class_valid = 0; -} - -/* * edac_workqueue_setup * initialize the edac work queue for polling operations */ @@ -154,21 +91,11 @@ static int __init edac_init(void) edac_pci_clear_parity_errors(); /* - * perform the registration of the /sys/devices/system/edac class object - */ - if (edac_register_sysfs_edac_name()) { - edac_printk(KERN_ERR, EDAC_MC, - "Error initializing 'edac' kobject\n"); - err = -ENODEV; - goto error; - } - - /* * now set up the mc_kset under the edac class object */ err = edac_sysfs_setup_mc_kset(); if (err) - goto sysfs_setup_fail; + goto error; /* Setup/Initialize the workq for this core */ err = edac_workqueue_setup(); @@ -183,9 +110,6 @@ static int __init edac_init(void) workq_fail: edac_sysfs_teardown_mc_kset(); -sysfs_setup_fail: - edac_unregister_sysfs_edac_name(); - error: return err; } @@ -201,7 +125,6 @@ static void __exit edac_exit(void) /* tear down the various subsystems */ edac_workqueue_teardown(); edac_sysfs_teardown_mc_kset(); - edac_unregister_sysfs_edac_name(); } /* diff --git a/drivers/edac/edac_module.h b/drivers/edac/edac_module.h index 233d4798c3aa..17aabb7b90ec 100644 --- a/drivers/edac/edac_module.h +++ b/drivers/edac/edac_module.h @@ -42,7 +42,6 @@ extern void edac_device_unregister_sysfs_main_kobj( struct edac_device_ctl_info *edac_dev); extern int edac_device_create_sysfs(struct edac_device_ctl_info *edac_dev); extern void edac_device_remove_sysfs(struct edac_device_ctl_info *edac_dev); -extern struct sysdev_class *edac_get_edac_class(void); /* edac core workqueue: single CPU mode */ extern struct workqueue_struct *edac_workqueue; diff --git a/drivers/edac/edac_pci_sysfs.c b/drivers/edac/edac_pci_sysfs.c index c39697df9cb4..023b01cb5175 100644 --- a/drivers/edac/edac_pci_sysfs.c +++ b/drivers/edac/edac_pci_sysfs.c @@ -7,7 +7,7 @@ * */ #include <linux/module.h> -#include <linux/sysdev.h> +#include <linux/edac.h> #include <linux/slab.h> #include <linux/ctype.h> @@ -354,7 +354,7 @@ static int edac_pci_main_kobj_setup(void) /* First time, so create the main kobject and its * controls and atributes */ - edac_class = edac_get_edac_class(); + edac_class = edac_get_sysfs_class(); if (edac_class == NULL) { debugf1("%s() no edac_class\n", __func__); err = -ENODEV; @@ -368,7 +368,7 @@ static int edac_pci_main_kobj_setup(void) if (!try_module_get(THIS_MODULE)) { debugf1("%s() try_module_get() failed\n", __func__); err = -ENODEV; - goto decrement_count_fail; + goto mod_get_fail; } edac_pci_top_main_kobj = kzalloc(sizeof(struct kobject), GFP_KERNEL); @@ -403,6 +403,9 @@ kobject_init_and_add_fail: kzalloc_fail: module_put(THIS_MODULE); +mod_get_fail: + edac_put_sysfs_class(); + decrement_count_fail: /* if are on this error exit, nothing to tear down */ atomic_dec(&edac_pci_sysfs_refcount); @@ -429,6 +432,7 @@ static void edac_pci_main_kobj_teardown(void) __func__); kobject_put(edac_pci_top_main_kobj); } + edac_put_sysfs_class(); } /* diff --git a/drivers/edac/edac_stub.c b/drivers/edac/edac_stub.c index 20b428aa155e..aab970760b75 100644 --- a/drivers/edac/edac_stub.c +++ b/drivers/edac/edac_stub.c @@ -3,10 +3,13 @@ * * Author: Dave Jiang <djiang@mvista.com> * - * 2007 (c) MontaVista Software, Inc. This file is licensed under - * the terms of the GNU General Public License version 2. This program - * is licensed "as is" without any warranty of any kind, whether express - * or implied. + * 2007 (c) MontaVista Software, Inc. + * 2010 (c) Advanced Micro Devices Inc. + * Borislav Petkov <borislav.petkov@amd.com> + * + * This file is licensed under the terms of the GNU General Public + * License version 2. This program is licensed "as is" without any + * warranty of any kind, whether express or implied. * */ #include <linux/module.h> @@ -23,6 +26,8 @@ EXPORT_SYMBOL_GPL(edac_handlers); int edac_err_assert = 0; EXPORT_SYMBOL_GPL(edac_err_assert); +static atomic_t edac_class_valid = ATOMIC_INIT(0); + /* * called to determine if there is an EDAC driver interested in * knowing an event (such as NMI) occurred @@ -44,3 +49,41 @@ void edac_atomic_assert_error(void) edac_err_assert++; } EXPORT_SYMBOL_GPL(edac_atomic_assert_error); + +/* + * sysfs object: /sys/devices/system/edac + * need to export to other files + */ +struct sysdev_class edac_class = { + .name = "edac", +}; +EXPORT_SYMBOL_GPL(edac_class); + +/* return pointer to the 'edac' node in sysfs */ +struct sysdev_class *edac_get_sysfs_class(void) +{ + int err = 0; + + if (atomic_read(&edac_class_valid)) + goto out; + + /* create the /sys/devices/system/edac directory */ + err = sysdev_class_register(&edac_class); + if (err) { + printk(KERN_ERR "Error registering toplevel EDAC sysfs dir\n"); + return NULL; + } + +out: + atomic_inc(&edac_class_valid); + return &edac_class; +} +EXPORT_SYMBOL_GPL(edac_get_sysfs_class); + +void edac_put_sysfs_class(void) +{ + /* last user unregisters it */ + if (atomic_dec_and_test(&edac_class_valid)) + sysdev_class_unregister(&edac_class); +} +EXPORT_SYMBOL_GPL(edac_put_sysfs_class); diff --git a/drivers/edac/mce_amd.c b/drivers/edac/mce_amd.c new file mode 100644 index 000000000000..c0181093b490 --- /dev/null +++ b/drivers/edac/mce_amd.c @@ -0,0 +1,680 @@ +#include <linux/module.h> +#include <linux/slab.h> + +#include "mce_amd.h" + +static struct amd_decoder_ops *fam_ops; + +static u8 nb_err_cpumask = 0xf; + +static bool report_gart_errors; +static void (*nb_bus_decoder)(int node_id, struct mce *m, u32 nbcfg); + +void amd_report_gart_errors(bool v) +{ + report_gart_errors = v; +} +EXPORT_SYMBOL_GPL(amd_report_gart_errors); + +void amd_register_ecc_decoder(void (*f)(int, struct mce *, u32)) +{ + nb_bus_decoder = f; +} +EXPORT_SYMBOL_GPL(amd_register_ecc_decoder); + +void amd_unregister_ecc_decoder(void (*f)(int, struct mce *, u32)) +{ + if (nb_bus_decoder) { + WARN_ON(nb_bus_decoder != f); + + nb_bus_decoder = NULL; + } +} +EXPORT_SYMBOL_GPL(amd_unregister_ecc_decoder); + +/* + * string representation for the different MCA reported error types, see F3x48 + * or MSR0000_0411. + */ + +/* transaction type */ +const char *tt_msgs[] = { "INSN", "DATA", "GEN", "RESV" }; +EXPORT_SYMBOL_GPL(tt_msgs); + +/* cache level */ +const char *ll_msgs[] = { "RESV", "L1", "L2", "L3/GEN" }; +EXPORT_SYMBOL_GPL(ll_msgs); + +/* memory transaction type */ +const char *rrrr_msgs[] = { + "GEN", "RD", "WR", "DRD", "DWR", "IRD", "PRF", "EV", "SNP" +}; +EXPORT_SYMBOL_GPL(rrrr_msgs); + +/* participating processor */ +const char *pp_msgs[] = { "SRC", "RES", "OBS", "GEN" }; +EXPORT_SYMBOL_GPL(pp_msgs); + +/* request timeout */ +const char *to_msgs[] = { "no timeout", "timed out" }; +EXPORT_SYMBOL_GPL(to_msgs); + +/* memory or i/o */ +const char *ii_msgs[] = { "MEM", "RESV", "IO", "GEN" }; +EXPORT_SYMBOL_GPL(ii_msgs); + +static const char *f10h_nb_mce_desc[] = { + "HT link data error", + "Protocol error (link, L3, probe filter, etc.)", + "Parity error in NB-internal arrays", + "Link Retry due to IO link transmission error", + "L3 ECC data cache error", + "ECC error in L3 cache tag", + "L3 LRU parity bits error", + "ECC Error in the Probe Filter directory" +}; + +static bool f12h_dc_mce(u16 ec) +{ + bool ret = false; + + if (MEM_ERROR(ec)) { + u8 ll = ec & 0x3; + ret = true; + + if (ll == LL_L2) + pr_cont("during L1 linefill from L2.\n"); + else if (ll == LL_L1) + pr_cont("Data/Tag %s error.\n", RRRR_MSG(ec)); + else + ret = false; + } + return ret; +} + +static bool f10h_dc_mce(u16 ec) +{ + u8 r4 = (ec >> 4) & 0xf; + u8 ll = ec & 0x3; + + if (r4 == R4_GEN && ll == LL_L1) { + pr_cont("during data scrub.\n"); + return true; + } + return f12h_dc_mce(ec); +} + +static bool k8_dc_mce(u16 ec) +{ + if (BUS_ERROR(ec)) { + pr_cont("during system linefill.\n"); + return true; + } + + return f10h_dc_mce(ec); +} + +static bool f14h_dc_mce(u16 ec) +{ + u8 r4 = (ec >> 4) & 0xf; + u8 ll = ec & 0x3; + u8 tt = (ec >> 2) & 0x3; + u8 ii = tt; + bool ret = true; + + if (MEM_ERROR(ec)) { + + if (tt != TT_DATA || ll != LL_L1) + return false; + + switch (r4) { + case R4_DRD: + case R4_DWR: + pr_cont("Data/Tag parity error due to %s.\n", + (r4 == R4_DRD ? "load/hw prf" : "store")); + break; + case R4_EVICT: + pr_cont("Copyback parity error on a tag miss.\n"); + break; + case R4_SNOOP: + pr_cont("Tag parity error during snoop.\n"); + break; + default: + ret = false; + } + } else if (BUS_ERROR(ec)) { + + if ((ii != II_MEM && ii != II_IO) || ll != LL_LG) + return false; + + pr_cont("System read data error on a "); + + switch (r4) { + case R4_RD: + pr_cont("TLB reload.\n"); + break; + case R4_DWR: + pr_cont("store.\n"); + break; + case R4_DRD: + pr_cont("load.\n"); + break; + default: + ret = false; + } + } else { + ret = false; + } + + return ret; +} + +static void amd_decode_dc_mce(struct mce *m) +{ + u16 ec = m->status & 0xffff; + u8 xec = (m->status >> 16) & 0xf; + + pr_emerg(HW_ERR "Data Cache Error: "); + + /* TLB error signatures are the same across families */ + if (TLB_ERROR(ec)) { + u8 tt = (ec >> 2) & 0x3; + + if (tt == TT_DATA) { + pr_cont("%s TLB %s.\n", LL_MSG(ec), + (xec ? "multimatch" : "parity error")); + return; + } + else + goto wrong_dc_mce; + } + + if (!fam_ops->dc_mce(ec)) + goto wrong_dc_mce; + + return; + +wrong_dc_mce: + pr_emerg(HW_ERR "Corrupted DC MCE info?\n"); +} + +static bool k8_ic_mce(u16 ec) +{ + u8 ll = ec & 0x3; + u8 r4 = (ec >> 4) & 0xf; + bool ret = true; + + if (!MEM_ERROR(ec)) + return false; + + if (ll == 0x2) + pr_cont("during a linefill from L2.\n"); + else if (ll == 0x1) { + switch (r4) { + case R4_IRD: + pr_cont("Parity error during data load.\n"); + break; + + case R4_EVICT: + pr_cont("Copyback Parity/Victim error.\n"); + break; + + case R4_SNOOP: + pr_cont("Tag Snoop error.\n"); + break; + + default: + ret = false; + break; + } + } else + ret = false; + + return ret; +} + +static bool f14h_ic_mce(u16 ec) +{ + u8 ll = ec & 0x3; + u8 tt = (ec >> 2) & 0x3; + u8 r4 = (ec >> 4) & 0xf; + bool ret = true; + + if (MEM_ERROR(ec)) { + if (tt != 0 || ll != 1) + ret = false; + + if (r4 == R4_IRD) + pr_cont("Data/tag array parity error for a tag hit.\n"); + else if (r4 == R4_SNOOP) + pr_cont("Tag error during snoop/victimization.\n"); + else + ret = false; + } + return ret; +} + +static void amd_decode_ic_mce(struct mce *m) +{ + u16 ec = m->status & 0xffff; + u8 xec = (m->status >> 16) & 0xf; + + pr_emerg(HW_ERR "Instruction Cache Error: "); + + if (TLB_ERROR(ec)) + pr_cont("%s TLB %s.\n", LL_MSG(ec), + (xec ? "multimatch" : "parity error")); + else if (BUS_ERROR(ec)) { + 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 (fam_ops->ic_mce(ec)) + ; + else + pr_emerg(HW_ERR "Corrupted IC MCE info?\n"); +} + +static void amd_decode_bu_mce(struct mce *m) +{ + u32 ec = m->status & 0xffff; + u32 xec = (m->status >> 16) & 0xf; + + pr_emerg(HW_ERR "Bus Unit Error"); + + if (xec == 0x1) + pr_cont(" in the write data buffers.\n"); + else if (xec == 0x3) + pr_cont(" in the victim data buffers.\n"); + else if (xec == 0x2 && MEM_ERROR(ec)) + pr_cont(": %s error in the L2 cache tags.\n", RRRR_MSG(ec)); + else if (xec == 0x0) { + if (TLB_ERROR(ec)) + pr_cont(": %s error in a Page Descriptor Cache or " + "Guest TLB.\n", TT_MSG(ec)); + else if (BUS_ERROR(ec)) + pr_cont(": %s/ECC error in data read from NB: %s.\n", + RRRR_MSG(ec), PP_MSG(ec)); + else if (MEM_ERROR(ec)) { + u8 rrrr = (ec >> 4) & 0xf; + + if (rrrr >= 0x7) + pr_cont(": %s error during data copyback.\n", + RRRR_MSG(ec)); + else if (rrrr <= 0x1) + pr_cont(": %s parity/ECC error during data " + "access from L2.\n", RRRR_MSG(ec)); + else + goto wrong_bu_mce; + } else + goto wrong_bu_mce; + } else + goto wrong_bu_mce; + + return; + +wrong_bu_mce: + pr_emerg(HW_ERR "Corrupted BU MCE info?\n"); +} + +static void amd_decode_ls_mce(struct mce *m) +{ + u16 ec = m->status & 0xffff; + u8 xec = (m->status >> 16) & 0xf; + + if (boot_cpu_data.x86 == 0x14) { + pr_emerg("You shouldn't be seeing an LS MCE on this cpu family," + " please report on LKML.\n"); + return; + } + + pr_emerg(HW_ERR "Load Store Error"); + + if (xec == 0x0) { + u8 r4 = (ec >> 4) & 0xf; + + if (!BUS_ERROR(ec) || (r4 != R4_DRD && r4 != R4_DWR)) + goto wrong_ls_mce; + + pr_cont(" during %s.\n", RRRR_MSG(ec)); + } else + goto wrong_ls_mce; + + return; + +wrong_ls_mce: + pr_emerg(HW_ERR "Corrupted LS MCE info?\n"); +} + +static bool k8_nb_mce(u16 ec, u8 xec) +{ + bool ret = true; + + switch (xec) { + case 0x1: + pr_cont("CRC error detected on HT link.\n"); + break; + + case 0x5: + pr_cont("Invalid GART PTE entry during GART table walk.\n"); + break; + + case 0x6: + pr_cont("Unsupported atomic RMW received from an IO link.\n"); + break; + + case 0x0: + case 0x8: + if (boot_cpu_data.x86 == 0x11) + return false; + + pr_cont("DRAM ECC error detected on the NB.\n"); + break; + + case 0xd: + pr_cont("Parity error on the DRAM addr/ctl signals.\n"); + break; + + default: + ret = false; + break; + } + + return ret; +} + +static bool f10h_nb_mce(u16 ec, u8 xec) +{ + bool ret = true; + u8 offset = 0; + + if (k8_nb_mce(ec, xec)) + return true; + + switch(xec) { + case 0xa ... 0xc: + offset = 10; + break; + + case 0xe: + offset = 11; + break; + + case 0xf: + if (TLB_ERROR(ec)) + pr_cont("GART Table Walk data error.\n"); + else if (BUS_ERROR(ec)) + pr_cont("DMA Exclusion Vector Table Walk error.\n"); + else + ret = false; + + goto out; + break; + + case 0x1c ... 0x1f: + offset = 24; + break; + + default: + ret = false; + + goto out; + break; + } + + pr_cont("%s.\n", f10h_nb_mce_desc[xec - offset]); + +out: + return ret; +} + +static bool nb_noop_mce(u16 ec, u8 xec) +{ + return false; +} + +void amd_decode_nb_mce(int node_id, struct mce *m, u32 nbcfg) +{ + u8 xec = (m->status >> 16) & 0x1f; + u16 ec = m->status & 0xffff; + u32 nbsh = (u32)(m->status >> 32); + + pr_emerg(HW_ERR "Northbridge Error, node %d: ", node_id); + + /* + * F10h, revD can disable ErrCpu[3:0] so check that first and also the + * value encoding has changed so interpret those differently + */ + if ((boot_cpu_data.x86 == 0x10) && + (boot_cpu_data.x86_model > 7)) { + if (nbsh & K8_NBSH_ERR_CPU_VAL) + pr_cont(", core: %u", (u8)(nbsh & nb_err_cpumask)); + } else { + u8 assoc_cpus = nbsh & nb_err_cpumask; + + if (assoc_cpus > 0) + pr_cont(", core: %d", fls(assoc_cpus) - 1); + } + + switch (xec) { + case 0x2: + pr_cont("Sync error (sync packets on HT link detected).\n"); + return; + + case 0x3: + pr_cont("HT Master abort.\n"); + return; + + case 0x4: + pr_cont("HT Target abort.\n"); + return; + + case 0x7: + pr_cont("NB Watchdog timeout.\n"); + return; + + case 0x9: + pr_cont("SVM DMA Exclusion Vector error.\n"); + return; + + default: + break; + } + + if (!fam_ops->nb_mce(ec, xec)) + goto wrong_nb_mce; + + if (boot_cpu_data.x86 == 0xf || boot_cpu_data.x86 == 0x10) + if ((xec == 0x8 || xec == 0x0) && nb_bus_decoder) + nb_bus_decoder(node_id, m, nbcfg); + + return; + +wrong_nb_mce: + pr_emerg(HW_ERR "Corrupted NB MCE info?\n"); +} +EXPORT_SYMBOL_GPL(amd_decode_nb_mce); + +static void amd_decode_fr_mce(struct mce *m) +{ + if (boot_cpu_data.x86 == 0xf || + boot_cpu_data.x86 == 0x11) + goto wrong_fr_mce; + + /* we have only one error signature so match all fields at once. */ + if ((m->status & 0xffff) == 0x0f0f) { + pr_emerg(HW_ERR "FR Error: CPU Watchdog timer expire.\n"); + return; + } + +wrong_fr_mce: + pr_emerg(HW_ERR "Corrupted FR MCE info?\n"); +} + +static inline void amd_decode_err_code(u16 ec) +{ + if (TLB_ERROR(ec)) { + pr_emerg(HW_ERR "Transaction: %s, Cache Level: %s\n", + TT_MSG(ec), LL_MSG(ec)); + } else if (MEM_ERROR(ec)) { + pr_emerg(HW_ERR "Transaction: %s, Type: %s, Cache Level: %s\n", + RRRR_MSG(ec), TT_MSG(ec), LL_MSG(ec)); + } else if (BUS_ERROR(ec)) { + pr_emerg(HW_ERR "Transaction: %s (%s), %s, Cache Level: %s, " + "Participating Processor: %s\n", + RRRR_MSG(ec), II_MSG(ec), TO_MSG(ec), LL_MSG(ec), + PP_MSG(ec)); + } else + pr_emerg(HW_ERR "Huh? Unknown MCE error 0x%x\n", ec); +} + +/* + * Filter out unwanted MCE signatures here. + */ +static bool amd_filter_mce(struct mce *m) +{ + u8 xec = (m->status >> 16) & 0x1f; + + /* + * NB GART TLB error reporting is disabled by default. + */ + if (m->bank == 4 && xec == 0x5 && !report_gart_errors) + return true; + + return false; +} + +int amd_decode_mce(struct notifier_block *nb, unsigned long val, void *data) +{ + struct mce *m = (struct mce *)data; + int node, ecc; + + if (amd_filter_mce(m)) + return NOTIFY_STOP; + + pr_emerg(HW_ERR "MC%d_STATUS: ", m->bank); + + pr_cont("%sorrected error, other errors lost: %s, " + "CPU context corrupt: %s", + ((m->status & MCI_STATUS_UC) ? "Unc" : "C"), + ((m->status & MCI_STATUS_OVER) ? "yes" : "no"), + ((m->status & MCI_STATUS_PCC) ? "yes" : "no")); + + /* do the two bits[14:13] together */ + ecc = (m->status >> 45) & 0x3; + if (ecc) + pr_cont(", %sECC Error", ((ecc == 2) ? "C" : "U")); + + pr_cont("\n"); + + switch (m->bank) { + case 0: + amd_decode_dc_mce(m); + break; + + case 1: + amd_decode_ic_mce(m); + break; + + case 2: + amd_decode_bu_mce(m); + break; + + case 3: + amd_decode_ls_mce(m); + break; + + case 4: + node = amd_get_nb_id(m->extcpu); + amd_decode_nb_mce(node, m, 0); + break; + + case 5: + amd_decode_fr_mce(m); + break; + + default: + break; + } + + amd_decode_err_code(m->status & 0xffff); + + return NOTIFY_STOP; +} +EXPORT_SYMBOL_GPL(amd_decode_mce); + +static struct notifier_block amd_mce_dec_nb = { + .notifier_call = amd_decode_mce, +}; + +static int __init mce_amd_init(void) +{ + if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD) + return 0; + + if ((boot_cpu_data.x86 < 0xf || boot_cpu_data.x86 > 0x12) && + (boot_cpu_data.x86 != 0x14 || boot_cpu_data.x86_model > 0xf)) + return 0; + + fam_ops = kzalloc(sizeof(struct amd_decoder_ops), GFP_KERNEL); + if (!fam_ops) + return -ENOMEM; + + switch (boot_cpu_data.x86) { + case 0xf: + fam_ops->dc_mce = k8_dc_mce; + fam_ops->ic_mce = k8_ic_mce; + fam_ops->nb_mce = k8_nb_mce; + break; + + case 0x10: + fam_ops->dc_mce = f10h_dc_mce; + fam_ops->ic_mce = k8_ic_mce; + fam_ops->nb_mce = f10h_nb_mce; + break; + + case 0x11: + fam_ops->dc_mce = k8_dc_mce; + fam_ops->ic_mce = k8_ic_mce; + fam_ops->nb_mce = f10h_nb_mce; + break; + + case 0x12: + fam_ops->dc_mce = f12h_dc_mce; + fam_ops->ic_mce = k8_ic_mce; + fam_ops->nb_mce = nb_noop_mce; + break; + + case 0x14: + nb_err_cpumask = 0x3; + fam_ops->dc_mce = f14h_dc_mce; + fam_ops->ic_mce = f14h_ic_mce; + fam_ops->nb_mce = nb_noop_mce; + break; + + default: + printk(KERN_WARNING "Huh? What family is that: %d?!\n", + boot_cpu_data.x86); + kfree(fam_ops); + return -EINVAL; + } + + pr_info("MCE: In-kernel MCE decoding enabled.\n"); + + atomic_notifier_chain_register(&x86_mce_decoder_chain, &amd_mce_dec_nb); + + return 0; +} +early_initcall(mce_amd_init); + +#ifdef MODULE +static void __exit mce_amd_exit(void) +{ + atomic_notifier_chain_unregister(&x86_mce_decoder_chain, &amd_mce_dec_nb); + kfree(fam_ops); +} + +MODULE_DESCRIPTION("AMD MCE decoder"); +MODULE_ALIAS("edac-mce-amd"); +MODULE_LICENSE("GPL"); +module_exit(mce_amd_exit); +#endif diff --git a/drivers/edac/edac_mce_amd.h b/drivers/edac/mce_amd.h index df23ee065f79..35f6e0e3b297 100644 --- a/drivers/edac/edac_mce_amd.h +++ b/drivers/edac/mce_amd.h @@ -1,11 +1,14 @@ #ifndef _EDAC_MCE_AMD_H #define _EDAC_MCE_AMD_H +#include <linux/notifier.h> + #include <asm/mce.h> +#define BIT_64(n) (U64_C(1) << (n)) + #define ERROR_CODE(x) ((x) & 0xffff) #define EXT_ERROR_CODE(x) (((x) >> 16) & 0x1f) -#define EXT_ERR_MSG(x) ext_msgs[EXT_ERROR_CODE(x)] #define LOW_SYNDROME(x) (((x) >> 15) & 0xff) #define HIGH_SYNDROME(x) (((x) >> 24) & 0xff) @@ -20,13 +23,14 @@ #define II_MSG(x) ii_msgs[II(x)] #define LL(x) (((x) >> 0) & 0x3) #define LL_MSG(x) ll_msgs[LL(x)] -#define RRRR(x) (((x) >> 4) & 0xf) -#define RRRR_MSG(x) rrrr_msgs[RRRR(x)] #define TO(x) (((x) >> 8) & 0x1) #define TO_MSG(x) to_msgs[TO(x)] #define PP(x) (((x) >> 9) & 0x3) #define PP_MSG(x) pp_msgs[PP(x)] +#define RRRR(x) (((x) >> 4) & 0xf) +#define RRRR_MSG(x) ((RRRR(x) < 9) ? rrrr_msgs[RRRR(x)] : "Wrong R4!") + #define K8_NBSH 0x4C #define K8_NBSH_VALID_BIT BIT(31) @@ -41,13 +45,45 @@ #define K8_NBSH_UECC BIT(13) #define K8_NBSH_ERR_SCRUBER BIT(8) +enum tt_ids { + TT_INSTR = 0, + TT_DATA, + TT_GEN, + TT_RESV, +}; + +enum ll_ids { + LL_RESV = 0, + LL_L1, + LL_L2, + LL_LG, +}; + +enum ii_ids { + II_MEM = 0, + II_RESV, + II_IO, + II_GEN, +}; + +enum rrrr_ids { + R4_GEN = 0, + R4_RD, + R4_WR, + R4_DRD, + R4_DWR, + R4_IRD, + R4_PREF, + R4_EVICT, + R4_SNOOP, +}; + extern const char *tt_msgs[]; extern const char *ll_msgs[]; extern const char *rrrr_msgs[]; extern const char *pp_msgs[]; extern const char *to_msgs[]; extern const char *ii_msgs[]; -extern const char *ext_msgs[]; /* * relevant NB regs @@ -60,10 +96,19 @@ struct err_regs { u32 nbeal; }; +/* + * per-family decoder ops + */ +struct amd_decoder_ops { + bool (*dc_mce)(u16); + bool (*ic_mce)(u16); + bool (*nb_mce)(u16, u8); +}; void amd_report_gart_errors(bool); -void amd_register_ecc_decoder(void (*f)(int, struct err_regs *)); -void amd_unregister_ecc_decoder(void (*f)(int, struct err_regs *)); -void amd_decode_nb_mce(int, struct err_regs *, int); +void amd_register_ecc_decoder(void (*f)(int, struct mce *, u32)); +void amd_unregister_ecc_decoder(void (*f)(int, struct mce *, u32)); +void amd_decode_nb_mce(int, struct mce *, u32); +int amd_decode_mce(struct notifier_block *nb, unsigned long val, void *data); #endif /* _EDAC_MCE_AMD_H */ diff --git a/drivers/edac/mce_amd_inj.c b/drivers/edac/mce_amd_inj.c new file mode 100644 index 000000000000..8d0688f36d4c --- /dev/null +++ b/drivers/edac/mce_amd_inj.c @@ -0,0 +1,171 @@ +/* + * A simple MCE injection facility for testing the MCE decoding code. This + * driver should be built as module so that it can be loaded on production + * kernels for testing purposes. + * + * This file may be distributed under the terms of the GNU General Public + * License version 2. + * + * Copyright (c) 2010: Borislav Petkov <borislav.petkov@amd.com> + * Advanced Micro Devices Inc. + */ + +#include <linux/kobject.h> +#include <linux/sysdev.h> +#include <linux/edac.h> +#include <asm/mce.h> + +#include "mce_amd.h" + +struct edac_mce_attr { + struct attribute attr; + ssize_t (*show) (struct kobject *kobj, struct edac_mce_attr *attr, char *buf); + ssize_t (*store)(struct kobject *kobj, struct edac_mce_attr *attr, + const char *buf, size_t count); +}; + +#define EDAC_MCE_ATTR(_name, _mode, _show, _store) \ +static struct edac_mce_attr mce_attr_##_name = __ATTR(_name, _mode, _show, _store) + +static struct kobject *mce_kobj; + +/* + * Collect all the MCi_XXX settings + */ +static struct mce i_mce; + +#define MCE_INJECT_STORE(reg) \ +static ssize_t edac_inject_##reg##_store(struct kobject *kobj, \ + struct edac_mce_attr *attr, \ + const char *data, size_t count)\ +{ \ + int ret = 0; \ + unsigned long value; \ + \ + ret = strict_strtoul(data, 16, &value); \ + if (ret < 0) \ + printk(KERN_ERR "Error writing MCE " #reg " field.\n"); \ + \ + i_mce.reg = value; \ + \ + return count; \ +} + +MCE_INJECT_STORE(status); +MCE_INJECT_STORE(misc); +MCE_INJECT_STORE(addr); + +#define MCE_INJECT_SHOW(reg) \ +static ssize_t edac_inject_##reg##_show(struct kobject *kobj, \ + struct edac_mce_attr *attr, \ + char *buf) \ +{ \ + return sprintf(buf, "0x%016llx\n", i_mce.reg); \ +} + +MCE_INJECT_SHOW(status); +MCE_INJECT_SHOW(misc); +MCE_INJECT_SHOW(addr); + +EDAC_MCE_ATTR(status, 0644, edac_inject_status_show, edac_inject_status_store); +EDAC_MCE_ATTR(misc, 0644, edac_inject_misc_show, edac_inject_misc_store); +EDAC_MCE_ATTR(addr, 0644, edac_inject_addr_show, edac_inject_addr_store); + +/* + * This denotes into which bank we're injecting and triggers + * the injection, at the same time. + */ +static ssize_t edac_inject_bank_store(struct kobject *kobj, + struct edac_mce_attr *attr, + const char *data, size_t count) +{ + int ret = 0; + unsigned long value; + + ret = strict_strtoul(data, 10, &value); + if (ret < 0) { + printk(KERN_ERR "Invalid bank value!\n"); + return -EINVAL; + } + + if (value > 5) { + printk(KERN_ERR "Non-existant MCE bank: %lu\n", value); + return -EINVAL; + } + + i_mce.bank = value; + + amd_decode_mce(NULL, 0, &i_mce); + + return count; +} + +static ssize_t edac_inject_bank_show(struct kobject *kobj, + struct edac_mce_attr *attr, char *buf) +{ + return sprintf(buf, "%d\n", i_mce.bank); +} + +EDAC_MCE_ATTR(bank, 0644, edac_inject_bank_show, edac_inject_bank_store); + +static struct edac_mce_attr *sysfs_attrs[] = { &mce_attr_status, &mce_attr_misc, + &mce_attr_addr, &mce_attr_bank +}; + +static int __init edac_init_mce_inject(void) +{ + struct sysdev_class *edac_class = NULL; + int i, err = 0; + + edac_class = edac_get_sysfs_class(); + if (!edac_class) + return -EINVAL; + + mce_kobj = kobject_create_and_add("mce", &edac_class->kset.kobj); + if (!mce_kobj) { + printk(KERN_ERR "Error creating a mce kset.\n"); + err = -ENOMEM; + goto err_mce_kobj; + } + + for (i = 0; i < ARRAY_SIZE(sysfs_attrs); i++) { + err = sysfs_create_file(mce_kobj, &sysfs_attrs[i]->attr); + if (err) { + printk(KERN_ERR "Error creating %s in sysfs.\n", + sysfs_attrs[i]->attr.name); + goto err_sysfs_create; + } + } + return 0; + +err_sysfs_create: + while (i-- >= 0) + sysfs_remove_file(mce_kobj, &sysfs_attrs[i]->attr); + + kobject_del(mce_kobj); + +err_mce_kobj: + edac_put_sysfs_class(); + + return err; +} + +static void __exit edac_exit_mce_inject(void) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(sysfs_attrs); i++) + sysfs_remove_file(mce_kobj, &sysfs_attrs[i]->attr); + + kobject_del(mce_kobj); + + edac_put_sysfs_class(); +} + +module_init(edac_init_mce_inject); +module_exit(edac_exit_mce_inject); + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("Borislav Petkov <borislav.petkov@amd.com>"); +MODULE_AUTHOR("AMD Inc."); +MODULE_DESCRIPTION("MCE injection facility for testing MCE decoding"); diff --git a/include/linux/edac.h b/include/linux/edac.h index 7cf92e8a4196..36c66443bdfd 100644 --- a/include/linux/edac.h +++ b/include/linux/edac.h @@ -13,6 +13,7 @@ #define _LINUX_EDAC_H_ #include <asm/atomic.h> +#include <linux/sysdev.h> #define EDAC_OPSTATE_INVAL -1 #define EDAC_OPSTATE_POLL 0 @@ -22,9 +23,12 @@ extern int edac_op_state; extern int edac_err_assert; extern atomic_t edac_handlers; +extern struct sysdev_class edac_class; extern int edac_handler_set(void); extern void edac_atomic_assert_error(void); +extern struct sysdev_class *edac_get_sysfs_class(void); +extern void edac_put_sysfs_class(void); static inline void opstate_init(void) { |