linux/drivers/edac/mce_amd.c

541 lines
12 KiB
C
Raw Normal View History

#include <linux/module.h>
#include <linux/slab.h>
#include "mce_amd.h"
static struct amd_decoder_ops *fam_ops;
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);
/*
* 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 bool f10h_dc_mce(u16 ec)
{
u8 r4 = (ec >> 4) & 0xf;
bool ret = false;
if (r4 == R4_GEN) {
pr_cont("during data scrub.\n");
return true;
}
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 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(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)
{
u32 ec = m->status & 0xffff;
u32 xec = (m->status >> 16) & 0xf;
pr_emerg(HW_ERR "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_emerg(HW_ERR "Corrupted LS MCE info?\n");
}
void amd_decode_nb_mce(int node_id, struct mce *m, u32 nbcfg)
{
u32 ec = m->status & 0xffff;
u32 nbsh = (u32)(m->status >> 32);
u32 nbsl = (u32)m->status;
/*
* GART TLB error reporting is disabled by default. Bail out early.
*/
if (TLB_ERROR(ec) && !report_gart_errors)
return;
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\n", (u8)(nbsh & 0xf));
} else {
u8 assoc_cpus = nbsh & 0xf;
if (assoc_cpus > 0)
pr_cont(", core: %d", fls(assoc_cpus) - 1);
pr_cont("\n");
}
pr_emerg(HW_ERR "%s.\n", EXT_ERR_MSG(nbsl));
if (BUS_ERROR(ec) && nb_bus_decoder)
nb_bus_decoder(node_id, m, nbcfg);
}
EXPORT_SYMBOL_GPL(amd_decode_nb_mce);
static void amd_decode_fr_mce(struct mce *m)
{
/* 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");
else
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);
}
int amd_decode_mce(struct notifier_block *nb, unsigned long val, void *data)
{
struct mce *m = (struct mce *)data;
int node, ecc;
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);
x86: EDAC: MCE: Fix MCE decoding callback logic Make decoding of MCEs happen only on AMD hardware by registering a non-default callback only on CPU families which support it. While looking at the interaction of decode_mce() with the other MCE code i also noticed a few other things and made the following cleanups/fixes: - Fixed the mce_decode() weak alias - a weak alias is really not good here, it should be a proper callback. A weak alias will be overriden if a piece of code is built into the kernel - not good, obviously. - The patch initializes the callback on AMD family 10h and 11h. - Added the more correct fallback printk of: No support for human readable MCE decoding on this CPU type. Transcribe the message and run it through 'mcelog --ascii' to decode. On CPUs that dont have a decoder. - Made the surrounding code more readable. Note that the callback allows us to have a default fallback - without having to check the CPU versions during the printout itself. When an EDAC module registers itself, it can install the decode-print function. (there's no unregister needed as this is core code.) version -v2 by Borislav Petkov: - add K8 to the set of supported CPUs - always build in edac_mce_amd since we use an early_initcall now - fix checkpatch warnings Signed-off-by: Borislav Petkov <borislav.petkov@amd.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andi Kleen <andi@firstfloor.org> LKML-Reference: <20091001141432.GA11410@aftab> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-01 22:14:32 +08:00
static struct notifier_block amd_mce_dec_nb = {
.notifier_call = amd_decode_mce,
};
x86: EDAC: MCE: Fix MCE decoding callback logic Make decoding of MCEs happen only on AMD hardware by registering a non-default callback only on CPU families which support it. While looking at the interaction of decode_mce() with the other MCE code i also noticed a few other things and made the following cleanups/fixes: - Fixed the mce_decode() weak alias - a weak alias is really not good here, it should be a proper callback. A weak alias will be overriden if a piece of code is built into the kernel - not good, obviously. - The patch initializes the callback on AMD family 10h and 11h. - Added the more correct fallback printk of: No support for human readable MCE decoding on this CPU type. Transcribe the message and run it through 'mcelog --ascii' to decode. On CPUs that dont have a decoder. - Made the surrounding code more readable. Note that the callback allows us to have a default fallback - without having to check the CPU versions during the printout itself. When an EDAC module registers itself, it can install the decode-print function. (there's no unregister needed as this is core code.) version -v2 by Borislav Petkov: - add K8 to the set of supported CPUs - always build in edac_mce_amd since we use an early_initcall now - fix checkpatch warnings Signed-off-by: Borislav Petkov <borislav.petkov@amd.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andi Kleen <andi@firstfloor.org> LKML-Reference: <20091001141432.GA11410@aftab> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-01 22:14:32 +08:00
static int __init mce_amd_init(void)
{
/*
* We can decode MCEs for K8, F10h and F11h CPUs:
x86: EDAC: MCE: Fix MCE decoding callback logic Make decoding of MCEs happen only on AMD hardware by registering a non-default callback only on CPU families which support it. While looking at the interaction of decode_mce() with the other MCE code i also noticed a few other things and made the following cleanups/fixes: - Fixed the mce_decode() weak alias - a weak alias is really not good here, it should be a proper callback. A weak alias will be overriden if a piece of code is built into the kernel - not good, obviously. - The patch initializes the callback on AMD family 10h and 11h. - Added the more correct fallback printk of: No support for human readable MCE decoding on this CPU type. Transcribe the message and run it through 'mcelog --ascii' to decode. On CPUs that dont have a decoder. - Made the surrounding code more readable. Note that the callback allows us to have a default fallback - without having to check the CPU versions during the printout itself. When an EDAC module registers itself, it can install the decode-print function. (there's no unregister needed as this is core code.) version -v2 by Borislav Petkov: - add K8 to the set of supported CPUs - always build in edac_mce_amd since we use an early_initcall now - fix checkpatch warnings Signed-off-by: Borislav Petkov <borislav.petkov@amd.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andi Kleen <andi@firstfloor.org> LKML-Reference: <20091001141432.GA11410@aftab> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-01 22:14:32 +08:00
*/
if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD)
return 0;
if (boot_cpu_data.x86 < 0xf || boot_cpu_data.x86 > 0x11)
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;
break;
case 0x10:
fam_ops->dc_mce = f10h_dc_mce;
fam_ops->ic_mce = k8_ic_mce;
break;
case 0x14:
fam_ops->dc_mce = f14h_dc_mce;
fam_ops->ic_mce = f14h_ic_mce;
break;
default:
printk(KERN_WARNING "Huh? What family is that: %d?!\n",
boot_cpu_data.x86);
kfree(fam_ops);
return -EINVAL;
}
atomic_notifier_chain_register(&x86_mce_decoder_chain, &amd_mce_dec_nb);
x86: EDAC: MCE: Fix MCE decoding callback logic Make decoding of MCEs happen only on AMD hardware by registering a non-default callback only on CPU families which support it. While looking at the interaction of decode_mce() with the other MCE code i also noticed a few other things and made the following cleanups/fixes: - Fixed the mce_decode() weak alias - a weak alias is really not good here, it should be a proper callback. A weak alias will be overriden if a piece of code is built into the kernel - not good, obviously. - The patch initializes the callback on AMD family 10h and 11h. - Added the more correct fallback printk of: No support for human readable MCE decoding on this CPU type. Transcribe the message and run it through 'mcelog --ascii' to decode. On CPUs that dont have a decoder. - Made the surrounding code more readable. Note that the callback allows us to have a default fallback - without having to check the CPU versions during the printout itself. When an EDAC module registers itself, it can install the decode-print function. (there's no unregister needed as this is core code.) version -v2 by Borislav Petkov: - add K8 to the set of supported CPUs - always build in edac_mce_amd since we use an early_initcall now - fix checkpatch warnings Signed-off-by: Borislav Petkov <borislav.petkov@amd.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andi Kleen <andi@firstfloor.org> LKML-Reference: <20091001141432.GA11410@aftab> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-01 22:14:32 +08:00
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