[U-Boot,v3,5/6] omap_gpmc: add support for hw assisted BCH8

diff --git a/doc/README.omap3 b/doc/README.omap3
index 0a37de0..56aca8e 100644
--- a/doc/README.omap3
+++ b/doc/README.omap3
@@ -145,6 +145,25 @@  int omap3_dma_wait_for_transfer(uint32_t chan)
 int omap3_dma_get_revision(uint32_t *minor, uint32_t *major)
 	Read silicon Revision of the DMA module
 
+NAND
+====
+
+Ther eare some OMAP3 devices out ther with NAND attached. Due to the fact that
+OMAP3 ROM code can only handle 1-bit hamming ECC for accessing first page
+(place where SPL lives) we require this setup for u-boot at least when reading
+the second progam within SPL.  A lot of newer NAND chips however require more
+than 1-bit ECC for the pages, some can live with 1-bit for the first page. To
+handle this we can switch to another ECC algorithm after reading the payload
+within SPL.
+
+BCH8
+----
+
+To enable hardware assisted BCH8 (8-bit BCH [Bose, Chaudhuri, Hocquenghem]) on
+OMAP3 devices we can use the BCH library in lib/bch.c. To do so add CONFIG_BCH
+to enable the library and CONFIG_NAND_OMAP_BCH8 to your board config.
+The NAND OOB layout is the same as in linux kernel, if the linux kernel BCH8
+implementation for OMAP3 works for you so the u-boot version should also.
 
 Acknowledgements
 ================
diff --git a/drivers/mtd/nand/omap_gpmc.c b/drivers/mtd/nand/omap_gpmc.c
index 0647828..de41e46 100644
--- a/drivers/mtd/nand/omap_gpmc.c
+++ b/drivers/mtd/nand/omap_gpmc.c
@@ -28,6 +28,7 @@ 
 #include <asm/arch/cpu.h>
 #include <asm/omap_gpmc.h>
 #include <linux/mtd/nand_ecc.h>
+#include <linux/bch.h>
 #include <linux/compiler.h>
 #include <nand.h>
 #ifdef CONFIG_AM33XX
@@ -37,6 +38,8 @@ 
 static uint8_t cs;
 static __maybe_unused struct nand_ecclayout hw_nand_oob =
 	GPMC_NAND_HW_ECC_LAYOUT;
+static __maybe_unused struct nand_ecclayout hw_bch8_nand_oob =
+	GPMC_NAND_HW_BCH8_ECC_LAYOUT;
 
 /*
  * omap_nand_hwcontrol - Set the address pointers corretly for the
@@ -239,13 +242,13 @@  static void __maybe_unused omap_enable_hwecc(struct mtd_info *mtd, int32_t mode)
 }
 
 /*
- * BCH8 support (needs ELM and thus AM33xx-only)
+ * Generic BCH interface
  */
-#ifdef CONFIG_AM33XX
 struct nand_bch_priv {
 	uint8_t mode;
 	uint8_t type;
 	uint8_t nibbles;
+	struct bch_control *control;
 };
 
 /* bch types */
@@ -253,21 +256,146 @@  struct nand_bch_priv {
 #define ECC_BCH8	1
 #define ECC_BCH16	2
 
+/* GPMC ecc engine settings */
+#define BCH_WRAPMODE_1		1	/* BCH wrap mode 1 */
+#define BCH_WRAPMODE_6		6	/* BCH wrap mode 6 */
+
 /* BCH nibbles for diff bch levels */
 #define NAND_ECC_HW_BCH ((uint8_t)(NAND_ECC_HW_OOB_FIRST) + 1)
 #define ECC_BCH4_NIBBLES	13
 #define ECC_BCH8_NIBBLES	26
 #define ECC_BCH16_NIBBLES	52
 
-static struct nand_ecclayout hw_bch8_nand_oob = GPMC_NAND_HW_BCH8_ECC_LAYOUT;
-
-static struct nand_bch_priv bch_priv = {
+/*
+ * This can be a single instance cause all current users have only one NAND
+ * with nearly the same setup (BCH8, some with ELM and others with sw BCH
+ * library).
+ * When some users with other BCH strength will exists this have to change!
+ */
+static __maybe_unused struct nand_bch_priv bch_priv = {
 	.mode = NAND_ECC_HW_BCH,
 	.type = ECC_BCH8,
-	.nibbles = ECC_BCH8_NIBBLES
+	.nibbles = ECC_BCH8_NIBBLES,
+	.control = NULL
 };
 
 /*
+ * omap_hwecc_init_bch - Initialize the BCH Hardware ECC for NAND flash in
+ *				GPMC controller
+ * @mtd:	MTD device structure
+ * @mode:	Read/Write mode
+ */
+__maybe_unused
+static void omap_hwecc_init_bch(struct nand_chip *chip, int32_t mode)
+{
+	uint32_t val;
+	uint32_t dev_width = (chip->options & NAND_BUSWIDTH_16) >> 1;
+#ifdef CONFIG_AM33XX
+	uint32_t unused_length = 0;
+#endif
+	uint32_t wr_mode = BCH_WRAPMODE_6;
+	struct nand_bch_priv *bch = chip->priv;
+
+	/* Clear the ecc result registers, select ecc reg as 1 */
+	writel(ECCCLEAR | ECCRESULTREG1, &gpmc_cfg->ecc_control);
+
+#ifdef CONFIG_AM33XX
+	wr_mode = BCH_WRAPMODE_1;
+
+	switch (bch->nibbles) {
+	case ECC_BCH4_NIBBLES:
+		unused_length = 3;
+		break;
+	case ECC_BCH8_NIBBLES:
+		unused_length = 2;
+		break;
+	case ECC_BCH16_NIBBLES:
+		unused_length = 0;
+		break;
+	}
+
+	/*
+	 * This is ecc_size_config for ELM mode.
+	 * Here we are using different settings for read and write access and
+	 * also depending on BCH strength.
+	 */
+	switch (mode) {
+	case NAND_ECC_WRITE:
+		/* write access only setup eccsize1 config */
+		val = ((unused_length + bch->nibbles) << 22);
+		break;
+
+	case NAND_ECC_READ:
+	default:
+		/*
+		 * by default eccsize0 selected for ecc1resultsize
+		 * eccsize0 config.
+		 */
+		val  = (bch->nibbles << 12);
+		/* eccsize1 config */
+		val |= (unused_length << 22);
+		break;
+	}
+#else
+	/*
+	 * This ecc_size_config setting is for BCH sw library.
+	 *
+	 * Note: we only support BCH8 currently with BCH sw library!
+	 * Should be really easy to adobt to BCH4, however some omap3 have
+	 * flaws with BCH4.
+	 *
+	 * Here we are using wrapping mode 6 both for reading and writing, with:
+	 *  size0 = 0  (no additional protected byte in spare area)
+	 *  size1 = 32 (skip 32 nibbles = 16 bytes per sector in spare area)
+	 */
+	val = (32 << 22) | (0 << 12);
+#endif
+	/* ecc size configuration */
+	writel(val, &gpmc_cfg->ecc_size_config);
+
+	/*
+	 * Configure the ecc engine in gpmc
+	 * We assume 512 Byte sector pages for access to NAND.
+	 */
+	val  = (1 << 16);		/* enable BCH mode */
+	val |= (bch->type << 12);	/* setup BCH type */
+	val |= (wr_mode << 8);		/* setup wrapping mode */
+	val |= (dev_width << 7);	/* setup device width (16 or 8 bit) */
+	val |= (cs << 1);		/* setup chip select to work on */
+	debug("set ECC_CONFIG=0x%08x\n", val);
+	writel(val, &gpmc_cfg->ecc_config);
+}
+
+/*
+ * omap_enable_ecc_bch - This function enables the bch h/w ecc functionality
+ * @mtd:	MTD device structure
+ * @mode:	Read/Write mode
+ */
+__maybe_unused
+static void omap_enable_ecc_bch(struct mtd_info *mtd, int32_t mode)
+{
+	struct nand_chip *chip = mtd->priv;
+
+	omap_hwecc_init_bch(chip, mode);
+	/* enable ecc */
+	writel((readl(&gpmc_cfg->ecc_config) | 0x1), &gpmc_cfg->ecc_config);
+}
+
+/*
+ * omap_ecc_disable - Disable H/W ECC calculation
+ *
+ * @mtd:	MTD device structure
+ */
+static void __maybe_unused omap_ecc_disable(struct mtd_info *mtd)
+{
+	writel((readl(&gpmc_cfg->ecc_config) & ~0x1), &gpmc_cfg->ecc_config);
+}
+
+/*
+ * BCH8 support (needs ELM and thus AM33xx-only)
+ */
+#ifdef CONFIG_AM33XX
+/*
  * omap_read_bch8_result - Read BCH result for BCH8 level
  *
  * @mtd:	MTD device structure
@@ -306,18 +434,6 @@  static void omap_read_bch8_result(struct mtd_info *mtd, uint8_t big_endian,
 }
 
 /*
- * omap_ecc_disable - Disable H/W ECC calculation
- *
- * @mtd:	MTD device structure
- *
- */
-static void omap_ecc_disable(struct mtd_info *mtd)
-{
-	writel((readl(&gpmc_cfg->ecc_config) & ~0x1),
-		&gpmc_cfg->ecc_config);
-}
-
-/*
  * omap_rotate_ecc_bch - Rotate the syndrome bytes
  *
  * @mtd:	MTD device structure
@@ -468,76 +584,6 @@  static int omap_correct_data_bch(struct mtd_info *mtd, uint8_t *dat,
 
 	return 0;
 }
-/*
- * omap_hwecc_init_bch - Initialize the BCH Hardware ECC for NAND flash in
- *				GPMC controller
- * @mtd:       MTD device structure
- * @mode:	Read/Write mode
- */
-static void omap_hwecc_init_bch(struct nand_chip *chip, int32_t mode)
-{
-	uint32_t val, dev_width = (chip->options & NAND_BUSWIDTH_16) >> 1;
-	uint32_t unused_length = 0;
-	struct nand_bch_priv *bch = chip->priv;
-
-	switch (bch->nibbles) {
-	case ECC_BCH4_NIBBLES:
-		unused_length = 3;
-		break;
-	case ECC_BCH8_NIBBLES:
-		unused_length = 2;
-		break;
-	case ECC_BCH16_NIBBLES:
-		unused_length = 0;
-		break;
-	}
-
-	/* Clear the ecc result registers, select ecc reg as 1 */
-	writel(ECCCLEAR | ECCRESULTREG1, &gpmc_cfg->ecc_control);
-
-	switch (mode) {
-	case NAND_ECC_WRITE:
-		/* eccsize1 config */
-		val = ((unused_length + bch->nibbles) << 22);
-		break;
-
-	case NAND_ECC_READ:
-	default:
-		/* by default eccsize0 selected for ecc1resultsize */
-		/* eccsize0 config */
-		val  = (bch->nibbles << 12);
-		/* eccsize1 config */
-		val |= (unused_length << 22);
-		break;
-	}
-	/* ecc size configuration */
-	writel(val, &gpmc_cfg->ecc_size_config);
-	/* by default 512bytes sector page is selected */
-	/* set bch mode */
-	val  = (1 << 16);
-	/* bch4 / bch8 / bch16 */
-	val |= (bch->type << 12);
-	/* set wrap mode to 1 */
-	val |= (1 << 8);
-	val |= (dev_width << 7);
-	val |= (cs << 1);
-	writel(val, &gpmc_cfg->ecc_config);
-}
-
-/*
- * omap_enable_ecc_bch- This function enables the bch h/w ecc functionality
- * @mtd:        MTD device structure
- * @mode:       Read/Write mode
- *
- */
-static void omap_enable_ecc_bch(struct mtd_info *mtd, int32_t mode)
-{
-	struct nand_chip *chip = mtd->priv;
-
-	omap_hwecc_init_bch(chip, mode);
-	/* enable ecc */
-	writel((readl(&gpmc_cfg->ecc_config) | 0x1), &gpmc_cfg->ecc_config);
-}
 
 /**
  * omap_read_page_bch - hardware ecc based page read function
@@ -602,6 +648,127 @@  static int omap_read_page_bch(struct mtd_info *mtd, struct nand_chip *chip,
 }
 #endif /* CONFIG_AM33XX */
 
+/*
+ * OMAP3 BCH8 support (with BCH library)
+ */
+#ifdef CONFIG_NAND_OMAP_BCH8
+/*
+ *  omap_calculate_ecc_bch - Read BCH ECC result
+ *
+ *  @mtd:	MTD device structure
+ *  @dat:	The pointer to data on which ecc is computed (unused here)
+ *  @ecc:	The ECC output buffer
+ */
+static int omap_calculate_ecc_bch(struct mtd_info *mtd, const uint8_t *dat,
+				uint8_t *ecc)
+{
+	int ret = 0;
+	size_t i;
+	unsigned long nsectors, val1, val2, val3, val4;
+
+	nsectors = ((readl(&gpmc_cfg->ecc_config) >> 4) & 0x7) + 1;
+
+	for (i = 0; i < nsectors; i++) {
+		/* Read hw-computed remainder */
+		val1 = readl(&gpmc_cfg->bch_result_0_3[i].bch_result_x[0]);
+		val2 = readl(&gpmc_cfg->bch_result_0_3[i].bch_result_x[1]);
+		val3 = readl(&gpmc_cfg->bch_result_0_3[i].bch_result_x[2]);
+		val4 = readl(&gpmc_cfg->bch_result_0_3[i].bch_result_x[3]);
+
+		/*
+		 * Add constant polynomial to remainder, in order to get an ecc
+		 * sequence of 0xFFs for a buffer filled with 0xFFs.
+		 */
+		*ecc++ = 0xef ^ (val4 & 0xFF);
+		*ecc++ = 0x51 ^ ((val3 >> 24) & 0xFF);
+		*ecc++ = 0x2e ^ ((val3 >> 16) & 0xFF);
+		*ecc++ = 0x09 ^ ((val3 >> 8) & 0xFF);
+		*ecc++ = 0xed ^ (val3 & 0xFF);
+		*ecc++ = 0x93 ^ ((val2 >> 24) & 0xFF);
+		*ecc++ = 0x9a ^ ((val2 >> 16) & 0xFF);
+		*ecc++ = 0xc2 ^ ((val2 >> 8) & 0xFF);
+		*ecc++ = 0x97 ^ (val2 & 0xFF);
+		*ecc++ = 0x79 ^ ((val1 >> 24) & 0xFF);
+		*ecc++ = 0xe5 ^ ((val1 >> 16) & 0xFF);
+		*ecc++ = 0x24 ^ ((val1 >> 8) & 0xFF);
+		*ecc++ = 0xb5 ^ (val1 & 0xFF);
+	}
+
+	/*
+	 * Stop reading anymore ECC vals and clear old results
+	 * enable will be called if more reads are required
+	 */
+	omap_ecc_disable(mtd);
+
+	return ret;
+}
+
+/**
+ * omap_correct_data_bch - Decode received data and correct errors
+ * @mtd: MTD device structure
+ * @data: page data
+ * @read_ecc: ecc read from nand flash
+ * @calc_ecc: ecc read from HW ECC registers
+ */
+static int omap_correct_data_bch(struct mtd_info *mtd, u_char *data,
+				 u_char *read_ecc, u_char *calc_ecc)
+{
+	int i, count;
+	/* cannot correct more than 8 errors */
+	unsigned int errloc[8];
+	struct nand_chip *chip = mtd->priv;
+	struct nand_bch_priv *chip_priv = chip->priv;
+	struct bch_control *bch = chip_priv->control;
+
+	count = decode_bch(bch, NULL, 512, read_ecc, calc_ecc, NULL, errloc);
+	if (count > 0) {
+		/* correct errors */
+		for (i = 0; i < count; i++) {
+			/* correct data only, not ecc bytes */
+			if (errloc[i] < 8*512)
+				data[errloc[i]/8] ^= 1 << (errloc[i] & 7);
+			printf("corrected bitflip %u\n", errloc[i]);
+#ifdef DEBUG
+			puts("read_ecc: ");
+			/*
+			 * BCH8 have 13 bytes of ECC; BCH4 needs adoption
+			 * here!
+			 */
+			for (i = 0; i < 13; i++)
+				printf("%02x ", read_ecc[i]);
+			puts("\n");
+			puts("calc_ecc: ");
+			for (i = 0; i < 13; i++)
+				printf("%02x ", calc_ecc[i]);
+			puts("\n");
+#endif
+		}
+	} else if (count < 0) {
+		puts("ecc unrecoverable error\n");
+	}
+	return count;
+}
+
+/**
+ * omap_free_bch - Release BCH ecc resources
+ * @mtd: MTD device structure
+ */
+static void __maybe_unused omap_free_bch(struct mtd_info *mtd)
+{
+	struct nand_chip *chip = mtd->priv;
+	struct nand_bch_priv *chip_priv = chip->priv;
+	struct bch_control *bch = NULL;
+
+	if (chip_priv)
+		bch = chip_priv->control;
+
+	if (bch) {
+		free_bch(bch);
+		chip_priv->control = NULL;
+	}
+}
+#endif /* CONFIG_NAND_OMAP_BCH8 */
+
 #ifndef CONFIG_SPL_BUILD
 /*
  * omap_nand_switch_ecc - switch the ECC operation between different engines
@@ -650,13 +817,17 @@  void omap_nand_switch_ecc(bool hardware, uint32_t eccstrength)
 			omap_hwecc_init(nand);
 			puts("1-bit hamming HW ECC selected\n");
 		}
-#ifdef CONFIG_AM33XX
+#if defined(CONFIG_AM33XX) || defined(CONFIG_NAND_OMAP_BCH8)
 		else if (eccstrength == 8) {
 			nand->ecc.mode = NAND_ECC_HW;
 			nand->ecc.layout = &hw_bch8_nand_oob;
 			nand->ecc.size = 512;
+#ifdef CONFIG_AM33XX
 			nand->ecc.bytes = 14;
 			nand->ecc.read_page = omap_read_page_bch;
+#else
+			nand->ecc.bytes = 13;
+#endif
 			nand->ecc.hwctl = omap_enable_ecc_bch;
 			nand->ecc.correct = omap_correct_data_bch;
 			nand->ecc.calculate = omap_calculate_ecc_bch;
@@ -739,16 +910,28 @@  int board_nand_init(struct nand_chip *nand)
 
 	nand->chip_delay = 100;
 
+#if defined(CONFIG_AM33XX) || defined(CONFIG_NAND_OMAP_BCH8)
 #ifdef CONFIG_AM33XX
+	/* AM33xx uses the ELM */
 	/* required in case of BCH */
 	elm_init();
-
+#else
+	/*
+	 * Whereas other OMAP based SoC do not have the ELM, they use the BCH
+	 * SW library.
+	 */
+	bch_priv.control = init_bch(13, 8, 0x201b /* hw polynominal */);
+	if (!bch_priv.control) {
+		puts("Could not init_bch()\n");
+		return -ENODEV;
+	}
+#endif
 	/* BCH info that will be correct for SPL or overridden otherwise. */
 	nand->priv = &bch_priv;
 #endif
 
 	/* Default ECC mode */
-#ifdef CONFIG_AM33XX
+#if defined(CONFIG_AM33XX) || defined(CONFIG_NAND_OMAP_BCH8)
 	nand->ecc.mode = NAND_ECC_HW;
 	nand->ecc.layout = &hw_bch8_nand_oob;
 	nand->ecc.size = CONFIG_SYS_NAND_ECCSIZE;
@@ -756,7 +939,9 @@  int board_nand_init(struct nand_chip *nand)
 	nand->ecc.hwctl = omap_enable_ecc_bch;
 	nand->ecc.correct = omap_correct_data_bch;
 	nand->ecc.calculate = omap_calculate_ecc_bch;
+#ifdef CONFIG_AM33XX
 	nand->ecc.read_page = omap_read_page_bch;
+#endif
 	omap_hwecc_init_bch(nand, NAND_ECC_READ);
 #else
 #if !defined(CONFIG_SPL_BUILD) || defined(CONFIG_SPL_NAND_SOFTECC)
diff --git a/lib/Makefile b/lib/Makefile
index 1bfd3ee..5d71b80 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -27,7 +27,6 @@  LIB	= $(obj)libgeneric.o
 
 ifndef CONFIG_SPL_BUILD
 COBJS-$(CONFIG_ADDR_MAP) += addr_map.o
-COBJS-$(CONFIG_BCH) += bch.o
 COBJS-$(CONFIG_AES) += aes.o
 COBJS-$(CONFIG_BZIP2) += bzlib.o
 COBJS-$(CONFIG_BZIP2) += bzlib_crctable.o
@@ -67,6 +66,7 @@  COBJS-$(CONFIG_SPL_NET_SUPPORT) += errno.o
 COBJS-$(CONFIG_SPL_NET_SUPPORT) += hashtable.o
 COBJS-$(CONFIG_SPL_NET_SUPPORT) += net_utils.o
 endif
+COBJS-$(CONFIG_BCH) += bch.o
 COBJS-y += crc32.o
 COBJS-y += ctype.o
 COBJS-y += div64.o