[deliverable/linux.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/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 <linux/slow-work.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"

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.mh_format = cpu_to_be32(GFS2_FORMAT_LH);
	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);
}

static int gfs2_recover_get_ref(struct slow_work *work)
{
	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
	if (test_and_set_bit(JDF_RECOVERY, &jd->jd_flags))
		return -EBUSY;
	return 0;
}

static void gfs2_recover_put_ref(struct slow_work *work)
{
	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
	clear_bit(JDF_RECOVERY, &jd->jd_flags);
	smp_mb__after_clear_bit();
	wake_up_bit(&jd->jd_flags, JDF_RECOVERY);
}

static void gfs2_recover_work(struct slow_work *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;

	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
		fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
			jd->jd_jid);

		/* 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);
	}

	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid)
		gfs2_glock_dq_uninit(&ji_gh);

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

	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid)
		gfs2_glock_dq_uninit(&j_gh);

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

fail_gunlock_tr:
	gfs2_glock_dq_uninit(&t_gh);
fail_gunlock_ji:
	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
		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);
}

struct slow_work_ops gfs2_recover_ops = {
	.get_ref = gfs2_recover_get_ref,
	.put_ref = gfs2_recover_put_ref,
	.execute = gfs2_recover_work,
};


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

int gfs2_recover_journal(struct gfs2_jdesc *jd)
{
	int rv;
	rv = slow_work_enqueue(&jd->jd_work);
	if (rv)
		return rv;
	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
b3b94faa DT	10	#include <linux/slab.h>
	11	#include <linux/spinlock.h>
	12	#include <linux/completion.h>
	13	#include <linux/buffer_head.h>
5c676f6d	14	#include <linux/gfs2_ondisk.h>
71b86f56	15	#include <linux/crc32.h>
fe64d517	16	#include <linux/slow-work.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 DT	29
	30	int gfs2_replay_read_block(struct gfs2_jdesc *jd, unsigned int blk,
	31	struct buffer_head **bh)
	32	{
feaa7bba	33	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
5c676f6d	34	struct gfs2_glock *gl = ip->i_gl;
b3b94faa	35	int new = 0;
cd915493 SW	36	u64 dblock;
cd915493 SW	37	u32 extlen;
b3b94faa DT	38	int error;
b3b94faa DT	39
feaa7bba	40	error = gfs2_extent_map(&ip->i_inode, blk, &new, &dblock, &extlen);
b3b94faa DT	41	if (error)
	42	return error;
	43	if (!dblock) {
5c676f6d	44	gfs2_consist_inode(ip);
b3b94faa DT	45	return -EIO;
	46	}
	47
7276b3b0	48	*bh = gfs2_meta_ra(gl, dblock, extlen);
b3b94faa DT	49
	50	return error;
	51	}
	52
cd915493	53	int gfs2_revoke_add(struct gfs2_sbd *sdp, u64 blkno, unsigned int where)
b3b94faa DT	54	{
	55	struct list_head *head = &sdp->sd_revoke_list;
	56	struct gfs2_revoke_replay *rr;
	57	int found = 0;
	58
	59	list_for_each_entry(rr, head, rr_list) {
	60	if (rr->rr_blkno == blkno) {
	61	found = 1;
	62	break;
	63	}
	64	}
	65
	66	if (found) {
	67	rr->rr_where = where;
	68	return 0;
	69	}
	70
16c5f06f	71	rr = kmalloc(sizeof(struct gfs2_revoke_replay), GFP_NOFS);
b3b94faa DT	72	if (!rr)
	73	return -ENOMEM;
	74
	75	rr->rr_blkno = blkno;
	76	rr->rr_where = where;
	77	list_add(&rr->rr_list, head);
	78
	79	return 1;
	80	}
	81
cd915493	82	int gfs2_revoke_check(struct gfs2_sbd *sdp, u64 blkno, unsigned int where)
b3b94faa DT	83	{
	84	struct gfs2_revoke_replay *rr;
	85	int wrap, a, b, revoke;
	86	int found = 0;
	87
	88	list_for_each_entry(rr, &sdp->sd_revoke_list, rr_list) {
	89	if (rr->rr_blkno == blkno) {
	90	found = 1;
	91	break;
	92	}
	93	}
	94
	95	if (!found)
	96	return 0;
	97
	98	wrap = (rr->rr_where < sdp->sd_replay_tail);
	99	a = (sdp->sd_replay_tail < where);
	100	b = (where < rr->rr_where);
	101	revoke = (wrap) ? (a \|\| b) : (a && b);
	102
	103	return revoke;
	104	}
	105
	106	void gfs2_revoke_clean(struct gfs2_sbd *sdp)
	107	{
	108	struct list_head *head = &sdp->sd_revoke_list;
	109	struct gfs2_revoke_replay *rr;
	110
	111	while (!list_empty(head)) {
	112	rr = list_entry(head->next, struct gfs2_revoke_replay, rr_list);
	113	list_del(&rr->rr_list);
	114	kfree(rr);
	115	}
	116	}
	117
bb8d8a6f SW	118	static int gfs2_log_header_in(struct gfs2_log_header_host lh, const void buf)
	119	{
	120	const struct gfs2_log_header *str = buf;
	121
	122	if (str->lh_header.mh_magic != cpu_to_be32(GFS2_MAGIC) \|\|
	123	str->lh_header.mh_type != cpu_to_be32(GFS2_METATYPE_LH))
	124	return 1;
	125
	126	lh->lh_sequence = be64_to_cpu(str->lh_sequence);
	127	lh->lh_flags = be32_to_cpu(str->lh_flags);
	128	lh->lh_tail = be32_to_cpu(str->lh_tail);
	129	lh->lh_blkno = be32_to_cpu(str->lh_blkno);
	130	lh->lh_hash = be32_to_cpu(str->lh_hash);
	131	return 0;
	132	}
	133
b3b94faa DT	134	/**
	135	* get_log_header - read the log header for a given segment
	136	* @jd: the journal
	137	* @blk: the block to look at
	138	* @lh: the log header to return
	139	*
	140	* Read the log header for a given segement in a given journal. Do a few
	141	* sanity checks on it.
	142	*
	143	* Returns: 0 on success,
	144	* 1 if the header was invalid or incomplete,
	145	* errno on error
	146	*/
	147
	148	static int get_log_header(struct gfs2_jdesc *jd, unsigned int blk,
55167622	149	struct gfs2_log_header_host *head)
b3b94faa DT	150	{
b3b94faa DT	151	struct buffer_head *bh;
f5a8cd02	152	struct gfs2_log_header_host uninitialized_var(lh);
887bc5d0	153	const u32 nothing = 0;
cd915493	154	u32 hash;
b3b94faa DT	155	int error;
	156
	157	error = gfs2_replay_read_block(jd, blk, &bh);
	158	if (error)
	159	return error;
	160
2a2c9824 SW	161	hash = crc32_le((u32)~0, bh->b_data, sizeof(struct gfs2_log_header) -
	162	sizeof(u32));
	163	hash = crc32_le(hash, (unsigned char const *)&nothing, sizeof(nothing));
	164	hash ^= (u32)~0;
bb8d8a6f	165	error = gfs2_log_header_in(&lh, bh->b_data);
b3b94faa DT	166	brelse(bh);
b3b94faa DT	167
bb8d8a6f	168	if (error \|\| lh.lh_blkno != blk \|\| lh.lh_hash != hash)
b3b94faa DT	169	return 1;
	170
	171	*head = lh;
	172
	173	return 0;
	174	}
	175
	176	/**
	177	* find_good_lh - find a good log header
	178	* @jd: the journal
	179	* @blk: the segment to start searching from
	180	* @lh: the log header to fill in
	181	* @forward: if true search forward in the log, else search backward
	182	*
	183	* Call get_log_header() to get a log header for a segment, but if the
	184	* segment is bad, either scan forward or backward until we find a good one.
	185	*
	186	* Returns: errno
	187	*/
	188
	189	static int find_good_lh(struct gfs2_jdesc jd, unsigned int blk,
55167622	190	struct gfs2_log_header_host *head)
b3b94faa DT	191	{
	192	unsigned int orig_blk = *blk;
	193	int error;
	194
	195	for (;;) {
	196	error = get_log_header(jd, *blk, head);
	197	if (error <= 0)
	198	return error;
	199
	200	if (++*blk == jd->jd_blocks)
	201	*blk = 0;
	202
	203	if (*blk == orig_blk) {
feaa7bba	204	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	205	return -EIO;
	206	}
	207	}
	208	}
	209
	210	/**
	211	* jhead_scan - make sure we've found the head of the log
	212	* @jd: the journal
	213	* @head: this is filled in with the log descriptor of the head
	214	*
	215	* At this point, seg and lh should be either the head of the log or just
	216	* before. Scan forward until we find the head.
	217	*
	218	* Returns: errno
	219	*/
	220
55167622	221	static int jhead_scan(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
b3b94faa DT	222	{
b3b94faa DT	223	unsigned int blk = head->lh_blkno;
55167622	224	struct gfs2_log_header_host lh;
b3b94faa DT	225	int error;
	226
	227	for (;;) {
	228	if (++blk == jd->jd_blocks)
	229	blk = 0;
	230
	231	error = get_log_header(jd, blk, &lh);
	232	if (error < 0)
	233	return error;
	234	if (error == 1)
	235	continue;
	236
	237	if (lh.lh_sequence == head->lh_sequence) {
feaa7bba	238	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	239	return -EIO;
	240	}
	241	if (lh.lh_sequence < head->lh_sequence)
	242	break;
	243
	244	*head = lh;
	245	}
	246
	247	return 0;
	248	}
	249
	250	/**
	251	* gfs2_find_jhead - find the head of a log
	252	* @jd: the journal
	253	* @head: the log descriptor for the head of the log is returned here
	254	*
	255	* Do a binary search of a journal and find the valid log entry with the
	256	* highest sequence number. (i.e. the log head)
	257	*
	258	* Returns: errno
	259	*/
	260
55167622	261	int gfs2_find_jhead(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
b3b94faa	262	{
55167622	263	struct gfs2_log_header_host lh_1, lh_m;
cd915493	264	u32 blk_1, blk_2, blk_m;
b3b94faa DT	265	int error;
	266
	267	blk_1 = 0;
	268	blk_2 = jd->jd_blocks - 1;
	269
	270	for (;;) {
	271	blk_m = (blk_1 + blk_2) / 2;
	272
	273	error = find_good_lh(jd, &blk_1, &lh_1);
	274	if (error)
	275	return error;
	276
	277	error = find_good_lh(jd, &blk_m, &lh_m);
	278	if (error)
	279	return error;
	280
	281	if (blk_1 == blk_m \|\| blk_m == blk_2)
	282	break;
	283
	284	if (lh_1.lh_sequence <= lh_m.lh_sequence)
	285	blk_1 = blk_m;
	286	else
	287	blk_2 = blk_m;
	288	}
	289
	290	error = jhead_scan(jd, &lh_1);
	291	if (error)
	292	return error;
	293
	294	*head = lh_1;
	295
	296	return error;
	297	}
	298
	299	/**
	300	* foreach_descriptor - go through the active part of the log
	301	* @jd: the journal
	302	* @start: the first log header in the active region
	303	* @end: the last log header (don't process the contents of this entry))
	304	*
	305	* Call a given function once for every log descriptor in the active
	306	* portion of the log.
	307	*
	308	* Returns: errno
	309	*/
	310
	311	static int foreach_descriptor(struct gfs2_jdesc *jd, unsigned int start,
	312	unsigned int end, int pass)
	313	{
feaa7bba	314	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa DT	315	struct buffer_head *bh;
	316	struct gfs2_log_descriptor *ld;
	317	int error = 0;
	318	u32 length;
	319	__be64 *ptr;
	320	unsigned int offset = sizeof(struct gfs2_log_descriptor);
a67cdbd4 SW	321	offset += sizeof(__be64) - 1;
a67cdbd4 SW	322	offset &= ~(sizeof(__be64) - 1);
b3b94faa DT	323
	324	while (start != end) {
	325	error = gfs2_replay_read_block(jd, start, &bh);
	326	if (error)
	327	return error;
	328	if (gfs2_meta_check(sdp, bh)) {
	329	brelse(bh);
	330	return -EIO;
	331	}
	332	ld = (struct gfs2_log_descriptor *)bh->b_data;
	333	length = be32_to_cpu(ld->ld_length);
	334
e3167ded	335	if (be32_to_cpu(ld->ld_header.mh_type) == GFS2_METATYPE_LH) {
55167622	336	struct gfs2_log_header_host lh;
b3b94faa DT	337	error = get_log_header(jd, start, &lh);
	338	if (!error) {
	339	gfs2_replay_incr_blk(sdp, &start);
88721877	340	brelse(bh);
b3b94faa DT	341	continue;
	342	}
	343	if (error == 1) {
feaa7bba	344	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	345	error = -EIO;
	346	}
	347	brelse(bh);
	348	return error;
	349	} else if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LD)) {
	350	brelse(bh);
	351	return -EIO;
	352	}
	353	ptr = (__be64 *)(bh->b_data + offset);
	354	error = lops_scan_elements(jd, start, ld, ptr, pass);
	355	if (error) {
	356	brelse(bh);
	357	return error;
	358	}
	359
	360	while (length--)
	361	gfs2_replay_incr_blk(sdp, &start);
	362
	363	brelse(bh);
	364	}
	365
	366	return 0;
	367	}
	368
	369	/**
	370	* clean_journal - mark a dirty journal as being clean
	371	* @sdp: the filesystem
	372	* @jd: the journal
	373	* @gl: the journal's glock
	374	* @head: the head journal to start from
	375	*
	376	* Returns: errno
	377	*/
	378
55167622	379	static int clean_journal(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
b3b94faa	380	{
feaa7bba SW	381	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	382	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa	383	unsigned int lblock;
b3b94faa	384	struct gfs2_log_header *lh;
cd915493	385	u32 hash;
b3b94faa DT	386	struct buffer_head *bh;
b3b94faa DT	387	int error;
23591256	388	struct buffer_head bh_map = { .b_state = 0, .b_blocknr = 0 };
fd88de56	389
b3b94faa DT	390	lblock = head->lh_blkno;
b3b94faa DT	391	gfs2_replay_incr_blk(sdp, &lblock);
23591256	392	bh_map.b_size = 1 << ip->i_inode.i_blkbits;
e9e1ef2b	393	error = gfs2_block_map(&ip->i_inode, lblock, &bh_map, 0);
b3b94faa DT	394	if (error)
b3b94faa DT	395	return error;
7a6bbacb	396	if (!bh_map.b_blocknr) {
b3b94faa DT	397	gfs2_consist_inode(ip);
	398	return -EIO;
	399	}
	400
7a6bbacb	401	bh = sb_getblk(sdp->sd_vfs, bh_map.b_blocknr);
b3b94faa DT	402	lock_buffer(bh);
	403	memset(bh->b_data, 0, bh->b_size);
	404	set_buffer_uptodate(bh);
	405	clear_buffer_dirty(bh);
	406	unlock_buffer(bh);
	407
	408	lh = (struct gfs2_log_header *)bh->b_data;
	409	memset(lh, 0, sizeof(struct gfs2_log_header));
	410	lh->lh_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
88721877	411	lh->lh_header.mh_type = cpu_to_be32(GFS2_METATYPE_LH);
e3167ded	412	lh->lh_header.mh_format = cpu_to_be32(GFS2_FORMAT_LH);
b3b94faa DT	413	lh->lh_sequence = cpu_to_be64(head->lh_sequence + 1);
	414	lh->lh_flags = cpu_to_be32(GFS2_LOG_HEAD_UNMOUNT);
	415	lh->lh_blkno = cpu_to_be32(lblock);
	416	hash = gfs2_disk_hash((const char *)lh, sizeof(struct gfs2_log_header));
	417	lh->lh_hash = cpu_to_be32(hash);
	418
	419	set_buffer_dirty(bh);
	420	if (sync_dirty_buffer(bh))
	421	gfs2_io_error_bh(sdp, bh);
	422	brelse(bh);
	423
	424	return error;
	425	}
	426
da755fdb	427
f057f6cd SW	428	static void gfs2_recovery_done(struct gfs2_sbd *sdp, unsigned int jid,
f057f6cd SW	429	unsigned int message)
da755fdb	430	{
f057f6cd SW	431	char env_jid[20];
	432	char env_status[20];
	433	char *envp[] = { env_jid, env_status, NULL };
	434	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
	435	ls->ls_recover_jid_done = jid;
	436	ls->ls_recover_jid_status = message;
	437	sprintf(env_jid, "JID=%d", jid);
	438	sprintf(env_status, "RECOVERY=%s",
	439	message == LM_RD_SUCCESS ? "Done" : "Failed");
	440	kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb SW	441	}
da755fdb SW	442
fe64d517 SW	443	static int gfs2_recover_get_ref(struct slow_work *work)
	444	{
	445	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
	446	if (test_and_set_bit(JDF_RECOVERY, &jd->jd_flags))
	447	return -EBUSY;
	448	return 0;
	449	}
b3b94faa	450
fe64d517 SW	451	static void gfs2_recover_put_ref(struct slow_work *work)
	452	{
	453	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
	454	clear_bit(JDF_RECOVERY, &jd->jd_flags);
	455	smp_mb__after_clear_bit();
	456	wake_up_bit(&jd->jd_flags, JDF_RECOVERY);
	457	}
	458
	459	static void gfs2_recover_work(struct slow_work *work)
b3b94faa	460	{
fe64d517	461	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
feaa7bba SW	462	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	463	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
55167622	464	struct gfs2_log_header_host head;
b3b94faa DT	465	struct gfs2_holder j_gh, ji_gh, t_gh;
	466	unsigned long t;
	467	int ro = 0;
	468	unsigned int pass;
	469	int error;
	470
5965b1f4 SW	471	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
	472	fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
	473	jd->jd_jid);
b3b94faa	474
c78bad11	475	/* Acquire the journal lock so we can do recovery */
b3b94faa	476
5965b1f4 SW	477	error = gfs2_glock_nq_num(sdp, jd->jd_jid, &gfs2_journal_glops,
	478	LM_ST_EXCLUSIVE,
	479	LM_FLAG_NOEXP \| LM_FLAG_TRY \| GL_NOCACHE,
	480	&j_gh);
	481	switch (error) {
	482	case 0:
	483	break;
907b9bce	484
5965b1f4 SW	485	case GLR_TRYFAILED:
	486	fs_info(sdp, "jid=%u: Busy\n", jd->jd_jid);
	487	error = 0;
907b9bce	488
5965b1f4 SW	489	default:
	490	goto fail;
	491	};
b3b94faa	492
5965b1f4	493	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8	494	LM_FLAG_NOEXP \| GL_NOCACHE, &ji_gh);
5965b1f4 SW	495	if (error)
	496	goto fail_gunlock_j;
	497	} else {
	498	fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
	499	}
b3b94faa DT	500
	501	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);
	502
	503	error = gfs2_jdesc_check(jd);
	504	if (error)
	505	goto fail_gunlock_ji;
	506
	507	error = gfs2_find_jhead(jd, &head);
	508	if (error)
	509	goto fail_gunlock_ji;
	510
	511	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
	512	fs_info(sdp, "jid=%u: Acquiring the transaction lock...\n",
	513	jd->jd_jid);
	514
	515	t = jiffies;
	516
	517	/* Acquire a shared hold on the transaction lock */
	518
59a1cc6b	519	error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED,
5965b1f4	520	LM_FLAG_NOEXP \| LM_FLAG_PRIORITY \|
6802e340	521	GL_NOCACHE, &t_gh);
b3b94faa DT	522	if (error)
	523	goto fail_gunlock_ji;
	524
	525	if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
	526	if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
	527	ro = 1;
	528	} else {
7bc5c414 AD	529	if (sdp->sd_vfs->s_flags & MS_RDONLY) {
	530	/* check if device itself is read-only */
	531	ro = bdev_read_only(sdp->sd_vfs->s_bdev);
	532	if (!ro) {
	533	fs_info(sdp, "recovery required on "
	534	"read-only filesystem.\n");
	535	fs_info(sdp, "write access will be "
	536	"enabled during recovery.\n");
	537	}
	538	}
b3b94faa DT	539	}
	540
	541	if (ro) {
7bc5c414 AD	542	fs_warn(sdp, "jid=%u: Can't replay: read-only block "
7bc5c414 AD	543	"device\n", jd->jd_jid);
b3b94faa DT	544	error = -EROFS;
	545	goto fail_gunlock_tr;
	546	}
	547
	548	fs_info(sdp, "jid=%u: Replaying journal...\n", jd->jd_jid);
	549
	550	for (pass = 0; pass < 2; pass++) {
	551	lops_before_scan(jd, &head, pass);
	552	error = foreach_descriptor(jd, head.lh_tail,
	553	head.lh_blkno, pass);
	554	lops_after_scan(jd, error, pass);
	555	if (error)
	556	goto fail_gunlock_tr;
	557	}
	558
	559	error = clean_journal(jd, &head);
	560	if (error)
	561	goto fail_gunlock_tr;
	562
	563	gfs2_glock_dq_uninit(&t_gh);
5c676f6d	564	t = DIV_ROUND_UP(jiffies - t, HZ);
b3b94faa DT	565	fs_info(sdp, "jid=%u: Journal replayed in %lus\n",
	566	jd->jd_jid, t);
	567	}
	568
5965b1f4 SW	569	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid)
5965b1f4 SW	570	gfs2_glock_dq_uninit(&ji_gh);
b3b94faa	571
f057f6cd	572	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);
b3b94faa	573
5965b1f4 SW	574	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid)
5965b1f4 SW	575	gfs2_glock_dq_uninit(&j_gh);
b3b94faa DT	576
b3b94faa DT	577	fs_info(sdp, "jid=%u: Done\n", jd->jd_jid);
fe64d517	578	return;
b3b94faa	579
5965b1f4	580	fail_gunlock_tr:
b3b94faa	581	gfs2_glock_dq_uninit(&t_gh);
5965b1f4 SW	582	fail_gunlock_ji:
	583	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
	584	gfs2_glock_dq_uninit(&ji_gh);
	585	fail_gunlock_j:
	586	gfs2_glock_dq_uninit(&j_gh);
	587	}
b3b94faa DT	588
	589	fs_info(sdp, "jid=%u: %s\n", jd->jd_jid, (error) ? "Failed" : "Done");
	590
5965b1f4	591	fail:
f057f6cd	592	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
b3b94faa DT	593	}
b3b94faa DT	594
fe64d517 SW	595	struct slow_work_ops gfs2_recover_ops = {
	596	.get_ref = gfs2_recover_get_ref,
	597	.put_ref = gfs2_recover_put_ref,
	598	.execute = gfs2_recover_work,
	599	};
2bfb6449	600
2bfb6449	601
fe64d517	602	static int gfs2_recovery_wait(void *word)
b3b94faa	603	{
fe64d517 SW	604	schedule();
fe64d517 SW	605	return 0;
b3b94faa DT	606	}
b3b94faa DT	607
fe64d517	608	int gfs2_recover_journal(struct gfs2_jdesc *jd)
9ac1b4d9	609	{
fe64d517 SW	610	int rv;
	611	rv = slow_work_enqueue(&jd->jd_work);
	612	if (rv)
	613	return rv;
	614	wait_on_bit(&jd->jd_flags, JDF_RECOVERY, gfs2_recovery_wait, TASK_UNINTERRUPTIBLE);
9ac1b4d9 SW	615	return 0;
	616	}
	617