tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
ghes_edac: Register at EDAC core the BIOS report
Register GHES at EDAC MC core, in order to avoid other
drivers to also handle errors and mangle with error data.
The edac core will warrant that just one driver will be used,
so the first one to register (BIOS first) will be the one that
will be reporting the hardware errors.
For now, the EDAC driver does nothing but to register at the
EDAC core, preventing the hardware-driven mechanism to
interfere with GHES.
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
+145
+7
MAINTAINERS
+7
MAINTAINERS
···
2803
2803
S: Maintained
2804
2804
F: drivers/edac/e7xxx_edac.c
2805
2805
2806
+
EDAC-GHES
2807
+
M: Mauro Carvalho Chehab <mchehab@redhat.com>
2808
+
L: linux-edac@vger.kernel.org
2809
+
W: bluesmoke.sourceforge.net
2810
+
S: Maintained
2811
+
F: drivers/edac/ghes-edac.c
2812
+
2806
2813
EDAC-I82443BXGX
2807
2814
M: Tim Small <tim@buttersideup.com>
2808
2815
L: linux-edac@vger.kernel.org
+23
drivers/edac/Kconfig
+23
drivers/edac/Kconfig
···
80
80
occurred so that a particular failing memory module can be
81
81
replaced. If unsure, select 'Y'.
82
82
83
+
config EDAC_GHES
84
+
bool "Output ACPI APEI/GHES BIOS detected errors via EDAC"
85
+
depends on ACPI_APEI_GHES && (EDAC_MM_EDAC=y)
86
+
default y
87
+
help
88
+
Not all machines support hardware-driven error report. Some of those
89
+
provide a BIOS-driven error report mechanism via ACPI, using the
90
+
APEI/GHES driver. By enabling this option, the error reports provided
91
+
by GHES are sent to userspace via the EDAC API.
92
+
93
+
When this option is enabled, it will disable the hardware-driven
94
+
mechanisms, if a GHES BIOS is detected, entering into the
95
+
"Firmware First" mode.
96
+
97
+
It should be noticed that keeping both GHES and a hardware-driven
98
+
error mechanism won't work well, as BIOS will race with OS, while
99
+
reading the error registers. So, if you want to not use "Firmware
100
+
first" GHES error mechanism, you should disable GHES either at
101
+
compilation time or by passing "ghes.disable=1" Kernel parameter
102
+
at boot time.
103
+
104
+
In doubt, say 'Y'.
105
+
83
106
config EDAC_AMD64
84
107
tristate "AMD64 (Opteron, Athlon64) K8, F10h"
85
108
depends on EDAC_MM_EDAC && AMD_NB && X86_64 && EDAC_DECODE_MCE
+1
drivers/edac/Makefile
+1
drivers/edac/Makefile
+114
drivers/edac/ghes_edac.c
+114
drivers/edac/ghes_edac.c
···
1
+
/*
2
+
* GHES/EDAC Linux driver
3
+
*
4
+
* This file may be distributed under the terms of the GNU General Public
5
+
* License version 2.
6
+
*
7
+
* Copyright (c) 2013 by Mauro Carvalho Chehab <mchehab@redhat.com>
8
+
*
9
+
* Red Hat Inc. http://www.redhat.com
10
+
*/
11
+
12
+
#include <acpi/ghes.h>
13
+
#include <linux/edac.h>
14
+
#include "edac_core.h"
15
+
16
+
#define GHES_PFX "ghes_edac: "
17
+
#define GHES_EDAC_REVISION " Ver: 1.0.0"
18
+
19
+
struct ghes_edac_pvt {
20
+
struct list_head list;
21
+
struct ghes *ghes;
22
+
struct mem_ctl_info *mci;
23
+
};
24
+
25
+
static LIST_HEAD(ghes_reglist);
26
+
static DEFINE_MUTEX(ghes_edac_lock);
27
+
static int ghes_edac_mc_num;
28
+
29
+
void ghes_edac_report_mem_error(struct ghes *ghes, int sev,
30
+
struct cper_sec_mem_err *mem_err)
31
+
{
32
+
}
33
+
EXPORT_SYMBOL_GPL(ghes_edac_report_mem_error);
34
+
35
+
int ghes_edac_register(struct ghes *ghes, struct device *dev)
36
+
{
37
+
int rc;
38
+
struct mem_ctl_info *mci;
39
+
struct edac_mc_layer layers[1];
40
+
struct csrow_info *csrow;
41
+
struct dimm_info *dimm;
42
+
struct ghes_edac_pvt *pvt;
43
+
44
+
layers[0].type = EDAC_MC_LAYER_ALL_MEM;
45
+
layers[0].size = 1;
46
+
layers[0].is_virt_csrow = true;
47
+
48
+
/*
49
+
* We need to serialize edac_mc_alloc() and edac_mc_add_mc(),
50
+
* to avoid duplicated memory controller numbers
51
+
*/
52
+
mutex_lock(&ghes_edac_lock);
53
+
mci = edac_mc_alloc(ghes_edac_mc_num, ARRAY_SIZE(layers), layers,
54
+
sizeof(*pvt));
55
+
if (!mci) {
56
+
pr_info(GHES_PFX "Can't allocate memory for EDAC data\n");
57
+
mutex_unlock(&ghes_edac_lock);
58
+
return -ENOMEM;
59
+
}
60
+
61
+
pvt = mci->pvt_info;
62
+
memset(pvt, 0, sizeof(*pvt));
63
+
list_add_tail(&pvt->list, &ghes_reglist);
64
+
pvt->ghes = ghes;
65
+
pvt->mci = mci;
66
+
mci->pdev = dev;
67
+
68
+
mci->mtype_cap = MEM_FLAG_EMPTY;
69
+
mci->edac_ctl_cap = EDAC_FLAG_NONE;
70
+
mci->edac_cap = EDAC_FLAG_NONE;
71
+
mci->mod_name = "ghes_edac.c";
72
+
mci->mod_ver = GHES_EDAC_REVISION;
73
+
mci->ctl_name = "ghes_edac";
74
+
mci->dev_name = "ghes";
75
+
76
+
csrow = mci->csrows[0];
77
+
dimm = csrow->channels[0]->dimm;
78
+
79
+
/* FIXME: FAKE DATA */
80
+
dimm->nr_pages = 1000;
81
+
dimm->grain = 128;
82
+
dimm->mtype = MEM_UNKNOWN;
83
+
dimm->dtype = DEV_UNKNOWN;
84
+
dimm->edac_mode = EDAC_SECDED;
85
+
86
+
rc = edac_mc_add_mc(mci);
87
+
if (rc < 0) {
88
+
pr_info(GHES_PFX "Can't register at EDAC core\n");
89
+
edac_mc_free(mci);
90
+
mutex_unlock(&ghes_edac_lock);
91
+
return -ENODEV;
92
+
}
93
+
94
+
ghes_edac_mc_num++;
95
+
mutex_unlock(&ghes_edac_lock);
96
+
return 0;
97
+
}
98
+
EXPORT_SYMBOL_GPL(ghes_edac_register);
99
+
100
+
void ghes_edac_unregister(struct ghes *ghes)
101
+
{
102
+
struct mem_ctl_info *mci;
103
+
struct ghes_edac_pvt *pvt;
104
+
105
+
list_for_each_entry(pvt, &ghes_reglist, list) {
106
+
if (ghes == pvt->ghes) {
107
+
mci = pvt->mci;
108
+
edac_mc_del_mc(mci->pdev);
109
+
edac_mc_free(mci);
110
+
list_del(&pvt->list);
111
+
}
112
+
}
113
+
}
114
+
EXPORT_SYMBOL_GPL(ghes_edac_unregister);