Memory errors could be printed with incorrect row values since the DIMM
size has outgrown the 16 bit row field in the CPER structure. UEFI
Specification Version 2.8 has increased the size of row by allowing it to
use the first 2 bits from a previously reserved space within the structure.
When needed, add the extension bits to the row value printed.
Based on UEFI 2.8 Table 299. Memory Error Record
Tested-by: Russ Anderson <[email protected]>
Signed-off-by: Alex Kluver <[email protected]>
---
drivers/edac/ghes_edac.c | 10 ++++++++--
drivers/firmware/efi/cper.c | 11 +++++++++--
include/linux/cper.h | 9 +++++++--
3 files changed, 24 insertions(+), 6 deletions(-)
diff --git a/drivers/edac/ghes_edac.c b/drivers/edac/ghes_edac.c
index cb3dab56a875..cfa3156300f5 100644
--- a/drivers/edac/ghes_edac.c
+++ b/drivers/edac/ghes_edac.c
@@ -337,8 +337,14 @@ void ghes_edac_report_mem_error(int sev, struct cper_sec_mem_err *mem_err)
p += sprintf(p, "rank:%d ", mem_err->rank);
if (mem_err->validation_bits & CPER_MEM_VALID_BANK)
p += sprintf(p, "bank:%d ", mem_err->bank);
- if (mem_err->validation_bits & CPER_MEM_VALID_ROW)
- p += sprintf(p, "row:%d ", mem_err->row);
+ if (mem_err->validation_bits & (CPER_MEM_VALID_ROW | CPER_MEM_VALID_ROW_EXT)) {
+ u32 row_extended = 0;
+ if (mem_err->validation_bits & CPER_MEM_VALID_ROW_EXT)
+ row_extended = (mem_err->extended & CPER_MEM_EXT_ROW_MASK)
+ <<CPER_MEM_EXT_ROW_SHIFT;
+ row_extended |= mem_err->row;
+ p += sprintf(p, "row:%d ", row_extended);
+ }
if (mem_err->validation_bits & CPER_MEM_VALID_COLUMN)
p += sprintf(p, "col:%d ", mem_err->column);
if (mem_err->validation_bits & CPER_MEM_VALID_BIT_POSITION)
diff --git a/drivers/firmware/efi/cper.c b/drivers/firmware/efi/cper.c
index f564e15fbc7e..5faaf6ecd659 100644
--- a/drivers/firmware/efi/cper.c
+++ b/drivers/firmware/efi/cper.c
@@ -234,8 +234,14 @@ static int cper_mem_err_location(struct cper_mem_err_compact *mem, char *msg)
n += scnprintf(msg + n, len - n, "bank: %d ", mem->bank);
if (mem->validation_bits & CPER_MEM_VALID_DEVICE)
n += scnprintf(msg + n, len - n, "device: %d ", mem->device);
- if (mem->validation_bits & CPER_MEM_VALID_ROW)
- n += scnprintf(msg + n, len - n, "row: %d ", mem->row);
+ if (mem->validation_bits & (CPER_MEM_VALID_ROW | CPER_MEM_VALID_ROW_EXT)) {
+ u32 row_extended = 0;
+ if (mem->validation_bits & CPER_MEM_VALID_ROW_EXT)
+ row_extended = (mem->extended & CPER_MEM_EXT_ROW_MASK)
+ <<CPER_MEM_EXT_ROW_SHIFT;
+ row_extended |= mem->row;
+ n += scnprintf(msg + n, len - n, "row: %d ", row_extended);
+ }
if (mem->validation_bits & CPER_MEM_VALID_COLUMN)
n += scnprintf(msg + n, len - n, "column: %d ", mem->column);
if (mem->validation_bits & CPER_MEM_VALID_BIT_POSITION)
@@ -292,6 +298,7 @@ void cper_mem_err_pack(const struct cper_sec_mem_err *mem,
cmem->requestor_id = mem->requestor_id;
cmem->responder_id = mem->responder_id;
cmem->target_id = mem->target_id;
+ cmem->extended = mem->extended;
cmem->rank = mem->rank;
cmem->mem_array_handle = mem->mem_array_handle;
cmem->mem_dev_handle = mem->mem_dev_handle;
diff --git a/include/linux/cper.h b/include/linux/cper.h
index 8537e9282a65..c8313753ee49 100644
--- a/include/linux/cper.h
+++ b/include/linux/cper.h
@@ -230,6 +230,10 @@ enum {
#define CPER_MEM_VALID_RANK_NUMBER 0x8000
#define CPER_MEM_VALID_CARD_HANDLE 0x10000
#define CPER_MEM_VALID_MODULE_HANDLE 0x20000
+#define CPER_MEM_VALID_ROW_EXT 0x40000
+
+#define CPER_MEM_EXT_ROW_MASK 0x3
+#define CPER_MEM_EXT_ROW_SHIFT 16
#define CPER_PCIE_VALID_PORT_TYPE 0x0001
#define CPER_PCIE_VALID_VERSION 0x0002
@@ -443,7 +447,7 @@ struct cper_sec_mem_err_old {
u8 error_type;
};
-/* Memory Error Section (UEFI >= v2.3), UEFI v2.7 sec N.2.5 */
+/* Memory Error Section (UEFI >= v2.3), UEFI v2.8 sec N.2.5 */
struct cper_sec_mem_err {
u64 validation_bits;
u64 error_status;
@@ -461,7 +465,7 @@ struct cper_sec_mem_err {
u64 responder_id;
u64 target_id;
u8 error_type;
- u8 reserved;
+ u8 extended;
u16 rank;
u16 mem_array_handle; /* "card handle" in UEFI 2.4 */
u16 mem_dev_handle; /* "module handle" in UEFI 2.4 */
@@ -483,6 +487,7 @@ struct cper_mem_err_compact {
u16 rank;
u16 mem_array_handle;
u16 mem_dev_handle;
+ u8 extended;
};
/* PCI Express Error Section, UEFI v2.7 sec N.2.7 */
On Mon, 27 Jul 2020 at 20:15, Alex Kluver <[email protected]> wrote:
>
> Memory errors could be printed with incorrect row values since the DIMM
> size has outgrown the 16 bit row field in the CPER structure. UEFI
> Specification Version 2.8 has increased the size of row by allowing it to
> use the first 2 bits from a previously reserved space within the structure.
>
> When needed, add the extension bits to the row value printed.
>
> Based on UEFI 2.8 Table 299. Memory Error Record
>
> Tested-by: Russ Anderson <[email protected]>
> Signed-off-by: Alex Kluver <[email protected]>
> ---
> drivers/edac/ghes_edac.c | 10 ++++++++--
> drivers/firmware/efi/cper.c | 11 +++++++++--
> include/linux/cper.h | 9 +++++++--
> 3 files changed, 24 insertions(+), 6 deletions(-)
>
Unless any of the EDAC folks object, I will take this as a fix via the EFI tree.
Thanks,
Ard.
> diff --git a/drivers/edac/ghes_edac.c b/drivers/edac/ghes_edac.c
> index cb3dab56a875..cfa3156300f5 100644
> --- a/drivers/edac/ghes_edac.c
> +++ b/drivers/edac/ghes_edac.c
> @@ -337,8 +337,14 @@ void ghes_edac_report_mem_error(int sev, struct cper_sec_mem_err *mem_err)
> p += sprintf(p, "rank:%d ", mem_err->rank);
> if (mem_err->validation_bits & CPER_MEM_VALID_BANK)
> p += sprintf(p, "bank:%d ", mem_err->bank);
> - if (mem_err->validation_bits & CPER_MEM_VALID_ROW)
> - p += sprintf(p, "row:%d ", mem_err->row);
> + if (mem_err->validation_bits & (CPER_MEM_VALID_ROW | CPER_MEM_VALID_ROW_EXT)) {
> + u32 row_extended = 0;
> + if (mem_err->validation_bits & CPER_MEM_VALID_ROW_EXT)
> + row_extended = (mem_err->extended & CPER_MEM_EXT_ROW_MASK)
> + <<CPER_MEM_EXT_ROW_SHIFT;
> + row_extended |= mem_err->row;
> + p += sprintf(p, "row:%d ", row_extended);
> + }
> if (mem_err->validation_bits & CPER_MEM_VALID_COLUMN)
> p += sprintf(p, "col:%d ", mem_err->column);
> if (mem_err->validation_bits & CPER_MEM_VALID_BIT_POSITION)
> diff --git a/drivers/firmware/efi/cper.c b/drivers/firmware/efi/cper.c
> index f564e15fbc7e..5faaf6ecd659 100644
> --- a/drivers/firmware/efi/cper.c
> +++ b/drivers/firmware/efi/cper.c
> @@ -234,8 +234,14 @@ static int cper_mem_err_location(struct cper_mem_err_compact *mem, char *msg)
> n += scnprintf(msg + n, len - n, "bank: %d ", mem->bank);
> if (mem->validation_bits & CPER_MEM_VALID_DEVICE)
> n += scnprintf(msg + n, len - n, "device: %d ", mem->device);
> - if (mem->validation_bits & CPER_MEM_VALID_ROW)
> - n += scnprintf(msg + n, len - n, "row: %d ", mem->row);
> + if (mem->validation_bits & (CPER_MEM_VALID_ROW | CPER_MEM_VALID_ROW_EXT)) {
> + u32 row_extended = 0;
> + if (mem->validation_bits & CPER_MEM_VALID_ROW_EXT)
> + row_extended = (mem->extended & CPER_MEM_EXT_ROW_MASK)
> + <<CPER_MEM_EXT_ROW_SHIFT;
> + row_extended |= mem->row;
> + n += scnprintf(msg + n, len - n, "row: %d ", row_extended);
> + }
> if (mem->validation_bits & CPER_MEM_VALID_COLUMN)
> n += scnprintf(msg + n, len - n, "column: %d ", mem->column);
> if (mem->validation_bits & CPER_MEM_VALID_BIT_POSITION)
> @@ -292,6 +298,7 @@ void cper_mem_err_pack(const struct cper_sec_mem_err *mem,
> cmem->requestor_id = mem->requestor_id;
> cmem->responder_id = mem->responder_id;
> cmem->target_id = mem->target_id;
> + cmem->extended = mem->extended;
> cmem->rank = mem->rank;
> cmem->mem_array_handle = mem->mem_array_handle;
> cmem->mem_dev_handle = mem->mem_dev_handle;
> diff --git a/include/linux/cper.h b/include/linux/cper.h
> index 8537e9282a65..c8313753ee49 100644
> --- a/include/linux/cper.h
> +++ b/include/linux/cper.h
> @@ -230,6 +230,10 @@ enum {
> #define CPER_MEM_VALID_RANK_NUMBER 0x8000
> #define CPER_MEM_VALID_CARD_HANDLE 0x10000
> #define CPER_MEM_VALID_MODULE_HANDLE 0x20000
> +#define CPER_MEM_VALID_ROW_EXT 0x40000
> +
> +#define CPER_MEM_EXT_ROW_MASK 0x3
> +#define CPER_MEM_EXT_ROW_SHIFT 16
>
> #define CPER_PCIE_VALID_PORT_TYPE 0x0001
> #define CPER_PCIE_VALID_VERSION 0x0002
> @@ -443,7 +447,7 @@ struct cper_sec_mem_err_old {
> u8 error_type;
> };
>
> -/* Memory Error Section (UEFI >= v2.3), UEFI v2.7 sec N.2.5 */
> +/* Memory Error Section (UEFI >= v2.3), UEFI v2.8 sec N.2.5 */
> struct cper_sec_mem_err {
> u64 validation_bits;
> u64 error_status;
> @@ -461,7 +465,7 @@ struct cper_sec_mem_err {
> u64 responder_id;
> u64 target_id;
> u8 error_type;
> - u8 reserved;
> + u8 extended;
> u16 rank;
> u16 mem_array_handle; /* "card handle" in UEFI 2.4 */
> u16 mem_dev_handle; /* "module handle" in UEFI 2.4 */
> @@ -483,6 +487,7 @@ struct cper_mem_err_compact {
> u16 rank;
> u16 mem_array_handle;
> u16 mem_dev_handle;
> + u8 extended;
> };
>
> /* PCI Express Error Section, UEFI v2.7 sec N.2.7 */
On Mon, Jul 27, 2020 at 01:14:45PM -0500, Alex Kluver wrote:
> Memory errors could be printed with incorrect row values since the DIMM
> size has outgrown the 16 bit row field in the CPER structure. UEFI
> Specification Version 2.8 has increased the size of row by allowing it to
> use the first 2 bits from a previously reserved space within the structure.
>
> When needed, add the extension bits to the row value printed.
>
> Based on UEFI 2.8 Table 299. Memory Error Record
>
> Tested-by: Russ Anderson <[email protected]>
> Signed-off-by: Alex Kluver <[email protected]>
> ---
> drivers/edac/ghes_edac.c | 10 ++++++++--
> drivers/firmware/efi/cper.c | 11 +++++++++--
> include/linux/cper.h | 9 +++++++--
> 3 files changed, 24 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/edac/ghes_edac.c b/drivers/edac/ghes_edac.c
> index cb3dab56a875..cfa3156300f5 100644
> --- a/drivers/edac/ghes_edac.c
> +++ b/drivers/edac/ghes_edac.c
> @@ -337,8 +337,14 @@ void ghes_edac_report_mem_error(int sev, struct cper_sec_mem_err *mem_err)
> p += sprintf(p, "rank:%d ", mem_err->rank);
> if (mem_err->validation_bits & CPER_MEM_VALID_BANK)
> p += sprintf(p, "bank:%d ", mem_err->bank);
> - if (mem_err->validation_bits & CPER_MEM_VALID_ROW)
> - p += sprintf(p, "row:%d ", mem_err->row);
> + if (mem_err->validation_bits & (CPER_MEM_VALID_ROW | CPER_MEM_VALID_ROW_EXT)) {
> + u32 row_extended = 0;
> + if (mem_err->validation_bits & CPER_MEM_VALID_ROW_EXT)
> + row_extended = (mem_err->extended & CPER_MEM_EXT_ROW_MASK)
> + <<CPER_MEM_EXT_ROW_SHIFT;
> + row_extended |= mem_err->row;
> + p += sprintf(p, "row:%d ", row_extended);
> + }
> if (mem_err->validation_bits & CPER_MEM_VALID_COLUMN)
> p += sprintf(p, "col:%d ", mem_err->column);
> if (mem_err->validation_bits & CPER_MEM_VALID_BIT_POSITION)
> diff --git a/drivers/firmware/efi/cper.c b/drivers/firmware/efi/cper.c
> index f564e15fbc7e..5faaf6ecd659 100644
> --- a/drivers/firmware/efi/cper.c
> +++ b/drivers/firmware/efi/cper.c
> @@ -234,8 +234,14 @@ static int cper_mem_err_location(struct cper_mem_err_compact *mem, char *msg)
> n += scnprintf(msg + n, len - n, "bank: %d ", mem->bank);
> if (mem->validation_bits & CPER_MEM_VALID_DEVICE)
> n += scnprintf(msg + n, len - n, "device: %d ", mem->device);
> - if (mem->validation_bits & CPER_MEM_VALID_ROW)
> - n += scnprintf(msg + n, len - n, "row: %d ", mem->row);
> + if (mem->validation_bits & (CPER_MEM_VALID_ROW | CPER_MEM_VALID_ROW_EXT)) {
> + u32 row_extended = 0;
> + if (mem->validation_bits & CPER_MEM_VALID_ROW_EXT)
> + row_extended = (mem->extended & CPER_MEM_EXT_ROW_MASK)
> + <<CPER_MEM_EXT_ROW_SHIFT;
This is not very readable.
> + row_extended |= mem->row;
> + n += scnprintf(msg + n, len - n, "row: %d ", row_extended);
> + }
Both those hunks contain duplicated code which kinda wants to be an
inline function in cper.h which returns row_extended and gets called by
both sites. And then the call site can look very simple:
if (mem_err->validation_bits & CPER_MEM_VALID_ROW)
row = mem_err->row;
/* add row extension */
row |= cper_get_mem_extension();
p += sprintf(p, "row:%d ", row);
with
static inline u32 cper_get_mem_extension(void)
{
if (!(mem_err->validation_bits & CPER_MEM_VALID_ROW_EXT))
return 0;
return (mem_err->extended & CPER_MEM_EXT_ROW_MASK) << CPER_MEM_EXT_ROW_SHIFT;
}
Something along those lines...
Thx.
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette
Yes, I am working on a resubmit. The updated patch will be resubmitted in a series of patches that include other updates to the cper memory record.
Thanks,
--Alex Kluver
-----Original Message-----
From: Borislav Petkov <[email protected]>
Sent: Saturday, August 15, 2020 4:34 AM
To: Kluver, Alex <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; Anderson, Russ <[email protected]>; Sivanich, Dimitri <[email protected]>
Subject: Re: [PATCH] edac,ghes,cper: Add Row Extension to Memory Error Record
On Mon, Jul 27, 2020 at 01:14:45PM -0500, Alex Kluver wrote:
> Memory errors could be printed with incorrect row values since the
> DIMM size has outgrown the 16 bit row field in the CPER structure.
> UEFI Specification Version 2.8 has increased the size of row by
> allowing it to use the first 2 bits from a previously reserved space within the structure.
>
> When needed, add the extension bits to the row value printed.
>
> Based on UEFI 2.8 Table 299. Memory Error Record
>
> Tested-by: Russ Anderson <[email protected]>
> Signed-off-by: Alex Kluver <[email protected]>
> ---
> drivers/edac/ghes_edac.c | 10 ++++++++--
> drivers/firmware/efi/cper.c | 11 +++++++++--
> include/linux/cper.h | 9 +++++++--
> 3 files changed, 24 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/edac/ghes_edac.c b/drivers/edac/ghes_edac.c index
> cb3dab56a875..cfa3156300f5 100644
> --- a/drivers/edac/ghes_edac.c
> +++ b/drivers/edac/ghes_edac.c
> @@ -337,8 +337,14 @@ void ghes_edac_report_mem_error(int sev, struct cper_sec_mem_err *mem_err)
> p += sprintf(p, "rank:%d ", mem_err->rank);
> if (mem_err->validation_bits & CPER_MEM_VALID_BANK)
> p += sprintf(p, "bank:%d ", mem_err->bank);
> - if (mem_err->validation_bits & CPER_MEM_VALID_ROW)
> - p += sprintf(p, "row:%d ", mem_err->row);
> + if (mem_err->validation_bits & (CPER_MEM_VALID_ROW | CPER_MEM_VALID_ROW_EXT)) {
> + u32 row_extended = 0;
> + if (mem_err->validation_bits & CPER_MEM_VALID_ROW_EXT)
> + row_extended = (mem_err->extended & CPER_MEM_EXT_ROW_MASK)
> + <<CPER_MEM_EXT_ROW_SHIFT;
> + row_extended |= mem_err->row;
> + p += sprintf(p, "row:%d ", row_extended);
> + }
> if (mem_err->validation_bits & CPER_MEM_VALID_COLUMN)
> p += sprintf(p, "col:%d ", mem_err->column);
> if (mem_err->validation_bits & CPER_MEM_VALID_BIT_POSITION) diff
> --git a/drivers/firmware/efi/cper.c b/drivers/firmware/efi/cper.c
> index f564e15fbc7e..5faaf6ecd659 100644
> --- a/drivers/firmware/efi/cper.c
> +++ b/drivers/firmware/efi/cper.c
> @@ -234,8 +234,14 @@ static int cper_mem_err_location(struct cper_mem_err_compact *mem, char *msg)
> n += scnprintf(msg + n, len - n, "bank: %d ", mem->bank);
> if (mem->validation_bits & CPER_MEM_VALID_DEVICE)
> n += scnprintf(msg + n, len - n, "device: %d ", mem->device);
> - if (mem->validation_bits & CPER_MEM_VALID_ROW)
> - n += scnprintf(msg + n, len - n, "row: %d ", mem->row);
> + if (mem->validation_bits & (CPER_MEM_VALID_ROW | CPER_MEM_VALID_ROW_EXT)) {
> + u32 row_extended = 0;
> + if (mem->validation_bits & CPER_MEM_VALID_ROW_EXT)
> + row_extended = (mem->extended & CPER_MEM_EXT_ROW_MASK)
> + <<CPER_MEM_EXT_ROW_SHIFT;
This is not very readable.
> + row_extended |= mem->row;
> + n += scnprintf(msg + n, len - n, "row: %d ", row_extended);
> + }
Both those hunks contain duplicated code which kinda wants to be an inline function in cper.h which returns row_extended and gets called by both sites. And then the call site can look very simple:
if (mem_err->validation_bits & CPER_MEM_VALID_ROW)
row = mem_err->row;
/* add row extension */
row |= cper_get_mem_extension();
p += sprintf(p, "row:%d ", row);
with
static inline u32 cper_get_mem_extension(void) {
if (!(mem_err->validation_bits & CPER_MEM_VALID_ROW_EXT))
return 0;
return (mem_err->extended & CPER_MEM_EXT_ROW_MASK) << CPER_MEM_EXT_ROW_SHIFT; }
Something along those lines...
Thx.
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette