Skip to content

L
linux
Commit b70ef010 authored by Borislav Petkov's avatar Borislav Petkov
Browse files
Options

EDAC: move MCE error descriptions to EDAC core 

This is in preparation of adding AMD-specific MCE decoding functionality
to the EDAC core. The error decoding macros originate from the AMD64
EDAC driver albeit in a simplified and cleaned up version here.

While at it, add macros to generate the error description strings and
use them in the error type decoders directly which removes a bunch of
code and makes the decoding functions much more readable. Also, fix
strings and shorten macro names.

Remove superfluous htlink_msgs.
Signed-off-by: default avatarBorislav Petkov <borislav.petkov@amd.com>
parent 74fca6a4
Hide whitespace changes
Inline Side-by-side
Showing with 122 additions and 148 deletions
drivers/edac/Makefile
View file @ b70ef010
......@@ -17,6 +17,10 @@ ifdef CONFIG_PCI
edac_core-objs += edac_pci.o edac_pci_sysfs.o
endif
ifdef CONFIG_CPU_SUP_AMD
edac_core-objs += edac_mce_amd.o
endif
obj-$(CONFIG_EDAC_AMD76X) += amd76x_edac.o
obj-$(CONFIG_EDAC_CPC925) += cpc925_edac.o
obj-$(CONFIG_EDAC_I5000) += i5000_edac.o
......@@ -32,7 +36,7 @@ obj-$(CONFIG_EDAC_X38) += x38_edac.o
obj-$(CONFIG_EDAC_I82860) += i82860_edac.o
obj-$(CONFIG_EDAC_R82600) += r82600_edac.o
amd64_edac_mod-y := amd64_edac_err_types.o amd64_edac.o
amd64_edac_mod-y := amd64_edac.o
amd64_edac_mod-$(CONFIG_EDAC_DEBUG) += amd64_edac_dbg.o
amd64_edac_mod-$(CONFIG_EDAC_AMD64_ERROR_INJECTION) += amd64_edac_inj.o
......
drivers/edac/amd64_edac.c
View file @ b70ef010
......@@ -18,6 +18,63 @@ struct amd64_pvt;
static struct mem_ctl_info *mci_lookup[MAX_NUMNODES];
static struct amd64_pvt *pvt_lookup[MAX_NUMNODES];
/*
* See F2x80 for K8 and F2x[1,0]80 for Fam10 and later. The table below is only
* for DDR2 DRAM mapping.
*/
u32 revf_quad_ddr2_shift[] = {
0, /* 0000b NULL DIMM (128mb) */
28, /* 0001b 256mb */
29, /* 0010b 512mb */
29, /* 0011b 512mb */
29, /* 0100b 512mb */
30, /* 0101b 1gb */
30, /* 0110b 1gb */
31, /* 0111b 2gb */
31, /* 1000b 2gb */
32, /* 1001b 4gb */
32, /* 1010b 4gb */
33, /* 1011b 8gb */
0, /* 1100b future */
0, /* 1101b future */
0, /* 1110b future */
0 /* 1111b future */
};
/*
* Valid scrub rates for the K8 hardware memory scrubber. We map the scrubbing
* bandwidth to a valid bit pattern. The 'set' operation finds the 'matching-
* or higher value'.
*
*FIXME: Produce a better mapping/linearisation.
*/
struct scrubrate scrubrates[] = {
{ 0x01, 1600000000UL},
{ 0x02, 800000000UL},
{ 0x03, 400000000UL},
{ 0x04, 200000000UL},
{ 0x05, 100000000UL},
{ 0x06, 50000000UL},
{ 0x07, 25000000UL},
{ 0x08, 12284069UL},
{ 0x09, 6274509UL},
{ 0x0A, 3121951UL},
{ 0x0B, 1560975UL},
{ 0x0C, 781440UL},
{ 0x0D, 390720UL},
{ 0x0E, 195300UL},
{ 0x0F, 97650UL},
{ 0x10, 48854UL},
{ 0x11, 24427UL},
{ 0x12, 12213UL},
{ 0x13, 6101UL},
{ 0x14, 3051UL},
{ 0x15, 1523UL},
{ 0x16, 761UL},
{ 0x00, 0UL}, /* scrubbing off */
};
/*
* Memory scrubber control interface. For K8, memory scrubbing is handled by
* hardware and can involve L2 cache, dcache as well as the main memory. With
......@@ -1101,8 +1158,8 @@ static void k8_map_sysaddr_to_csrow(struct mem_ctl_info *mci,
u32 page, offset;
/* Extract the syndrome parts and form a 16-bit syndrome */
syndrome = EXTRACT_HIGH_SYNDROME(info->nbsl) << 8;
syndrome |= EXTRACT_LOW_SYNDROME(info->nbsh);
syndrome = HIGH_SYNDROME(info->nbsl) << 8;
syndrome |= LOW_SYNDROME(info->nbsh);
/* CHIPKILL enabled */
if (info->nbcfg & K8_NBCFG_CHIPKILL) {
......@@ -1701,8 +1758,8 @@ static void f10_map_sysaddr_to_csrow(struct mem_ctl_info *mci,
if (csrow >= 0) {
error_address_to_page_and_offset(sys_addr, &page, &offset);
syndrome = EXTRACT_HIGH_SYNDROME(info->nbsl) << 8;
syndrome |= EXTRACT_LOW_SYNDROME(info->nbsh);
syndrome = HIGH_SYNDROME(info->nbsl) << 8;
syndrome |= LOW_SYNDROME(info->nbsh);
/*
* Is CHIPKILL on? If so, then we can attempt to use the
......@@ -2155,36 +2212,22 @@ static int amd64_get_error_info(struct mem_ctl_info *mci,
static inline void amd64_decode_gart_tlb_error(struct mem_ctl_info *mci,
struct amd64_error_info_regs *info)
{
u32 err_code;
u32 ec_tt; /* error code transaction type (2b) */
u32 ec_ll; /* error code cache level (2b) */
err_code = EXTRACT_ERROR_CODE(info->nbsl);
ec_ll = EXTRACT_LL_CODE(err_code);
ec_tt = EXTRACT_TT_CODE(err_code);
u32 ec = ERROR_CODE(info->nbsl);
amd64_mc_printk(mci, KERN_ERR,
"GART TLB event: transaction type(%s), "
"cache level(%s)\n", tt_msgs[ec_tt], ll_msgs[ec_ll]);
"cache level(%s)\n", TT_MSG(ec), LL_MSG(ec));
}
static inline void amd64_decode_mem_cache_error(struct mem_ctl_info *mci,
struct amd64_error_info_regs *info)
{
u32 err_code;
u32 ec_rrrr; /* error code memory transaction (4b) */
u32 ec_tt; /* error code transaction type (2b) */
u32 ec_ll; /* error code cache level (2b) */
err_code = EXTRACT_ERROR_CODE(info->nbsl);
ec_ll = EXTRACT_LL_CODE(err_code);
ec_tt = EXTRACT_TT_CODE(err_code);
ec_rrrr = EXTRACT_RRRR_CODE(err_code);
u32 ec = ERROR_CODE(info->nbsl);
amd64_mc_printk(mci, KERN_ERR,
"cache hierarchy error: memory transaction type(%s), "
"transaction type(%s), cache level(%s)\n",
rrrr_msgs[ec_rrrr], tt_msgs[ec_tt], ll_msgs[ec_ll]);
RRRR_MSG(ec), TT_MSG(ec), LL_MSG(ec));
}
......@@ -2264,21 +2307,8 @@ static void amd64_handle_ue(struct mem_ctl_info *mci,
static void amd64_decode_bus_error(struct mem_ctl_info *mci,
struct amd64_error_info_regs *info)
{
u32 err_code, ext_ec;
u32 ec_pp; /* error code participating processor (2p) */
u32 ec_to; /* error code timed out (1b) */
u32 ec_rrrr; /* error code memory transaction (4b) */
u32 ec_ii; /* error code memory or I/O (2b) */
u32 ec_ll; /* error code cache level (2b) */
ext_ec = EXTRACT_EXT_ERROR_CODE(info->nbsl);
err_code = EXTRACT_ERROR_CODE(info->nbsl);
ec_ll = EXTRACT_LL_CODE(err_code);
ec_ii = EXTRACT_II_CODE(err_code);
ec_rrrr = EXTRACT_RRRR_CODE(err_code);
ec_to = EXTRACT_TO_CODE(err_code);
ec_pp = EXTRACT_PP_CODE(err_code);
u32 ec = ERROR_CODE(info->nbsl);
u32 xec = EXT_ERROR_CODE(info->nbsl);
amd64_mc_printk(mci, KERN_ERR,
"BUS ERROR:\n"
......@@ -2286,20 +2316,17 @@ static void amd64_decode_bus_error(struct mem_ctl_info *mci,
" participating processor(%s)\n"
" memory transaction type(%s)\n"
" cache level(%s) Error Found by: %s\n",
to_msgs[ec_to],
ii_msgs[ec_ii],
pp_msgs[ec_pp],
rrrr_msgs[ec_rrrr],
ll_msgs[ec_ll],
TO_MSG(ec), II_MSG(ec), PP_MSG(ec), RRRR_MSG(ec), LL_MSG(ec),
(info->nbsh & K8_NBSH_ERR_SCRUBER) ?
"Scrubber" : "Normal Operation");
/* If this was an 'observed' error, early out */
if (ec_pp == K8_NBSL_PP_OBS)
return; /* We aren't the node involved */
/* Bail early out if this was an 'observed' error */
if (PP(ec) == K8_NBSL_PP_OBS)
return;
/* Parse out the extended error code for ECC events */
switch (ext_ec) {
switch (xec) {
/* F10 changed to one Extended ECC error code */
case F10_NBSL_EXT_ERR_RES: /* Reserved field */
case F10_NBSL_EXT_ERR_ECC: /* F10 ECC ext err code */
......@@ -2379,7 +2406,7 @@ int amd64_process_error_info(struct mem_ctl_info *mci,
(regs->nbsh & K8_NBSH_CORE3) ? "True" : "False");
err_code = EXTRACT_ERROR_CODE(regs->nbsl);
err_code = ERROR_CODE(regs->nbsl);
/* Determine which error type:
* 1) GART errors - non-fatal, developmental events
......@@ -2387,7 +2414,7 @@ int amd64_process_error_info(struct mem_ctl_info *mci,
* 3) BUS errors
* 4) Unknown error
*/
if (TEST_TLB_ERROR(err_code)) {
if (TLB_ERROR(err_code)) {
/*
* GART errors are intended to help graphics driver developers
* to detect bad GART PTEs. It is recommended by AMD to disable
......@@ -2411,10 +2438,10 @@ int amd64_process_error_info(struct mem_ctl_info *mci,
debugf1("GART TLB error\n");
amd64_decode_gart_tlb_error(mci, info);
} else if (TEST_MEM_ERROR(err_code)) {
} else if (MEM_ERROR(err_code)) {
debugf1("Memory/Cache error\n");
amd64_decode_mem_cache_error(mci, info);
} else if (TEST_BUS_ERROR(err_code)) {
} else if (BUS_ERROR(err_code)) {
debugf1("Bus (Link/DRAM) error\n");
amd64_decode_bus_error(mci, info);
} else {
......@@ -2424,21 +2451,10 @@ int amd64_process_error_info(struct mem_ctl_info *mci,
err_code);
}
ext_ec = EXTRACT_EXT_ERROR_CODE(regs->nbsl);
ext_ec = EXT_ERROR_CODE(regs->nbsl);
amd64_mc_printk(mci, KERN_ERR,
"ExtErr=(0x%x) %s\n", ext_ec, ext_msgs[ext_ec]);
if (((ext_ec >= F10_NBSL_EXT_ERR_CRC &&
ext_ec <= F10_NBSL_EXT_ERR_TGT) ||
(ext_ec == F10_NBSL_EXT_ERR_RMW)) &&
EXTRACT_LDT_LINK(info->nbsh)) {
amd64_mc_printk(mci, KERN_ERR,
"Error on hypertransport link: %s\n",
htlink_msgs[
EXTRACT_LDT_LINK(info->nbsh)]);
}
/*
* Check the UE bit of the NB status high register, if set generate some
* logs. If NOT a GART error, then process the event as a NO-INFO event.
......
drivers/edac/amd64_edac.h
View file @ b70ef010
......@@ -72,6 +72,7 @@
#include <linux/edac.h>
#include <asm/msr.h>
#include "edac_core.h"
#include "edac_mce_amd.h"
#define amd64_printk(level, fmt, arg...) \
edac_printk(level, "amd64", fmt, ##arg)
......@@ -303,9 +304,6 @@ enum {
#define K8_NBSL 0x48
#define EXTRACT_HIGH_SYNDROME(x) (((x) >> 24) & 0xff)
#define EXTRACT_EXT_ERROR_CODE(x) (((x) >> 16) & 0x1f)
/* Family F10h: Normalized Extended Error Codes */
#define F10_NBSL_EXT_ERR_RES 0x0
#define F10_NBSL_EXT_ERR_CRC 0x1
......@@ -348,17 +346,6 @@ enum {
#define K8_NBSL_EXT_ERR_CHIPKILL_ECC 0x8
#define K8_NBSL_EXT_ERR_DRAM_PARITY 0xD
#define EXTRACT_ERROR_CODE(x) ((x) & 0xffff)
#define TEST_TLB_ERROR(x) (((x) & 0xFFF0) == 0x0010)
#define TEST_MEM_ERROR(x) (((x) & 0xFF00) == 0x0100)
#define TEST_BUS_ERROR(x) (((x) & 0xF800) == 0x0800)
#define EXTRACT_TT_CODE(x) (((x) >> 2) & 0x3)
#define EXTRACT_II_CODE(x) (((x) >> 2) & 0x3)
#define EXTRACT_LL_CODE(x) (((x) >> 0) & 0x3)
#define EXTRACT_RRRR_CODE(x) (((x) >> 4) & 0xf)
#define EXTRACT_TO_CODE(x) (((x) >> 8) & 0x1)
#define EXTRACT_PP_CODE(x) (((x) >> 9) & 0x3)
/*
* The following are for BUS type errors AFTER values have been normalized by
* shifting right
......@@ -386,9 +373,7 @@ enum {
#define K8_NBSH_CORE1 BIT(1)
#define K8_NBSH_CORE0 BIT(0)
#define EXTRACT_LDT_LINK(x) (((x) >> 4) & 0x7)
#define EXTRACT_ERR_CPU_MAP(x) ((x) & 0xF)
#define EXTRACT_LOW_SYNDROME(x) (((x) >> 15) & 0xff)
#define K8_NBEAL 0x50
......
drivers/edac/amd64_edac_err_types.c → drivers/edac/edac_mce_amd.c
View file @ b70ef010
#include "amd64_edac.h"
/*
* See F2x80 for K8 and F2x[1,0]80 for Fam10 and later. The table below is only
* for DDR2 DRAM mapping.
*/
u32 revf_quad_ddr2_shift[] = {
0, /* 0000b NULL DIMM (128mb) */
28, /* 0001b 256mb */
29, /* 0010b 512mb */
29, /* 0011b 512mb */
29, /* 0100b 512mb */
30, /* 0101b 1gb */
30, /* 0110b 1gb */
31, /* 0111b 2gb */
31, /* 1000b 2gb */
32, /* 1001b 4gb */
32, /* 1010b 4gb */
33, /* 1011b 8gb */
0, /* 1100b future */
0, /* 1101b future */
0, /* 1110b future */
0 /* 1111b future */
};
/*
* Valid scrub rates for the K8 hardware memory scrubber. We map the scrubbing
* bandwidth to a valid bit pattern. The 'set' operation finds the 'matching-
* or higher value'.
*
*FIXME: Produce a better mapping/linearisation.
*/
struct scrubrate scrubrates[] = {
{ 0x01, 1600000000UL},
{ 0x02, 800000000UL},
{ 0x03, 400000000UL},
{ 0x04, 200000000UL},
{ 0x05, 100000000UL},
{ 0x06, 50000000UL},
{ 0x07, 25000000UL},
{ 0x08, 12284069UL},
{ 0x09, 6274509UL},
{ 0x0A, 3121951UL},
{ 0x0B, 1560975UL},
{ 0x0C, 781440UL},
{ 0x0D, 390720UL},
{ 0x0E, 195300UL},
{ 0x0F, 97650UL},
{ 0x10, 48854UL},
{ 0x11, 24427UL},
{ 0x12, 12213UL},
{ 0x13, 6101UL},
{ 0x14, 3051UL},
{ 0x15, 1523UL},
{ 0x16, 761UL},
{ 0x00, 0UL}, /* scrubbing off */
};
#include <linux/module.h>
#include "edac_mce_amd.h"
/*
* string representation for the different MCA reported error types, see F3x48
......@@ -67,6 +11,7 @@ const char *tt_msgs[] = { /* transaction type */
"generic",
"reserved"
};
EXPORT_SYMBOL_GPL(tt_msgs);
const char *ll_msgs[] = { /* cache level */
"L0",
......@@ -74,6 +19,7 @@ const char *ll_msgs[] = { /* cache level */
"L2",
"L3/generic"
};
EXPORT_SYMBOL_GPL(ll_msgs);
const char *rrrr_msgs[] = {
"generic",
......@@ -93,6 +39,7 @@ const char *rrrr_msgs[] = {
"reserved RRRR= 14",
"reserved RRRR= 15"
};
EXPORT_SYMBOL_GPL(rrrr_msgs);
const char *pp_msgs[] = { /* participating processor */
"local node originated (SRC)",
......@@ -100,11 +47,13 @@ const char *pp_msgs[] = { /* participating processor */
"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",
......@@ -112,6 +61,7 @@ const char *ii_msgs[] = { /* memory or i/o */
"i/o access",
"generic"
};
EXPORT_SYMBOL_GPL(ii_msgs);
/* Map the 5 bits of Extended Error code to the string table. */
const char *ext_msgs[] = { /* extended error */
......@@ -148,14 +98,4 @@ const char *ext_msgs[] = { /* extended error */
"L3 Cache LRU error", /* 1_1110b */
"Res 0x1FF error" /* 1_1111b */
};
const char *htlink_msgs[] = {
"none",
"1",
"2",
"1 2",
"3",
"1 3",
"2 3",
"1 2 3"
};
EXPORT_SYMBOL_GPL(ext_msgs);
drivers/edac/edac_mce_amd.h 0 → 100644
View file @ b70ef010
#define ERROR_CODE(x) ((x) & 0xffff)
#define EXT_ERROR_CODE(x) (((x) >> 16) & 0x1f)
#define LOW_SYNDROME(x) (((x) >> 15) & 0xff)
#define HIGH_SYNDROME(x) (((x) >> 24) & 0xff)
#define TLB_ERROR(x) (((x) & 0xFFF0) == 0x0010)
#define MEM_ERROR(x) (((x) & 0xFF00) == 0x0100)
#define BUS_ERROR(x) (((x) & 0xF800) == 0x0800)
#define TT(x) (((x) >> 2) & 0x3)
#define TT_MSG(x) tt_msgs[TT(x)]
#define II(x) (((x) >> 2) & 0x3)
#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)]
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[];
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7