On 09/17/2012 07:39:35 PM, Scott Wood wrote:
On Mon, Aug 13, 2012 at 10:49:42PM +0200, Benoît Thébaudeau wrote:
On the NFC IP 1.1, the 32-bit ecc_status_result value comes from 2 consecutive 16-bit registers. This patch reads all the fields of
this value,
which makes a difference for 4-kiB NF pages.
Signed-off-by: Benoît Thébaudeau benoit.thebaudeau@advansee.com Cc: Scott Wood scottwood@freescale.com Cc: Stefano Babic sbabic@denx.de
.../drivers/mtd/nand/mxc_nand.c | 2 +- .../include/fsl_nfc.h | 3 +-- .../nand_spl/nand_boot_fsl_nfc.c | 4 ++++ 3 files changed, 6 insertions(+), 3 deletions(-)
diff --git u-boot-4d3c95f.orig/drivers/mtd/nand/mxc_nand.c
u-boot-4d3c95f/drivers/mtd/nand/mxc_nand.c
index 62d8c6b..4564b25 100644 --- u-boot-4d3c95f.orig/drivers/mtd/nand/mxc_nand.c +++ u-boot-4d3c95f/drivers/mtd/nand/mxc_nand.c @@ -653,7 +653,7 @@ static int mxc_nand_correct_data(struct
mtd_info *mtd, u_char *dat,
{ struct nand_chip *nand_chip = mtd->priv; struct mxc_nand_host *host = nand_chip->priv;
- uint16_t ecc_status = readw(&host->regs->ecc_status_result);
- uint32_t ecc_status = readl(&host->regs->ecc_status_result); int subpages = mtd->writesize / nand_chip->subpagesize; int pg2blk_shift = nand_chip->phys_erase_shift - nand_chip->page_shift;
diff --git u-boot-4d3c95f.orig/include/fsl_nfc.h
u-boot-4d3c95f/include/fsl_nfc.h
index 1c16c71..c248cef 100644 --- u-boot-4d3c95f.orig/include/fsl_nfc.h +++ u-boot-4d3c95f/include/fsl_nfc.h @@ -94,8 +94,7 @@ struct fsl_nfc_regs { u16 flash_addr; u16 flash_cmd; u16 config;
- u16 ecc_status_result;
- u16 ecc_status_result2;
- u32 ecc_status_result; u16 spare_area_size; u16 wrprot; u16 reserved3[2];
diff --git u-boot-4d3c95f.orig/nand_spl/nand_boot_fsl_nfc.c
u-boot-4d3c95f/nand_spl/nand_boot_fsl_nfc.c
index 842943c..4c5a7fe 100644 --- u-boot-4d3c95f.orig/nand_spl/nand_boot_fsl_nfc.c +++ u-boot-4d3c95f/nand_spl/nand_boot_fsl_nfc.c @@ -139,7 +139,11 @@ static void nfc_nand_data_output(void)
static int nfc_nand_check_ecc(void) { +#if defined(MXC_NFC_V1) return readw(&nfc->ecc_status_result); +#elif defined(MXC_NFC_V1_1)
- return readl(&nfc->ecc_status_result);
+#endif }
On MXC_NFC_V1 you're using readw() on a u32 -- that's not OK, even if the ARM accessors let you get away with it. Use a union or #ifdef the struct definition, or just always readl() if you can get away with it, or do two readw()s on V1_1.
Never mind, didn't notice that the struct definition is under #ifdef MXC_NFC_V1_1.
The ARM accessors still should be fixed to do type checking...
-Scott