ASR_BASE
Change-Id: Icf3719cc0afe3eeb3edc7fa80a2eb5199ca9dda1
diff --git a/marvell/linux/drivers/mtd/nand/raw/nand_ecc.c b/marvell/linux/drivers/mtd/nand/raw/nand_ecc.c
new file mode 100644
index 0000000..b6a46b1
--- /dev/null
+++ b/marvell/linux/drivers/mtd/nand/raw/nand_ecc.c
@@ -0,0 +1,484 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * This file contains an ECC algorithm that detects and corrects 1 bit
+ * errors in a 256 byte block of data.
+ *
+ * Copyright © 2008 Koninklijke Philips Electronics NV.
+ * Author: Frans Meulenbroeks
+ *
+ * Completely replaces the previous ECC implementation which was written by:
+ * Steven J. Hill (sjhill@realitydiluted.com)
+ * Thomas Gleixner (tglx@linutronix.de)
+ *
+ * Information on how this algorithm works and how it was developed
+ * can be found in Documentation/driver-api/mtd/nand_ecc.rst
+ */
+
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/rawnand.h>
+#include <linux/mtd/nand_ecc.h>
+#include <asm/byteorder.h>
+
+/*
+ * invparity is a 256 byte table that contains the odd parity
+ * for each byte. So if the number of bits in a byte is even,
+ * the array element is 1, and when the number of bits is odd
+ * the array eleemnt is 0.
+ */
+static const char invparity[256] = {
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+ 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1
+};
+
+/*
+ * bitsperbyte contains the number of bits per byte
+ * this is only used for testing and repairing parity
+ * (a precalculated value slightly improves performance)
+ */
+static const char bitsperbyte[256] = {
+ 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4,
+ 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+ 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+ 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+ 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+ 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+ 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+ 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+ 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+ 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+ 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+ 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+ 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+ 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+ 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+ 4, 5, 5, 6, 5, 6, 6, 7, 5, 6, 6, 7, 6, 7, 7, 8,
+};
+
+/*
+ * addressbits is a lookup table to filter out the bits from the xor-ed
+ * ECC data that identify the faulty location.
+ * this is only used for repairing parity
+ * see the comments in nand_correct_data for more details
+ */
+static const char addressbits[256] = {
+ 0x00, 0x00, 0x01, 0x01, 0x00, 0x00, 0x01, 0x01,
+ 0x02, 0x02, 0x03, 0x03, 0x02, 0x02, 0x03, 0x03,
+ 0x00, 0x00, 0x01, 0x01, 0x00, 0x00, 0x01, 0x01,
+ 0x02, 0x02, 0x03, 0x03, 0x02, 0x02, 0x03, 0x03,
+ 0x04, 0x04, 0x05, 0x05, 0x04, 0x04, 0x05, 0x05,
+ 0x06, 0x06, 0x07, 0x07, 0x06, 0x06, 0x07, 0x07,
+ 0x04, 0x04, 0x05, 0x05, 0x04, 0x04, 0x05, 0x05,
+ 0x06, 0x06, 0x07, 0x07, 0x06, 0x06, 0x07, 0x07,
+ 0x00, 0x00, 0x01, 0x01, 0x00, 0x00, 0x01, 0x01,
+ 0x02, 0x02, 0x03, 0x03, 0x02, 0x02, 0x03, 0x03,
+ 0x00, 0x00, 0x01, 0x01, 0x00, 0x00, 0x01, 0x01,
+ 0x02, 0x02, 0x03, 0x03, 0x02, 0x02, 0x03, 0x03,
+ 0x04, 0x04, 0x05, 0x05, 0x04, 0x04, 0x05, 0x05,
+ 0x06, 0x06, 0x07, 0x07, 0x06, 0x06, 0x07, 0x07,
+ 0x04, 0x04, 0x05, 0x05, 0x04, 0x04, 0x05, 0x05,
+ 0x06, 0x06, 0x07, 0x07, 0x06, 0x06, 0x07, 0x07,
+ 0x08, 0x08, 0x09, 0x09, 0x08, 0x08, 0x09, 0x09,
+ 0x0a, 0x0a, 0x0b, 0x0b, 0x0a, 0x0a, 0x0b, 0x0b,
+ 0x08, 0x08, 0x09, 0x09, 0x08, 0x08, 0x09, 0x09,
+ 0x0a, 0x0a, 0x0b, 0x0b, 0x0a, 0x0a, 0x0b, 0x0b,
+ 0x0c, 0x0c, 0x0d, 0x0d, 0x0c, 0x0c, 0x0d, 0x0d,
+ 0x0e, 0x0e, 0x0f, 0x0f, 0x0e, 0x0e, 0x0f, 0x0f,
+ 0x0c, 0x0c, 0x0d, 0x0d, 0x0c, 0x0c, 0x0d, 0x0d,
+ 0x0e, 0x0e, 0x0f, 0x0f, 0x0e, 0x0e, 0x0f, 0x0f,
+ 0x08, 0x08, 0x09, 0x09, 0x08, 0x08, 0x09, 0x09,
+ 0x0a, 0x0a, 0x0b, 0x0b, 0x0a, 0x0a, 0x0b, 0x0b,
+ 0x08, 0x08, 0x09, 0x09, 0x08, 0x08, 0x09, 0x09,
+ 0x0a, 0x0a, 0x0b, 0x0b, 0x0a, 0x0a, 0x0b, 0x0b,
+ 0x0c, 0x0c, 0x0d, 0x0d, 0x0c, 0x0c, 0x0d, 0x0d,
+ 0x0e, 0x0e, 0x0f, 0x0f, 0x0e, 0x0e, 0x0f, 0x0f,
+ 0x0c, 0x0c, 0x0d, 0x0d, 0x0c, 0x0c, 0x0d, 0x0d,
+ 0x0e, 0x0e, 0x0f, 0x0f, 0x0e, 0x0e, 0x0f, 0x0f
+};
+
+/**
+ * __nand_calculate_ecc - [NAND Interface] Calculate 3-byte ECC for 256/512-byte
+ * block
+ * @buf: input buffer with raw data
+ * @eccsize: data bytes per ECC step (256 or 512)
+ * @code: output buffer with ECC
+ * @sm_order: Smart Media byte ordering
+ */
+void __nand_calculate_ecc(const unsigned char *buf, unsigned int eccsize,
+ unsigned char *code, bool sm_order)
+{
+ int i;
+ const uint32_t *bp = (uint32_t *)buf;
+ /* 256 or 512 bytes/ecc */
+ const uint32_t eccsize_mult = eccsize >> 8;
+ uint32_t cur; /* current value in buffer */
+ /* rp0..rp15..rp17 are the various accumulated parities (per byte) */
+ uint32_t rp0, rp1, rp2, rp3, rp4, rp5, rp6, rp7;
+ uint32_t rp8, rp9, rp10, rp11, rp12, rp13, rp14, rp15, rp16;
+ uint32_t rp17;
+ uint32_t par; /* the cumulative parity for all data */
+ uint32_t tmppar; /* the cumulative parity for this iteration;
+ for rp12, rp14 and rp16 at the end of the
+ loop */
+
+ par = 0;
+ rp4 = 0;
+ rp6 = 0;
+ rp8 = 0;
+ rp10 = 0;
+ rp12 = 0;
+ rp14 = 0;
+ rp16 = 0;
+
+ /*
+ * The loop is unrolled a number of times;
+ * This avoids if statements to decide on which rp value to update
+ * Also we process the data by longwords.
+ * Note: passing unaligned data might give a performance penalty.
+ * It is assumed that the buffers are aligned.
+ * tmppar is the cumulative sum of this iteration.
+ * needed for calculating rp12, rp14, rp16 and par
+ * also used as a performance improvement for rp6, rp8 and rp10
+ */
+ for (i = 0; i < eccsize_mult << 2; i++) {
+ cur = *bp++;
+ tmppar = cur;
+ rp4 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp6 ^= tmppar;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp4 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp8 ^= tmppar;
+
+ cur = *bp++;
+ tmppar ^= cur;
+ rp4 ^= cur;
+ rp6 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp6 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp4 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp10 ^= tmppar;
+
+ cur = *bp++;
+ tmppar ^= cur;
+ rp4 ^= cur;
+ rp6 ^= cur;
+ rp8 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp6 ^= cur;
+ rp8 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp4 ^= cur;
+ rp8 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp8 ^= cur;
+
+ cur = *bp++;
+ tmppar ^= cur;
+ rp4 ^= cur;
+ rp6 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp6 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+ rp4 ^= cur;
+ cur = *bp++;
+ tmppar ^= cur;
+
+ par ^= tmppar;
+ if ((i & 0x1) == 0)
+ rp12 ^= tmppar;
+ if ((i & 0x2) == 0)
+ rp14 ^= tmppar;
+ if (eccsize_mult == 2 && (i & 0x4) == 0)
+ rp16 ^= tmppar;
+ }
+
+ /*
+ * handle the fact that we use longword operations
+ * we'll bring rp4..rp14..rp16 back to single byte entities by
+ * shifting and xoring first fold the upper and lower 16 bits,
+ * then the upper and lower 8 bits.
+ */
+ rp4 ^= (rp4 >> 16);
+ rp4 ^= (rp4 >> 8);
+ rp4 &= 0xff;
+ rp6 ^= (rp6 >> 16);
+ rp6 ^= (rp6 >> 8);
+ rp6 &= 0xff;
+ rp8 ^= (rp8 >> 16);
+ rp8 ^= (rp8 >> 8);
+ rp8 &= 0xff;
+ rp10 ^= (rp10 >> 16);
+ rp10 ^= (rp10 >> 8);
+ rp10 &= 0xff;
+ rp12 ^= (rp12 >> 16);
+ rp12 ^= (rp12 >> 8);
+ rp12 &= 0xff;
+ rp14 ^= (rp14 >> 16);
+ rp14 ^= (rp14 >> 8);
+ rp14 &= 0xff;
+ if (eccsize_mult == 2) {
+ rp16 ^= (rp16 >> 16);
+ rp16 ^= (rp16 >> 8);
+ rp16 &= 0xff;
+ }
+
+ /*
+ * we also need to calculate the row parity for rp0..rp3
+ * This is present in par, because par is now
+ * rp3 rp3 rp2 rp2 in little endian and
+ * rp2 rp2 rp3 rp3 in big endian
+ * as well as
+ * rp1 rp0 rp1 rp0 in little endian and
+ * rp0 rp1 rp0 rp1 in big endian
+ * First calculate rp2 and rp3
+ */
+#ifdef __BIG_ENDIAN
+ rp2 = (par >> 16);
+ rp2 ^= (rp2 >> 8);
+ rp2 &= 0xff;
+ rp3 = par & 0xffff;
+ rp3 ^= (rp3 >> 8);
+ rp3 &= 0xff;
+#else
+ rp3 = (par >> 16);
+ rp3 ^= (rp3 >> 8);
+ rp3 &= 0xff;
+ rp2 = par & 0xffff;
+ rp2 ^= (rp2 >> 8);
+ rp2 &= 0xff;
+#endif
+
+ /* reduce par to 16 bits then calculate rp1 and rp0 */
+ par ^= (par >> 16);
+#ifdef __BIG_ENDIAN
+ rp0 = (par >> 8) & 0xff;
+ rp1 = (par & 0xff);
+#else
+ rp1 = (par >> 8) & 0xff;
+ rp0 = (par & 0xff);
+#endif
+
+ /* finally reduce par to 8 bits */
+ par ^= (par >> 8);
+ par &= 0xff;
+
+ /*
+ * and calculate rp5..rp15..rp17
+ * note that par = rp4 ^ rp5 and due to the commutative property
+ * of the ^ operator we can say:
+ * rp5 = (par ^ rp4);
+ * The & 0xff seems superfluous, but benchmarking learned that
+ * leaving it out gives slightly worse results. No idea why, probably
+ * it has to do with the way the pipeline in pentium is organized.
+ */
+ rp5 = (par ^ rp4) & 0xff;
+ rp7 = (par ^ rp6) & 0xff;
+ rp9 = (par ^ rp8) & 0xff;
+ rp11 = (par ^ rp10) & 0xff;
+ rp13 = (par ^ rp12) & 0xff;
+ rp15 = (par ^ rp14) & 0xff;
+ if (eccsize_mult == 2)
+ rp17 = (par ^ rp16) & 0xff;
+
+ /*
+ * Finally calculate the ECC bits.
+ * Again here it might seem that there are performance optimisations
+ * possible, but benchmarks showed that on the system this is developed
+ * the code below is the fastest
+ */
+ if (sm_order) {
+ code[0] = (invparity[rp7] << 7) | (invparity[rp6] << 6) |
+ (invparity[rp5] << 5) | (invparity[rp4] << 4) |
+ (invparity[rp3] << 3) | (invparity[rp2] << 2) |
+ (invparity[rp1] << 1) | (invparity[rp0]);
+ code[1] = (invparity[rp15] << 7) | (invparity[rp14] << 6) |
+ (invparity[rp13] << 5) | (invparity[rp12] << 4) |
+ (invparity[rp11] << 3) | (invparity[rp10] << 2) |
+ (invparity[rp9] << 1) | (invparity[rp8]);
+ } else {
+ code[1] = (invparity[rp7] << 7) | (invparity[rp6] << 6) |
+ (invparity[rp5] << 5) | (invparity[rp4] << 4) |
+ (invparity[rp3] << 3) | (invparity[rp2] << 2) |
+ (invparity[rp1] << 1) | (invparity[rp0]);
+ code[0] = (invparity[rp15] << 7) | (invparity[rp14] << 6) |
+ (invparity[rp13] << 5) | (invparity[rp12] << 4) |
+ (invparity[rp11] << 3) | (invparity[rp10] << 2) |
+ (invparity[rp9] << 1) | (invparity[rp8]);
+ }
+
+ if (eccsize_mult == 1)
+ code[2] =
+ (invparity[par & 0xf0] << 7) |
+ (invparity[par & 0x0f] << 6) |
+ (invparity[par & 0xcc] << 5) |
+ (invparity[par & 0x33] << 4) |
+ (invparity[par & 0xaa] << 3) |
+ (invparity[par & 0x55] << 2) |
+ 3;
+ else
+ code[2] =
+ (invparity[par & 0xf0] << 7) |
+ (invparity[par & 0x0f] << 6) |
+ (invparity[par & 0xcc] << 5) |
+ (invparity[par & 0x33] << 4) |
+ (invparity[par & 0xaa] << 3) |
+ (invparity[par & 0x55] << 2) |
+ (invparity[rp17] << 1) |
+ (invparity[rp16] << 0);
+}
+EXPORT_SYMBOL(__nand_calculate_ecc);
+
+/**
+ * nand_calculate_ecc - [NAND Interface] Calculate 3-byte ECC for 256/512-byte
+ * block
+ * @chip: NAND chip object
+ * @buf: input buffer with raw data
+ * @code: output buffer with ECC
+ */
+int nand_calculate_ecc(struct nand_chip *chip, const unsigned char *buf,
+ unsigned char *code)
+{
+ bool sm_order = chip->ecc.options & NAND_ECC_SOFT_HAMMING_SM_ORDER;
+
+ __nand_calculate_ecc(buf, chip->ecc.size, code, sm_order);
+
+ return 0;
+}
+EXPORT_SYMBOL(nand_calculate_ecc);
+
+/**
+ * __nand_correct_data - [NAND Interface] Detect and correct bit error(s)
+ * @buf: raw data read from the chip
+ * @read_ecc: ECC from the chip
+ * @calc_ecc: the ECC calculated from raw data
+ * @eccsize: data bytes per ECC step (256 or 512)
+ * @sm_order: Smart Media byte order
+ *
+ * Detect and correct a 1 bit error for eccsize byte block
+ */
+int __nand_correct_data(unsigned char *buf,
+ unsigned char *read_ecc, unsigned char *calc_ecc,
+ unsigned int eccsize, bool sm_order)
+{
+ unsigned char b0, b1, b2, bit_addr;
+ unsigned int byte_addr;
+ /* 256 or 512 bytes/ecc */
+ const uint32_t eccsize_mult = eccsize >> 8;
+
+ /*
+ * b0 to b2 indicate which bit is faulty (if any)
+ * we might need the xor result more than once,
+ * so keep them in a local var
+ */
+ if (sm_order) {
+ b0 = read_ecc[0] ^ calc_ecc[0];
+ b1 = read_ecc[1] ^ calc_ecc[1];
+ } else {
+ b0 = read_ecc[1] ^ calc_ecc[1];
+ b1 = read_ecc[0] ^ calc_ecc[0];
+ }
+
+ b2 = read_ecc[2] ^ calc_ecc[2];
+
+ /* check if there are any bitfaults */
+
+ /* repeated if statements are slightly more efficient than switch ... */
+ /* ordered in order of likelihood */
+
+ if ((b0 | b1 | b2) == 0)
+ return 0; /* no error */
+
+ if ((((b0 ^ (b0 >> 1)) & 0x55) == 0x55) &&
+ (((b1 ^ (b1 >> 1)) & 0x55) == 0x55) &&
+ ((eccsize_mult == 1 && ((b2 ^ (b2 >> 1)) & 0x54) == 0x54) ||
+ (eccsize_mult == 2 && ((b2 ^ (b2 >> 1)) & 0x55) == 0x55))) {
+ /* single bit error */
+ /*
+ * rp17/rp15/13/11/9/7/5/3/1 indicate which byte is the faulty
+ * byte, cp 5/3/1 indicate the faulty bit.
+ * A lookup table (called addressbits) is used to filter
+ * the bits from the byte they are in.
+ * A marginal optimisation is possible by having three
+ * different lookup tables.
+ * One as we have now (for b0), one for b2
+ * (that would avoid the >> 1), and one for b1 (with all values
+ * << 4). However it was felt that introducing two more tables
+ * hardly justify the gain.
+ *
+ * The b2 shift is there to get rid of the lowest two bits.
+ * We could also do addressbits[b2] >> 1 but for the
+ * performance it does not make any difference
+ */
+ if (eccsize_mult == 1)
+ byte_addr = (addressbits[b1] << 4) + addressbits[b0];
+ else
+ byte_addr = (addressbits[b2 & 0x3] << 8) +
+ (addressbits[b1] << 4) + addressbits[b0];
+ bit_addr = addressbits[b2 >> 2];
+ /* flip the bit */
+ buf[byte_addr] ^= (1 << bit_addr);
+ return 1;
+
+ }
+ /* count nr of bits; use table lookup, faster than calculating it */
+ if ((bitsperbyte[b0] + bitsperbyte[b1] + bitsperbyte[b2]) == 1)
+ return 1; /* error in ECC data; no action needed */
+
+ pr_err("%s: uncorrectable ECC error\n", __func__);
+ return -EBADMSG;
+}
+EXPORT_SYMBOL(__nand_correct_data);
+
+/**
+ * nand_correct_data - [NAND Interface] Detect and correct bit error(s)
+ * @chip: NAND chip object
+ * @buf: raw data read from the chip
+ * @read_ecc: ECC from the chip
+ * @calc_ecc: the ECC calculated from raw data
+ *
+ * Detect and correct a 1 bit error for 256/512 byte block
+ */
+int nand_correct_data(struct nand_chip *chip, unsigned char *buf,
+ unsigned char *read_ecc, unsigned char *calc_ecc)
+{
+ bool sm_order = chip->ecc.options & NAND_ECC_SOFT_HAMMING_SM_ORDER;
+
+ return __nand_correct_data(buf, read_ecc, calc_ecc, chip->ecc.size,
+ sm_order);
+}
+EXPORT_SYMBOL(nand_correct_data);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Frans Meulenbroeks <fransmeulenbroeks@gmail.com>");
+MODULE_DESCRIPTION("Generic NAND ECC support");