recovery.c

linux 内核源代码

/*
 * linux/fs/jbd2/recovery.c
 *
 * Written by Stephen C. Tweedie <[e-mail]>, 1999
 *
 * Copyright 1999-2000 Red Hat Software --- All Rights Reserved
 *
 * This file is part of the Linux kernel and is made available under
 * the terms of the GNU General Public License, version 2, or at your
 * option, any later version, incorporated herein by reference.
 *
 * Journal recovery routines for the generic filesystem journaling code;
 * part of the ext2fs journaling system.
 */

#ifndef __KERNEL__
#include "jfs_user.h"
#else
#include <linux/time.h>
#include <linux/fs.h>
#include <linux/jbd2.h>
#include <linux/errno.h>
#include <linux/slab.h>
#endif

/*
 * Maintain information about the progress of the recovery job, so that
 * the different passes can carry information between them.
 */
struct recovery_info
{
	tid_t		start_transaction;
	tid_t		end_transaction;

	int		nr_replays;
	int		nr_revokes;
	int		nr_revoke_hits;
};

enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY};
static int do_one_pass(journal_t *journal,
				struct recovery_info *info, enum passtype pass);
static int scan_revoke_records(journal_t *, struct buffer_head *,
				tid_t, struct recovery_info *);

#ifdef __KERNEL__

/* Release readahead buffers after use */
static void journal_brelse_array(struct buffer_head *b[], int n)
{
	while (--n >= 0)
		brelse (b[n]);
}


/*
 * When reading from the journal, we are going through the block device
 * layer directly and so there is no readahead being done for us.  We
 * need to implement any readahead ourselves if we want it to happen at
 * all.  Recovery is basically one long sequential read, so make sure we
 * do the IO in reasonably large chunks.
 *
 * This is not so critical that we need to be enormously clever about
 * the readahead size, though.  128K is a purely arbitrary, good-enough
 * fixed value.
 */

#define MAXBUF 8
static int do_readahead(journal_t *journal, unsigned int start)
{
	int err;
	unsigned int max, nbufs, next;
	unsigned long long blocknr;
	struct buffer_head *bh;

	struct buffer_head * bufs[MAXBUF];

	/* Do up to 128K of readahead */
	max = start + (128 * 1024 / journal->j_blocksize);
	if (max > journal->j_maxlen)
		max = journal->j_maxlen;

	/* Do the readahead itself.  We'll submit MAXBUF buffer_heads at
	 * a time to the block device IO layer. */

	nbufs = 0;

	for (next = start; next < max; next++) {
		err = jbd2_journal_bmap(journal, next, &blocknr);

		if (err) {
			printk (KERN_ERR "JBD: bad block at offset %u\n",
				next);
			goto failed;
		}

		bh = __getblk(journal->j_dev, blocknr, journal->j_blocksize);
		if (!bh) {
			err = -ENOMEM;
			goto failed;
		}

		if (!buffer_uptodate(bh) && !buffer_locked(bh)) {
			bufs[nbufs++] = bh;
			if (nbufs == MAXBUF) {
				ll_rw_block(READ, nbufs, bufs);
				journal_brelse_array(bufs, nbufs);
				nbufs = 0;
			}
		} else
			brelse(bh);
	}

	if (nbufs)
		ll_rw_block(READ, nbufs, bufs);
	err = 0;

failed:
	if (nbufs)
		journal_brelse_array(bufs, nbufs);
	return err;
}

#endif /* __KERNEL__ */


/*
 * Read a block from the journal
 */

static int jread(struct buffer_head **bhp, journal_t *journal,
		 unsigned int offset)
{
	int err;
	unsigned long long blocknr;
	struct buffer_head *bh;

	*bhp = NULL;

	if (offset >= journal->j_maxlen) {
		printk(KERN_ERR "JBD: corrupted journal superblock\n");
		return -EIO;
	}

	err = jbd2_journal_bmap(journal, offset, &blocknr);

	if (err) {
		printk (KERN_ERR "JBD: bad block at offset %u\n",
			offset);
		return err;
	}

	bh = __getblk(journal->j_dev, blocknr, journal->j_blocksize);
	if (!bh)
		return -ENOMEM;

	if (!buffer_uptodate(bh)) {
		/* If this is a brand new buffer, start readahead.
                   Otherwise, we assume we are already reading it.  */
		if (!buffer_req(bh))
			do_readahead(journal, offset);
		wait_on_buffer(bh);
	}

	if (!buffer_uptodate(bh)) {
		printk (KERN_ERR "JBD: Failed to read block at offset %u\n",
			offset);
		brelse(bh);
		return -EIO;
	}

	*bhp = bh;
	return 0;
}


/*
 * Count the number of in-use tags in a journal descriptor block.
 */

static int count_tags(journal_t *journal, struct buffer_head *bh)
{
	char *			tagp;
	journal_block_tag_t *	tag;
	int			nr = 0, size = journal->j_blocksize;
	int			tag_bytes = journal_tag_bytes(journal);

	tagp = &bh->b_data[sizeof(journal_header_t)];

	while ((tagp - bh->b_data + tag_bytes) <= size) {
		tag = (journal_block_tag_t *) tagp;

		nr++;
		tagp += tag_bytes;
		if (!(tag->t_flags & cpu_to_be32(JBD2_FLAG_SAME_UUID)))
			tagp += 16;

		if (tag->t_flags & cpu_to_be32(JBD2_FLAG_LAST_TAG))
			break;
	}

	return nr;
}


/* Make sure we wrap around the log correctly! */
#define wrap(journal, var)						\
do {									\
	if (var >= (journal)->j_last)					\
		var -= ((journal)->j_last - (journal)->j_first);	\
} while (0)

/**
 * jbd2_journal_recover - recovers a on-disk journal
 * @journal: the journal to recover
 *
 * The primary function for recovering the log contents when mounting a
 * journaled device.
 *
 * Recovery is done in three passes.  In the first pass, we look for the
 * end of the log.  In the second, we assemble the list of revoke
 * blocks.  In the third and final pass, we replay any un-revoked blocks
 * in the log.
 */
int jbd2_journal_recover(journal_t *journal)
{
	int			err;
	journal_superblock_t *	sb;

	struct recovery_info	info;

	memset(&info, 0, sizeof(info));
	sb = journal->j_superblock;

	/*
	 * The journal superblock's s_start field (the current log head)
	 * is always zero if, and only if, the journal was cleanly
	 * unmounted.
	 */

	if (!sb->s_start) {
		jbd_debug(1, "No recovery required, last transaction %d\n",
			  be32_to_cpu(sb->s_sequence));
		journal->j_transaction_sequence = be32_to_cpu(sb->s_sequence) + 1;
		return 0;
	}

	err = do_one_pass(journal, &info, PASS_SCAN);
	if (!err)
		err = do_one_pass(journal, &info, PASS_REVOKE);
	if (!err)
		err = do_one_pass(journal, &info, PASS_REPLAY);

	jbd_debug(1, "JBD: recovery, exit status %d, "
		  "recovered transactions %u to %u\n",
		  err, info.start_transaction, info.end_transaction);
	jbd_debug(1, "JBD: Replayed %d and revoked %d/%d blocks\n",
		  info.nr_replays, info.nr_revoke_hits, info.nr_revokes);

	/* Restart the log at the next transaction ID, thus invalidating
	 * any existing commit records in the log. */
	journal->j_transaction_sequence = ++info.end_transaction;

	jbd2_journal_clear_revoke(journal);
	sync_blockdev(journal->j_fs_dev);
	return err;
}

/**
 * jbd2_journal_skip_recovery - Start journal and wipe exiting records
 * @journal: journal to startup
 *
 * Locate any valid recovery information from the journal and set up the
 * journal structures in memory to ignore it (presumably because the
 * caller has evidence that it is out of date).
 * This function does'nt appear to be exorted..
 *
 * We perform one pass over the journal to allow us to tell the user how
 * much recovery information is being erased, and to let us initialise
 * the journal transaction sequence numbers to the next unused ID.
 */
int jbd2_journal_skip_recovery(journal_t *journal)
{
	int			err;
	journal_superblock_t *	sb;

	struct recovery_info	info;

	memset (&info, 0, sizeof(info));
	sb = journal->j_superblock;

	err = do_one_pass(journal, &info, PASS_SCAN);

	if (err) {
		printk(KERN_ERR "JBD: error %d scanning journal\n", err);
		++journal->j_transaction_sequence;
	} else {
#ifdef CONFIG_JBD2_DEBUG
		int dropped = info.end_transaction - be32_to_cpu(sb->s_sequence);
#endif
		jbd_debug(1,
			  "JBD: ignoring %d transaction%s from the journal.\n",
			  dropped, (dropped == 1) ? "" : "s");
		journal->j_transaction_sequence = ++info.end_transaction;
	}