[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-arm-kernel mailing list