[PATCH 1/6] mtd: nand-imx: Fix v1 controller ECC code
Sascha Hauer
s.hauer at pengutronix.de
Thu Mar 17 04:08:34 PDT 2016
The driver returns wrong values for the ECC correction. For 2k pages
the controller reads and corrects in chunks of 512 bytes. The ECC status
register values are overwritten with each each new chunk read, so after
reading for chunks the .correct callback wil only return the ECC errors
for the last 512 byte chunk. ECC errors in the other three chunks remain
undetected.
Fix this by accumulating the ECC status while reading the chunks and
return the accumulated value in the .correct callback. Also return
-EBADMSG for a bad message and not -1.
Signed-off-by: Sascha Hauer <s.hauer at pengutronix.de>
---
drivers/mtd/nand/nand_imx.c | 33 +++++++++++++++++++++------------
1 file changed, 21 insertions(+), 12 deletions(-)
diff --git a/drivers/mtd/nand/nand_imx.c b/drivers/mtd/nand/nand_imx.c
index 6ea82a1..f0b4afb 100644
--- a/drivers/mtd/nand/nand_imx.c
+++ b/drivers/mtd/nand/nand_imx.c
@@ -100,6 +100,7 @@ struct imx_nand_host {
unsigned int buf_start;
int spare_len;
int eccsize;
+ int eccstatus_v1;
int hw_ecc;
int data_width;
@@ -335,12 +336,17 @@ static void send_page_v1_v2(struct imx_nand_host *host,
{
int bufs, i;
+ host->eccstatus_v1 = 0;
+
if (nfc_is_v1() && host->pagesize_2k)
bufs = 4;
else
bufs = 1;
for (i = 0; i < bufs; i++) {
+ u16 status;
+ int errors;
+
/* NANDFC buffer 0 is used for page read/write */
writew(i, host->regs + NFC_V1_V2_BUF_ADDR);
@@ -348,6 +354,14 @@ static void send_page_v1_v2(struct imx_nand_host *host,
/* Wait for operation to complete */
wait_op_done(host);
+
+ status = readw(host->regs + NFC_V1_ECC_STATUS_RESULT);
+ errors = max(status & 0x3, status >> 2);
+
+ if (errors == 1 && host->eccstatus_v1 >= 0)
+ host->eccstatus_v1++;
+ if (errors == 2)
+ host->eccstatus_v1 = -EBADMSG;
}
}
@@ -489,20 +503,15 @@ static int imx_nand_correct_data_v1(struct mtd_info *mtd, u_char * dat,
struct nand_chip *nand_chip = mtd->priv;
struct imx_nand_host *host = nand_chip->priv;
- /*
- * 1-Bit errors are automatically corrected in HW. No need for
- * additional correction. 2-Bit errors cannot be corrected by
- * HW ECC, so we need to return failure
- */
- u16 ecc_status = readw(host->regs + NFC_V1_ECC_STATUS_RESULT);
+ if (host->eccstatus_v1 < 0)
+ return host->eccstatus_v1;
- if (((ecc_status & 0x3) == 2) || ((ecc_status >> 2) == 2)) {
- MTD_DEBUG(MTD_DEBUG_LEVEL0,
- "MXC_NAND: HWECC uncorrectable 2-bit ECC error\n");
- return -1;
- }
+ mtd->ecc_stats.corrected += host->eccstatus_v1;
- return 0;
+ if (host->eccstatus_v1 > 0)
+ return 1;
+ else
+ return 0;
}
static int imx_nand_correct_data_v2_v3(struct mtd_info *mtd, u_char *dat,
--
2.7.0
More information about the barebox
mailing list