aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--MAINTAINERS7
-rw-r--r--drivers/edac/Kconfig23
-rw-r--r--drivers/edac/Makefile1
-rw-r--r--drivers/edac/ghes_edac.c114
4 files changed, 145 insertions, 0 deletions
diff --git a/MAINTAINERS b/MAINTAINERS
index 35a56bcd5e75..889644d9772b 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -2803,6 +2803,13 @@ W: bluesmoke.sourceforge.net
2803S: Maintained 2803S: Maintained
2804F: drivers/edac/e7xxx_edac.c 2804F: drivers/edac/e7xxx_edac.c
2805 2805
2806EDAC-GHES
2807M: Mauro Carvalho Chehab <mchehab@redhat.com>
2808L: linux-edac@vger.kernel.org
2809W: bluesmoke.sourceforge.net
2810S: Maintained
2811F: drivers/edac/ghes-edac.c
2812
2806EDAC-I82443BXGX 2813EDAC-I82443BXGX
2807M: Tim Small <tim@buttersideup.com> 2814M: Tim Small <tim@buttersideup.com>
2808L: linux-edac@vger.kernel.org 2815L: linux-edac@vger.kernel.org
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
index 66719925970f..7e38e5e576e8 100644
--- a/drivers/edac/Kconfig
+++ b/drivers/edac/Kconfig
@@ -80,6 +80,29 @@ config EDAC_MM_EDAC
80 occurred so that a particular failing memory module can be 80 occurred so that a particular failing memory module can be
81 replaced. If unsure, select 'Y'. 81 replaced. If unsure, select 'Y'.
82 82
83config 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
83config EDAC_AMD64 106config EDAC_AMD64
84 tristate "AMD64 (Opteron, Athlon64) K8, F10h" 107 tristate "AMD64 (Opteron, Athlon64) K8, F10h"
85 depends on EDAC_MM_EDAC && AMD_NB && X86_64 && EDAC_DECODE_MCE 108 depends on EDAC_MM_EDAC && AMD_NB && X86_64 && EDAC_DECODE_MCE
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
index 5608a9ba61b7..4154ed6a02c6 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -16,6 +16,7 @@ ifdef CONFIG_PCI
16edac_core-y += edac_pci.o edac_pci_sysfs.o 16edac_core-y += edac_pci.o edac_pci_sysfs.o
17endif 17endif
18 18
19obj-$(CONFIG_EDAC_GHES) += ghes_edac.o
19obj-$(CONFIG_EDAC_MCE_INJ) += mce_amd_inj.o 20obj-$(CONFIG_EDAC_MCE_INJ) += mce_amd_inj.o
20 21
21edac_mce_amd-y := mce_amd.o 22edac_mce_amd-y := mce_amd.o
diff --git a/drivers/edac/ghes_edac.c b/drivers/edac/ghes_edac.c
new file mode 100644
index 000000000000..d8e54b496e0f
--- /dev/null
+++ b/drivers/edac/ghes_edac.c
@@ -0,0 +1,114 @@
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
19struct ghes_edac_pvt {
20 struct list_head list;
21 struct ghes *ghes;
22 struct mem_ctl_info *mci;
23};
24
25static LIST_HEAD(ghes_reglist);
26static DEFINE_MUTEX(ghes_edac_lock);
27static int ghes_edac_mc_num;
28
29void ghes_edac_report_mem_error(struct ghes *ghes, int sev,
30 struct cper_sec_mem_err *mem_err)
31{
32}
33EXPORT_SYMBOL_GPL(ghes_edac_report_mem_error);
34
35int 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}
98EXPORT_SYMBOL_GPL(ghes_edac_register);
99
100void 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}
114EXPORT_SYMBOL_GPL(ghes_edac_unregister);