Since the DQ-bus width is now available in the driver it can be utilized
to calculate the exact bit-position corrected by the ECC engine for any
Synopsys memory controller setup. A corrected error syndrome is exposed by
the ECCSTAT.corrected_bit_num field. A particular erroneous bit position
is described in the lookup table [1] which also contains a dependency
between the field value and the DQ-bus widths. The syndrome values table
basically represents a standard lookup table for the Hamming
(64,8)/(32,7)/(16,6) codes (the error-correcting bits placed at the
power-of-two positions) except that the zero value means error in the
ecc[0] bit.
So using the offsets from that table introduce a new inline method
snps_get_bitpos() which would provide the actual CE bit-position. The
method will be called if a corrected error is detected.
[1] DesignWare® Cores Enhanced Universal DDR Memory Controller (uMCTL2)
Databook, Version 3.91a, October 2020, p.426-427
Signed-off-by: Serge Semin <[email protected]>
---
drivers/edac/synopsys_edac.c | 28 +++++++++++++++++++++++++++-
1 file changed, 27 insertions(+), 1 deletion(-)
diff --git a/drivers/edac/synopsys_edac.c b/drivers/edac/synopsys_edac.c
index e10778cead63..e08cb30b7a7d 100644
--- a/drivers/edac/synopsys_edac.c
+++ b/drivers/edac/synopsys_edac.c
@@ -10,6 +10,7 @@
#include <linux/bits.h>
#include <linux/edac.h>
#include <linux/fs.h>
+#include <linux/log2.h>
#include <linux/module.h>
#include <linux/platform_device.h>
#include <linux/seq_file.h>
@@ -301,6 +302,7 @@ struct snps_ddrc_info {
* @col: Column number.
* @bank: Bank number.
* @bankgrp: Bank group number.
+ * @syndrome: Error syndrome.
* @bitpos: Bit position.
* @data: Data causing the error.
*/
@@ -309,6 +311,7 @@ struct snps_ecc_error_info {
u32 col;
u32 bank;
u32 bankgrp;
+ u32 syndrome;
u32 bitpos;
u32 data;
};
@@ -359,6 +362,27 @@ struct snps_edac_priv {
#endif
};
+/**
+ * snps_get_bitpos - Get DQ-bus corrected bit position.
+ * @syndrome: Error syndrome.
+ * @dq_width: Controller DQ-bus width.
+ *
+ * Return: actual corrected DQ-bus bit position starting from 0.
+ */
+static inline u32 snps_get_bitpos(u32 syndrome, enum snps_dq_width dq_width)
+{
+ /* ecc[0] bit */
+ if (syndrome == 0)
+ return BITS_PER_BYTE << dq_width;
+
+ /* ecc[1:x] bit */
+ if (is_power_of_2(syndrome))
+ return (BITS_PER_BYTE << dq_width) + ilog2(syndrome) + 1;
+
+ /* data[0:y] bit */
+ return syndrome - ilog2(syndrome) - 2;
+}
+
/**
* snps_get_error_info - Get the current ECC error info.
* @priv: DDR memory controller private instance data.
@@ -379,7 +403,7 @@ static int snps_get_error_info(struct snps_edac_priv *priv)
if (!regval)
return 1;
- p->ceinfo.bitpos = FIELD_GET(ECC_STAT_BITNUM_MASK, regval);
+ p->ceinfo.syndrome = FIELD_GET(ECC_STAT_BITNUM_MASK, regval);
regval = readl(base + ECC_ERRCNT_OFST);
p->ce_cnt = FIELD_GET(ECC_ERRCNT_CECNT_MASK, regval);
@@ -387,6 +411,8 @@ static int snps_get_error_info(struct snps_edac_priv *priv)
if (!p->ce_cnt)
goto ue_err;
+ p->ceinfo.bitpos = snps_get_bitpos(p->ceinfo.syndrome, priv->info.dq_width);
+
regval = readl(base + ECC_CEADDR0_OFST);
p->ceinfo.row = FIELD_GET(ECC_CEADDR0_ROW_MASK, regval);
--
2.41.0