[net-next-2.6.git] / fs / gfs2 / recovery.c

/*
 * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
 * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
 *
 * This copyrighted material is made available to anyone wishing to use,
 * modify, copy, or redistribute it subject to the terms and conditions
 * of the GNU General Public License version 2.
 */

#include <linux/module.h>
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/completion.h>
#include <linux/buffer_head.h>
#include <linux/gfs2_ondisk.h>
#include <linux/crc32.h>

#include "gfs2.h"
#include "incore.h"
#include "bmap.h"
#include "glock.h"
#include "glops.h"
#include "lops.h"
#include "meta_io.h"
#include "recovery.h"
#include "super.h"
#include "util.h"
#include "dir.h"

struct workqueue_struct *gfs_recovery_wq;

int gfs2_replay_read_block(struct gfs2_jdesc *jd, unsigned int blk,
			   struct buffer_head **bh)
{
	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
	struct gfs2_glock *gl = ip->i_gl;
	int new = 0;
	u64 dblock;
	u32 extlen;
	int error;

	error = gfs2_extent_map(&ip->i_inode, blk, &new, &dblock, &extlen);
	if (error)
		return error;
	if (!dblock) {
		gfs2_consist_inode(ip);
		return -EIO;
	}

	*bh = gfs2_meta_ra(gl, dblock, extlen);

	return error;
}

int gfs2_revoke_add(struct gfs2_sbd *sdp, u64 blkno, unsigned int where)
{
	struct list_head *head = &sdp->sd_revoke_list;
	struct gfs2_revoke_replay *rr;
	int found = 0;

	list_for_each_entry(rr, head, rr_list) {
		if (rr->rr_blkno == blkno) {
			found = 1;
			break;
		}
	}

	if (found) {
		rr->rr_where = where;
		return 0;
	}

	rr = kmalloc(sizeof(struct gfs2_revoke_replay), GFP_NOFS);
	if (!rr)
		return -ENOMEM;

	rr->rr_blkno = blkno;
	rr->rr_where = where;
	list_add(&rr->rr_list, head);

	return 1;
}

int gfs2_revoke_check(struct gfs2_sbd *sdp, u64 blkno, unsigned int where)
{
	struct gfs2_revoke_replay *rr;
	int wrap, a, b, revoke;
	int found = 0;

	list_for_each_entry(rr, &sdp->sd_revoke_list, rr_list) {
		if (rr->rr_blkno == blkno) {
			found = 1;
			break;
		}
	}

	if (!found)
		return 0;

	wrap = (rr->rr_where < sdp->sd_replay_tail);
	a = (sdp->sd_replay_tail < where);
	b = (where < rr->rr_where);
	revoke = (wrap) ? (a || b) : (a && b);

	return revoke;
}

void gfs2_revoke_clean(struct gfs2_sbd *sdp)
{
	struct list_head *head = &sdp->sd_revoke_list;
	struct gfs2_revoke_replay *rr;

	while (!list_empty(head)) {
		rr = list_entry(head->next, struct gfs2_revoke_replay, rr_list);
		list_del(&rr->rr_list);
		kfree(rr);
	}
}

static int gfs2_log_header_in(struct gfs2_log_header_host *lh, const void *buf)
{
	const struct gfs2_log_header *str = buf;

	if (str->lh_header.mh_magic != cpu_to_be32(GFS2_MAGIC) ||
	    str->lh_header.mh_type != cpu_to_be32(GFS2_METATYPE_LH))
		return 1;

	lh->lh_sequence = be64_to_cpu(str->lh_sequence);
	lh->lh_flags = be32_to_cpu(str->lh_flags);
	lh->lh_tail = be32_to_cpu(str->lh_tail);
	lh->lh_blkno = be32_to_cpu(str->lh_blkno);
	lh->lh_hash = be32_to_cpu(str->lh_hash);
	return 0;
}

/**
 * get_log_header - read the log header for a given segment
 * @jd: the journal
 * @blk: the block to look at
 * @lh: the log header to return
 *
 * Read the log header for a given segement in a given journal.  Do a few
 * sanity checks on it.
 *
 * Returns: 0 on success,
 *          1 if the header was invalid or incomplete,
 *          errno on error
 */

static int get_log_header(struct gfs2_jdesc *jd, unsigned int blk,
			  struct gfs2_log_header_host *head)
{
	struct buffer_head *bh;
	struct gfs2_log_header_host uninitialized_var(lh);
	const u32 nothing = 0;
	u32 hash;
	int error;

	error = gfs2_replay_read_block(jd, blk, &bh);
	if (error)
		return error;

	hash = crc32_le((u32)~0, bh->b_data, sizeof(struct gfs2_log_header) -
					     sizeof(u32));
	hash = crc32_le(hash, (unsigned char const *)&nothing, sizeof(nothing));
	hash ^= (u32)~0;
	error = gfs2_log_header_in(&lh, bh->b_data);
	brelse(bh);

	if (error || lh.lh_blkno != blk || lh.lh_hash != hash)
		return 1;

	*head = lh;

	return 0;
}

/**
 * find_good_lh - find a good log header
 * @jd: the journal
 * @blk: the segment to start searching from
 * @lh: the log header to fill in
 * @forward: if true search forward in the log, else search backward
 *
 * Call get_log_header() to get a log header for a segment, but if the
 * segment is bad, either scan forward or backward until we find a good one.
 *
 * Returns: errno
 */

static int find_good_lh(struct gfs2_jdesc *jd, unsigned int *blk,
			struct gfs2_log_header_host *head)
{
	unsigned int orig_blk = *blk;
	int error;

	for (;;) {
		error = get_log_header(jd, *blk, head);
		if (error <= 0)
			return error;

		if (++*blk == jd->jd_blocks)
			*blk = 0;

		if (*blk == orig_blk) {
			gfs2_consist_inode(GFS2_I(jd->jd_inode));
			return -EIO;
		}
	}
}

/**
 * jhead_scan - make sure we've found the head of the log
 * @jd: the journal
 * @head: this is filled in with the log descriptor of the head
 *
 * At this point, seg and lh should be either the head of the log or just
 * before.  Scan forward until we find the head.
 *
 * Returns: errno
 */

static int jhead_scan(struct gfs2_jdesc *jd, struct gfs2_log_header_host *head)
{
	unsigned int blk = head->lh_blkno;
	struct gfs2_log_header_host lh;
	int error;

	for (;;) {
		if (++blk == jd->jd_blocks)
			blk = 0;

		error = get_log_header(jd, blk, &lh);
		if (error < 0)
			return error;
		if (error == 1)
			continue;

		if (lh.lh_sequence == head->lh_sequence) {
			gfs2_consist_inode(GFS2_I(jd->jd_inode));
			return -EIO;
		}
		if (lh.lh_sequence < head->lh_sequence)
			break;

		*head = lh;
	}

	return 0;
}

/**
 * gfs2_find_jhead - find the head of a log
 * @jd: the journal
 * @head: the log descriptor for the head of the log is returned here
 *
 * Do a binary search of a journal and find the valid log entry with the
 * highest sequence number.  (i.e. the log head)
 *
 * Returns: errno
 */

int gfs2_find_jhead(struct gfs2_jdesc *jd, struct gfs2_log_header_host *head)
{
	struct gfs2_log_header_host lh_1, lh_m;
	u32 blk_1, blk_2, blk_m;
	int error;

	blk_1 = 0;
	blk_2 = jd->jd_blocks - 1;

	for (;;) {
		blk_m = (blk_1 + blk_2) / 2;

		error = find_good_lh(jd, &blk_1, &lh_1);
		if (error)
			return error;

		error = find_good_lh(jd, &blk_m, &lh_m);
		if (error)
			return error;

		if (blk_1 == blk_m || blk_m == blk_2)
			break;

		if (lh_1.lh_sequence <= lh_m.lh_sequence)
			blk_1 = blk_m;
		else
			blk_2 = blk_m;
	}

	error = jhead_scan(jd, &lh_1);
	if (error)
		return error;

	*head = lh_1;

	return error;
}

/**
 * foreach_descriptor - go through the active part of the log
 * @jd: the journal
 * @start: the first log header in the active region
 * @end: the last log header (don't process the contents of this entry))
 *
 * Call a given function once for every log descriptor in the active
 * portion of the log.
 *
 * Returns: errno
 */

static int foreach_descriptor(struct gfs2_jdesc *jd, unsigned int start,
			      unsigned int end, int pass)
{
	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	struct buffer_head *bh;
	struct gfs2_log_descriptor *ld;
	int error = 0;
	u32 length;
	__be64 *ptr;
	unsigned int offset = sizeof(struct gfs2_log_descriptor);
	offset += sizeof(__be64) - 1;
	offset &= ~(sizeof(__be64) - 1);

	while (start != end) {
		error = gfs2_replay_read_block(jd, start, &bh);
		if (error)
			return error;
		if (gfs2_meta_check(sdp, bh)) {
			brelse(bh);
			return -EIO;
		}
		ld = (struct gfs2_log_descriptor *)bh->b_data;
		length = be32_to_cpu(ld->ld_length);

		if (be32_to_cpu(ld->ld_header.mh_type) == GFS2_METATYPE_LH) {
			struct gfs2_log_header_host lh;
			error = get_log_header(jd, start, &lh);
			if (!error) {
				gfs2_replay_incr_blk(sdp, &start);
				brelse(bh);
				continue;
			}
			if (error == 1) {
				gfs2_consist_inode(GFS2_I(jd->jd_inode));
				error = -EIO;
			}
			brelse(bh);
			return error;
		} else if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LD)) {
			brelse(bh);
			return -EIO;
		}
		ptr = (__be64 *)(bh->b_data + offset);
		error = lops_scan_elements(jd, start, ld, ptr, pass);
		if (error) {
			brelse(bh);
			return error;
		}

		while (length--)
			gfs2_replay_incr_blk(sdp, &start);

		brelse(bh);
	}

	return 0;
}

/**
 * clean_journal - mark a dirty journal as being clean
 * @sdp: the filesystem
 * @jd: the journal
 * @gl: the journal's glock
 * @head: the head journal to start from
 *
 * Returns: errno
 */

static int clean_journal(struct gfs2_jdesc *jd, struct gfs2_log_header_host *head)
{
	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	unsigned int lblock;
	struct gfs2_log_header *lh;
	u32 hash;
	struct buffer_head *bh;
	int error;
	struct buffer_head bh_map = { .b_state = 0, .b_blocknr = 0 };

	lblock = head->lh_blkno;
	gfs2_replay_incr_blk(sdp, &lblock);
	bh_map.b_size = 1 << ip->i_inode.i_blkbits;
	error = gfs2_block_map(&ip->i_inode, lblock, &bh_map, 0);
	if (error)
		return error;
	if (!bh_map.b_blocknr) {
		gfs2_consist_inode(ip);
		return -EIO;
	}

	bh = sb_getblk(sdp->sd_vfs, bh_map.b_blocknr);
	lock_buffer(bh);
	memset(bh->b_data, 0, bh->b_size);
	set_buffer_uptodate(bh);
	clear_buffer_dirty(bh);
	unlock_buffer(bh);

	lh = (struct gfs2_log_header *)bh->b_data;
	memset(lh, 0, sizeof(struct gfs2_log_header));
	lh->lh_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
	lh->lh_header.mh_type = cpu_to_be32(GFS2_METATYPE_LH);
	lh->lh_header.__pad0 = cpu_to_be64(0);
	lh->lh_header.mh_format = cpu_to_be32(GFS2_FORMAT_LH);
	lh->lh_header.mh_jid = cpu_to_be32(sdp->sd_jdesc->jd_jid);
	lh->lh_sequence = cpu_to_be64(head->lh_sequence + 1);
	lh->lh_flags = cpu_to_be32(GFS2_LOG_HEAD_UNMOUNT);
	lh->lh_blkno = cpu_to_be32(lblock);
	hash = gfs2_disk_hash((const char *)lh, sizeof(struct gfs2_log_header));
	lh->lh_hash = cpu_to_be32(hash);

	set_buffer_dirty(bh);
	if (sync_dirty_buffer(bh))
		gfs2_io_error_bh(sdp, bh);
	brelse(bh);

	return error;
}


static void gfs2_recovery_done(struct gfs2_sbd *sdp, unsigned int jid,
                               unsigned int message)
{
	char env_jid[20];
	char env_status[20];
	char *envp[] = { env_jid, env_status, NULL };
	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
        ls->ls_recover_jid_done = jid;
        ls->ls_recover_jid_status = message;
	sprintf(env_jid, "JID=%d", jid);
	sprintf(env_status, "RECOVERY=%s",
		message == LM_RD_SUCCESS ? "Done" : "Failed");
        kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
}

void gfs2_recover_func(struct work_struct *work)
{
	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	struct gfs2_log_header_host head;
	struct gfs2_holder j_gh, ji_gh, t_gh;
	unsigned long t;
	int ro = 0;
	unsigned int pass;
	int error;
	int jlocked = 0;

	if (sdp->sd_args.ar_spectator ||
	    (jd->jd_jid != sdp->sd_lockstruct.ls_jid)) {
		fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
			jd->jd_jid);
		jlocked = 1;
		/* Acquire the journal lock so we can do recovery */

		error = gfs2_glock_nq_num(sdp, jd->jd_jid, &gfs2_journal_glops,
					  LM_ST_EXCLUSIVE,
					  LM_FLAG_NOEXP | LM_FLAG_TRY | GL_NOCACHE,
					  &j_gh);
		switch (error) {
		case 0:
			break;

		case GLR_TRYFAILED:
			fs_info(sdp, "jid=%u: Busy\n", jd->jd_jid);
			error = 0;

		default:
			goto fail;
		};

		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
					   LM_FLAG_NOEXP | GL_NOCACHE, &ji_gh);
		if (error)
			goto fail_gunlock_j;
	} else {
		fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
	}

	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);

	error = gfs2_jdesc_check(jd);
	if (error)
		goto fail_gunlock_ji;

	error = gfs2_find_jhead(jd, &head);
	if (error)
		goto fail_gunlock_ji;

	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
		fs_info(sdp, "jid=%u: Acquiring the transaction lock...\n",
			jd->jd_jid);

		t = jiffies;

		/* Acquire a shared hold on the transaction lock */

		error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED,
					   LM_FLAG_NOEXP | LM_FLAG_PRIORITY |
					   GL_NOCACHE, &t_gh);
		if (error)
			goto fail_gunlock_ji;

		if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
			if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
				ro = 1;
		} else {
			if (sdp->sd_vfs->s_flags & MS_RDONLY) {
				/* check if device itself is read-only */
				ro = bdev_read_only(sdp->sd_vfs->s_bdev);
				if (!ro) {
					fs_info(sdp, "recovery required on "
						"read-only filesystem.\n");
					fs_info(sdp, "write access will be "
						"enabled during recovery.\n");
				}
			}
		}

		if (ro) {
			fs_warn(sdp, "jid=%u: Can't replay: read-only block "
				"device\n", jd->jd_jid);
			error = -EROFS;
			goto fail_gunlock_tr;
		}

		fs_info(sdp, "jid=%u: Replaying journal...\n", jd->jd_jid);

		for (pass = 0; pass < 2; pass++) {
			lops_before_scan(jd, &head, pass);
			error = foreach_descriptor(jd, head.lh_tail,
						   head.lh_blkno, pass);
			lops_after_scan(jd, error, pass);
			if (error)
				goto fail_gunlock_tr;
		}

		error = clean_journal(jd, &head);
		if (error)
			goto fail_gunlock_tr;

		gfs2_glock_dq_uninit(&t_gh);
		t = DIV_ROUND_UP(jiffies - t, HZ);
		fs_info(sdp, "jid=%u: Journal replayed in %lus\n",
			jd->jd_jid, t);
	}

	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);

	if (jlocked) {
		gfs2_glock_dq_uninit(&ji_gh);
		gfs2_glock_dq_uninit(&j_gh);
	}

	fs_info(sdp, "jid=%u: Done\n", jd->jd_jid);
	goto done;

fail_gunlock_tr:
	gfs2_glock_dq_uninit(&t_gh);
fail_gunlock_ji:
	if (jlocked) {
		gfs2_glock_dq_uninit(&ji_gh);
fail_gunlock_j:
		gfs2_glock_dq_uninit(&j_gh);
	}

	fs_info(sdp, "jid=%u: %s\n", jd->jd_jid, (error) ? "Failed" : "Done");
fail:
	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
done:
	clear_bit(JDF_RECOVERY, &jd->jd_flags);
	smp_mb__after_clear_bit();
	wake_up_bit(&jd->jd_flags, JDF_RECOVERY);
}

static int gfs2_recovery_wait(void *word)
{
	schedule();
	return 0;
}

int gfs2_recover_journal(struct gfs2_jdesc *jd, bool wait)
{
	int rv;

	if (test_and_set_bit(JDF_RECOVERY, &jd->jd_flags))
		return -EBUSY;

	/* we have JDF_RECOVERY, queue should always succeed */
	rv = queue_work(gfs_recovery_wq, &jd->jd_work);
	BUG_ON(!rv);

	if (wait)
		wait_on_bit(&jd->jd_flags, JDF_RECOVERY, gfs2_recovery_wait,
			    TASK_UNINTERRUPTIBLE);

	return 0;
}
Commit	Line	Data
b3b94faa DT	1	/*
b3b94faa DT	2	* Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3a8a9a10	3	* Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
b3b94faa DT	4	*
	5	* This copyrighted material is made available to anyone wishing to use,
	6	* modify, copy, or redistribute it subject to the terms and conditions
e9fc2aa0	7	* of the GNU General Public License version 2.
b3b94faa DT	8	*/
b3b94faa DT	9
1c2ea8a2	10	#include <linux/module.h>
b3b94faa DT	11	#include <linux/slab.h>
	12	#include <linux/spinlock.h>
	13	#include <linux/completion.h>
	14	#include <linux/buffer_head.h>
5c676f6d	15	#include <linux/gfs2_ondisk.h>
71b86f56	16	#include <linux/crc32.h>
b3b94faa DT	17
b3b94faa DT	18	#include "gfs2.h"
5c676f6d	19	#include "incore.h"
b3b94faa DT	20	#include "bmap.h"
	21	#include "glock.h"
	22	#include "glops.h"
b3b94faa DT	23	#include "lops.h"
	24	#include "meta_io.h"
	25	#include "recovery.h"
	26	#include "super.h"
5c676f6d	27	#include "util.h"
71b86f56	28	#include "dir.h"
b3b94faa	29
6ecd7c2d TH	30	struct workqueue_struct *gfs_recovery_wq;
6ecd7c2d TH	31
b3b94faa DT	32	int gfs2_replay_read_block(struct gfs2_jdesc *jd, unsigned int blk,
	33	struct buffer_head **bh)
	34	{
feaa7bba	35	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
5c676f6d	36	struct gfs2_glock *gl = ip->i_gl;
b3b94faa	37	int new = 0;
cd915493 SW	38	u64 dblock;
cd915493 SW	39	u32 extlen;
b3b94faa DT	40	int error;
b3b94faa DT	41
feaa7bba	42	error = gfs2_extent_map(&ip->i_inode, blk, &new, &dblock, &extlen);
b3b94faa DT	43	if (error)
	44	return error;
	45	if (!dblock) {
5c676f6d	46	gfs2_consist_inode(ip);
b3b94faa DT	47	return -EIO;
	48	}
	49
7276b3b0	50	*bh = gfs2_meta_ra(gl, dblock, extlen);
b3b94faa DT	51
	52	return error;
	53	}
	54
cd915493	55	int gfs2_revoke_add(struct gfs2_sbd *sdp, u64 blkno, unsigned int where)
b3b94faa DT	56	{
	57	struct list_head *head = &sdp->sd_revoke_list;
	58	struct gfs2_revoke_replay *rr;
	59	int found = 0;
	60
	61	list_for_each_entry(rr, head, rr_list) {
	62	if (rr->rr_blkno == blkno) {
	63	found = 1;
	64	break;
	65	}
	66	}
	67
	68	if (found) {
	69	rr->rr_where = where;
	70	return 0;
	71	}
	72
16c5f06f	73	rr = kmalloc(sizeof(struct gfs2_revoke_replay), GFP_NOFS);
b3b94faa DT	74	if (!rr)
	75	return -ENOMEM;
	76
	77	rr->rr_blkno = blkno;
	78	rr->rr_where = where;
	79	list_add(&rr->rr_list, head);
	80
	81	return 1;
	82	}
	83
cd915493	84	int gfs2_revoke_check(struct gfs2_sbd *sdp, u64 blkno, unsigned int where)
b3b94faa DT	85	{
	86	struct gfs2_revoke_replay *rr;
	87	int wrap, a, b, revoke;
	88	int found = 0;
	89
	90	list_for_each_entry(rr, &sdp->sd_revoke_list, rr_list) {
	91	if (rr->rr_blkno == blkno) {
	92	found = 1;
	93	break;
	94	}
	95	}
	96
	97	if (!found)
	98	return 0;
	99
	100	wrap = (rr->rr_where < sdp->sd_replay_tail);
	101	a = (sdp->sd_replay_tail < where);
	102	b = (where < rr->rr_where);
	103	revoke = (wrap) ? (a \|\| b) : (a && b);
	104
	105	return revoke;
	106	}
	107
	108	void gfs2_revoke_clean(struct gfs2_sbd *sdp)
	109	{
	110	struct list_head *head = &sdp->sd_revoke_list;
	111	struct gfs2_revoke_replay *rr;
	112
	113	while (!list_empty(head)) {
	114	rr = list_entry(head->next, struct gfs2_revoke_replay, rr_list);
	115	list_del(&rr->rr_list);
	116	kfree(rr);
	117	}
	118	}
	119
bb8d8a6f SW	120	static int gfs2_log_header_in(struct gfs2_log_header_host lh, const void buf)
	121	{
	122	const struct gfs2_log_header *str = buf;
	123
	124	if (str->lh_header.mh_magic != cpu_to_be32(GFS2_MAGIC) \|\|
	125	str->lh_header.mh_type != cpu_to_be32(GFS2_METATYPE_LH))
	126	return 1;
	127
	128	lh->lh_sequence = be64_to_cpu(str->lh_sequence);
	129	lh->lh_flags = be32_to_cpu(str->lh_flags);
	130	lh->lh_tail = be32_to_cpu(str->lh_tail);
	131	lh->lh_blkno = be32_to_cpu(str->lh_blkno);
	132	lh->lh_hash = be32_to_cpu(str->lh_hash);
	133	return 0;
	134	}
	135
b3b94faa DT	136	/**
	137	* get_log_header - read the log header for a given segment
	138	* @jd: the journal
	139	* @blk: the block to look at
	140	* @lh: the log header to return
	141	*
	142	* Read the log header for a given segement in a given journal. Do a few
	143	* sanity checks on it.
	144	*
	145	* Returns: 0 on success,
	146	* 1 if the header was invalid or incomplete,
	147	* errno on error
	148	*/
	149
	150	static int get_log_header(struct gfs2_jdesc *jd, unsigned int blk,
55167622	151	struct gfs2_log_header_host *head)
b3b94faa DT	152	{
b3b94faa DT	153	struct buffer_head *bh;
f5a8cd02	154	struct gfs2_log_header_host uninitialized_var(lh);
887bc5d0	155	const u32 nothing = 0;
cd915493	156	u32 hash;
b3b94faa DT	157	int error;
	158
	159	error = gfs2_replay_read_block(jd, blk, &bh);
	160	if (error)
	161	return error;
	162
2a2c9824 SW	163	hash = crc32_le((u32)~0, bh->b_data, sizeof(struct gfs2_log_header) -
	164	sizeof(u32));
	165	hash = crc32_le(hash, (unsigned char const *)&nothing, sizeof(nothing));
	166	hash ^= (u32)~0;
bb8d8a6f	167	error = gfs2_log_header_in(&lh, bh->b_data);
b3b94faa DT	168	brelse(bh);
b3b94faa DT	169
bb8d8a6f	170	if (error \|\| lh.lh_blkno != blk \|\| lh.lh_hash != hash)
b3b94faa DT	171	return 1;
	172
	173	*head = lh;
	174
	175	return 0;
	176	}
	177
	178	/**
	179	* find_good_lh - find a good log header
	180	* @jd: the journal
	181	* @blk: the segment to start searching from
	182	* @lh: the log header to fill in
	183	* @forward: if true search forward in the log, else search backward
	184	*
	185	* Call get_log_header() to get a log header for a segment, but if the
	186	* segment is bad, either scan forward or backward until we find a good one.
	187	*
	188	* Returns: errno
	189	*/
	190
	191	static int find_good_lh(struct gfs2_jdesc jd, unsigned int blk,
55167622	192	struct gfs2_log_header_host *head)
b3b94faa DT	193	{
	194	unsigned int orig_blk = *blk;
	195	int error;
	196
	197	for (;;) {
	198	error = get_log_header(jd, *blk, head);
	199	if (error <= 0)
	200	return error;
	201
	202	if (++*blk == jd->jd_blocks)
	203	*blk = 0;
	204
	205	if (*blk == orig_blk) {
feaa7bba	206	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	207	return -EIO;
	208	}
	209	}
	210	}
	211
	212	/**
	213	* jhead_scan - make sure we've found the head of the log
	214	* @jd: the journal
	215	* @head: this is filled in with the log descriptor of the head
	216	*
	217	* At this point, seg and lh should be either the head of the log or just
	218	* before. Scan forward until we find the head.
	219	*
	220	* Returns: errno
	221	*/
	222
55167622	223	static int jhead_scan(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
b3b94faa DT	224	{
b3b94faa DT	225	unsigned int blk = head->lh_blkno;
55167622	226	struct gfs2_log_header_host lh;
b3b94faa DT	227	int error;
	228
	229	for (;;) {
	230	if (++blk == jd->jd_blocks)
	231	blk = 0;
	232
	233	error = get_log_header(jd, blk, &lh);
	234	if (error < 0)
	235	return error;
	236	if (error == 1)
	237	continue;
	238
	239	if (lh.lh_sequence == head->lh_sequence) {
feaa7bba	240	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	241	return -EIO;
	242	}
	243	if (lh.lh_sequence < head->lh_sequence)
	244	break;
	245
	246	*head = lh;
	247	}
	248
	249	return 0;
	250	}
	251
	252	/**
	253	* gfs2_find_jhead - find the head of a log
	254	* @jd: the journal
	255	* @head: the log descriptor for the head of the log is returned here
	256	*
	257	* Do a binary search of a journal and find the valid log entry with the
	258	* highest sequence number. (i.e. the log head)
	259	*
	260	* Returns: errno
	261	*/
	262
55167622	263	int gfs2_find_jhead(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
b3b94faa	264	{
55167622	265	struct gfs2_log_header_host lh_1, lh_m;
cd915493	266	u32 blk_1, blk_2, blk_m;
b3b94faa DT	267	int error;
	268
	269	blk_1 = 0;
	270	blk_2 = jd->jd_blocks - 1;
	271
	272	for (;;) {
	273	blk_m = (blk_1 + blk_2) / 2;
	274
	275	error = find_good_lh(jd, &blk_1, &lh_1);
	276	if (error)
	277	return error;
	278
	279	error = find_good_lh(jd, &blk_m, &lh_m);
	280	if (error)
	281	return error;
	282
	283	if (blk_1 == blk_m \|\| blk_m == blk_2)
	284	break;
	285
	286	if (lh_1.lh_sequence <= lh_m.lh_sequence)
	287	blk_1 = blk_m;
	288	else
	289	blk_2 = blk_m;
	290	}
	291
	292	error = jhead_scan(jd, &lh_1);
	293	if (error)
	294	return error;
	295
	296	*head = lh_1;
	297
	298	return error;
	299	}
	300
	301	/**
	302	* foreach_descriptor - go through the active part of the log
	303	* @jd: the journal
	304	* @start: the first log header in the active region
	305	* @end: the last log header (don't process the contents of this entry))
	306	*
	307	* Call a given function once for every log descriptor in the active
	308	* portion of the log.
	309	*
	310	* Returns: errno
	311	*/
	312
	313	static int foreach_descriptor(struct gfs2_jdesc *jd, unsigned int start,
	314	unsigned int end, int pass)
	315	{
feaa7bba	316	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa DT	317	struct buffer_head *bh;
	318	struct gfs2_log_descriptor *ld;
	319	int error = 0;
	320	u32 length;
	321	__be64 *ptr;
	322	unsigned int offset = sizeof(struct gfs2_log_descriptor);
a67cdbd4 SW	323	offset += sizeof(__be64) - 1;
a67cdbd4 SW	324	offset &= ~(sizeof(__be64) - 1);
b3b94faa DT	325
	326	while (start != end) {
	327	error = gfs2_replay_read_block(jd, start, &bh);
	328	if (error)
	329	return error;
	330	if (gfs2_meta_check(sdp, bh)) {
	331	brelse(bh);
	332	return -EIO;
	333	}
	334	ld = (struct gfs2_log_descriptor *)bh->b_data;
	335	length = be32_to_cpu(ld->ld_length);
	336
e3167ded	337	if (be32_to_cpu(ld->ld_header.mh_type) == GFS2_METATYPE_LH) {
55167622	338	struct gfs2_log_header_host lh;
b3b94faa DT	339	error = get_log_header(jd, start, &lh);
	340	if (!error) {
	341	gfs2_replay_incr_blk(sdp, &start);
88721877	342	brelse(bh);
b3b94faa DT	343	continue;
	344	}
	345	if (error == 1) {
feaa7bba	346	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	347	error = -EIO;
	348	}
	349	brelse(bh);
	350	return error;
	351	} else if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LD)) {
	352	brelse(bh);
	353	return -EIO;
	354	}
	355	ptr = (__be64 *)(bh->b_data + offset);
	356	error = lops_scan_elements(jd, start, ld, ptr, pass);
	357	if (error) {
	358	brelse(bh);
	359	return error;
	360	}
	361
	362	while (length--)
	363	gfs2_replay_incr_blk(sdp, &start);
	364
	365	brelse(bh);
	366	}
	367
	368	return 0;
	369	}
	370
	371	/**
	372	* clean_journal - mark a dirty journal as being clean
	373	* @sdp: the filesystem
	374	* @jd: the journal
	375	* @gl: the journal's glock
	376	* @head: the head journal to start from
	377	*
	378	* Returns: errno
	379	*/
	380
55167622	381	static int clean_journal(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
b3b94faa	382	{
feaa7bba SW	383	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	384	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa	385	unsigned int lblock;
b3b94faa	386	struct gfs2_log_header *lh;
cd915493	387	u32 hash;
b3b94faa DT	388	struct buffer_head *bh;
b3b94faa DT	389	int error;
23591256	390	struct buffer_head bh_map = { .b_state = 0, .b_blocknr = 0 };
fd88de56	391
b3b94faa DT	392	lblock = head->lh_blkno;
b3b94faa DT	393	gfs2_replay_incr_blk(sdp, &lblock);
23591256	394	bh_map.b_size = 1 << ip->i_inode.i_blkbits;
e9e1ef2b	395	error = gfs2_block_map(&ip->i_inode, lblock, &bh_map, 0);
b3b94faa DT	396	if (error)
b3b94faa DT	397	return error;
7a6bbacb	398	if (!bh_map.b_blocknr) {
b3b94faa DT	399	gfs2_consist_inode(ip);
	400	return -EIO;
	401	}
	402
7a6bbacb	403	bh = sb_getblk(sdp->sd_vfs, bh_map.b_blocknr);
b3b94faa DT	404	lock_buffer(bh);
	405	memset(bh->b_data, 0, bh->b_size);
	406	set_buffer_uptodate(bh);
	407	clear_buffer_dirty(bh);
	408	unlock_buffer(bh);
	409
	410	lh = (struct gfs2_log_header *)bh->b_data;
	411	memset(lh, 0, sizeof(struct gfs2_log_header));
	412	lh->lh_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
88721877	413	lh->lh_header.mh_type = cpu_to_be32(GFS2_METATYPE_LH);
0ab7d13f	414	lh->lh_header.__pad0 = cpu_to_be64(0);
e3167ded	415	lh->lh_header.mh_format = cpu_to_be32(GFS2_FORMAT_LH);
0ab7d13f	416	lh->lh_header.mh_jid = cpu_to_be32(sdp->sd_jdesc->jd_jid);
b3b94faa DT	417	lh->lh_sequence = cpu_to_be64(head->lh_sequence + 1);
	418	lh->lh_flags = cpu_to_be32(GFS2_LOG_HEAD_UNMOUNT);
	419	lh->lh_blkno = cpu_to_be32(lblock);
	420	hash = gfs2_disk_hash((const char *)lh, sizeof(struct gfs2_log_header));
	421	lh->lh_hash = cpu_to_be32(hash);
	422
	423	set_buffer_dirty(bh);
	424	if (sync_dirty_buffer(bh))
	425	gfs2_io_error_bh(sdp, bh);
	426	brelse(bh);
	427
	428	return error;
	429	}
	430
da755fdb	431
f057f6cd SW	432	static void gfs2_recovery_done(struct gfs2_sbd *sdp, unsigned int jid,
f057f6cd SW	433	unsigned int message)
da755fdb	434	{
f057f6cd SW	435	char env_jid[20];
	436	char env_status[20];
	437	char *envp[] = { env_jid, env_status, NULL };
	438	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
	439	ls->ls_recover_jid_done = jid;
	440	ls->ls_recover_jid_status = message;
	441	sprintf(env_jid, "JID=%d", jid);
	442	sprintf(env_status, "RECOVERY=%s",
	443	message == LM_RD_SUCCESS ? "Done" : "Failed");
	444	kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb SW	445	}
da755fdb SW	446
6ecd7c2d	447	void gfs2_recover_func(struct work_struct *work)
b3b94faa	448	{
fe64d517	449	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
feaa7bba SW	450	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	451	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
55167622	452	struct gfs2_log_header_host head;
b3b94faa DT	453	struct gfs2_holder j_gh, ji_gh, t_gh;
	454	unsigned long t;
	455	int ro = 0;
	456	unsigned int pass;
	457	int error;
c741c455	458	int jlocked = 0;
b3b94faa	459
d0795f91 SW	460	if (sdp->sd_args.ar_spectator \|\|
d0795f91 SW	461	(jd->jd_jid != sdp->sd_lockstruct.ls_jid)) {
5965b1f4 SW	462	fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
5965b1f4 SW	463	jd->jd_jid);
c741c455	464	jlocked = 1;
c78bad11	465	/* Acquire the journal lock so we can do recovery */
b3b94faa	466
5965b1f4 SW	467	error = gfs2_glock_nq_num(sdp, jd->jd_jid, &gfs2_journal_glops,
	468	LM_ST_EXCLUSIVE,
	469	LM_FLAG_NOEXP \| LM_FLAG_TRY \| GL_NOCACHE,
	470	&j_gh);
	471	switch (error) {
	472	case 0:
	473	break;
907b9bce	474
5965b1f4 SW	475	case GLR_TRYFAILED:
	476	fs_info(sdp, "jid=%u: Busy\n", jd->jd_jid);
	477	error = 0;
907b9bce	478
5965b1f4 SW	479	default:
	480	goto fail;
	481	};
b3b94faa	482
5965b1f4	483	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8	484	LM_FLAG_NOEXP \| GL_NOCACHE, &ji_gh);
5965b1f4 SW	485	if (error)
	486	goto fail_gunlock_j;
	487	} else {
	488	fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
	489	}
b3b94faa DT	490
	491	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);
	492
	493	error = gfs2_jdesc_check(jd);
	494	if (error)
	495	goto fail_gunlock_ji;
	496
	497	error = gfs2_find_jhead(jd, &head);
	498	if (error)
	499	goto fail_gunlock_ji;
	500
	501	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
	502	fs_info(sdp, "jid=%u: Acquiring the transaction lock...\n",
	503	jd->jd_jid);
	504
	505	t = jiffies;
	506
	507	/* Acquire a shared hold on the transaction lock */
	508
59a1cc6b	509	error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED,
5965b1f4	510	LM_FLAG_NOEXP \| LM_FLAG_PRIORITY \|
6802e340	511	GL_NOCACHE, &t_gh);
b3b94faa DT	512	if (error)
	513	goto fail_gunlock_ji;
	514
	515	if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
	516	if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
	517	ro = 1;
	518	} else {
7bc5c414 AD	519	if (sdp->sd_vfs->s_flags & MS_RDONLY) {
	520	/* check if device itself is read-only */
	521	ro = bdev_read_only(sdp->sd_vfs->s_bdev);
	522	if (!ro) {
	523	fs_info(sdp, "recovery required on "
	524	"read-only filesystem.\n");
	525	fs_info(sdp, "write access will be "
	526	"enabled during recovery.\n");
	527	}
	528	}
b3b94faa DT	529	}
	530
	531	if (ro) {
7bc5c414 AD	532	fs_warn(sdp, "jid=%u: Can't replay: read-only block "
7bc5c414 AD	533	"device\n", jd->jd_jid);
b3b94faa DT	534	error = -EROFS;
	535	goto fail_gunlock_tr;
	536	}
	537
	538	fs_info(sdp, "jid=%u: Replaying journal...\n", jd->jd_jid);
	539
	540	for (pass = 0; pass < 2; pass++) {
	541	lops_before_scan(jd, &head, pass);
	542	error = foreach_descriptor(jd, head.lh_tail,
	543	head.lh_blkno, pass);
	544	lops_after_scan(jd, error, pass);
	545	if (error)
	546	goto fail_gunlock_tr;
	547	}
	548
	549	error = clean_journal(jd, &head);
	550	if (error)
	551	goto fail_gunlock_tr;
	552
	553	gfs2_glock_dq_uninit(&t_gh);
5c676f6d	554	t = DIV_ROUND_UP(jiffies - t, HZ);
b3b94faa DT	555	fs_info(sdp, "jid=%u: Journal replayed in %lus\n",
	556	jd->jd_jid, t);
	557	}
	558
f057f6cd	559	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);
b3b94faa	560
c741c455 SW	561	if (jlocked) {
c741c455 SW	562	gfs2_glock_dq_uninit(&ji_gh);
5965b1f4	563	gfs2_glock_dq_uninit(&j_gh);
c741c455	564	}
b3b94faa DT	565
b3b94faa DT	566	fs_info(sdp, "jid=%u: Done\n", jd->jd_jid);
6ecd7c2d	567	goto done;
b3b94faa	568
5965b1f4	569	fail_gunlock_tr:
b3b94faa	570	gfs2_glock_dq_uninit(&t_gh);
5965b1f4	571	fail_gunlock_ji:
c741c455	572	if (jlocked) {
5965b1f4 SW	573	gfs2_glock_dq_uninit(&ji_gh);
	574	fail_gunlock_j:
	575	gfs2_glock_dq_uninit(&j_gh);
	576	}
b3b94faa DT	577
b3b94faa DT	578	fs_info(sdp, "jid=%u: %s\n", jd->jd_jid, (error) ? "Failed" : "Done");
5965b1f4	579	fail:
f057f6cd	580	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
6ecd7c2d TH	581	done:
	582	clear_bit(JDF_RECOVERY, &jd->jd_flags);
	583	smp_mb__after_clear_bit();
	584	wake_up_bit(&jd->jd_flags, JDF_RECOVERY);
b3b94faa DT	585	}
b3b94faa DT	586
fe64d517	587	static int gfs2_recovery_wait(void *word)
b3b94faa	588	{
fe64d517 SW	589	schedule();
fe64d517 SW	590	return 0;
b3b94faa DT	591	}
b3b94faa DT	592
6ecd7c2d	593	int gfs2_recover_journal(struct gfs2_jdesc *jd, bool wait)
9ac1b4d9	594	{
fe64d517	595	int rv;
6ecd7c2d TH	596
	597	if (test_and_set_bit(JDF_RECOVERY, &jd->jd_flags))
	598	return -EBUSY;
	599
	600	/* we have JDF_RECOVERY, queue should always succeed */
	601	rv = queue_work(gfs_recovery_wq, &jd->jd_work);
	602	BUG_ON(!rv);
	603
	604	if (wait)
	605	wait_on_bit(&jd->jd_flags, JDF_RECOVERY, gfs2_recovery_wait,
	606	TASK_UNINTERRUPTIBLE);
	607
9ac1b4d9 SW	608	return 0;
	609	}
	610