[PATCH] omap3: nand: bch ecc support added
Sukumar Ghorai
s-ghorai at ti.com
Thu Jan 20 05:18:00 EST 2011
bch error correction (t=4 and t=8) for 512 bytes support added.
Tested in omap-3630 es-1.1 silicon.
Need to select the bch-ecc from board file. E.g.
arch/arm/mach-omap2/board-flash.c: board_nand_init()
board_nand_data.ecc_opt = OMAP_ECC_BCH4_CODE_HW
This patch has dependency on -
http://www.mail-archive.com/linux-omap@vger.kernel.org/msg42658.html
Signed-off-by: Sukumar Ghorai <s-ghorai at ti.com>
---
arch/arm/mach-omap2/gpmc.c | 126 ++++++++---
arch/arm/plat-omap/include/plat/gpmc.h | 6 +-
drivers/mtd/nand/Makefile | 1 +
drivers/mtd/nand/omap2.c | 119 ++++++++--
drivers/mtd/nand/omap_bch_decoder.c | 393 ++++++++++++++++++++++++++++++++
5 files changed, 583 insertions(+), 62 deletions(-)
create mode 100644 drivers/mtd/nand/omap_bch_decoder.c
diff --git a/arch/arm/mach-omap2/gpmc.c b/arch/arm/mach-omap2/gpmc.c
index 29c9732..91cfdca 100644
--- a/arch/arm/mach-omap2/gpmc.c
+++ b/arch/arm/mach-omap2/gpmc.c
@@ -48,6 +48,7 @@
#define GPMC_ECC_CONTROL 0x1f8
#define GPMC_ECC_SIZE_CONFIG 0x1fc
#define GPMC_ECC1_RESULT 0x200
+#define GPMC_ECC_BCH_RESULT_0 0x240
#define GPMC_CS0_OFFSET 0x60
#define GPMC_CS_SIZE 0x30
@@ -94,7 +95,6 @@ static struct resource gpmc_mem_root;
static struct resource gpmc_cs_mem[GPMC_CS_NUM];
static DEFINE_SPINLOCK(gpmc_mem_lock);
static unsigned int gpmc_cs_map; /* flag for cs which are initialized */
-static int gpmc_ecc_used = -EINVAL; /* cs using ecc engine */
static void __iomem *gpmc_base;
@@ -832,52 +832,77 @@ void omap3_gpmc_restore_context(void)
/**
* gpmc_enable_hwecc - enable hardware ecc functionality
+ * @ecc_type: ecc type e.g. Hamming, BCH
* @cs: chip select number
* @mode: read/write mode
* @dev_width: device bus width(1 for x16, 0 for x8)
* @ecc_size: bytes for which ECC will be generated
*/
-int gpmc_enable_hwecc(int cs, int mode, int dev_width, int ecc_size)
+int gpmc_enable_hwecc(int ecc_type, int cs, int mode,
+ int dev_width, int ecc_size)
{
- unsigned int val;
-
- /* check if ecc module is in used */
- if (gpmc_ecc_used != -EINVAL)
- return -EINVAL;
-
- gpmc_ecc_used = cs;
-
- /* clear ecc and enable bits */
- val = ((0x00000001<<8) | 0x00000001);
- gpmc_write_reg(GPMC_ECC_CONTROL, val);
-
- /* program ecc and result sizes */
- val = ((((ecc_size >> 1) - 1) << 22) | (0x0000000F));
- gpmc_write_reg(GPMC_ECC_SIZE_CONFIG, val);
+ unsigned int bch_mod = 0, bch_wrapmode = 0, eccsize1 = 0, eccsize0 = 0;
+ unsigned int ecc_conf_val = 0, ecc_size_conf_val = 0;
switch (mode) {
case GPMC_ECC_READ:
- gpmc_write_reg(GPMC_ECC_CONTROL, 0x101);
+ if (ecc_type == OMAP_ECC_BCH4_CODE_HW) {
+ eccsize1 = 0xD; eccsize0 = 0x48;
+ bch_mod = 0;
+ bch_wrapmode = 0x09;
+ } else if (ecc_type == OMAP_ECC_BCH8_CODE_HW) {
+ eccsize1 = 0x1A; eccsize0 = 0x18;
+ bch_mod = 1;
+ bch_wrapmode = 0x04;
+ } else
+ eccsize1 = ((ecc_size >> 1) - 1) << 22;
break;
+
case GPMC_ECC_READSYN:
- gpmc_write_reg(GPMC_ECC_CONTROL, 0x100);
break;
+
case GPMC_ECC_WRITE:
- gpmc_write_reg(GPMC_ECC_CONTROL, 0x101);
+ if (ecc_type == OMAP_ECC_BCH4_CODE_HW) {
+ eccsize1 = 0x20; eccsize0 = 0x00;
+ bch_mod = 0;
+ bch_wrapmode = 0x06;
+ } else if (ecc_type == OMAP_ECC_BCH8_CODE_HW) {
+ eccsize1 = 0x20; eccsize0 = 0x00;
+ bch_mod = 1;
+ bch_wrapmode = 0x06;
+ } else
+ eccsize1 = ((ecc_size >> 1) - 1) << 22;
break;
+
default:
printk(KERN_INFO "Error: Unrecognized Mode[%d]!\n", mode);
break;
}
- /* (ECC 16 or 8 bit col) | ( CS ) | ECC Enable */
- val = (dev_width << 7) | (cs << 1) | (0x1);
- gpmc_write_reg(GPMC_ECC_CONFIG, val);
+ /* clear ecc and enable bits */
+ if ((ecc_type == OMAP_ECC_BCH4_CODE_HW) ||
+ (ecc_type == OMAP_ECC_BCH8_CODE_HW)) {
+ gpmc_write_reg(GPMC_ECC_CONTROL, 0x00000001);
+ ecc_size_conf_val = (eccsize1 << 22) | (eccsize0 << 12);
+ ecc_conf_val = ((0x01 << 16) | (bch_mod << 12)
+ | (bch_wrapmode << 8) | (dev_width << 7)
+ | (0x03 << 4) | (cs << 1) | (0x1));
+ } else {
+ gpmc_write_reg(GPMC_ECC_CONTROL, 0x00000101);
+ ecc_size_conf_val = (eccsize1 << 22) | 0x0000000F;
+ ecc_conf_val = (dev_width << 7) | (cs << 1) | (0x1);
+ }
+
+ gpmc_write_reg(GPMC_ECC_SIZE_CONFIG, ecc_size_conf_val);
+ gpmc_write_reg(GPMC_ECC_CONFIG, ecc_conf_val);
+ gpmc_write_reg(GPMC_ECC_CONTROL, 0x00000101);
+
return 0;
}
/**
* gpmc_calculate_ecc - generate non-inverted ecc bytes
+ * @ecc_type: ecc type e.g. Hamming, BCH
* @cs: chip select number
* @dat: data pointer over which ecc is computed
* @ecc_code: ecc code buffer
@@ -888,20 +913,51 @@ int gpmc_enable_hwecc(int cs, int mode, int dev_width, int ecc_size)
* an erased page will produce an ECC mismatch between generated and read
* ECC bytes that has to be dealt with separately.
*/
-int gpmc_calculate_ecc(int cs, const u_char *dat, u_char *ecc_code)
+int gpmc_calculate_ecc(int ecc_type, int cs,
+ const u_char *dat, u_char *ecc_code)
{
- unsigned int val = 0x0;
-
- if (gpmc_ecc_used != cs)
- return -EINVAL;
+ unsigned int reg;
+ unsigned int val1 = 0x0, val2 = 0x0;
+ unsigned int val3 = 0x0, val4 = 0x0;
+ int i;
- /* read ecc result */
- val = gpmc_read_reg(GPMC_ECC1_RESULT);
- *ecc_code++ = val; /* P128e, ..., P1e */
- *ecc_code++ = val >> 16; /* P128o, ..., P1o */
- /* P2048o, P1024o, P512o, P256o, P2048e, P1024e, P512e, P256e */
- *ecc_code++ = ((val >> 8) & 0x0f) | ((val >> 20) & 0xf0);
+ if ((ecc_type == OMAP_ECC_BCH4_CODE_HW) ||
+ (ecc_type == OMAP_ECC_BCH8_CODE_HW)) {
+ for (i = 0; i < 4; i++) {
+ /*
+ * Reading HW ECC_BCH_Results
+ * 0x240-0x24C, 0x250-0x25C, 0x260-0x26C, 0x270-0x27C
+ */
+ reg = GPMC_ECC_BCH_RESULT_0 + (0x10 * i);
+ val1 = gpmc_read_reg(reg);
+ val2 = gpmc_read_reg(reg + 4);
+ if (ecc_type == OMAP_ECC_BCH8_CODE_HW) {
+ val3 = gpmc_read_reg(reg + 8);
+ val4 = gpmc_read_reg(reg + 12);
+
+ *ecc_code++ = (val4 & 0xFF);
+ *ecc_code++ = ((val3 >> 24) & 0xFF);
+ *ecc_code++ = ((val3 >> 16) & 0xFF);
+ *ecc_code++ = ((val3 >> 8) & 0xFF);
+ *ecc_code++ = (val3 & 0xFF);
+ *ecc_code++ = ((val2 >> 24) & 0xFF);
+ }
+ *ecc_code++ = ((val2 >> 16) & 0xFF);
+ *ecc_code++ = ((val2 >> 8) & 0xFF);
+ *ecc_code++ = (val2 & 0xFF);
+ *ecc_code++ = ((val1 >> 24) & 0xFF);
+ *ecc_code++ = ((val1 >> 16) & 0xFF);
+ *ecc_code++ = ((val1 >> 8) & 0xFF);
+ *ecc_code++ = (val1 & 0xFF);
+ }
+ } else {
+ /* read ecc result */
+ val1 = gpmc_read_reg(GPMC_ECC1_RESULT);
+ *ecc_code++ = val1; /* P128e, ..., P1e */
+ *ecc_code++ = val1 >> 16; /* P128o, ..., P1o */
+ /* P2048o, P1024o, P512o, P256o, P2048e, P1024e, P512e, P256e */
+ *ecc_code++ = ((val1 >> 8) & 0x0f) | ((val1 >> 20) & 0xf0);
+ }
- gpmc_ecc_used = -EINVAL;
return 0;
}
diff --git a/arch/arm/plat-omap/include/plat/gpmc.h b/arch/arm/plat-omap/include/plat/gpmc.h
index 49aea09..838c185 100644
--- a/arch/arm/plat-omap/include/plat/gpmc.h
+++ b/arch/arm/plat-omap/include/plat/gpmc.h
@@ -92,6 +92,8 @@ enum omap_ecc {
OMAP_ECC_HAMMING_CODE_HW, /* gpmc to detect the error */
/* 1-bit ecc: stored at begining of spare area as romcode */
OMAP_ECC_HAMMING_CODE_HW_ROMCODE, /* gpmc method & romcode layout */
+ OMAP_ECC_BCH4_CODE_HW, /* gpmc bch detection & s/w method correction */
+ OMAP_ECC_BCH8_CODE_HW, /* gpmc bch detection & s/w method correction */
};
/*
@@ -156,6 +158,6 @@ extern int gpmc_cs_configure(int cs, int cmd, int wval);
extern int gpmc_nand_read(int cs, int cmd);
extern int gpmc_nand_write(int cs, int cmd, int wval);
-int gpmc_enable_hwecc(int cs, int mode, int dev_width, int ecc_size);
-int gpmc_calculate_ecc(int cs, const u_char *dat, u_char *ecc_code);
+int gpmc_enable_hwecc(int ecc, int cs, int mode, int dev_width, int ecc_size);
+int gpmc_calculate_ecc(int ecc, int cs, const u_char *dat, u_char *ecc_code);
#endif
diff --git a/drivers/mtd/nand/Makefile b/drivers/mtd/nand/Makefile
index 8ad6fae..ae02711 100644
--- a/drivers/mtd/nand/Makefile
+++ b/drivers/mtd/nand/Makefile
@@ -29,6 +29,7 @@ obj-$(CONFIG_MTD_NAND_NDFC) += ndfc.o
obj-$(CONFIG_MTD_NAND_ATMEL) += atmel_nand.o
obj-$(CONFIG_MTD_NAND_GPIO) += gpio.o
obj-$(CONFIG_MTD_NAND_OMAP2) += omap2.o
+obj-$(CONFIG_MTD_NAND_OMAP2) += omap_bch_decoder.o
obj-$(CONFIG_MTD_NAND_CM_X270) += cmx270_nand.o
obj-$(CONFIG_MTD_NAND_PXA3xx) += pxa3xx_nand.o
obj-$(CONFIG_MTD_NAND_TMIO) += tmio_nand.o
diff --git a/drivers/mtd/nand/omap2.c b/drivers/mtd/nand/omap2.c
index 4e33972..14c7dfe 100644
--- a/drivers/mtd/nand/omap2.c
+++ b/drivers/mtd/nand/omap2.c
@@ -98,6 +98,8 @@
static const char *part_probes[] = { "cmdlinepart", NULL };
#endif
+int decode_bch(int select_4_8, unsigned char *ecc, unsigned int *err_loc);
+
/* oob info generated runtime depending on ecc algorithm and layout selected */
static struct nand_ecclayout omap_oobinfo;
/* Define some generic bad / good block scan pattern which are used
@@ -130,7 +132,8 @@ struct omap_nand_info {
OMAP_NAND_IO_WRITE, /* write */
} iomode;
u_char *buf;
- int buf_len;
+ int buf_len;
+ int ecc_opt;
};
/**
@@ -529,7 +532,6 @@ static void omap_read_buf_irq_pref(struct mtd_info *mtd, u_char *buf, int len)
struct omap_nand_info *info = container_of(mtd,
struct omap_nand_info, mtd);
int ret = 0;
-
if (len <= mtd->oobsize) {
omap_read_buf_pref(mtd, buf, len);
return;
@@ -803,6 +805,8 @@ static int omap_correct_data(struct mtd_info *mtd, u_char *dat,
struct omap_nand_info *info = container_of(mtd, struct omap_nand_info,
mtd);
int blockCnt = 0, i = 0, ret = 0;
+ int j, eccsize, eccflag, count;
+ unsigned int err_loc[8];
/* Ex NAND_ECC_HW12_2048 */
if ((info->nand.ecc.mode == NAND_ECC_HW) &&
@@ -811,16 +815,57 @@ static int omap_correct_data(struct mtd_info *mtd, u_char *dat,
else
blockCnt = 1;
- for (i = 0; i < blockCnt; i++) {
- if (memcmp(read_ecc, calc_ecc, 3) != 0) {
- ret = omap_compare_ecc(read_ecc, calc_ecc, dat);
- if (ret < 0)
- return ret;
+ switch (info->ecc_opt) {
+ case OMAP_ECC_HAMMING_CODE_HW:
+ case OMAP_ECC_HAMMING_CODE_HW_ROMCODE:
+ for (i = 0; i < blockCnt; i++) {
+ if (memcmp(read_ecc, calc_ecc, 3) != 0) {
+ ret = omap_compare_ecc(read_ecc, calc_ecc, dat);
+ if (ret < 0)
+ return ret;
+ }
+ read_ecc += 3;
+ calc_ecc += 3;
+ dat += 512;
}
- read_ecc += 3;
- calc_ecc += 3;
- dat += 512;
+ break;
+
+ case OMAP_ECC_BCH4_CODE_HW:
+ eccsize = 7;
+ gpmc_calculate_ecc(info->ecc_opt, info->gpmc_cs, dat, calc_ecc);
+ for (i = 0; i < blockCnt; i++) {
+ /* check if any ecc error */
+ eccflag = 0;
+ for (j = 0; (j < eccsize) && (eccflag == 0); j++)
+ if (calc_ecc[j] != 0)
+ eccflag = 1;
+
+ if (eccflag == 1) {
+ eccflag = 0;
+ for (j = 0; (j < eccsize) &&
+ (eccflag == 0); j++)
+ if (read_ecc[j] != 0xFF)
+ eccflag = 1;
+ }
+
+ count = 0;
+ if (eccflag == 1)
+ count = decode_bch(0, calc_ecc, err_loc);
+
+ for (j = 0; j < count; j++) {
+ if (err_loc[j] < 4096)
+ dat[err_loc[j] >> 3] ^=
+ 1 << (err_loc[j] & 7);
+ /* else, not interested to correct ecc */
+ }
+
+ calc_ecc = calc_ecc + eccsize;
+ read_ecc = read_ecc + eccsize;
+ dat += 512;
+ }
+ break;
}
+
return 0;
}
@@ -841,7 +886,7 @@ static int omap_calculate_ecc(struct mtd_info *mtd, const u_char *dat,
{
struct omap_nand_info *info = container_of(mtd, struct omap_nand_info,
mtd);
- return gpmc_calculate_ecc(info->gpmc_cs, dat, ecc_code);
+ return gpmc_calculate_ecc(info->ecc_opt, info->gpmc_cs, dat, ecc_code);
}
/**
@@ -856,7 +901,8 @@ static void omap_enable_hwecc(struct mtd_info *mtd, int mode)
struct nand_chip *chip = mtd->priv;
unsigned int dev_width = (chip->options & NAND_BUSWIDTH_16) ? 1 : 0;
- gpmc_enable_hwecc(info->gpmc_cs, mode, dev_width, info->nand.ecc.size);
+ gpmc_enable_hwecc(info->ecc_opt, info->gpmc_cs, mode,
+ dev_width, info->nand.ecc.size);
}
/**
@@ -953,6 +999,7 @@ static int __devinit omap_nand_probe(struct platform_device *pdev)
info->mtd.priv = &info->nand;
info->mtd.name = dev_name(&pdev->dev);
info->mtd.owner = THIS_MODULE;
+ info->ecc_opt = pdata->ecc_opt;
info->nand.options = pdata->devsize;
info->nand.options |= NAND_SKIP_BBTSCAN;
@@ -991,7 +1038,6 @@ static int __devinit omap_nand_probe(struct platform_device *pdev)
info->nand.waitfunc = omap_wait;
info->nand.chip_delay = 50;
}
-
switch (pdata->xfer_type) {
case NAND_OMAP_PREFETCH_POLLED:
info->nand.read_buf = omap_read_buf_pref;
@@ -1052,10 +1098,17 @@ static int __devinit omap_nand_probe(struct platform_device *pdev)
/* selsect the ecc type */
if (pdata->ecc_opt == OMAP_ECC_HAMMING_CODE_DEFAULT)
info->nand.ecc.mode = NAND_ECC_SOFT;
- else if ((pdata->ecc_opt == OMAP_ECC_HAMMING_CODE_HW) ||
- (pdata->ecc_opt == OMAP_ECC_HAMMING_CODE_HW_ROMCODE)) {
- info->nand.ecc.bytes = 3;
- info->nand.ecc.size = 512;
+ else {
+ if (pdata->ecc_opt == OMAP_ECC_BCH4_CODE_HW) {
+ info->nand.ecc.bytes = 4*7;
+ info->nand.ecc.size = 4*512;
+ } else if (pdata->ecc_opt == OMAP_ECC_BCH8_CODE_HW) {
+ info->nand.ecc.bytes = 13;
+ info->nand.ecc.size = 4*512;
+ } else {
+ info->nand.ecc.bytes = 3;
+ info->nand.ecc.size = 512;
+ }
info->nand.ecc.calculate = omap_calculate_ecc;
info->nand.ecc.hwctl = omap_enable_hwecc;
info->nand.ecc.correct = omap_correct_data;
@@ -1073,8 +1126,8 @@ static int __devinit omap_nand_probe(struct platform_device *pdev)
}
}
- /* rom code layout */
- if (pdata->ecc_opt == OMAP_ECC_HAMMING_CODE_HW_ROMCODE) {
+ /* select ecc lyout */
+ if (info->nand.ecc.mode != NAND_ECC_SOFT) {
if (info->nand.options & NAND_BUSWIDTH_16)
offset = 2;
@@ -1082,15 +1135,31 @@ static int __devinit omap_nand_probe(struct platform_device *pdev)
offset = 1;
info->nand.badblock_pattern = &bb_descrip_flashbased;
}
- omap_oobinfo.eccbytes = 3 * (info->mtd.oobsize/16);
- for (i = 0; i < omap_oobinfo.eccbytes; i++)
- omap_oobinfo.eccpos[i] = i+offset;
- omap_oobinfo.oobfree->offset = offset + omap_oobinfo.eccbytes;
- omap_oobinfo.oobfree->length = info->mtd.oobsize -
- (offset + omap_oobinfo.eccbytes);
+ if (info->mtd.oobsize == 64)
+ omap_oobinfo.eccbytes = info->nand.ecc.bytes *
+ 2048/info->nand.ecc.size;
+ else
+ omap_oobinfo.eccbytes = info->nand.ecc.bytes;
+
+ if (pdata->ecc_opt == OMAP_ECC_HAMMING_CODE_HW_ROMCODE) {
+ for (i = 0; i < omap_oobinfo.eccbytes; i++)
+ omap_oobinfo.eccpos[i] = i + offset;
+ omap_oobinfo.oobfree->offset =
+ offset + omap_oobinfo.eccbytes;
+ omap_oobinfo.oobfree->length = info->mtd.oobsize -
+ offset - omap_oobinfo.eccbytes;
+ } else {
+ omap_oobinfo.oobfree->offset = offset;
+ omap_oobinfo.oobfree->length = info->mtd.oobsize -
+ offset - omap_oobinfo.eccbytes;
+ offset = info->mtd.oobsize - omap_oobinfo.eccbytes;
+ for (i = 0; i < omap_oobinfo.eccbytes; i++)
+ omap_oobinfo.eccpos[i] = i + offset;
+ }
info->nand.ecc.layout = &omap_oobinfo;
+
}
#ifdef CONFIG_MTD_PARTITIONS
diff --git a/drivers/mtd/nand/omap_bch_decoder.c b/drivers/mtd/nand/omap_bch_decoder.c
new file mode 100644
index 0000000..da42bda
--- /dev/null
+++ b/drivers/mtd/nand/omap_bch_decoder.c
@@ -0,0 +1,393 @@
+/*
+ * drivers/mtd/nand/omap_omap_bch_decoder.c
+ *
+ * Whole BCH ECC Decoder (Post hardware generated syndrome decoding)
+ *
+ * Copyright (c) 2007 Texas Instruments
+ *
+ * Author: Sukumar Ghorai <s-ghorai at ti.com
+ * Michael Fillinger <m-fillinger at ti.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#undef DEBUG
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+
+#define mm 13
+#define kk_shorten 4096
+#define nn 8191 /* Length of codeword, n = 2**mm - 1 */
+
+#define PPP 0x201B /* Primary Polynomial : x^13 + x^4 + x^3 + x + 1 */
+#define P 0x001B /* With omitted x^13 */
+#define POLY 12 /* degree of the primary Polynomial less one */
+
+/**
+ * mpy_mod_gf - GALOIS field multiplier
+ * Input : A(x), B(x)
+ * Output : A(x)*B(x) mod P(x)
+ */
+static unsigned int mpy_mod_gf(unsigned int a, unsigned int b)
+{
+ unsigned int R = 0;
+ unsigned int R1 = 0;
+ unsigned int k = 0;
+
+ for (k = 0; k < mm; k++) {
+
+ R = (R << 1) & 0x1FFE;
+ if (R1 == 1)
+ R ^= P;
+
+ if (((a >> (POLY - k)) & 1) == 1)
+ R ^= b;
+
+ if (k < POLY)
+ R1 = (R >> POLY) & 1;
+ }
+ return R;
+}
+
+/**
+ * chien - CHIEN search
+ *
+ * @location - Error location vector pointer
+ *
+ * Inputs : ELP(z)
+ * No. of found errors
+ * Size of input codeword
+ * Outputs : Up to 8 locations
+ * No. of errors
+ */
+static int chien(unsigned int select_4_8, int err_nums,
+ unsigned int err[], unsigned int *location)
+{
+ int i, count; /* Number of dectected errors */
+ /* Contains accumulation of evaluation at x^i (i:1->8) */
+ unsigned int gammas[8] = {0};
+ unsigned int alpha;
+ unsigned int bit, ecc_bits;
+ unsigned int elp_sum;
+
+ ecc_bits = (select_4_8 == 0) ? 52 : 104;
+
+ /* Start evaluation at Alpha**8192 and decreasing */
+ for (i = 0; i < 8; i++)
+ gammas[i] = err[i];
+
+ count = 0;
+ for (i = 1; (i <= nn) && (count < err_nums); i++) {
+
+ /* Result of evaluation at root */
+ elp_sum = 1 ^ gammas[0] ^ gammas[1] ^
+ gammas[2] ^ gammas[3] ^
+ gammas[4] ^ gammas[5] ^
+ gammas[6] ^ gammas[7];
+
+ alpha = PPP >> 1;
+ gammas[0] = mpy_mod_gf(gammas[0], alpha);
+ alpha = mpy_mod_gf(alpha, (PPP >> 1)); /* x alphha^-2 */
+ gammas[1] = mpy_mod_gf(gammas[1], alpha);
+ alpha = mpy_mod_gf(alpha, (PPP >> 1)); /* x alphha^-2 */
+ gammas[2] = mpy_mod_gf(gammas[2], alpha);
+ alpha = mpy_mod_gf(alpha, (PPP >> 1)); /* x alphha^-3 */
+ gammas[3] = mpy_mod_gf(gammas[3], alpha);
+ alpha = mpy_mod_gf(alpha, (PPP >> 1)); /* x alphha^-4 */
+ gammas[4] = mpy_mod_gf(gammas[4], alpha);
+ alpha = mpy_mod_gf(alpha, (PPP >> 1)); /* x alphha^-5 */
+ gammas[5] = mpy_mod_gf(gammas[5], alpha);
+ alpha = mpy_mod_gf(alpha, (PPP >> 1)); /* x alphha^-6 */
+ gammas[6] = mpy_mod_gf(gammas[6], alpha);
+ alpha = mpy_mod_gf(alpha, (PPP >> 1)); /* x alphha^-7 */
+ gammas[7] = mpy_mod_gf(gammas[7], alpha);
+
+ if (elp_sum == 0) {
+ /* calculate bit position in main data area */
+ bit = ((i-1) & ~7)|(7-((i-1) & 7));
+ if (i >= 2 * ecc_bits)
+ location[count++] =
+ kk_shorten - (bit - 2 * ecc_bits) - 1;
+ }
+ }
+
+ /* Failure: No. of detected errors != No. or corrected errors */
+ if (count != err_nums) {
+ count = -1;
+ printk(KERN_ERR "BCH decoding failed\n");
+ }
+ for (i = 0; i < count; i++)
+ pr_debug("%d ", location[i]);
+
+ return count;
+}
+
+/* synd : 16 Syndromes
+ * return: gamaas - Coefficients to the error polynomial
+ * return: : Number of detected errors
+*/
+static unsigned int berlekamp(unsigned int select_4_8,
+ unsigned int synd[], unsigned int err[])
+{
+ int loop, iteration;
+ unsigned int LL = 0; /* Detected errors */
+ unsigned int d = 0; /* Distance between Syndromes and ELP[n](z) */
+ unsigned int invd = 0; /* Inverse of d */
+ /* Intermediate ELP[n](z).
+ * Final ELP[n](z) is Error Location Polynomial
+ */
+ unsigned int gammas[16] = {0};
+ /* Intermediate normalized ELP[n](z) : D[n](z) */
+ unsigned int D[16] = {0};
+ /* Temporary value that holds an ELP[n](z) coefficient */
+ unsigned int next_gamma = 0;
+
+ int e = 0;
+ unsigned int sign = 0;
+ unsigned int u = 0;
+ unsigned int v = 0;
+ unsigned int C1 = 0, C2 = 0;
+ unsigned int ss = 0;
+ unsigned int tmp_v = 0, tmp_s = 0;
+ unsigned int tmp_poly;
+
+ /*-------------- Step 0 ------------------*/
+ for (loop = 0; loop < 16; loop++)
+ gammas[loop] = 0;
+ gammas[0] = 1;
+ D[1] = 1;
+
+ iteration = 0;
+ LL = 0;
+ while ((iteration < ((select_4_8+1)*2*4)) &&
+ (LL <= ((select_4_8+1)*4))) {
+
+ pr_debug("\nIteration.............%d\n", iteration);
+ d = 0;
+ /* Step: 0 */
+ for (loop = 0; loop <= LL; loop++) {
+ tmp_poly = mpy_mod_gf(
+ gammas[loop], synd[iteration - loop]);
+ d ^= tmp_poly;
+ pr_debug("%02d. s=0 LL=%x poly %x\n",
+ loop, LL, tmp_poly);
+ }
+
+ /* Step 1: 1 cycle only to perform inversion */
+ v = d << 1;
+ e = -1;
+ sign = 1;
+ ss = 0x2000;
+ invd = 0;
+ u = PPP;
+ for (loop = 0; (d != 0) && (loop <= (2 * POLY)); loop++) {
+ pr_debug("%02d. s=1 LL=%x poly NULL\n",
+ loop, LL);
+ C1 = (v >> 13) & 1;
+ C2 = C1 & sign;
+
+ sign ^= C2 ^ (e == 0);
+
+ tmp_v = v;
+ tmp_s = ss;
+
+ if (C1 == 1) {
+ v ^= u;
+ ss ^= invd;
+ }
+ v = (v << 1) & 0x3FFF;
+ if (C2 == 1) {
+ u = tmp_v;
+ invd = tmp_s;
+ e = -e;
+ }
+ invd >>= 1;
+ e--;
+ }
+
+ for (loop = 0; (d != 0) && (loop <= (iteration + 1)); loop++) {
+ /* Step 2
+ * Interleaved with Step 3, if L<(n-k)
+ * invd: Update of ELP[n](z) = ELP[n-1](z) - d.D[n-1](z)
+ */
+
+ /* Holds value of ELP coefficient until precedent
+ * value does not have to be used anymore
+ */
+ tmp_poly = mpy_mod_gf(d, D[loop]);
+ pr_debug("%02d. s=2 LL=%x poly %x\n",
+ loop, LL, tmp_poly);
+
+ next_gamma = gammas[loop] ^ tmp_poly;
+ if ((2 * LL) < (iteration + 1)) {
+ /* Interleaving with Step 3
+ * for parallelized update of ELP(z) and D(z)
+ */
+ } else {
+ /* Update of ELP(z) only -> stay in Step 2 */
+ gammas[loop] = next_gamma;
+ if (loop == (iteration + 1)) {
+ /* to step 4 */
+ break;
+ }
+ }
+
+ /* Step 3
+ * Always interleaved with Step 2 (case when L<(n-k))
+ * Update of D[n-1](z) = ELP[n-1](z)/d
+ */
+ D[loop] = mpy_mod_gf(gammas[loop], invd);
+ pr_debug("%02d. s=3 LL=%x poly %x\n",
+ loop, LL, D[loop]);
+
+ /* Can safely update ELP[n](z) */
+ gammas[loop] = next_gamma;
+
+ if (loop == (iteration + 1)) {
+ /* If update finished */
+ LL = iteration - LL + 1;
+ /* to step 4 */
+ break;
+ }
+ /* Else, interleaving to step 2*/
+ }
+
+ /* Step 4: Update D(z): i:0->L */
+ /* Final update of D[n](z) = D[n](z).z*/
+ for (loop = 0; loop < 15; loop++) /* Left Shift */
+ D[15 - loop] = D[14 - loop];
+
+ D[0] = 0;
+
+ iteration++;
+ } /* while */
+
+ /* Processing finished, copy ELP to final registers : 0->2t-1*/
+ for (loop = 0; loop < 8; loop++)
+ err[loop] = gammas[loop+1];
+
+ pr_debug("\n Err poly:");
+ for (loop = 0; loop < 8; loop++)
+ pr_debug("0x%x ", err[loop]);
+
+ return LL;
+}
+
+/*
+ * syndrome - Generate syndrome components from hw generate syndrome
+ * r(x) = c(x) + e(x)
+ * s(x) = c(x) mod g(x) + e(x) mod g(x) = e(x) mod g(x)
+ * so receiver checks if the syndrome s(x) = r(x) mod g(x) is equal to zero.
+ * unsigned int s[16]; - Syndromes
+ */
+static void syndrome(unsigned int select_4_8,
+ unsigned char *ecc, unsigned int syn[])
+{
+ unsigned int k, l, t;
+ unsigned int alpha_bit, R_bit;
+ int ecc_pos, ecc_min;
+
+ /* 2t-1 = 15 (for t=8) minimal polynomials of the first 15 powers of a
+ * primitive elemmants of GF(m); Even powers minimal polynomials are
+ * duplicate of odd powers' minimal polynomials.
+ * Odd powers of alpha (1 to 15)
+ */
+ unsigned int pow_alpha[8] = {0x0002, 0x0008, 0x0020, 0x0080,
+ 0x0200, 0x0800, 0x001B, 0x006C};
+
+ pr_debug("\n ECC[0..n]: ");
+ for (k = 0; k < 13; k++)
+ pr_debug("0x%x ", ecc[k]);
+
+ if (select_4_8 == 0) {
+ t = 4;
+ ecc_pos = 55; /* bits(52-bits): 55->4 */
+ ecc_min = 4;
+ } else {
+ t = 8;
+ ecc_pos = 103; /* bits: 103->0 */
+ ecc_min = 0;
+ }
+
+ /* total numbber of syndrom to be used is 2t */
+ /* Step1: calculate the odd syndrome(s) */
+ R_bit = ((ecc[ecc_pos/8] >> (7 - ecc_pos%8)) & 1);
+ ecc_pos--;
+ for (k = 0; k < t; k++)
+ syn[2 * k] = R_bit;
+
+ while (ecc_pos >= ecc_min) {
+ R_bit = ((ecc[ecc_pos/8] >> (7 - ecc_pos%8)) & 1);
+ ecc_pos--;
+
+ for (k = 0; k < t; k++) {
+ /* Accumulate value of x^i at alpha^(2k+1) */
+ if (R_bit == 1)
+ syn[2*k] ^= pow_alpha[k];
+
+ /* Compute a**(2k+1), using LSFR */
+ for (l = 0; l < (2 * k + 1); l++) {
+ alpha_bit = (pow_alpha[k] >> POLY) & 1;
+ pow_alpha[k] = (pow_alpha[k] << 1) & 0x1FFF;
+ if (alpha_bit == 1)
+ pow_alpha[k] ^= P;
+ }
+ }
+ }
+
+ /* Step2: calculate the even syndrome(s)
+ * Compute S(a), where a is an even power of alpha
+ * Evenry even power of primitive element has the same minimal
+ * polynomial as some odd power of elemets.
+ * And based on S(a^2) = S^2(a)
+ */
+ for (k = 0; k < t; k++)
+ syn[2*k+1] = mpy_mod_gf(syn[k], syn[k]);
+
+ pr_debug("\n Syndromes: ");
+ for (k = 0; k < 16; k++)
+ pr_debug("0x%x ", syn[k]);
+}
+
+/**
+ * decode_bch - BCH decoder for 4- and 8-bit error correction
+ *
+ * @ecc - ECC syndrome generated by hw BCH engine
+ * @err_loc - pointer to error location array
+ *
+ * This function does post sydrome generation (hw generated) decoding
+ * for:-
+ * Dimension of Galoise Field: m = 13
+ * Length of codeword: n = 2**m - 1
+ * Number of errors that can be corrected: 4- or 8-bits
+ * Length of information bit: kk = nn - rr
+ */
+int decode_bch(int select_4_8, unsigned char *ecc, unsigned int *err_loc)
+{
+ int no_of_err;
+ unsigned int syn[16] = {0,}; /* 16 Syndromes */
+ unsigned int err_poly[8] = {0,};
+ /* Coefficients to the error polynomial
+ * ELP(x) = 1 + err0.x + err1.x^2 + ... + err7.x^8
+ */
+
+ /* Decoting involes three steps
+ * 1. Compute the syndrom from teh received codeword,
+ * 2. Find the error location polynomial from a set of equations
+ * derived from the syndrome,
+ * 3. Use the error location polynomial to identify errants bits,
+ *
+ * And correcttion done by bit flips using error locaiton and expected
+ * to be outseide of this implementation.
+ */
+ syndrome(select_4_8, ecc, syn);
+ no_of_err = berlekamp(select_4_8, syn, err_poly);
+ if (no_of_err <= (4 << select_4_8))
+ no_of_err = chien(select_4_8, no_of_err, err_poly, err_loc);
+
+ return no_of_err;
+}
+EXPORT_SYMBOL(decode_bch);
+
--
1.7.0.4
More information about the linux-mtd
mailing list