aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorBorislav Petkov <borislav.petkov@amd.com>2010-09-02 12:33:24 -0400
committerBorislav Petkov <bp@amd64.org>2010-10-21 08:47:59 -0400
commit9cdeb404a1870c5022915e576dbdc3cde21af5bf (patch)
tree19843bf914f56f362a7c9cdf86c27a648caa475c
parent30e1f7a8122145f44f45c95366e27b6bb0b08428 (diff)
EDAC, MCE: Rework MCE injection
Add sysfs injection facilities for testing of the MCE decoding code. Remove large parts of amd64_edac_dbg.c, as a result, which did only NB MCE injection anyway and the new injection code supports that functionality already. Add an injection module so that MCE decoding code in production kernels like those in RHEL and SLES can be tested. Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
-rw-r--r--drivers/edac/Kconfig14
-rw-r--r--drivers/edac/Makefile2
-rw-r--r--drivers/edac/amd64_edac.h2
-rw-r--r--drivers/edac/amd64_edac_dbg.c213
-rw-r--r--drivers/edac/edac_mce_amd.c4
-rw-r--r--drivers/edac/edac_mce_amd.h4
-rw-r--r--drivers/edac/mce_amd_inj.c171
7 files changed, 203 insertions, 207 deletions
diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
index 70bb350de996..3bb3a671baa0 100644
--- a/drivers/edac/Kconfig
+++ b/drivers/edac/Kconfig
@@ -39,7 +39,7 @@ config EDAC_DEBUG
39 there're four debug levels (x=0,1,2,3 from low to high). 39 there're four debug levels (x=0,1,2,3 from low to high).
40 Usually you should select 'N'. 40 Usually you should select 'N'.
41 41
42 config EDAC_DECODE_MCE 42config EDAC_DECODE_MCE
43 tristate "Decode MCEs in human-readable form (only on AMD for now)" 43 tristate "Decode MCEs in human-readable form (only on AMD for now)"
44 depends on CPU_SUP_AMD && X86_MCE 44 depends on CPU_SUP_AMD && X86_MCE
45 default y 45 default y
@@ -51,6 +51,16 @@ config EDAC_DEBUG
51 which occur really early upon boot, before the module infrastructure 51 which occur really early upon boot, before the module infrastructure
52 has been initialized. 52 has been initialized.
53 53
54config EDAC_MCE_INJ
55 tristate "Simple MCE injection interface over /sysfs"
56 depends on EDAC_DECODE_MCE
57 default n
58 help
59 This is a simple interface to inject MCEs over /sysfs and test
60 the MCE decoding code in EDAC.
61
62 This is currently AMD-only.
63
54config EDAC_MM_EDAC 64config EDAC_MM_EDAC
55 tristate "Main Memory EDAC (Error Detection And Correction) reporting" 65 tristate "Main Memory EDAC (Error Detection And Correction) reporting"
56 help 66 help
@@ -72,7 +82,7 @@ config EDAC_AMD64
72 Families of Memory Controllers (K8, F10h and F11h) 82 Families of Memory Controllers (K8, F10h and F11h)
73 83
74config EDAC_AMD64_ERROR_INJECTION 84config EDAC_AMD64_ERROR_INJECTION
75 bool "Sysfs Error Injection facilities" 85 bool "Sysfs HW Error injection facilities"
76 depends on EDAC_AMD64 86 depends on EDAC_AMD64
77 help 87 help
78 Recent Opterons (Family 10h and later) provide for Memory Error 88 Recent Opterons (Family 10h and later) provide for Memory Error
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
index ca6b1bb24ccc..5c38ad38f3a3 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -17,6 +17,8 @@ ifdef CONFIG_PCI
17edac_core-objs += edac_pci.o edac_pci_sysfs.o 17edac_core-objs += edac_pci.o edac_pci_sysfs.o
18endif 18endif
19 19
20obj-$(CONFIG_EDAC_MCE_INJ) += mce_amd_inj.o
21
20obj-$(CONFIG_EDAC_DECODE_MCE) += edac_mce_amd.o 22obj-$(CONFIG_EDAC_DECODE_MCE) += edac_mce_amd.o
21 23
22obj-$(CONFIG_EDAC_AMD76X) += amd76x_edac.o 24obj-$(CONFIG_EDAC_AMD76X) += amd76x_edac.o
diff --git a/drivers/edac/amd64_edac.h b/drivers/edac/amd64_edac.h
index 613b9381e71a..67d9ceb4b839 100644
--- a/drivers/edac/amd64_edac.h
+++ b/drivers/edac/amd64_edac.h
@@ -486,7 +486,7 @@ extern const char *ext_msgs[32];
486extern const char *htlink_msgs[8]; 486extern const char *htlink_msgs[8];
487 487
488#ifdef CONFIG_EDAC_DEBUG 488#ifdef CONFIG_EDAC_DEBUG
489#define NUM_DBG_ATTRS 9 489#define NUM_DBG_ATTRS 5
490#else 490#else
491#define NUM_DBG_ATTRS 0 491#define NUM_DBG_ATTRS 0
492#endif 492#endif
diff --git a/drivers/edac/amd64_edac_dbg.c b/drivers/edac/amd64_edac_dbg.c
index f6d5695de5b6..e3562288f4ce 100644
--- a/drivers/edac/amd64_edac_dbg.c
+++ b/drivers/edac/amd64_edac_dbg.c
@@ -1,173 +1,16 @@
1#include "amd64_edac.h" 1#include "amd64_edac.h"
2 2
3/* 3#define EDAC_DCT_ATTR_SHOW(reg) \
4 * accept a hex value and store it into the virtual error register file, field: 4static ssize_t amd64_##reg##_show(struct mem_ctl_info *mci, char *data) \
5 * nbeal and nbeah. Assume virtual error values have already been set for: NBSL, 5{ \
6 * NBSH and NBCFG. Then proceed to map the error values to a MC, CSROW and 6 struct amd64_pvt *pvt = mci->pvt_info; \
7 * CHANNEL 7 return sprintf(data, "0x%016llx\n", (u64)pvt->reg); \
8 */
9static ssize_t amd64_nbea_store(struct mem_ctl_info *mci, const char *data,
10 size_t count)
11{
12 struct amd64_pvt *pvt = mci->pvt_info;
13 u64 value;
14 int ret = 0;
15 struct mce m;
16
17 ret = strict_strtoull(data, 16, &value);
18 if (ret != -EINVAL) {
19 struct err_regs *regs = &pvt->ctl_error_info;
20
21 debugf0("received NBEA= 0x%llx\n", value);
22
23 /* place the value into the virtual error packet */
24 pvt->ctl_error_info.nbeal = (u32) value;
25 value >>= 32;
26 pvt->ctl_error_info.nbeah = (u32) value;
27
28 m.addr = value;
29 m.status = regs->nbsl | ((u64)regs->nbsh << 32);
30
31 /* Process the Mapping request */
32 /* TODO: Add race prevention */
33 amd_decode_nb_mce(pvt->mc_node_id, &m, regs->nbcfg);
34
35 return count;
36 }
37 return ret;
38} 8}
39 9
40/* display back what the last NBEA (MCA NB Address (MC4_ADDR)) was written */ 10EDAC_DCT_ATTR_SHOW(dhar);
41static ssize_t amd64_nbea_show(struct mem_ctl_info *mci, char *data) 11EDAC_DCT_ATTR_SHOW(dbam0);
42{ 12EDAC_DCT_ATTR_SHOW(top_mem);
43 struct amd64_pvt *pvt = mci->pvt_info; 13EDAC_DCT_ATTR_SHOW(top_mem2);
44 u64 value;
45
46 value = pvt->ctl_error_info.nbeah;
47 value <<= 32;
48 value |= pvt->ctl_error_info.nbeal;
49
50 return sprintf(data, "%llx\n", value);
51}
52
53/* store the NBSL (MCA NB Status Low (MC4_STATUS)) value user desires */
54static ssize_t amd64_nbsl_store(struct mem_ctl_info *mci, const char *data,
55 size_t count)
56{
57 struct amd64_pvt *pvt = mci->pvt_info;
58 unsigned long value;
59 int ret = 0;
60
61 ret = strict_strtoul(data, 16, &value);
62 if (ret != -EINVAL) {
63 debugf0("received NBSL= 0x%lx\n", value);
64
65 pvt->ctl_error_info.nbsl = (u32) value;
66
67 return count;
68 }
69 return ret;
70}
71
72/* display back what the last NBSL value written */
73static ssize_t amd64_nbsl_show(struct mem_ctl_info *mci, char *data)
74{
75 struct amd64_pvt *pvt = mci->pvt_info;
76 u32 value;
77
78 value = pvt->ctl_error_info.nbsl;
79
80 return sprintf(data, "%x\n", value);
81}
82
83/* store the NBSH (MCA NB Status High) value user desires */
84static ssize_t amd64_nbsh_store(struct mem_ctl_info *mci, const char *data,
85 size_t count)
86{
87 struct amd64_pvt *pvt = mci->pvt_info;
88 unsigned long value;
89 int ret = 0;
90
91 ret = strict_strtoul(data, 16, &value);
92 if (ret != -EINVAL) {
93 debugf0("received NBSH= 0x%lx\n", value);
94
95 pvt->ctl_error_info.nbsh = (u32) value;
96
97 return count;
98 }
99 return ret;
100}
101
102/* display back what the last NBSH value written */
103static ssize_t amd64_nbsh_show(struct mem_ctl_info *mci, char *data)
104{
105 struct amd64_pvt *pvt = mci->pvt_info;
106 u32 value;
107