[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / gfs2 / recovery.c

/*
 * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
 * Copyright (C) 2004-2005 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 v.2.
 */

#include <linux/sched.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 "lm_interface.h"
#include "incore.h"
#include "bmap.h"
#include "glock.h"
#include "glops.h"
#include "lm.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 = jd->jd_inode->u.generic_ip;
	struct gfs2_glock *gl = ip->i_gl;
	int new = 0;
	uint64_t dblock;
	uint32_t extlen;
	int error;

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

	gfs2_meta_ra(gl, dblock, extlen);
	error = gfs2_meta_read(gl, dblock, DIO_START | DIO_WAIT, bh);

	return error;
}

int gfs2_revoke_add(struct gfs2_sbd *sdp, uint64_t 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_KERNEL);
	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, uint64_t 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);
	}
}

/**
 * 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 *head)
{
	struct buffer_head *bh;
	struct gfs2_log_header lh;
	uint32_t hash;
	int error;

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

	memcpy(&lh, bh->b_data, sizeof(struct gfs2_log_header));
	lh.lh_hash = 0;
	hash = gfs2_disk_hash((char *)&lh, sizeof(struct gfs2_log_header));
	gfs2_log_header_in(&lh, bh->b_data);

	brelse(bh);

	if (lh.lh_header.mh_magic != GFS2_MAGIC ||
	    lh.lh_header.mh_type != GFS2_METATYPE_LH ||
	    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 *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(jd->jd_inode->u.generic_ip);
			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 *head)
{
	unsigned int blk = head->lh_blkno;
	struct gfs2_log_header 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(jd->jd_inode->u.generic_ip);
			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 *head)
{
	struct gfs2_log_header lh_1, lh_m;
	uint32_t 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_inode *ip = jd->jd_inode->u.generic_ip;
	struct gfs2_sbd *sdp = ip->i_sbd;
	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 lh;
			error = get_log_header(jd, start, &lh);
			if (!error) {
				gfs2_replay_incr_blk(sdp, &start);
				continue;
			}
			if (error == 1) {
				gfs2_consist_inode(jd->jd_inode->u.generic_ip);
				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 *head)
{
	struct gfs2_inode *ip = jd->jd_inode->u.generic_ip;
	struct gfs2_sbd *sdp = ip->i_sbd;
	unsigned int lblock;
	int new = 0;
	uint64_t dblock;
	struct gfs2_log_header *lh;
	uint32_t hash;
	struct buffer_head *bh;
	int error;
	int boundary;

	lblock = head->lh_blkno;
	gfs2_replay_incr_blk(sdp, &lblock);
	error = gfs2_block_map(ip->i_vnode, lblock, &new, &dblock, &boundary);
	if (error)
		return error;
	if (!dblock) {
		gfs2_consist_inode(ip);
		return -EIO;
	}

	bh = sb_getblk(sdp->sd_vfs, dblock);
	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_be16(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;
}

/**
 * gfs2_recover_journal - recovery a given journal
 * @jd: the struct gfs2_jdesc describing the journal
 *
 * Acquire the journal's lock, check to see if the journal is clean, and
 * do recovery if necessary.
 *
 * Returns: errno
 */

int gfs2_recover_journal(struct gfs2_jdesc *jd)
{
	struct gfs2_inode *ip = jd->jd_inode->u.generic_ip;
	struct gfs2_sbd *sdp = ip->i_sbd;
	struct gfs2_log_header 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);

		/* Aquire 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, &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_NOCANCEL | 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)
				ro = 1;
		}

		if (ro) {
			fs_warn(sdp, "jid=%u: Can't replay: read-only FS\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_lm_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 0;

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_lm_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
	return error;
}

/**
 * gfs2_check_journals - Recover any dirty journals
 * @sdp: the filesystem
 *
 */

void gfs2_check_journals(struct gfs2_sbd *sdp)
{
	struct gfs2_jdesc *jd;

	for (;;) {
		jd = gfs2_jdesc_find_dirty(sdp);
		if (!jd)
			break;

		if (jd != sdp->sd_jdesc)
			gfs2_recover_journal(jd);
	}
}
Commit	Line	Data
b3b94faa DT	1	/*
	2	* Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
	3	* Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
	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
	7	* of the GNU General Public License v.2.
	8	*/
	9
	10	#include <linux/sched.h>
	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 SW	19	#include "lm_interface.h"
5c676f6d SW	20	#include "incore.h"
b3b94faa DT	21	#include "bmap.h"
	22	#include "glock.h"
	23	#include "glops.h"
	24	#include "lm.h"
	25	#include "lops.h"
	26	#include "meta_io.h"
	27	#include "recovery.h"
	28	#include "super.h"
5c676f6d	29	#include "util.h"
71b86f56	30	#include "dir.h"
b3b94faa DT	31
	32	int gfs2_replay_read_block(struct gfs2_jdesc *jd, unsigned int blk,
	33	struct buffer_head **bh)
	34	{
5c676f6d SW	35	struct gfs2_inode *ip = jd->jd_inode->u.generic_ip;
5c676f6d SW	36	struct gfs2_glock *gl = ip->i_gl;
b3b94faa DT	37	int new = 0;
	38	uint64_t dblock;
	39	uint32_t extlen;
	40	int error;
	41
fd88de56	42	error = gfs2_extent_map(ip->i_vnode, 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
	50	gfs2_meta_ra(gl, dblock, extlen);
	51	error = gfs2_meta_read(gl, dblock, DIO_START \| DIO_WAIT, bh);
	52
	53	return error;
	54	}
	55
	56	int gfs2_revoke_add(struct gfs2_sbd *sdp, uint64_t blkno, unsigned int where)
	57	{
	58	struct list_head *head = &sdp->sd_revoke_list;
	59	struct gfs2_revoke_replay *rr;
	60	int found = 0;
	61
	62	list_for_each_entry(rr, head, rr_list) {
	63	if (rr->rr_blkno == blkno) {
	64	found = 1;
	65	break;
	66	}
	67	}
	68
	69	if (found) {
	70	rr->rr_where = where;
	71	return 0;
	72	}
	73
	74	rr = kmalloc(sizeof(struct gfs2_revoke_replay), GFP_KERNEL);
	75	if (!rr)
	76	return -ENOMEM;
	77
	78	rr->rr_blkno = blkno;
	79	rr->rr_where = where;
	80	list_add(&rr->rr_list, head);
	81
	82	return 1;
	83	}
	84
	85	int gfs2_revoke_check(struct gfs2_sbd *sdp, uint64_t blkno, unsigned int where)
	86	{
	87	struct gfs2_revoke_replay *rr;
	88	int wrap, a, b, revoke;
	89	int found = 0;
	90
	91	list_for_each_entry(rr, &sdp->sd_revoke_list, rr_list) {
	92	if (rr->rr_blkno == blkno) {
	93	found = 1;
	94	break;
	95	}
	96	}
	97
	98	if (!found)
	99	return 0;
	100
	101	wrap = (rr->rr_where < sdp->sd_replay_tail);
	102	a = (sdp->sd_replay_tail < where);
	103	b = (where < rr->rr_where);
	104	revoke = (wrap) ? (a \|\| b) : (a && b);
	105
	106	return revoke;
	107	}
	108
	109	void gfs2_revoke_clean(struct gfs2_sbd *sdp)
	110	{
111	struct list_head *head = &sdp->sd_revoke_list;
112	struct gfs2_revoke_replay *rr;
113
114	while (!list_empty(head)) {
115	rr = list_entry(head->next, struct gfs2_revoke_replay, rr_list);
116	list_del(&rr->rr_list);
117	kfree(rr);
118	}
119	}
120
121	/**
122	* get_log_header - read the log header for a given segment
123	* @jd: the journal
124	* @blk: the block to look at
125	* @lh: the log header to return
126	*
127	* Read the log header for a given segement in a given journal. Do a few
128	* sanity checks on it.
129	*
130	* Returns: 0 on success,
131	* 1 if the header was invalid or incomplete,
132	* errno on error
133	*/
134
135	static int get_log_header(struct gfs2_jdesc *jd, unsigned int blk,
136	struct gfs2_log_header *head)
137	{
138	struct buffer_head *bh;
139	struct gfs2_log_header lh;
140	uint32_t hash;
141	int error;
142
143	error = gfs2_replay_read_block(jd, blk, &bh);
144	if (error)
145	return error;
146
147	memcpy(&lh, bh->b_data, sizeof(struct gfs2_log_header));
148	lh.lh_hash = 0;
149	hash = gfs2_disk_hash((char *)&lh, sizeof(struct gfs2_log_header));
150	gfs2_log_header_in(&lh, bh->b_data);
151
152	brelse(bh);
153
154	if (lh.lh_header.mh_magic != GFS2_MAGIC \|\|
155	lh.lh_header.mh_type != GFS2_METATYPE_LH \|\|
156	lh.lh_blkno != blk \|\|
157	lh.lh_hash != hash)
158	return 1;
159
160	*head = lh;
161
162	return 0;
163	}
164
165	/**
166	* find_good_lh - find a good log header
167	* @jd: the journal
168	* @blk: the segment to start searching from
169	* @lh: the log header to fill in
170	* @forward: if true search forward in the log, else search backward
171	*
172	* Call get_log_header() to get a log header for a segment, but if the
173	* segment is bad, either scan forward or backward until we find a good one.
174	*
175	* Returns: errno
176	*/
177
178	static int find_good_lh(struct gfs2_jdesc jd, unsigned int blk,
179	struct gfs2_log_header *head)
180	{
181	unsigned int orig_blk = *blk;
182	int error;
183
184	for (;;) {
185	error = get_log_header(jd, *blk, head);
186	if (error <= 0)
187	return error;
188
189	if (++*blk == jd->jd_blocks)
190	*blk = 0;
191
192	if (*blk == orig_blk) {
5c676f6d	193	gfs2_consist_inode(jd->jd_inode->u.generic_ip);
b3b94faa DT	194	return -EIO;
	195	}
	196	}
	197	}
	198
	199	/**
	200	* jhead_scan - make sure we've found the head of the log
	201	* @jd: the journal
	202	* @head: this is filled in with the log descriptor of the head
	203	*
	204	* At this point, seg and lh should be either the head of the log or just
	205	* before. Scan forward until we find the head.
	206	*
	207	* Returns: errno
	208	*/
	209
	210	static int jhead_scan(struct gfs2_jdesc jd, struct gfs2_log_header head)
	211	{
	212	unsigned int blk = head->lh_blkno;
	213	struct gfs2_log_header lh;
	214	int error;
	215
	216	for (;;) {
	217	if (++blk == jd->jd_blocks)
	218	blk = 0;
	219
	220	error = get_log_header(jd, blk, &lh);
	221	if (error < 0)
	222	return error;
	223	if (error == 1)
	224	continue;
	225
	226	if (lh.lh_sequence == head->lh_sequence) {
5c676f6d	227	gfs2_consist_inode(jd->jd_inode->u.generic_ip);
b3b94faa DT	228	return -EIO;
	229	}
	230	if (lh.lh_sequence < head->lh_sequence)
	231	break;
	232
	233	*head = lh;
	234	}
	235
	236	return 0;
	237	}
	238
	239	/**
	240	* gfs2_find_jhead - find the head of a log
	241	* @jd: the journal
	242	* @head: the log descriptor for the head of the log is returned here
	243	*
	244	* Do a binary search of a journal and find the valid log entry with the
	245	* highest sequence number. (i.e. the log head)
	246	*
	247	* Returns: errno
	248	*/
	249
	250	int gfs2_find_jhead(struct gfs2_jdesc jd, struct gfs2_log_header head)
	251	{
	252	struct gfs2_log_header lh_1, lh_m;
	253	uint32_t blk_1, blk_2, blk_m;
	254	int error;
	255
	256	blk_1 = 0;
	257	blk_2 = jd->jd_blocks - 1;
	258
	259	for (;;) {
	260	blk_m = (blk_1 + blk_2) / 2;
	261
	262	error = find_good_lh(jd, &blk_1, &lh_1);
	263	if (error)
	264	return error;
	265
	266	error = find_good_lh(jd, &blk_m, &lh_m);
	267	if (error)
	268	return error;
	269
	270	if (blk_1 == blk_m \|\| blk_m == blk_2)
	271	break;
	272
	273	if (lh_1.lh_sequence <= lh_m.lh_sequence)
	274	blk_1 = blk_m;
	275	else
	276	blk_2 = blk_m;
	277	}
	278
	279	error = jhead_scan(jd, &lh_1);
	280	if (error)
	281	return error;
	282
	283	*head = lh_1;
	284
	285	return error;
	286	}
	287
	288	/**
	289	* foreach_descriptor - go through the active part of the log
	290	* @jd: the journal
	291	* @start: the first log header in the active region
292	* @end: the last log header (don't process the contents of this entry))
293	*
294	* Call a given function once for every log descriptor in the active
295	* portion of the log.
296	*
297	* Returns: errno
298	*/
299
300	static int foreach_descriptor(struct gfs2_jdesc *jd, unsigned int start,
301	unsigned int end, int pass)
302	{
5c676f6d SW	303	struct gfs2_inode *ip = jd->jd_inode->u.generic_ip;
5c676f6d SW	304	struct gfs2_sbd *sdp = ip->i_sbd;
b3b94faa DT	305	struct buffer_head *bh;
	306	struct gfs2_log_descriptor *ld;
	307	int error = 0;
	308	u32 length;
	309	__be64 *ptr;
	310	unsigned int offset = sizeof(struct gfs2_log_descriptor);
	311	offset += (sizeof(__be64)-1);
	312	offset &= ~(sizeof(__be64)-1);
	313
	314	while (start != end) {
	315	error = gfs2_replay_read_block(jd, start, &bh);
	316	if (error)
	317	return error;
	318	if (gfs2_meta_check(sdp, bh)) {
	319	brelse(bh);
	320	return -EIO;
	321	}
	322	ld = (struct gfs2_log_descriptor *)bh->b_data;
	323	length = be32_to_cpu(ld->ld_length);
	324
e3167ded	325	if (be32_to_cpu(ld->ld_header.mh_type) == GFS2_METATYPE_LH) {
b3b94faa DT	326	struct gfs2_log_header lh;
	327	error = get_log_header(jd, start, &lh);
	328	if (!error) {
	329	gfs2_replay_incr_blk(sdp, &start);
	330	continue;
	331	}
	332	if (error == 1) {
5c676f6d	333	gfs2_consist_inode(jd->jd_inode->u.generic_ip);
b3b94faa DT	334	error = -EIO;
	335	}
	336	brelse(bh);
	337	return error;
	338	} else if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LD)) {
	339	brelse(bh);
	340	return -EIO;
	341	}
	342	ptr = (__be64 *)(bh->b_data + offset);
	343	error = lops_scan_elements(jd, start, ld, ptr, pass);
	344	if (error) {
	345	brelse(bh);
	346	return error;
	347	}
	348
	349	while (length--)
	350	gfs2_replay_incr_blk(sdp, &start);
	351
	352	brelse(bh);
	353	}
	354
	355	return 0;
	356	}
	357
	358	/**
	359	* clean_journal - mark a dirty journal as being clean
	360	* @sdp: the filesystem
	361	* @jd: the journal
	362	* @gl: the journal's glock
	363	* @head: the head journal to start from
	364	*
	365	* Returns: errno
	366	*/
	367
	368	static int clean_journal(struct gfs2_jdesc jd, struct gfs2_log_header head)
	369	{
5c676f6d	370	struct gfs2_inode *ip = jd->jd_inode->u.generic_ip;
b3b94faa DT	371	struct gfs2_sbd *sdp = ip->i_sbd;
	372	unsigned int lblock;
	373	int new = 0;
	374	uint64_t dblock;
	375	struct gfs2_log_header *lh;
	376	uint32_t hash;
	377	struct buffer_head *bh;
	378	int error;
fd88de56 SW	379	int boundary;
fd88de56 SW	380
b3b94faa DT	381	lblock = head->lh_blkno;
b3b94faa DT	382	gfs2_replay_incr_blk(sdp, &lblock);
fd88de56	383	error = gfs2_block_map(ip->i_vnode, lblock, &new, &dblock, &boundary);
b3b94faa DT	384	if (error)
	385	return error;
	386	if (!dblock) {
	387	gfs2_consist_inode(ip);
	388	return -EIO;
	389	}
	390
	391	bh = sb_getblk(sdp->sd_vfs, dblock);
	392	lock_buffer(bh);
	393	memset(bh->b_data, 0, bh->b_size);
	394	set_buffer_uptodate(bh);
	395	clear_buffer_dirty(bh);
	396	unlock_buffer(bh);
	397
	398	lh = (struct gfs2_log_header *)bh->b_data;
	399	memset(lh, 0, sizeof(struct gfs2_log_header));
	400	lh->lh_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
	401	lh->lh_header.mh_type = cpu_to_be16(GFS2_METATYPE_LH);
e3167ded	402	lh->lh_header.mh_format = cpu_to_be32(GFS2_FORMAT_LH);
b3b94faa DT	403	lh->lh_sequence = cpu_to_be64(head->lh_sequence + 1);
	404	lh->lh_flags = cpu_to_be32(GFS2_LOG_HEAD_UNMOUNT);
	405	lh->lh_blkno = cpu_to_be32(lblock);
	406	hash = gfs2_disk_hash((const char *)lh, sizeof(struct gfs2_log_header));
	407	lh->lh_hash = cpu_to_be32(hash);
	408
	409	set_buffer_dirty(bh);
	410	if (sync_dirty_buffer(bh))
	411	gfs2_io_error_bh(sdp, bh);
	412	brelse(bh);
	413
	414	return error;
	415	}
	416
	417	/**
	418	* gfs2_recover_journal - recovery a given journal
	419	* @jd: the struct gfs2_jdesc describing the journal
b3b94faa DT	420	*
	421	* Acquire the journal's lock, check to see if the journal is clean, and
	422	* do recovery if necessary.
	423	*
	424	* Returns: errno
	425	*/
	426
c63e31c2	427	int gfs2_recover_journal(struct gfs2_jdesc *jd)
b3b94faa	428	{
5c676f6d SW	429	struct gfs2_inode *ip = jd->jd_inode->u.generic_ip;
5c676f6d SW	430	struct gfs2_sbd *sdp = ip->i_sbd;
b3b94faa DT	431	struct gfs2_log_header head;
	432	struct gfs2_holder j_gh, ji_gh, t_gh;
	433	unsigned long t;
	434	int ro = 0;
	435	unsigned int pass;
	436	int error;
	437
5965b1f4 SW	438	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
	439	fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
	440	jd->jd_jid);
b3b94faa	441
5965b1f4	442	/* Aquire the journal lock so we can do recovery */
b3b94faa	443
5965b1f4 SW	444	error = gfs2_glock_nq_num(sdp, jd->jd_jid, &gfs2_journal_glops,
	445	LM_ST_EXCLUSIVE,
	446	LM_FLAG_NOEXP \| LM_FLAG_TRY \| GL_NOCACHE,
	447	&j_gh);
	448	switch (error) {
	449	case 0:
	450	break;
	451
	452	case GLR_TRYFAILED:
	453	fs_info(sdp, "jid=%u: Busy\n", jd->jd_jid);
	454	error = 0;
	455
	456	default:
	457	goto fail;
	458	};
b3b94faa	459
5965b1f4 SW	460	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
	461	LM_FLAG_NOEXP, &ji_gh);
	462	if (error)
	463	goto fail_gunlock_j;
	464	} else {
	465	fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
	466	}
b3b94faa DT	467
	468	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);
	469
	470	error = gfs2_jdesc_check(jd);
	471	if (error)
	472	goto fail_gunlock_ji;
	473
	474	error = gfs2_find_jhead(jd, &head);
	475	if (error)
	476	goto fail_gunlock_ji;
	477
	478	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
	479	fs_info(sdp, "jid=%u: Acquiring the transaction lock...\n",
	480	jd->jd_jid);
	481
	482	t = jiffies;
	483
	484	/* Acquire a shared hold on the transaction lock */
	485
	486	error = gfs2_glock_nq_init(sdp->sd_trans_gl,
	487	LM_ST_SHARED,
5965b1f4	488	LM_FLAG_NOEXP \| LM_FLAG_PRIORITY \|
579b78a4	489	GL_NOCANCEL \| GL_NOCACHE,
b3b94faa DT	490	&t_gh);
	491	if (error)
	492	goto fail_gunlock_ji;
	493
	494	if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
	495	if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
	496	ro = 1;
	497	} else {
	498	if (sdp->sd_vfs->s_flags & MS_RDONLY)
	499	ro = 1;
	500	}
	501
	502	if (ro) {
	503	fs_warn(sdp, "jid=%u: Can't replay: read-only FS\n",
	504	jd->jd_jid);
	505	error = -EROFS;
	506	goto fail_gunlock_tr;
	507	}
	508
	509	fs_info(sdp, "jid=%u: Replaying journal...\n", jd->jd_jid);
	510
	511	for (pass = 0; pass < 2; pass++) {
	512	lops_before_scan(jd, &head, pass);
	513	error = foreach_descriptor(jd, head.lh_tail,
	514	head.lh_blkno, pass);
	515	lops_after_scan(jd, error, pass);
	516	if (error)
	517	goto fail_gunlock_tr;
	518	}
	519
	520	error = clean_journal(jd, &head);
	521	if (error)
	522	goto fail_gunlock_tr;
	523
	524	gfs2_glock_dq_uninit(&t_gh);
5c676f6d	525	t = DIV_ROUND_UP(jiffies - t, HZ);
b3b94faa DT	526	fs_info(sdp, "jid=%u: Journal replayed in %lus\n",
	527	jd->jd_jid, t);
	528	}
	529
5965b1f4 SW	530	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid)
5965b1f4 SW	531	gfs2_glock_dq_uninit(&ji_gh);
b3b94faa DT	532
	533	gfs2_lm_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);
	534
5965b1f4 SW	535	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid)
5965b1f4 SW	536	gfs2_glock_dq_uninit(&j_gh);
b3b94faa DT	537
b3b94faa DT	538	fs_info(sdp, "jid=%u: Done\n", jd->jd_jid);
b3b94faa DT	539	return 0;
b3b94faa DT	540
5965b1f4	541	fail_gunlock_tr:
b3b94faa	542	gfs2_glock_dq_uninit(&t_gh);
5965b1f4 SW	543	fail_gunlock_ji:
	544	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
	545	gfs2_glock_dq_uninit(&ji_gh);
	546	fail_gunlock_j:
	547	gfs2_glock_dq_uninit(&j_gh);
	548	}
b3b94faa DT	549
	550	fs_info(sdp, "jid=%u: %s\n", jd->jd_jid, (error) ? "Failed" : "Done");
	551
5965b1f4	552	fail:
b3b94faa	553	gfs2_lm_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
b3b94faa DT	554	return error;
	555	}
	556
	557	/**
	558	* gfs2_check_journals - Recover any dirty journals
	559	* @sdp: the filesystem
	560	*
	561	*/
	562
	563	void gfs2_check_journals(struct gfs2_sbd *sdp)
	564	{
	565	struct gfs2_jdesc *jd;
	566
	567	for (;;) {
	568	jd = gfs2_jdesc_find_dirty(sdp);
	569	if (!jd)
	570	break;
	571
	572	if (jd != sdp->sd_jdesc)
c63e31c2	573	gfs2_recover_journal(jd);
b3b94faa DT	574	}
	575	}
	576