io.c

基于linux-2.6.28的mtd驱动

/*
 * Copyright (c) International Business Machines Corp., 2006
 * Copyright (c) Nokia Corporation, 2006, 2007
 *
 * This program 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.
 *
 * This program 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 this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
 *
 * Author: Artem Bityutskiy (Битюцкий Артём)
 */

/*
 * UBI input/output sub-system.
 *
 * This sub-system provides a uniform way to work with all kinds of the
 * underlying MTD devices. It also implements handy functions for reading and
 * writing UBI headers.
 *
 * We are trying to have a paranoid mindset and not to trust to what we read
 * from the flash media in order to be more secure and robust. So this
 * sub-system validates every single header it reads from the flash media.
 *
 * Some words about how the eraseblock headers are stored.
 *
 * The erase counter header is always stored at offset zero. By default, the
 * VID header is stored after the EC header at the closest aligned offset
 * (i.e. aligned to the minimum I/O unit size). Data starts next to the VID
 * header at the closest aligned offset. But this default layout may be
 * changed. For example, for different reasons (e.g., optimization) UBI may be
 * asked to put the VID header at further offset, and even at an unaligned
 * offset. Of course, if the offset of the VID header is unaligned, UBI adds
 * proper padding in front of it. Data offset may also be changed but it has to
 * be aligned.
 *
 * About minimal I/O units. In general, UBI assumes flash device model where
 * there is only one minimal I/O unit size. E.g., in case of NOR flash it is 1,
 * in case of NAND flash it is a NAND page, etc. This is reported by MTD in the
 * @ubi->mtd->writesize field. But as an exception, UBI admits of using another
 * (smaller) minimal I/O unit size for EC and VID headers to make it possible
 * to do different optimizations.
 *
 * This is extremely useful in case of NAND flashes which admit of several
 * write operations to one NAND page. In this case UBI can fit EC and VID
 * headers at one NAND page. Thus, UBI may use "sub-page" size as the minimal
 * I/O unit for the headers (the @ubi->hdrs_min_io_size field). But it still
 * reports NAND page size (@ubi->min_io_size) as a minimal I/O unit for the UBI
 * users.
 *
 * Example: some Samsung NANDs with 2KiB pages allow 4x 512-byte writes, so
 * although the minimal I/O unit is 2K, UBI uses 512 bytes for EC and VID
 * headers.
 *
 * Q: why not just to treat sub-page as a minimal I/O unit of this flash
 * device, e.g., make @ubi->min_io_size = 512 in the example above?
 *
 * A: because when writing a sub-page, MTD still writes a full 2K page but the
 * bytes which are no relevant to the sub-page are 0xFF. So, basically, writing
 * 4x512 sub-pages is 4 times slower then writing one 2KiB NAND page. Thus, we
 * prefer to use sub-pages only for EV and VID headers.
 *
 * As it was noted above, the VID header may start at a non-aligned offset.
 * For example, in case of a 2KiB page NAND flash with a 512 bytes sub-page,
 * the VID header may reside at offset 1984 which is the last 64 bytes of the
 * last sub-page (EC header is always at offset zero). This causes some
 * difficulties when reading and writing VID headers.
 *
 * Suppose we have a 64-byte buffer and we read a VID header at it. We change
 * the data and want to write this VID header out. As we can only write in
 * 512-byte chunks, we have to allocate one more buffer and copy our VID header
 * to offset 448 of this buffer.
 *
 * The I/O sub-system does the following trick in order to avoid this extra
 * copy. It always allocates a @ubi->vid_hdr_alsize bytes buffer for the VID
 * header and returns a pointer to offset @ubi->vid_hdr_shift of this buffer.
 * When the VID header is being written out, it shifts the VID header pointer
 * back and writes the whole sub-page.
 */

#include <linux/crc32.h>
#include <linux/err.h>
#include "ubi.h"

#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
static int paranoid_check_not_bad(const struct ubi_device *ubi, int pnum);
static int paranoid_check_peb_ec_hdr(const struct ubi_device *ubi, int pnum);
static int paranoid_check_ec_hdr(const struct ubi_device *ubi, int pnum,
				 const struct ubi_ec_hdr *ec_hdr);
static int paranoid_check_peb_vid_hdr(const struct ubi_device *ubi, int pnum);
static int paranoid_check_vid_hdr(const struct ubi_device *ubi, int pnum,
				  const struct ubi_vid_hdr *vid_hdr);
static int paranoid_check_all_ff(struct ubi_device *ubi, int pnum, int offset,
				 int len);
#else
#define paranoid_check_not_bad(ubi, pnum) 0
#define paranoid_check_peb_ec_hdr(ubi, pnum)  0
#define paranoid_check_ec_hdr(ubi, pnum, ec_hdr)  0
#define paranoid_check_peb_vid_hdr(ubi, pnum) 0
#define paranoid_check_vid_hdr(ubi, pnum, vid_hdr) 0
#define paranoid_check_all_ff(ubi, pnum, offset, len) 0
#endif

/**
 * ubi_io_read - read data from a physical eraseblock.
 * @ubi: UBI device description object
 * @buf: buffer where to store the read data
 * @pnum: physical eraseblock number to read from
 * @offset: offset within the physical eraseblock from where to read
 * @len: how many bytes to read
 *
 * This function reads data from offset @offset of physical eraseblock @pnum
 * and stores the read data in the @buf buffer. The following return codes are
 * possible:
 *
 * o %0 if all the requested data were successfully read;
 * o %UBI_IO_BITFLIPS if all the requested data were successfully read, but
 *   correctable bit-flips were detected; this is harmless but may indicate
 *   that this eraseblock may become bad soon (but do not have to);
 * o %-EBADMSG if the MTD subsystem reported about data integrity problems, for
 *   example it can be an ECC error in case of NAND; this most probably means
 *   that the data is corrupted;
 * o %-EIO if some I/O error occurred;
 * o other negative error codes in case of other errors.
 */
int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
		int len)
{
	int err, retries = 0;
	size_t read;
	loff_t addr;

	dbg_io("read %d bytes from PEB %d:%d", len, pnum, offset);

	ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
	ubi_assert(offset >= 0 && offset + len <= ubi->peb_size);
	ubi_assert(len > 0);

	err = paranoid_check_not_bad(ubi, pnum);
	if (err)
		return err > 0 ? -EINVAL : err;

	addr = (loff_t)pnum * ubi->peb_size + offset;
retry:
	err = ubi->mtd->read(ubi->mtd, addr, len, &read, buf);
	if (err) {
		if (err == -EUCLEAN) {
			/*
			 * -EUCLEAN is reported if there was a bit-flip which
			 * was corrected, so this is harmless.
			 *
			 * We do not report about it here unless debugging is
			 * enabled. A corresponding message will be printed
			 * later, when it is has been scrubbed.
			 */
			dbg_msg("fixable bit-flip detected at PEB %d", pnum);
			ubi_assert(len == read);
			return UBI_IO_BITFLIPS;
		}

		if (read != len && retries++ < UBI_IO_RETRIES) {
			dbg_io("error %d while reading %d bytes from PEB %d:%d,"
			       " read only %zd bytes, retry",
			       err, len, pnum, offset, read);
			yield();
			goto retry;
		}

		ubi_err("error %d while reading %d bytes from PEB %d:%d, "
			"read %zd bytes", err, len, pnum, offset, read);
		ubi_dbg_dump_stack();

		/*
		 * The driver should never return -EBADMSG if it failed to read
		 * all the requested data. But some buggy drivers might do
		 * this, so we change it to -EIO.
		 */
		if (read != len && err == -EBADMSG) {
			ubi_assert(0);
			err = -EIO;
		}
	} else {
		ubi_assert(len == read);

		if (ubi_dbg_is_bitflip()) {
			dbg_gen("bit-flip (emulated)");
			err = UBI_IO_BITFLIPS;
		}
	}

	return err;
}

/**
 * ubi_io_write - write data to a physical eraseblock.
 * @ubi: UBI device description object
 * @buf: buffer with the data to write
 * @pnum: physical eraseblock number to write to
 * @offset: offset within the physical eraseblock where to write
 * @len: how many bytes to write
 *
 * This function writes @len bytes of data from buffer @buf to offset @offset
 * of physical eraseblock @pnum. If all the data were successfully written,
 * zero is returned. If an error occurred, this function returns a negative
 * error code. If %-EIO is returned, the physical eraseblock most probably went
 * bad.
 *
 * Note, in case of an error, it is possible that something was still written
 * to the flash media, but may be some garbage.
 */
int ubi_io_write(struct ubi_device *ubi, const void *buf, int pnum, int offset,
		 int len)
{
	int err;
	size_t written;
	loff_t addr;

	dbg_io("write %d bytes to PEB %d:%d", len, pnum, offset);

	ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
	ubi_assert(offset >= 0 && offset + len <= ubi->peb_size);
	ubi_assert(offset % ubi->hdrs_min_io_size == 0);
	ubi_assert(len > 0 && len % ubi->hdrs_min_io_size == 0);

	if (ubi->ro_mode) {
		ubi_err("read-only mode");
		return -EROFS;
	}

	/* The below has to be compiled out if paranoid checks are disabled */

	err = paranoid_check_not_bad(ubi, pnum);
	if (err)
		return err > 0 ? -EINVAL : err;

	/* The area we are writing to has to contain all 0xFF bytes */
	err = paranoid_check_all_ff(ubi, pnum, offset, len);
	if (err)
		return err > 0 ? -EINVAL : err;

	if (offset >= ubi->leb_start) {
		/*
		 * We write to the data area of the physical eraseblock. Make
		 * sure it has valid EC and VID headers.
		 */
		err = paranoid_check_peb_ec_hdr(ubi, pnum);
		if (err)
			return err > 0 ? -EINVAL : err;
		err = paranoid_check_peb_vid_hdr(ubi, pnum);
		if (err)
			return err > 0 ? -EINVAL : err;
	}

	if (ubi_dbg_is_write_failure()) {
		dbg_err("cannot write %d bytes to PEB %d:%d "
			"(emulated)", len, pnum, offset);
		ubi_dbg_dump_stack();
		return -EIO;
	}

	addr = (loff_t)pnum * ubi->peb_size + offset;
	err = ubi->mtd->write(ubi->mtd, addr, len, &written, buf);
	if (err) {
		ubi_err("error %d while writing %d bytes to PEB %d:%d, written"
			" %zd bytes", err, len, pnum, offset, written);
		ubi_dbg_dump_stack();
	} else
		ubi_assert(written == len);

	return err;
}

/**
 * erase_callback - MTD erasure call-back.
 * @ei: MTD erase information object.
 *
 * Note, even though MTD erase interface is asynchronous, all the current
 * implementations are synchronous anyway.
 */
static void erase_callback(struct erase_info *ei)
{
	wake_up_interruptible((wait_queue_head_t *)ei->priv);
}

/**
 * do_sync_erase - synchronously erase a physical eraseblock.
 * @ubi: UBI device description object
 * @pnum: the physical eraseblock number to erase
 *
 * This function synchronously erases physical eraseblock @pnum and returns
 * zero in case of success and a negative error code in case of failure. If
 * %-EIO is returned, the physical eraseblock most probably went bad.
 */
static int do_sync_erase(struct ubi_device *ubi, int pnum)
{
	int err, retries = 0;
	struct erase_info ei;
	wait_queue_head_t wq;

	dbg_io("erase PEB %d", pnum);

retry:
	init_waitqueue_head(&wq);
	memset(&ei, 0, sizeof(struct erase_info));

	ei.mtd      = ubi->mtd;
	ei.addr     = (loff_t)pnum * ubi->peb_size;
	ei.len      = ubi->peb_size;
	ei.callback = erase_callback;
	ei.priv     = (unsigned long)&wq;

	err = ubi->mtd->erase(ubi->mtd, &ei);
	if (err) {
		if (retries++ < UBI_IO_RETRIES) {
			dbg_io("error %d while erasing PEB %d, retry",
			       err, pnum);
			yield();
			goto retry;
		}
		ubi_err("cannot erase PEB %d, error %d", pnum, err);
		ubi_dbg_dump_stack();
		return err;
	}

	err = wait_event_interruptible(wq, ei.state == MTD_ERASE_DONE ||
					   ei.state == MTD_ERASE_FAILED);
	if (err) {
		ubi_err("interrupted PEB %d erasure", pnum);
		return -EINTR;
	}

	if (ei.state == MTD_ERASE_FAILED) {
		if (retries++ < UBI_IO_RETRIES) {
			dbg_io("error while erasing PEB %d, retry", pnum);
			yield();
			goto retry;
		}
		ubi_err("cannot erase PEB %d", pnum);
		ubi_dbg_dump_stack();
		return -EIO;
	}

	err = paranoid_check_all_ff(ubi, pnum, 0, ubi->peb_size);
	if (err)
		return err > 0 ? -EINVAL : err;

	if (ubi_dbg_is_erase_failure() && !err) {
		dbg_err("cannot erase PEB %d (emulated)", pnum);
		return -EIO;
	}

	return 0;
}

/**
 * check_pattern - check if buffer contains only a certain byte pattern.
 * @buf: buffer to check
 * @patt: the pattern to check
 * @size: buffer size in bytes
 *
 * This function returns %1 in there are only @patt bytes in @buf, and %0 if
 * something else was also found.
 */
static int check_pattern(const void *buf, uint8_t patt, int size)
{
	int i;

	for (i = 0; i < size; i++)
		if (((const uint8_t *)buf)[i] != patt)
			return 0;
	return 1;
}

/* Patterns to write to a physical eraseblock when torturing it */
static uint8_t patterns[] = {0xa5, 0x5a, 0x0};

/**
 * torture_peb - test a supposedly bad physical eraseblock.
 * @ubi: UBI device description object
 * @pnum: the physical eraseblock number to test
 *
 * This function returns %-EIO if the physical eraseblock did not pass the
 * test, a positive number of erase operations done if the test was
 * successfully passed, and other negative error codes in case of other errors.
 */
static int torture_peb(struct ubi_device *ubi, int pnum)
{
	int err, i, patt_count;

	ubi_msg("run torture test for PEB %d", pnum);
	patt_count = ARRAY_SIZE(patterns);
	ubi_assert(patt_count > 0);

	mutex_lock(&ubi->buf_mutex);
	for (i = 0; i < patt_count; i++) {
		err = do_sync_erase(ubi, pnum);
		if (err)
			goto out;

		/* Make sure the PEB contains only 0xFF bytes */
		err = ubi_io_read(ubi, ubi->peb_buf1, pnum, 0, ubi->peb_size);
		if (err)
			goto out;

		err = check_pattern(ubi->peb_buf1, 0xFF, ubi->peb_size);
		if (err == 0) {
			ubi_err("erased PEB %d, but a non-0xFF byte found",
				pnum);
			err = -EIO;
			goto out;
		}

		/* Write a pattern and check it */
		memset(ubi->peb_buf1, patterns[i], ubi->peb_size);
		err = ubi_io_write(ubi, ubi->peb_buf1, pnum, 0, ubi->peb_size);
		if (err)
			goto out;