summaryrefslogtreecommitdiff
path: root/target/linux/ubicom32/files/drivers/mtd/devices/nand-spi-er.c
diff options
context:
space:
mode:
Diffstat (limited to 'target/linux/ubicom32/files/drivers/mtd/devices/nand-spi-er.c')
-rw-r--r--target/linux/ubicom32/files/drivers/mtd/devices/nand-spi-er.c1017
1 files changed, 1017 insertions, 0 deletions
diff --git a/target/linux/ubicom32/files/drivers/mtd/devices/nand-spi-er.c b/target/linux/ubicom32/files/drivers/mtd/devices/nand-spi-er.c
new file mode 100644
index 0000000000..73938c8827
--- /dev/null
+++ b/target/linux/ubicom32/files/drivers/mtd/devices/nand-spi-er.c
@@ -0,0 +1,1017 @@
+/*
+ * Micron SPI-ER NAND Flash Memory
+ *
+ * (C) Copyright 2009, Ubicom, Inc.
+ *
+ * This file is part of the Ubicom32 Linux Kernel Port.
+ *
+ * The Ubicom32 Linux Kernel Port is free software: you can redistribute
+ * it and/or modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation, either version 2 of the
+ * License, or (at your option) any later version.
+ *
+ * The Ubicom32 Linux Kernel Port is distributed in the hope that it
+ * will be useful, but WITHOUT ANY WARRANTY; without even the implied
+ * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
+ * the GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with the Ubicom32 Linux Kernel Port. If not,
+ * see <http://www.gnu.org/licenses/>.
+*/
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/device.h>
+#include <linux/mutex.h>
+#include <linux/err.h>
+
+#include <linux/spi/spi.h>
+#include <linux/spi/flash.h>
+
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/partitions.h>
+
+#define NAND_SPI_ER_BLOCK_FROM_ROW(row) (row >> 6)
+
+#define NAND_SPI_ER_STATUS_P_FAIL (1 << 3)
+#define NAND_SPI_ER_STATUS_E_FAIL (1 << 2)
+#define NAND_SPI_ER_STATUS_OIP (1 << 0)
+
+#define NAND_SPI_ER_LAST_ROW_INVALID 0xFFFFFFFF
+#define NAND_SPI_ER_BAD_BLOCK_MARK_OFFSET 0x08
+
+struct nand_spi_er_device {
+ const char *name;
+
+ uint8_t id0;
+ uint8_t id1;
+
+ unsigned int blocks;
+ unsigned int pages_per_block;
+ unsigned int page_size;
+ unsigned int write_size;
+ unsigned int erase_size;
+};
+
+struct nand_spi_er {
+ char name[24];
+
+ const struct nand_spi_er_device *device;
+
+ struct mutex lock;
+ struct spi_device *spi;
+
+ struct mtd_info mtd;
+
+ unsigned int last_row; /* the last row we fetched */
+
+ /*
+ * Bad block table (MUST be last in strcuture)
+ */
+ unsigned long nbb;
+ unsigned long bbt[0];
+};
+
+const struct nand_spi_er_device nand_spi_er_devices[] = {
+ {
+ name: "MT29F1G01ZDC",
+ id0: 0x2C,
+ id1: 0x12,
+ blocks: 1024,
+ pages_per_block: 64,
+ page_size: 2048,
+ write_size: 512,
+ erase_size: 64 * 2048,
+ },
+ {
+ name: "MT29F1G01ZDC",
+ id0: 0x2C,
+ id1: 0x13,
+ blocks: 1024,
+ pages_per_block: 64,
+ page_size: 2048,
+ write_size: 512,
+ erase_size: 64 * 2048,
+ },
+};
+
+static int read_only = 0;
+module_param(read_only, int, 0);
+MODULE_PARM_DESC(read_only, "Leave device locked");
+
+/*
+ * nand_spi_er_get_feature
+ * Get Feature register
+ */
+static int nand_spi_er_get_feature(struct nand_spi_er *chip, int reg, uint8_t *data)
+{
+ uint8_t txbuf[2];
+ uint8_t rxbuf[1];
+ int res;
+
+ txbuf[0] = 0x0F;
+ txbuf[1] = reg;
+ res = spi_write_then_read(chip->spi, txbuf, 2, rxbuf, 1);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed get feature res=%d\n", chip->name, res);
+ return res;
+ }
+ *data = rxbuf[0];
+ return 0;
+}
+
+/*
+ * nand_spi_er_busywait
+ * Wait until the chip is not busy
+ */
+static int nand_spi_er_busywait(struct nand_spi_er *chip, uint8_t *data)
+{
+ int i;
+
+ for (i = 0; i < 100; i++) {
+ int res = nand_spi_er_get_feature(chip, 0xC0, data);
+ if (res) {
+ return res;
+ }
+ if (!(*data & NAND_SPI_ER_STATUS_OIP)) {
+ break;
+ }
+ }
+
+ return 0;
+}
+
+/*
+ * nand_spi_er_erase
+ * Erase a block, parameters must be block aligned
+ */
+static int nand_spi_er_erase(struct mtd_info *mtd, struct erase_info *instr)
+{
+ struct nand_spi_er *chip = mtd->priv;
+ struct spi_device *spi = chip->spi;
+ uint8_t txbuf[4];
+ int res;
+
+ DEBUG(MTD_DEBUG_LEVEL3, "%s: erase addr:%x len:%x\n", chip->name, instr->addr, instr->len);
+
+ if ((instr->addr + instr->len) > mtd->size) {
+ return -EINVAL;
+ }
+
+ if (instr->addr & (chip->device->erase_size - 1)) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: erase address is not aligned %x\n", chip->name, instr->addr);
+ return -EINVAL;
+ }
+
+ if (instr->len & (chip->device->erase_size - 1)) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: erase len is not aligned %x\n", chip->name, instr->len);
+ return -EINVAL;
+ }
+
+ mutex_lock(&chip->lock);
+ chip->last_row = NAND_SPI_ER_LAST_ROW_INVALID;
+
+ while (instr->len) {
+ uint32_t block = instr->addr >> 17;
+ uint32_t row = block << 6;
+ uint8_t stat;
+ DEBUG(MTD_DEBUG_LEVEL3, "%s: block erase row:%x block:%x addr:%x rem:%x\n", chip->name, row, block, instr->addr, instr->len);
+
+ /*
+ * Write enable
+ */
+ txbuf[0] = 0x06;
+ res = spi_write(spi, txbuf, 1);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed write enable res=%d\n", chip->name, res);
+ mutex_unlock(&chip->lock);
+ return res;
+ }
+
+ /*
+ * Test for bad block
+ */
+ if (test_bit(block, chip->bbt)) {
+ instr->fail_addr = block << 17;
+ instr->state = MTD_ERASE_FAILED;
+ res = -EBADMSG;
+ goto done;
+ }
+
+ /*
+ * Block erase
+ */
+ txbuf[0] = 0xD8;
+ txbuf[1] = 0x00;
+ txbuf[2] = row >> 8;
+ txbuf[3] = row & 0xFF;
+ res = spi_write(spi, txbuf, 4);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed block erase res=%d\n", chip->name, res);
+ instr->fail_addr = block << 17;
+ instr->state = MTD_ERASE_FAILED;
+ goto done;
+ }
+
+ /*
+ * Wait
+ */
+ res = nand_spi_er_busywait(chip, &stat);
+ if (res || (stat & NAND_SPI_ER_STATUS_OIP)) {
+ instr->fail_addr = block << 17;
+ instr->state = MTD_ERASE_FAILED;
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: chip is busy or nonresponsive res=%d stat=%02x\n", chip->name, res, stat);
+ if (res) {
+ goto done;
+ }
+
+ /*
+ * Chip is stuck?
+ */
+ res = -EIO;
+ goto done;
+ }
+
+ /*
+ * Check the status register
+ */
+ if (stat & NAND_SPI_ER_STATUS_E_FAIL) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: E_FAIL signalled (%02x)\n", chip->name, stat);
+ instr->fail_addr = block << 17;
+ instr->state = MTD_ERASE_FAILED;
+ goto done;
+ }
+
+ /*
+ * Next
+ */
+ block++;
+ instr->len -= chip->device->erase_size;
+ instr->addr += chip->device->erase_size;
+ }
+
+ instr->state = MTD_ERASE_DONE;
+
+ mutex_unlock(&chip->lock);
+ return 0;
+
+done:
+ /*
+ * Write disable
+ */
+ txbuf[0] = 0x04;
+ res = spi_write(spi, txbuf, 1);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed write disable res=%d\n", chip->name, res);
+ }
+
+ mutex_unlock(&chip->lock);
+
+ mtd_erase_callback(instr);
+ return 0;
+}
+
+/*
+ * nand_spi_er_read
+ *
+ * return -EUCLEAN: ecc error recovered
+ * return -EBADMSG: ecc error not recovered
+*/
+static int nand_spi_er_read(struct mtd_info *mtd, loff_t from, size_t len,
+ size_t *retlen, u_char *buf)
+{
+ struct nand_spi_er *chip = mtd->priv;
+ struct spi_device *spi = chip->spi;
+
+ uint32_t row;
+ uint32_t column;
+ int retval = 0;
+
+ *retlen = 0;
+ DEBUG(MTD_DEBUG_LEVEL2, "%s: read block from %llx len %d into %p\n", chip->name, from, len, buf);
+
+ /*
+ * Zero length reads, nothing to do
+ */
+ if (len == 0) {
+ return 0;
+ }
+
+ /*
+ * Reject reads which go over the end of the flash
+ */
+ if ((from + len) > mtd->size) {
+ return -EINVAL;
+ }
+
+ /*
+ * Get the row and column address to start at
+ */
+ row = from >> 11;
+ column = from & 0x7FF;
+ DEBUG(MTD_DEBUG_LEVEL3, "%s: row=%x %d column=%x %d last_row=%x %d\n", chip->name, row, row, column, column, chip->last_row, chip->last_row);
+
+ /*
+ * Read the data from the chip
+ */
+ mutex_lock(&chip->lock);
+ while (len) {
+ uint8_t stat;
+ uint8_t txbuf[4];
+ struct spi_message message;
+ struct spi_transfer x[2];
+ int res;
+ size_t toread;
+
+ /*
+ * Figure out how much to read
+ *
+ * If we are reading from the middle of a page then the most we
+ * can read is to the end of the page
+ */
+ toread = len;
+ if (toread > (chip->device->page_size - column)) {
+ toread = chip->device->page_size - column;
+ }
+
+ DEBUG(MTD_DEBUG_LEVEL3, "%s: buf=%p toread=%x row=%x column=%x last_row=%x\n", chip->name, buf, toread, row, column, chip->last_row);
+
+ if (chip->last_row != row) {
+ /*
+ * Check if the block is bad
+ */
+ if (test_bit(NAND_SPI_ER_BLOCK_FROM_ROW(row), chip->bbt)) {
+ mutex_unlock(&chip->lock);
+ return -EBADMSG;
+ }
+
+ /*
+ * Load the appropriate page
+ */
+ txbuf[0] = 0x13;
+ txbuf[1] = 0x00;
+ txbuf[2] = row >> 8;
+ txbuf[3] = row & 0xFF;
+ res = spi_write(spi, txbuf, 4);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed page load res=%d\n", chip->name, res);
+ mutex_unlock(&chip->lock);
+ return res;
+ }
+
+ /*
+ * Wait
+ */
+ res = nand_spi_er_busywait(chip, &stat);
+ if (res || (stat & NAND_SPI_ER_STATUS_OIP)) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: chip is busy or nonresponsive res=%d stat=%02x\n", chip->name, res, stat);
+ if (res) {
+ mutex_unlock(&chip->lock);
+ return res;
+ }
+
+ /*
+ * Chip is stuck?
+ */
+ mutex_unlock(&chip->lock);
+ return -EIO;
+ }
+
+ /*
+ * Check the ECC bits
+ */
+ stat >>= 4;
+ if (stat == 1) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: ECC recovered, row=%x\n", chip->name, row);
+ retval = -EUCLEAN;
+ }
+ if (stat == 2) {
+ DEBUG(MTD_DEBUG_LEVEL0, "%s: failed ECC, row=%x\n", chip->name, row);
+ chip->last_row = NAND_SPI_ER_LAST_ROW_INVALID;
+ mutex_unlock(&chip->lock);
+ return -EBADMSG;
+ }
+
+ }
+
+ chip->last_row = row;
+
+ /*
+ * Read out the data
+ */
+ spi_message_init(&message);
+ memset(x, 0, sizeof(x));
+
+ txbuf[0] = 0x03;
+ txbuf[1] = column >> 8;
+ txbuf[2] = column & 0xFF;
+ txbuf[3] = 0;
+ x[0].tx_buf = txbuf;
+ x[0].len = 4;
+ spi_message_add_tail(&x[0], &message);
+
+ x[1].rx_buf = buf;
+ x[1].len = toread;
+ spi_message_add_tail(&x[1], &message);
+
+ res = spi_sync(spi, &message);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed data read res=%d\n", chip->name, res);
+ mutex_unlock(&chip->lock);
+ return res;
+ }
+ buf += toread;
+ len -= toread;
+ *retlen += toread;
+
+ /*
+ * For the next page, increment the row and always start at column 0
+ */
+ column = 0;
+ row++;
+ }
+
+ mutex_unlock(&chip->lock);
+ return retval;
+}
+
+/*
+ * nand_spi_er_write
+ */
+#define NOT_ALIGNED(x) ((x & (device->write_size - 1)) != 0)
+static int nand_spi_er_write(struct mtd_info *mtd, loff_t to, size_t len,
+ size_t *retlen, const u_char *buf)
+{
+ struct nand_spi_er *chip = mtd->priv;
+ struct spi_device *spi = chip->spi;
+ const struct nand_spi_er_device *device = chip->device;
+ uint32_t row;
+ uint32_t col;
+ uint8_t txbuf[4];
+ int res;
+ size_t towrite;
+
+ DEBUG(MTD_DEBUG_LEVEL2, "%s: write block to %llx len %d from %p\n", chip->name, to, len, buf);
+
+ *retlen = 0;
+
+ /*
+ * nothing to write
+ */
+ if (!len) {
+ return 0;
+ }
+
+ /*
+ * Reject writes which go over the end of the flash
+ */
+ if ((to + len) > mtd->size) {
+ return -EINVAL;
+ }
+
+ /*
+ * Check to see if everything is page aligned
+ */
+ if (NOT_ALIGNED(to) || NOT_ALIGNED(len)) {
+ printk(KERN_NOTICE "nand_spi_er_write: Attempt to write non page aligned data\n");
+ return -EINVAL;
+ }
+
+ mutex_lock(&chip->lock);
+ chip->last_row = NAND_SPI_ER_LAST_ROW_INVALID;
+
+ /*
+ * If the first write is a partial write then write at most the number of
+ * bytes to get us page aligned and then the remainder will be
+ * page aligned. The last bit may be a partial page as well.
+ */
+ col = to & (device->page_size - 1);
+ towrite = device->page_size - col;
+ if (towrite > len) {
+ towrite = len;
+ }
+
+ /*
+ * Write the data
+ */
+ row = to >> 11;
+ while (len) {
+ struct spi_message message;
+ struct spi_transfer x[2];
+ uint8_t stat;
+
+ DEBUG(MTD_DEBUG_LEVEL3, "%s: write %p to row:%x col:%x len:%x rem:%x\n", chip->name, buf, row, col, towrite, len);
+
+ /*
+ * Write enable
+ */
+ txbuf[0] = 0x06;
+ res = spi_write(spi, txbuf, 1);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed write enable res=%d\n", chip->name, res);
+ mutex_unlock(&chip->lock);
+ return res;
+ }
+
+ /*
+ * Write the data into the cache
+ */
+ spi_message_init(&message);
+ memset(x, 0, sizeof(x));
+ txbuf[0] = 0x02;
+ txbuf[1] = col >> 8;
+ txbuf[2] = col & 0xFF;
+ x[0].tx_buf = txbuf;
+ x[0].len = 3;
+ spi_message_add_tail(&x[0], &message);
+ x[1].tx_buf = buf;
+ x[1].len = towrite;
+ spi_message_add_tail(&x[1], &message);
+ res = spi_sync(spi, &message);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed cache write res=%d\n", chip->name, res);
+ goto done;
+ }
+
+ /*
+ * Program execute
+ */
+ txbuf[0] = 0x10;
+ txbuf[1] = 0x00;
+ txbuf[2] = row >> 8;
+ txbuf[3] = row & 0xFF;
+ res = spi_write(spi, txbuf, 4);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed prog execute res=%d\n", chip->name, res);
+ goto done;
+ }
+
+ /*
+ * Wait
+ */
+ res = nand_spi_er_busywait(chip, &stat);
+ if (res || (stat & NAND_SPI_ER_STATUS_OIP)) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: chip is busy or nonresponsive res=%d stat=%02x\n", chip->name, res, stat);
+ if (res) {
+ goto done;
+ }
+
+ /*
+ * Chip is stuck?
+ */
+ res = -EIO;
+ goto done;
+ }
+
+ if (stat & (1 << 3)) {
+ res = -EBADMSG;
+ goto done;
+ }
+
+ row++;
+ buf += towrite;
+ len -= towrite;
+ *retlen += towrite;
+
+ /*
+ * At this point, we are always page aligned so start at column 0.
+ * Note we may not have a full page to write at the end, hence the
+ * check if towrite > len.
+ */
+ col = 0;
+ towrite = device->page_size;
+ if (towrite > len) {
+ towrite = len;
+ }
+ }
+
+ mutex_unlock(&chip->lock);
+ return res;
+
+done:
+ /*
+ * Write disable
+ */
+ txbuf[0] = 0x04;
+ res = spi_write(spi, txbuf, 1);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed write disable res=%d\n", chip->name, res);
+ }
+
+ mutex_unlock(&chip->lock);
+
+ return res;
+}
+
+/*
+ * nand_spi_er_isbad
+ */
+static int nand_spi_er_isbad(struct mtd_info *mtd, loff_t ofs)
+{
+ struct nand_spi_er *chip = mtd->priv;
+ uint32_t block;
+
+ if (ofs & (chip->device->erase_size - 1)) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: address not aligned %llx\n", chip->name, ofs);
+ return -EINVAL;
+ }
+
+ block = ofs >> 17;
+
+ return test_bit(block, chip->bbt);
+}
+
+/*
+ * nand_spi_er_markbad
+ */
+static int nand_spi_er_markbad(struct mtd_info *mtd, loff_t ofs)
+{
+ struct nand_spi_er *chip = mtd->priv;
+ struct spi_device *spi = chip->spi;
+ uint32_t block;
+ uint32_t row;
+ uint8_t txbuf[7];
+ int res;
+ uint8_t stat;
+
+ if (ofs & (chip->device->erase_size - 1)) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: address not aligned %llx\n", chip->name, ofs);
+ return -EINVAL;
+ }
+
+ block = ofs >> 17;
+
+ /*
+ * If it's already marked bad, no need to mark it
+ */
+ if (test_bit(block, chip->bbt)) {
+ return 0;
+ }
+
+ /*
+ * Mark it in our cache
+ */
+ __set_bit(block, chip->bbt);
+
+ /*
+ * Write the user bad block mark. If it fails, then we really
+ * can't do anything about it.
+ */
+ mutex_lock(&chip->lock);
+ chip->last_row = NAND_SPI_ER_LAST_ROW_INVALID;
+
+ /*
+ * Write enable
+ */
+ txbuf[0] = 0x06;
+ res = spi_write(spi, txbuf, 1);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed write enable res=%d\n", chip->name, res);
+ mutex_unlock(&chip->lock);
+ return res;
+ }
+
+ /*
+ * Write the mark
+ */
+ txbuf[0] = 0x84;
+ txbuf[1] = 0x08;
+ txbuf[2] = NAND_SPI_ER_BAD_BLOCK_MARK_OFFSET;
+ txbuf[3] = 0xde;
+ txbuf[4] = 0xad;
+ txbuf[5] = 0xbe;
+ txbuf[6] = 0xef;
+ res = spi_write(spi, txbuf, 7);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed write mark res=%d\n", chip->name, res);
+ goto done;
+ }
+
+ /*
+ * Program execute
+ */
+ row = ofs >> 11;
+ txbuf[0] = 0x10;
+ txbuf[1] = 0x00;
+ txbuf[2] = row >> 8;
+ txbuf[3] = row & 0xFF;
+ res = spi_write(spi, txbuf, 4);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed program execute res=%d\n", chip->name, res);
+ goto done;
+ }
+
+ /*
+ * Wait
+ */
+ res = nand_spi_er_busywait(chip, &stat);
+ if (res || (stat & NAND_SPI_ER_STATUS_OIP)) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: chip is busy or nonresponsive res=%d stat=%02x\n", chip->name, res, stat);
+ if (res) {
+ goto done;
+ }
+
+ /*
+ * Chip is stuck?
+ */
+ res = -EIO;
+ goto done;
+ }
+
+ if (stat & (1 << 3)) {
+ res = -EBADMSG;
+ }
+
+done:
+ /*
+ * Write disable
+ */
+ txbuf[0] = 0x04;
+ res = spi_write(spi, txbuf, 1);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed write disable res=%d\n", chip->name, res);
+ }
+
+ mutex_unlock(&chip->lock);
+
+ return res;
+}
+
+/*
+ * nand_spi_er_read_bbt
+ */
+static int nand_spi_er_read_bbt(struct nand_spi_er *chip)
+{
+ int j;
+ for (j = 0; j < chip->device->blocks; j++) {
+ uint8_t txbuf[4];
+ uint8_t rxbuf[16];
+ uint32_t bbmark;
+ int res;
+ unsigned short row = j << 6;
+ uint8_t stat;
+
+ /*
+ * Read Page
+ */
+ txbuf[0] = 0x13;
+ txbuf[1] = 0x00;
+ txbuf[2] = row >> 8;
+ txbuf[3] = row & 0xFF;
+ res = spi_write(chip->spi, txbuf, 4);
+ if (res) {
+ return res;
+ }
+
+ /*
+ * Wait
+ */
+ res = nand_spi_er_busywait(chip, &stat);
+ if (res || (stat & NAND_SPI_ER_STATUS_OIP)) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: chip is busy or nonresponsive res=%d stat=%02x\n", chip->name, res, stat);
+ if (res) {
+ return res;
+ }
+
+ /*
+ * Chip is stuck?
+ */
+ return -EIO;
+ }
+
+ /*
+ * Check factory bad block mark
+ */
+ txbuf[0] = 0x03;
+ txbuf[1] = 0x08;
+ txbuf[2] = 0x00;
+ txbuf[3] = 0x00;
+ res = spi_write_then_read(chip->spi, txbuf, 4, rxbuf, 16);
+ if (res) {
+ return res;
+ }
+ if (rxbuf[0] != 0xFF) {
+ chip->nbb++;
+ __set_bit(j, chip->bbt);
+ continue;
+ }
+
+ memcpy(&bbmark, &rxbuf[8], 4);
+ if (bbmark == 0xdeadbeef) {
+ chip->nbb++;
+ __set_bit(j, chip->bbt);
+ }
+ }
+
+#if defined(CONFIG_MTD_DEBUG) && (MTD_DEBUG_LEVEL3 <= CONFIG_MTD_DEBUG_VERBOSE)
+ printk("%s: Bad Block Table:", chip->name);
+ for (j = 0; j < chip->device->blocks; j++) {
+ if ((j % 64) == 0) {
+ printk("\n%s: block %03x: ", chip->name, j);
+ }
+ printk("%c", test_bit(j, chip->bbt) ? 'X' : '.');
+ }
+ printk("\n%s: Bad Block Numbers: ", chip->name);
+ for (j = 0; j < chip->device->blocks; j++) {
+ if (test_bit(j, chip->bbt)) {
+ printk("%x ", j);
+ }
+ }
+ printk("\n");
+#endif
+
+ return 0;
+}
+
+#ifndef MODULE
+/*
+ * Called at boot time:
+ *
+ * nand_spi_er=read_only
+ * if read_only specified then do not unlock device
+ */
+static int __init nand_spi_er_setup(char *str)
+{
+ if (str && (strncasecmp(str, "read_only", 9) == 0)) {
+ read_only = 1;
+ }
+ return 0;
+}
+
+__setup("nand_spi_er=", nand_spi_er_setup);
+#endif
+
+/*
+ * nand_spi_er_probe
+ * Detect and initialize nand_spi_er device.
+ */
+static int __devinit nand_spi_er_probe(struct spi_device *spi)
+{
+ uint8_t txbuf[3];
+ uint8_t rxbuf[2];
+ int i;
+ int res;
+ size_t bbt_bytes;
+ struct nand_spi_er *chip;
+ const struct nand_spi_er_device *device;
+
+ res = spi_setup(spi);
+ if (res) {
+ return res;
+ }
+
+ /*
+ * Reset
+ */
+ for (i = 0; i < 2; i++) {
+ txbuf[0] = 0xFF;
+ res = spi_write(spi, txbuf, 1);
+ if (res) {
+ return res;
+ }
+ udelay(250);
+ }
+ udelay(1000);
+
+ /*
+ * Read ID
+ */
+ txbuf[0] = 0x9F;
+ txbuf[1] = 0x00;
+ res = spi_write_then_read(spi, txbuf, 2, rxbuf, 2);
+ if (res) {
+ return res;
+ }
+
+ device = nand_spi_er_devices;
+ for (i = 0; i < ARRAY_SIZE(nand_spi_er_devices); i++) {
+ if ((device->id0 == rxbuf[0]) && (device->id1 == rxbuf[1])) {
+ break;
+ }
+ device++;
+ }
+ if (i == ARRAY_SIZE(nand_spi_er_devices)) {
+ return -ENODEV;
+ }
+
+ /*
+ * Initialize our chip structure
+ */
+ bbt_bytes = DIV_ROUND_UP(device->blocks, BITS_PER_BYTE);
+ chip = kzalloc(sizeof(struct nand_spi_er) + bbt_bytes, GFP_KERNEL);
+ if (!chip) {
+ return -ENOMEM;
+ }
+ snprintf(chip->name, sizeof(chip->name), "%s.%d.%d", device->name, spi->master->bus_num, spi->chip_select);
+
+ chip->spi = spi;
+ chip->device = device;
+ chip->last_row = NAND_SPI_ER_LAST_ROW_INVALID;
+
+ mutex_init(&chip->lock);
+
+ chip->mtd.type = MTD_NANDFLASH;
+ chip->mtd.flags = MTD_WRITEABLE;
+
+ /*
+ * #blocks * block size * n blocks
+ */
+ chip->mtd.size = device->blocks * device->pages_per_block * device->page_size;
+ chip->mtd.erasesize = device->erase_size;
+
+ /*
+ * 1 page, optionally we can support partial write (512)
+ */
+ chip->mtd.writesize = device->write_size;
+ chip->mtd.name = device->name;
+ chip->mtd.erase = nand_spi_er_erase;
+ chip->mtd.read = nand_spi_er_read;
+ chip->mtd.write = nand_spi_er_write;
+ chip->mtd.block_isbad = nand_spi_er_isbad;
+ chip->mtd.block_markbad = nand_spi_er_markbad;
+ chip->mtd.priv = chip;
+
+ /*
+ * Cache the bad block table
+ */
+ res = nand_spi_er_read_bbt(chip);
+ if (res) {
+ kfree(chip);
+ return res;
+ }
+
+ /*
+ * Un/lock the chip
+ */
+ txbuf[0] = 0x1F;
+ txbuf[1] = 0xA0;
+ if (read_only) {
+ txbuf[2] = 0x38;
+ } else {
+ txbuf[2] = 0x00;
+ }
+ res = spi_write(spi, txbuf, 3);
+ if (res) {
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: failed lock operation res=%d\n", chip->name, res);
+ mutex_unlock(&chip->lock);
+ return res;
+ }
+
+ spi_set_drvdata(spi, chip);
+
+ printk(KERN_INFO "%s: added device %s size: %u KBytes %u bad blocks %s\n", spi->dev.bus_id, chip->mtd.name, DIV_ROUND_UP(chip->mtd.size, 1024), chip->nbb, read_only ? "[read only]" : "");
+ return add_mtd_device(&chip->mtd);
+}
+
+/*
+ * nand_spi_er_remove
+ */
+static int __devexit nand_spi_er_remove(struct spi_device *spi)
+{
+ struct nand_spi_er *chip = spi_get_drvdata(spi);
+ int status = 0;
+
+ DEBUG(MTD_DEBUG_LEVEL1, "%s: remove\n", spi->dev.bus_id);
+ status = del_mtd_device(&chip->mtd);
+ if (status == 0)
+ kfree(chip);
+ return status;
+}
+
+static struct spi_driver nand_spi_er_driver = {
+ .driver = {
+ .name = "nand-spi-er",
+ .bus = &spi_bus_type,
+ .owner = THIS_MODULE,
+ },
+
+ .probe = nand_spi_er_probe,
+ .remove = __devexit_p(nand_spi_er_remove),
+
+ /* FIXME: investigate suspend and resume... */
+};
+
+/*
+ * nand_spi_er_init
+ */
+static int __init nand_spi_er_init(void)
+{
+ return spi_register_driver(&nand_spi_er_driver);
+}
+module_init(nand_spi_er_init);
+
+/*
+ * nand_spi_er_exit
+ */
+static void __exit nand_spi_er_exit(void)
+{
+ spi_unregister_driver(&nand_spi_er_driver);
+}
+module_exit(nand_spi_er_exit);
+
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Patrick Tjin");
+MODULE_DESCRIPTION("MTD nand_spi_er driver");