提交 888ab8e6 编写于 作者: B Borislav Petkov 提交者: Borislav Petkov

EDAC, MCE: Adjust DC decoders to F14h

Add a per-family data cache decoders. Since there is a certain overlap
between the different DC MCE signatures, reuse functionality between the
families as far as possible.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
上级 47ca08a4
#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);
......@@ -97,41 +101,116 @@ const char *ext_msgs[] = {
};
EXPORT_SYMBOL_GPL(ext_msgs);
static void amd_decode_dc_mce(struct mce *m)
static bool f10h_dc_mce(u16 ec)
{
u32 ec = m->status & 0xffff;
u32 xec = (m->status >> 16) & 0xf;
u8 r4 = (ec >> 4) & 0xf;
bool ret = false;
pr_emerg(HW_ERR "Data Cache Error: ");
if (r4 == R4_GEN) {
pr_cont("during data scrub.\n");
return true;
}
if (xec == 1 && TLB_ERROR(ec))
pr_cont(": %s TLB multimatch.\n", LL_MSG(ec));
else if (xec == 0) {
if (m->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;
if (MEM_ERROR(ec)) {
u8 ll = ec & 0x3;
ret = true;
/* see F10h BKDG (31116), Table 92. */
if (ll == 0x1) {
if (tt != 0x1)
goto wrong_dc_mce;
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;
}
pr_cont(": Data/Tag %s error.\n", RRRR_MSG(ec));
static bool k8_dc_mce(u16 ec)
{
if (BUS_ERROR(ec)) {
pr_cont("during system linefill.\n");
return true;
}
} 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");
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;
} else
}
if (!fam_ops->dc_mce(ec))
goto wrong_dc_mce;
return;
......@@ -395,6 +474,30 @@ static int __init mce_amd_init(void)
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;
break;
case 0x10:
fam_ops->dc_mce = f10h_dc_mce;
break;
case 0x14:
fam_ops->dc_mce = f14h_dc_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);
return 0;
......@@ -405,6 +508,7 @@ early_initcall(mce_amd_init);
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");
......
......@@ -44,6 +44,39 @@
#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[];
......@@ -63,6 +96,13 @@ struct err_regs {
u32 nbeal;
};
/*
* per-family decoder ops
*/
struct amd_decoder_ops {
bool (*dc_mce)(u16);
};
void amd_report_gart_errors(bool);
void amd_register_ecc_decoder(void (*f)(int, struct mce *, u32));
void amd_unregister_ecc_decoder(void (*f)(int, struct mce *, u32));
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册